CELL_101015

Volume 143

www.cell.com

Number 2

October 15, 2010

Volume 143

www.cell.com

Number 2

October 15, 2010

Five Flavors of Chromatin

ISGylation and Immunity

Five Flavors of Chromatin

ISGylation and Immunity

VVo

lum

e

Leading EdgeCell Volume 143 Number 2, October 15, 2010

IN THIS ISSUE

SELECT

177 GWAS Gets Functional

PREVIEWS

181 Insider Influence on ErbB Activity B.-Z. Shilo

183 Chromatin in Multicolor D. Sch€ubeler

184 The Myc Connection: ES Cells and Cancer M.E. Rothenberg, M.F. Clarke, and M. Diehn

MINIREVIEW

187 Emerging Role of ISG15

in Antiviral Immunity

B. Skaug and Z.J. Chen

PRIMER

191 Biological Applications

of Protein Splicing

M. Vila-Perello and T.W. Muir

SNAPSHOT

326 Network Motifs O. Shoval and U. Alon

ArticlesCell Volume 143 Number 2, October 15, 2010

201 Cytohesins Are Cytoplasmic

ErbB Receptor Activators

A. Bill, A. Schmitz, B. Albertoni, J.-N. Song,

L.C. Heukamp, D. Walrafen, F. Thorwirth, P.J. Verveer,

S. Zimmer, L. Meffert, A. Schreiber, S. Chatterjee,

R.K. Thomas, R.T. Ullrich, T. Lang, and M. Famulok

212 Systematic Protein Location

Mapping Reveals Five Principal

Chromatin Types in Drosophila Cells

G.J. Filion, J.G. van Bemmel, U. Braunschweig,

W. Talhout, J. Kind, L.D. Ward, W. Brugman,

I.J. de Castro, R.M. Kerkhoven, H.J. Bussemaker,

and B. van Steensel

225 The Solution Structure of the ADAR2

dsRBM-RNA Complex Reveals a Sequence-

Specific Readout of the Minor Groove

R. Stefl, F.C. Oberstrass, J.L. Hood, M. Jourdan,

M. Zimmermann, L. Skrisovska, C. Maris, L. Peng,

C. Hofr, R.B. Emeson, and F.H.-T. Allain

238 Exon Junction Complex Subunits Are

Required to Splice Drosophila MAP

Kinase, a Large Heterochromatic Gene

J.-Y. Roignant and J.E. Treisman

251 The Exon Junction Complex Controls

the Splicing of mapk and Other Long

Intron-Containing Transcripts in Drosophila

D. Ashton-Beaucage, C.M. Udell, H. Lavoie,

C. Baril, M. Lefrancois, P. Chagnon, P. Gendron,

O. Caron-Lizotte, E. Bonneil, P. Thibault,

and M. Therrien

263 Patronin Regulates the

Microtubule Network by

Protecting Microtubule Minus Ends

S.S. Goodwin and R.D. Vale

275 Structural Basis for Actin Assembly,

Activation of ATP Hydrolysis,

and Delayed Phosphate Release

K. Murakami, T. Yasunaga, T.Q.P. Noguchi,

Y. Gomibuchi, K.X. Ngo, T.Q.P. Uyeda,

and T. Wakabayashi

288 Nuclear Size Is Regulated

by Importin a and Ntf2 in Xenopus

D.L. Levy and R. Heald

(continued)

299 TGF-b and Insulin Signaling Regulate

Reproductive Aging via Oocyte and

Germline Quality Maintenance

S. Luo, G.A. Kleemann, J.M. Ashraf, W.M. Shaw,

and C.T. Murphy

313 A Myc Network Accounts for Similarities

between Embryonic Stem and Cancer

Cell Transcription Programs

J. Kim, A.J. Woo, J. Chu, J.W. Snow, Y. Fujiwara,

C.G. Kim, A.B. Cantor, and S.H. Orkin

ANNOUNCEMENTS

POSITIONS AVAILABLE

On the cover: Chromatin organization and distribution are important for the regulation of

gene expression. In this issue, Filion et al. (pp. 212–224) report a global view of chromatin

diversity and domain organization in Drosophila by identifying five principal types of chro-

matin. In line with the meaning of the Greek word ‘‘chroma’’ (color), the authors refer to

the chromatin types by the colors GREEN, BLUE, BLACK, YELLOW, and RED (each repre-

sented by colored candy). The cover was designed by U. Braunschweig, W. Talhout, and

J.G. van Bemmel.

Leading Edge

In This Issue

EGF Receptor’s Cytoplasmic ConspiratorPAGE 201

Signaling by ErbB receptors (ErbBRs) requires the activation of their cytoplasmic kinase domains by ligand binding. Now,Bill et al. find that cytohesins contribute to activation. These cytoplasmic proteins bind and promote a rearragement of thedimerized receptor’s intracellular domains. Cytohesins are overexpressed in human lung adenocarcinomas, and a cytohesininhibitor reduces EGFR-dependent lung cancer cell proliferation in mice. Thus these findings establish cytohesins as patho-physiological targets in the ErbBR pathway.

When Splicing Gets Tough, EJC Gets GoingPAGE 251 and PAGE 238

The exon junction complex (EJC) binds to newly spliced transcripts and controlsmRNA surveillance, nuclear export, and translational efficiency. Two relatedstudies now report a role for the Drosophila EJC in the splicing process itself.Studying mutations in a core EJC subunit, Roignant and Treisman show thatthe EJC influences splicing of introns within theMAP kinase gene. Ashton-Beau-cage et al. identify EJC subunits in a screen for factors modulating RAS/MAPKsignaling and show that these subunits influence the splicing and expressionlevels of genes in this pathway. The findings suggest that other genes with char-acteristics similar to the MAP kinase gene—large introns and a heterochromaticlocation—may also rely on the EJC for splicing.

Chromatin Takes FivePAGE 212

Chromatin composition and distribution are important for the regulation of gene expression. By analyzing binding maps of 53proteins, Filion et al. demonstrate that the Drosophila genome is packaged into five principal chromatin types. In addition toHP1- and Polycomb-associated heterochromatin, they characterized two types of transcriptionally active euchromatin regu-lating distinct classes of genes and a repressive chromatin type covering about half the genome and lacking classic hetero-chromatic markers. These results provide a global view of chromatin diversity and domain organization in a metazoan cell.

Reading into dsRNA BindingPAGE 225

Stefl et al. investigate how the correct RNA substrates are recognized for site-specific editing. Structural analysis of the editingenzyme ADAR2 bound to an RNA substrate revealed an unexpected mode of sequence-specific RNA recognition by the twodouble-stranded RNA-binding motifs (dsRBMs) of ADAR2. The dsRBMs make specific contacts with bases in the minorgroove that are important for editing function. The authors suggest that recognition of specific RNA sequences is likelya feature of the other members of the dsRBM family of proteins that are involved in numerous aspects of posttranscriptionalgene regulation.

Peering into Actin PolymersPAGE 275

ATP hydrolysis triggered by actin assembly promotes filament turnover. In thisissue, Murakami et al. present the cryo-electron microscopic structure of filamen-tous actin (F-actin) at a resolution sufficient to visualize some a-helical backbonesand large side chains. The structure indicates that the conserved proline-rich loopadopts a bent conformation as a prerequisite for ATP hydrolysis and that thisconformation triggers a phosphate-release pathway. Combining the cryo-EManalysis with crystal structures of monomeric G-actin mutated in this loop, theauthors propose a molecular mechanism for actin polymerization and associatedATPase activation.

Cell 143, October 15, 2010 ª2010 Elsevier Inc. 173

Patronin Patrols Microtuble DynamicsPAGE 263

Tubulin assembles into polar filaments with dynamic plus and minus ends. However, in living cells, most microtubule minusends are static. Goodwin and Vale identify Patronin as a protein that stabilizes minus ends and protects them from depoly-merization in vivo. They further show that purified Patronin binds selectively to minus ends and shields them from a microtu-bule depolymerase, Kinesin-13. Patronin contributes to proper organization of the microtubule cytoskeleton and formation ofthemitotic spindle, indicating that these structures are regulated by competing actions of destabilizing and stabilizing proteinsacting on minus ends.

Why Is Your Nucleus Bigger Than Mine?PAGE 288

The size of the nucleus varies among different cell types, species, anddisease states, but mechanisms of nuclear size regulation are poorly under-stood. In this issue, Levy and Heald demonstrate that two nuclear importfactors account for differences in nuclear size between two related frogspecies, and that a similar mechanism accounts for changes in nuclearsize during early frog development. These findings provide a context toinvestigate nuclear size regulation in other systems and to elucidate theinterplay between nuclear size and function.

Signaling that Sets the Biological ClockPAGE 299

Women lose reproductive capacity relatively early in their life span asoocytes degrade. C. elegans also stop reproducing midway through their

life span. Luo et al. demonstrate that the Insulin/IGF-1 and Sma/Mab TGF-b signaling pathways determine reproductivespan in C. elegans through the regulation of oocyte quality. Chromosome segregation, cell-cycle, and DNA repair genesare upregulated in TGF-b mutant oocytes and are critical for oocyte quality maintenance. Expression of genes involved inthese processes also declines in aged mammalian oocytes, suggesting conserved mechanisms of oocyte quality mainte-nance.

Myc Bridges ES Cells and CancerPAGE 313

The transcriptional programs of embryonic stem (ES) cells and cancersexhibit similarities, but the basis for this connection has been unclear. Kimet al. report that the ES cell transcription program can be subdivided intothree functionally separable regulatory modules: a Mycmodule, a Polycombmodule, and a ‘‘Core’’ module that is centered on core pluripotency factors.Assessment of cancer gene expression signatures reveals that the Mycmodule, independent of the Coremodule, is active in various cancers. Thesefindings suggest that the Myc regulatory network is primarily responsible forthe similarities in gene expression between ES cells and cancer cells.


Leading Edge

In This Issue

EGF Receptor’s Cytoplasmic ConspiratorPAGE 201

Signaling by ErbB receptors (ErbBRs) requires the activation of their cytoplasmic kinase domains by ligand binding. Now,Bill et al. find that cytohesins contribute to activation. These cytoplasmic proteins bind and promote a rearragement of thedimerized receptor’s intracellular domains. Cytohesins are overexpressed in human lung adenocarcinomas, and a cytohesininhibitor reduces EGFR-dependent lung cancer cell proliferation in mice. Thus these findings establish cytohesins as patho-physiological targets in the ErbBR pathway.

When Splicing Gets Tough, EJC Gets GoingPAGE 251 and PAGE 238

The exon junction complex (EJC) binds to newly spliced transcripts and controlsmRNA surveillance, nuclear export, and translational efficiency. Two relatedstudies now report a role for the Drosophila EJC in the splicing process itself.Studying mutations in a core EJC subunit, Roignant and Treisman show thatthe EJC influences splicing of introns within theMAP kinase gene. Ashton-Beau-cage et al. identify EJC subunits in a screen for factors modulating RAS/MAPKsignaling and show that these subunits influence the splicing and expressionlevels of genes in this pathway. The findings suggest that other genes with char-acteristics similar to the MAP kinase gene—large introns and a heterochromaticlocation—may also rely on the EJC for splicing.

Chromatin Takes FivePAGE 212

Chromatin composition and distribution are important for the regulation of gene expression. By analyzing binding maps of 53proteins, Filion et al. demonstrate that the Drosophila genome is packaged into five principal chromatin types. In addition toHP1- and Polycomb-associated heterochromatin, they characterized two types of transcriptionally active euchromatin regu-lating distinct classes of genes and a repressive chromatin type covering about half the genome and lacking classic hetero-chromatic markers. These results provide a global view of chromatin diversity and domain organization in a metazoan cell.

Reading into dsRNA BindingPAGE 225

Stefl et al. investigate how the correct RNA substrates are recognized for site-specific editing. Structural analysis of the editingenzyme ADAR2 bound to an RNA substrate revealed an unexpected mode of sequence-specific RNA recognition by the twodouble-stranded RNA-binding motifs (dsRBMs) of ADAR2. The dsRBMs make specific contacts with bases in the minorgroove that are important for editing function. The authors suggest that recognition of specific RNA sequences is likelya feature of the other members of the dsRBM family of proteins that are involved in numerous aspects of posttranscriptionalgene regulation.

Peering into Actin PolymersPAGE 275

ATP hydrolysis triggered by actin assembly promotes filament turnover. In thisissue, Murakami et al. present the cryo-electron microscopic structure of filamen-tous actin (F-actin) at a resolution sufficient to visualize some a-helical backbonesand large side chains. The structure indicates that the conserved proline-rich loopadopts a bent conformation as a prerequisite for ATP hydrolysis and that thisconformation triggers a phosphate-release pathway. Combining the cryo-EManalysis with crystal structures of monomeric G-actin mutated in this loop, theauthors propose a molecular mechanism for actin polymerization and associatedATPase activation.


Leading Edge

Previews

Insider Influence on ErbB Activity

Ben-Zion Shilo1,*1Department of Molecular Genetics, Weizmann Institute of Science, Rehovot 76100, Israel

*Correspondence: [email protected]

DOI 10.1016/j.cell.2010.09.042

The receptor tyrosine kinase ErbB is activated by ligand-induced dimerization, leading to trans-

phosphorylation of the cytoplasmic kinase domains. Bill et al. (2010) now demonstrate that trans-

phosphorylation can be modulated from within the cell by the cytoplasmic protein cytohesin,

providing new insights into ErbB-dependent processes during normal development and cancer.

Receptor tyrosine kinases (RTKs) are

a large family of single-pass transmem-

brane receptors that convert extracellular

information, conveyed by ligands, to

the activation of intracellular signaling

cascades. Ligand binding induces

receptor dimerization and leads to trans-

phosphorylation of tyrosines in the

cytoplasmic kinase domains. The phos-

phorylation creates docking sites for SH2

domains, thus recruiting to the activated

receptor complex proteins that trigger

intracellular signal transductioncascades.

A variety of positive and negative regula-

tory interactions affect the signaling

outcome. The mechanisms include ubiq-

uitination and phoshorylation of the

receptor, formation of nonproductive

receptor dimers, trapping of the extracel-

lular ligand, andmodulationof the intracel-

lular cascade. However, until now, no

cytoplasmic components were known to

directly affect the process of ligand-

induced receptor phosphorylation. Bill

et al. (2010) now show that cytohesins,

guanine nucleotide exchange factors

(GEFs), bind to the cytoplasmic domain

of ErbB receptor dimers and facilitate

conformational changes that promote

transphosphorylation and signaling

activity.

These findings add to an increasingly

complex and nuanced understanding of

the mechanisms of receptor tyrosine

kinase activation. It was originally pro-

posed that communication between the

extracellular and intracellular domains of

RTKs is sequential, that is, dimerization

of the extracellular domains, triggered by

ligand binding, leads to dimerization of

the intracellular domains and subsequent

kinase activation (Yarden and Schles-

singer, 1987). However, further work has

since revealed several surprising new

features of RTK dimerization, a process

that has been examined in the greatest

detail for the epidermal growth factor

(EGF) receptor/ErbB family. In the case

of the Drosophila EGF receptor dimer,

binding of the first ligand molecule

induces a conformational change that

reduces the affinity for binding of the

second ligand molecule (Alvarado et al.,

2010). The cytoplasmic juxtamembrane

region also plays a role in activation of

the kinase (Red Brewer et al., 2009; Thiel

and Carpenter, 2007). Ligand binding

relieves an inhibitory association between

the juxtamembrane region and the kinase

domain, facilitating dimerization between

the two juxtamembrane domains that

stabilizes the kinase domain dimer (Jura

et al., 2009). Finally, recent work shows

that kinase domains exist in an autoinhi-

bited state. Activation of the kinase

requires generation of an asymmetric

dimer, where the C-terminal lobe of one

kinase molecule activates the second

kinase domain (Zhang et al., 2006). An

extreme case of dimer asymmetry occurs

in the formation of active heterodimeric

ErbB complexes that comprise one

receptor with an active kinase domain

and one with a catalytically dead kinase

domain (ErbB3).

One question, therefore, is whether and

how cells regulate these additional steps

in the activation of RTKs. A recent study

on Dok-7, an SH2-containing adaptor

protein for the MuSK RTK, suggested

that Dok-7 facilitates MuSK activity by

promoting the juxtaposition of the

two kinase domains, forming a positive

feedback loop. This loop enhances

receptor activation in distinct domains

along the muscle plasma membrane

at neuromuscular junctions (Bergamin

et al., 2010; Inoue et al., 2009).

The paper by Bill et al. (2010) identifies

a new scenario in which cytoplasmic

components impinge on the process of

RTK activation. The work demonstrates

that cytohesins play a critical role of facil-

itating ErbB receptor family activation.

Cytohesin proteins were previously char-

acterized as guanine nucleotide ex-

change factors for ADP ribosylation

factors (ARFs). Interestingly, GEF activity

is dispensable for their role in fa-

cilitating ErbB activation. Bill et al. (2010)

show that the level of cytohesins directly

affects the signaling outcome of ErbB

receptors (Figure 1). Although overex-

pression of cytohesins does not affect

EGF receptor clustering or endocytosis,

it leads to an increase in the phosphoryla-

tion of EGF receptor dimers. Conversely,

inhibition of cytohesin with the specific

inhibitor SecinH3 reduces the phosphory-

lation of dimerized receptors. Further-

more, fluorescence resonance energy

transfer (FRET) studies of EGF receptor

dimers tagged with a fluorescent protein

suggest that the addition of cytohesin

leads to conformational changes in the

cytoplasmic domains. These changes

affect kinase activation, presumably by

facilitating structural changes required

for formation of an asymmetric kinase

dimer.

This study is important because it iden-

tifies a new way for the signal-receiving

cell to modify the RTK signal early in the

signaling process, at the level of receptor

phosphorylation. And as may be ex-

pected, cancer cells point theway topath-

ological abrogation of this circuit. Elevated

EGF receptor/ErbB signaling is character-

istic of many cancers. Bill et al. show that


there is an increase in ErbB transphos-

phorylation in human lung adenocarci-

nomas with elevated levels of cytohesins,

without a corresponding increase in

receptor protein levels. These observa-

tions raise the possibility of attenuating

ErbB activity in tumors by antagonizing

cytohesins. Indeed, Bill et al. show that

addition of SecinH3 reduced the prolifera-

tion of an EGF receptor-dependent lung

cancer cell line. When the same cells

were injected into mice, treatment of the

mice with the inhibitor also resulted in

reduced proliferation of tumor cells.

Similar modulations of RTK activity

from within the cell may also occur during

normal organismal development to adjust

the responsiveness of tissues to a given

RTK signaling pathway. In Drosophila,

for example, immunohistochemical stain-

ing for the activated form of MAP kinase

has shown dramatic differences among

tissues in the range of EGF receptor

signaling activity around a ligand source

(Gabay et al., 1997). In some tissues, the

range of signaling is clearly modulated

by the level of active ligand. However, it

now seems feasible that the local level of

cytohesins or other, yet to be identified

cytoplasmic modulators also determines

the sensitivity of individual tissues to

EGF receptor activation.

REFERENCES

Alvarado, D., Klein, D.E., and Lemmon, M.A.

(2010). Cell 142, 568–579.

Bill, A., Schmitz, A., Albertoni, B., Song, J.-N., Heu-

kamp, L.C., Walrafen, D., Thorwith, F., Verveer,

P.J., Zimmer, S., et al. (2010). Cell 143, this issue,

201–211.

Bergamin, E., Hallock, P.T., Burden, S.J., and Hub-

bard, S.R. (2010). Mol. Cell 39, 100–109.

Gabay, L., Seger, R., and Shilo, B.Z. (1997).

Science 277, 1103–1106.

Inoue, A., Setoguchi, K., Matsubara, Y., Okada, K.,

Sato, N., Iwakura, Y., Higuchi, O., and Yamanashi,

Y. (2009). Sci. Signal. 2, ra7.

Jura, N., Endres, N.F., Engel, K., Deindl, S., Das,

R., Lamers, M.H., Wemmer, D.E., Zhang, X., and

Kuriyan, J. (2009). Cell 137, 1293–1307.

Red Brewer, M., Choi, S.H., Alvarado, D., Morav-

cevic, K., Pozzi, A., Lemmon, M.A., and Carpenter,

G. (2009). Mol. Cell 34, 641–651.

Thiel, K.W., and Carpenter, G. (2007). Proc. Natl.

Acad. Sci. USA 104, 19238–19243.

Yarden, Y., and Schlessinger, J. (1987). Biochem-

istry 26, 1434–1442.

Zhang, X., Gureasko, J., Shen, K., Cole, P.A., and

Kuriyan, J. (2006). Cell 125, 1137–1149.

Figure 1. Cytohesin Levels Modulate ErbB Dimer Phosphorylation(A) Upon ligand binding, ErbB receptor tyrosine kinases form dimers. In order to activate the kinase

domain and trigger transphosphorylation, conformational changes that induce formation of asymmetric

dimersmust take place. Direct binding of cytohesin to the cytoplasmic domain facilitates these conforma-

tional changes.

(B) Overexpression of cytohesins, which occurs in some lung adenocarcinomas, elevates EGF receptor

phosphorylation.

(C) Reduction in cytohesin levels, for example following treatment with the drug SecinH3 or cytohesin RNA

interference, leads to a decrease in receptor phosphorylation.

182 Cell 143, October 15, 2010 ª2010 Elsevier Inc.

Leading Edge

Previews

Chromatin in Multicolor

Dirk Schubeler1,*1Friedrich Miescher Institute for Biomedical Research, 4058 Basel, Switzerland


DOI 10.1016/j.cell.2010.09.045

Chromatin consists of DNA and a large number of associated proteins. Filion et al. (2010) provide

a genome-wide analysis of the location of 53 chromatin proteins in Drosophila, revealing important

principles underlying chromatin regulation and providing colorful insights into their organization.

Chromatin is the complex of DNA and

protein that makes up eukaryotic chromo-

somes. The word is derived from the

Greek word for color (chroma) because

the nuclear material can be visualized

through staining and detected by micros-

copy. Emil Heitz was the first to describe

that chromatin comes in different forms

when he noticed that part of the chromo-

somal material of moss stayed con-

densed throughout the cell cycle (Heitz,

1928). He named the condensed part

heterochromatin and named the part

that decondensed in the interphase

nucleus euchromatin. This binary division

has since dominated the discussion of

chromatin states even though the large

number of protein constituents and mul-

tiple chemical modifications suggest

greater combinatorial complexity (Felsen-

feld and Groudine, 2003). Chromatin con-

sists of nucleosomal histones and a large

number of other proteins, some of which

show specificity for certain histone modi-

fications. With a few exceptions, such as

the well-characterized heterochromatin

protein 1 (HP1) (Eissenberg et al., 1990),

the binding characteristics and chromo-

somal location of most nonhistone chro-

matin proteins remains unknown. In a

tour de force, Filion et al. (2010) now

determine the genomic location of 53

such proteins in the Drosophila genome.

Although the number of chromatin sub-

types revealed by the analysis is surpris-

ingly small, their findings suggest that it

is time to rethink the classical binary divi-

sion of chromatin into euchromatin and

heterochromatin.

To determine the genomic location of

each of the 53 proteins, the authors

employ the ‘‘DamID’’ method, which they

previously developed (van Steensel

et al., 2001). The method entails fusing

each protein to a bacterial DNA adenine

methyltransferase and expressing it in a

cultured Drosophila cell line. Local DNA

methyltransferase activity is then used as

an indicator of protein binding.

The result of these efforts is a large data

set, revealing the genomic locations of 53

chromatin proteins. This in itself is a useful

resource for the research community, as

the local binding preferences and co-

occurrence with other proteins generate

testable hypotheses on the function and

recruitment of each chromatin protein.

However, Filion et al. go further and ask

whether meta-analysis of the binding

data can reveal different classes of chro-

matin. To identify groups of proteins that

tend to colocalize, they perform principal

component analysis (PCA), a computa-

tional method that reduces the dimen-

sionality of multivariate data in order to

identify uncorrelated variables called prin-

cipal components. The authors show that

three principal components are neces-

sary and sufficient to identify five distinct

states of chromatin, to which the authors

assign colors (BLACK, GREEN, BLUE,

RED, and YELLOW). The identification of

five chromatin types suggests non-

random localization of at least a subset

of proteins. This idea is not too surprising:

some form of regularity is expected, given

that the overall process of chromatin

structure assembly is nonrandom, and

many chromatin processes such as tran-

scription or replication initiation entail

sets of specialized proteins. Importantly,

however, many proteins are present in

several chromatin types, and it is thus

their unique combination rather than

exclusive binding that characterizes

each chromatin state, which in turn sug-

gests a flexible, not rigid, organization of

chromatin.

Of the five types of chromatin, BLACK

regions are most prevalent, encompass-

ing close to 50% of the genome, and

are enriched in inactive genes. Surpris-

ingly, however, proteins associated with

BLACK regions are not known to mediate

chromatin repression, suggesting that

either the absence of activators is suffi-

cient to ensure the off-state of a gene or

yet to be identified repressive pathways

are at work. The latter possibility is sup-

ported by the authors’ observation that

transgenes inserted into regions of

BLACK chromatin are frequently silenced.

Proteins corresponding to the known

pathways of gene repression, including

the Polycomb and HP1 pathways, are

found in BLUE and GREEN chromatin.

YELLOWand RED regions both contain

active genes but differ in several ways.

YELLOW regions harbor histone H3 lysine

trimethylation (H3K36me3), a chromatin

mark specific to transcriptional elonga-

tion. RED regions do not exhibit this

mark even though they contain many

regulatory factors, including several DNA-

binding proteins. RED furthermore har-

bors more developmental genes than

YELLOW, raising the possibility that dis-

tinct forms of gene regulation account

for the observed chromatin states. Among

the 53 chromatin proteins analyzed in this

paper are five DNA-binding factors, GAF,

CTCF, JRA, MNT, and SU(HW). With the

exception of SU(HW), all of them show

preferential binding in RED chromatin

even though their binding sites occur

throughout the genome. This finding sug-

gests a role for RED chromatin in directing

these factors to a subset of binding sites.

Intriguingly, this preferential binding does

not simply reflect greater DNA accessi-

bility in RED chromatin, as the authors

show that an unrelated transcription

factor, GAL4 from budding yeast, can

find its correct binding motif in any type

of chromatin. An alternative explanation


for the specificity of the Drosophila pro-

teins could be interactions with other

chromatin components that are also en-

riched in RED chromatin.

Follow-up experiments will improve our

understanding of these intriguing new

colors of chromatin and their interplay

with DNA-binding factors. Combined

with data on other epigenomic variables

such as replication initiation (Gilbert,

2001), repair (Groth et al., 2007), nucleo-

somal turnover (Henikoff, 2008), and

three-dimensional genome organization

(Cockell and Gasser, 1999), these results

will lead to a more comprehensive picture

of chromatin architecture and function.

Clearly, it is time to say goodbye to the

black and white world of heterochromatin

and euchromatin.

REFERENCES

Cockell, M., and Gasser, S.M. (1999). Curr. Opin.

Genet. Dev. 9, 199–205.

Eissenberg, J.C., James, T.C., Foster-Hartnett,

D.M., Hartnett, T., Ngan, V., and Elgin, S.C.

(1990). Proc. Natl. Acad. Sci. USA 87, 9923–9927.

Felsenfeld, G., and Groudine, M. (2003). Nature

421, 448–453.

Filion, G.J., van Bemmel, J.G., Braunschweig, U.,

Talhout, W., Kind, J., Ward, L.D., Brugman, W.,

de Castro, I.J., Kerkhoven, R.M., Bussemaker,

H.J., et al. (2010). Cell 143, this issue, 212–224.

Gilbert, D.M. (2001). Science 294, 96–100.

Groth, A., Rocha, W., Verreault, A., and Almouzni,

G. (2007). Cell 128, 721–733.

Heitz, E. (1928). Jahrb Wiss Botanik 69, 762–818.

Henikoff, S. (2008). Nat. Rev. Genet. 9, 15–26.

van Steensel, B., Delrow, J., and Henikoff, S.

(2001). Nat. Genet. 27, 304–308.

The Myc Connection: ES Cells and Cancer

Michael E. Rothenberg,1,4 Michael F. Clarke,2,4 and Maximilian Diehn3,4,*1Division of Gastroenterology and Hepatology, Department of Medicine2Division of Oncology, Department of Medicine3Department of Radiation Oncology4Cancer Center and Institute for Stem Cell Biology and Regenerative Medicine

Stanford University School of Medicine, Palo Alto, CA 94305, USA


DOI 10.1016/j.cell.2010.09.046

Gene profiling experiments have revealed similarities between cancer and embryonic stem (ES)

cells. Kim et al. (2010) dissect the gene expression signature of ES cells into three functional

modules and find that the Myc module, including genes targeted by Myc-interacting proteins,

accounts for most of the similarity between ES and cancer cells.

Modern techniques in stem cell biology in

the postgenomic era have led to dramatic

advances in our understanding of the

molecular underpinnings of both embry-

onic stem (ES) cells and cancer. Several

essential ‘‘core’’ pluripotency genes regu-

lating the ES cell fate (including Oct4,

Sox2, and Nanog) have been defined in

both mice and humans, and biologists

are now using gene expression profiling

experiments to discover genome-wide

‘‘signatures’’ for ES and cancer cells.

Intriguing similarities between ES cells

and cancer have arisen in such experi-

ments, suggesting that cancers and ES

cells may share fundamental mechanisms

for self-renewal and differentiation (Ben-

Porath et al., 2008; Somervaille et al.,

2009; Wong et al., 2008). On the other

hand, the similarity in gene expression

between some cancers and ES cells has

beenpuzzlingbecause a core ‘‘stemness’’

signature that is shared between ES cells

and other tissue stem cells has remained

elusive (Fortunel et al., 2003). In addition,

most human tumors do not exhibit true

pluripotency. So, how can we explain the

similarities in gene expression patterns

between ES and cancer cells?

In this issue of Cell, Kim et al. address

this question by carefully scrutinizing the

ES cell signature and breaking it down

into several functional units. Using this

approach, the authors show that the

connections between ES cells and cancer

are largely due to Myc, the well-studied

proto-oncogene that regulates many

aspects of gene expression, proliferation,

and differentiation in adult tissues (Kim

et al., 2010).

Using a powerful, highly stringent, and

innovative in vivo biotinylation technique

to probe protein-protein and protein-

DNA interactions (Kim et al., 2009), the

authors begin by defining a Myc-centered

protein interaction network in mouse ES

cells. They show that this Myc complex

likely interacts with the NuA4 histone

acetyltransferase (HAT) complex, a highly

conserved protein complex involved in

diverse functions, including histone acet-

ylation. This suggests an important role

for Myc in epigenetic regulation in ES

cells. The authors then use chromatin

immunoprecipitation (ChIP) to define

the transcriptional targets of this Myc

complex. Myc targets with the most


Myc-associated factors bound to their

regulatory regions are positively associ-

ated with epigenetic marks of active chro-

matin—H3 and H4 histone acetylation

and H3K4 trimethylation—consistent

with their data suggesting a connection

between Myc and epigenetic regulation.

Kim et al. then use this Myc complex

ChIP data set and other previously pub-

lished ChIP experiments to obtain a more

complete characterization of the targets

of important transcription factors in ES

cells. From there, they define three sepa-

rate target gene modules based on factor

co-occupancy in the regulatory regions of

those target genes (Figure 1). Together,

these modules constitute the ES gene

expression signature: a Polycomb cluster

(genes bound by the Polycomb complex

factors), a Core cluster (genes targeted

by the core pluripotency factors Oct4,

Sox2, and Nanog), and a Myc cluster

(genes targeted by the Myc-interacting

proteins). These modules appear to be

functionally significant, as they behave

independently in different scenarios, such

as during ES cell differentiation. Although

previous studies had suggested that the

Myc pathway is a major component of

the link between ES cells and some

cancers (Wong et al., 2008), it remained

unclear whether Myc activates funda-

mental core ES cell programs such as

pluripotency and self-renewal in both

contexts or whether the Myc pathway is

coincidentally utilized for other reasons

by both ES cells and some cancers. The

current study by Kim et al. clarifies this

point and suggests the latter to be the

case.

After defining these three separate

gene expression submodules, the authors

analyze gene expression data from sev-

eral different cancers in bothmice and hu-

mans to obtain a more precise under-

standing of how the ES cell signature

relates to gene expression changes in

cancer. This analysis shows that the

Myc module is highly expressed and

dominant in multiple scenarios: Myc-

transformed human epithelial cancers,

several mouse myeloid leukemias, some

human bladder cancers, and some

human breast cancers. Of interest, the

Core ES cell module is not significantly

expressed in these situations. Thus, in

the end,Myc–rather than the core pluripo-

tency factors or the Polycomb proteins—

seems to be the common thread that ties

ES cells to cancer. But what is Myc’s

precise role in ES cells and these cancers,

particularly as it relates to self-renewal,

a hallmark of stem cells? Is it inhibiting

differentiation (Prochownik and Kukow-

ska, 1986), regulating apoptosis, con-

trolling proliferation, or performing some

other function or some combination of

functions?

Although Myc may affect self-renewal

capacity in ES cells and cancer, it may

not be a central player in this process.

For example, although Myc can increase

the efficiency of the generation of induced

pluripotent stem (iPS) cells, it is not strictly

required for reprogramming (Jaenisch

and Young, 2008). In agreement with this

finding, Kim et al. convincingly demon-

strate that theMycmodule is independent

of the core pluripotency module in ES and

iPS cells. Similarly, they show that, in the

normal mouse hematopoietic system,

the Myc module appears to segregate

away from the property of long-term

self-renewal. Specifically, the Myc mod-

ule is upregulated in highly proliferative

short-term hematopoietic stem cells

(bearing the marker profile Lin cKit+

Sca1+CD34+)—which are more akin to

progenitors, given that they lack sus-

tained self-renewal—and not in themostly

quiescent long-term self-renewing hema-

topoietic stem cells (Lin-cKit+Sca1+

CD34 ), which do not exhibit Myc module

expression. Thus, in the hematopoietic

lineage, the proliferating progenitor is

actually the cell that upregulates Myc

targets rather than the self-renewing

stem cell. This suggests that the presence

Figure 1. Components of the ES Cell SignatureKim et al. (2010) analyze the regulatory regions of target genes for transcription factor co-occupancy.

By analyzing chromatin immunoprecipitation (ChIP) data from their own experiments on the Myc protein

complex in embryonic stem (ES) cells and other published ChIP experiments on different transcription

factors in ES cells, they separate the ES cell transcriptional signature (far left) into three distinct modules

(indicated in the gray box): the Myc module (green), the Polycomb module (blue), and the Core module

(red). The authors then analyze the expression levels of these modules in various scenarios. High expres-

sion of the Myc module (green bars) is a shared property of ES cells, induced pluripotent stem (iPS) cells,

short-term hematopoietic stem cells (ST-HSCs), and various cancers. However, long-term hematopoietic

stem cells (LT-HSCs) and differentiated ES cells exhibit low Myc module expression. Of note, the Core

(red) module—those genes targeted by Oct4, Sox2, and Nanog—is only predominant in ES and iPS cells.


of the Myc module in gene expression

signatures from ES cell populations and

poor prognosis cancers may be more of

a reflection of the active proliferation

occurring in both rather than self-renewal.

In cancer, Myc’s relationship to self-

renewal is complex (Arvanitis and Felsher,

2006), and sometimes Myc expression

maycorrelatewith self-renewal. For exam-

ple, the authors show that, in various

mouse models of acute myelogenous

leukemia (AML), the activity of the Myc

module trends with the frequency of

self-renewing leukemia-initiating cells.

Although these findings could potentially

be explained by the closer resemblance

of the leukemia-initiating cells in AML to

progenitors rather than hematopoietic

stem cells (Majeti et al., 2007), further

work on Myc is needed to decipher its

precise role(s) in the regulation of prolifera-

tion, apoptosis, or differentiation in various

stem cell settings, including ES cells, iPS

cells, adult stem cells, and cancer cells.

Ultimately, by focusing on factor co-

occupancy of target genes in ES cells

and thereby taking a modular look at

gene expression in ES cells and cancer,

this paper helps us to understand the

basis for their similarities. It illustrates the

important role of Myc and will likely spur

cancer biologists to further clarify the

precise role of Myc in tumor biology, a

question with potential therapeutic ramifi-

cations (Arvanitis and Felsher, 2006).

REFERENCES

Arvanitis, C., and Felsher, D.W. (2006). Semin.

Cancer Biol. 16, 313–317.

Ben-Porath, I., Thomson, M.W., Carey, V.J., Ge,

R., Bell, G.W., Regev, A., and Weinberg, R.A.

(2008). Nat. Genet. 40, 499–507.

Fortunel, N.O., Otu, H.H., Ng, H.H., Chen, J., Mu,

X., Chevassut, T., Li, X., Joseph, M., Bailey, C.,

Hatzfeld, J.A., et al. (2003). Science 302, 393.

Jaenisch, R., and Young, R. (2008). Cell 132,

567–582.

Kim, J., Cantor, A.B., Orkin, S.H., and Wang, J.

(2009). Nat. Protoc. 4, 506–517.

Kim, J., Woo, A.J., Chu, J., Snow, J.W., Fujiwara,

Y., Kim, C.G., Cantor, A.B., and Orkin, S.H.

(2010). Cell 143, this issue, 313–324.

Majeti, R., Park, C.Y., and Weissman, I.L. (2007).

Cell Stem Cell 1, 635–645.

Prochownik, E.V., and Kukowska, J. (1986). Nature

322, 848–850.

Somervaille, T.C., Matheny, C.J., Spencer, G.J.,

Iwasaki, M., Rinn, J.L., Witten, D.M., Chang, H.Y.,

Shurtleff, S.A., Downing, J.R., and Cleary, M.L.

(2009). Cell Stem Cell 4, 129–140.

Wong, D.J., Liu, H., Ridky, T.W., Cassarino, D.,

Segal, E., and Chang, H.Y. (2008). Cell Stem Cell

2, 333–344.


Leading Edge

Minireview

Emerging Role of ISG15

in Antiviral Immunity

Brian Skaug1 and Zhijian J. Chen1,2,*1Department of Molecular Biology2Howard Hughes Medical Institute

University of Texas Southwestern Medical Center, Dallas, TX 75390-9148, USA


DOI 10.1016/j.cell.2010.09.033

Cells express a plethora of interferon-stimulated genes (ISGs) in response to viral infection. Among

these is ISG15, a ubiquitin-like protein (UBL) that can be covalently attached to both host and viral

proteins. Here we review recent advances toward understanding the role and mechanism of ISG15

modification in antiviral defense.

Introduction

Secretion of type I interferons (IFNs) from virus-infected cells

is a hallmark of antiviral immunity. Cells that receive these sig-

nals increase expression of interferon-stimulated genes (ISGs),

preparing the cells for impending infection. ISG15, a 15 kDa

ubiquitin-like protein (UBL), has recently emerged as an impor-

tant tool in the struggle against many viral pathogens (reviewed

by Jeon et al., 2010). The ISG15 structure consists of two ubiq-

uitin-likemoieties linked by a short hinge. Like ubiquitin and other

UBLs, ISG15 is attached to target proteins through a C-terminal

Gly-Gly motif. Conjugation of ISG15, commonly referred to as

ISGylation, is a three-step enzymatic cascade (Figure 1A).

The ISG15 E1 enzyme is UBE1L, which specifically activates

ISG15 but not ubiquitin, and the E2 enzyme is UBCH8.

The predominant E3 enzyme appears to be the HECT domain

protein HERC5 because RNA interference against HERC5 abol-

ishes most IFN-induced ISGylation. In addition, coexpression of

UBE1L, UBCH8, HERC5, and ISG15 is sufficient to produce

a level of ISGylation similar to that of IFN stimulation. However,

biochemical evidence that HERC5 directly transfers ISG15 to

substrates is still lacking. Like other UBLs, addition of ISG15 is

reversible; indeed, UBP43 was identified as a deISGylation

enzyme. Notably, expression of UBE1L, UBCH8, HERC5, and

UBP43 is also induced by IFN.

The function of ISG15 since its discovery in the 1980s

remained enigmatic until very recently. Over the past few years,

significant advances have led to a clearer understanding of the

physiological function of ISG15 and several potential antiviral

mechanisms.

Genetic Evidence Linking ISG15 and Antiviral Immunity

The robust induction of ISG15 in response to IFN treatment or

viral infection implies a role for ISG15 in antiviral defense, yet

initial analyses of mice lacking ISG15 or UBE1L revealed no

apparent defect in defense against vesicular stomatitis virus

(VSV) and lymphocytic choriomeningitis virus (LCMV) (Kim

et al., 2006; Osiak et al., 2005). Nevertheless, a growing body

of work strongly suggests a role for ISG15 in defense against

many viral pathogens. ISG15 overexpression in cell culture has

broad antiviral effects, such as suppressing the replication of

HIV and the budding of Ebola VP40 virus-like particles. Also

consistent with a role for ISG15 in antiviral defense, several

viruses express proteins that antagonize the ISGylation machi-

nery (reviewed by Jeon et al., 2010). Here we focus primarily

on recent results from mouse models of viral infection and the

interaction between the influenza B nonstructural protein 1

(NS1B) and the ISGylation machinery.

Functional Insight from Mouse Models of Viral Infection

Strong evidence that ISG15 protects mammals from viral infec-

tion came from studies using a recombinant chimeric Sindbis

virus system (Lenschow et al., 2005). Exogenous expression of

ISG15 inmice lacking the IFN-a and -b receptors confers protec-

tion against systemic infection and lethality. Importantly, muta-

tion of the two C-terminal glycine residues of ISG15 to alanines

(GG > AA) abrogates this protective effect, suggesting that

ISG15 conjugation is important for protection against Sindbis

virus. In addition, mice lacking ISG15 succumb more readily

than wild-type mice to infection with several viruses, including

Sindbis virus, influenza A and B viruses, herpes simplex virus

type 1 (HSV-1), and murine gammaherpesvirus 68 (gHV68). The

impaired defense against Sindbis virus is rescued in ISG15

knockout mice by expressing wild-type ISG15, but not the GG >

AA mutant (Lenschow et al., 2007). Consistent with a critical role

of ISG15conjugation in antiviral defense,mice lackingUBE1Lare

susceptible to infectionwith Sindbis virus, andmutation of ISG15

Arg151, a residue critical for interaction with UBE1L, abrogates

the protective effect of ISG15 (Giannakopoulos et al., 2009).

UBE1L-deficient mice are also susceptible to infection with influ-

enza B (Lai et al., 2009). Taken together, these results implicate

ISG15 conjugation as a key component of mammalian antiviral

immunity. Interestingly, bone marrow transplantation experi-

ments show that ISG15 exerts its antiviral function exclusively

in cells of nonhematopoetic origin (Lai et al., 2009).

Species Specificity in the ISGylation System

Two reports this year have introduced the intriguing prospect

of species specificity in the ISG15 system, including key differ-

ences between mice and humans (Sridharan et al., 2010; Ver-

steeg et al., 2010). The influenza NS1B protein can antagonize


host cell ISGylation, one of the earliest indications that the ISGy-

lation system might be antiviral (Yuan and Krug, 2001). Indeed,

NS1B can bind directly to ISG15 (Chang et al., 2008). However,

as mentioned above, ISG15- and UBE1L-deficient mice are

more susceptible to influenza B than their wild-type counter-

parts. This finding suggests that in wild-type mice NS1B fails

to protect the virus from ISGylation. A potential explanation for

this finding has recently been uncovered; NS1B cannot bind to

mouse ISG15. The binding of NS1B to human ISG15 involves

residues within the N terminus and the short hinge region of

ISG15. The five residues in this hinge region are highly conserved

among primates but divergent in other mammalian species

including mouse and dog. Indeed, NS1B can only bind to

ISG15 from humans and nonhuman primates. Remarkably,

substitution of residues from the human hinge region with the

corresponding mouse residues abolishes this binding (Sridharan

et al., 2010). Consistent with the species selectivity of the NS1B-

ISG15 interaction, NS1B cannot antagonize mouse ISGylation

(Versteeg et al., 2010). Substitution of the N terminus of mouse

ISG15 with the human N terminus restores the NS1B-ISG15

interaction. This report also reveals that HERC6 is the apparent

E3 protein in mice, whereas mouse HERC5 does not support

ISGylation. These findings warrant careful attention in studies

utilizing mice or mouse cells to study the role, and mechanism

(s) of action, of ISG15. It will be of interest to determine the

extent to which the species specificity of ISG15 and ISGylation

machinery contributes to the different responses among mam-

mals to viral infection.

Biochemical Mechanisms of Antiviral Defense by ISG15

Proteomics studies have identified more than 150 proteins as

putative ISGylation targets, a few of which have been validated

under conditions of endogenous expression (Zhao et al., 2005).

Notably, several of the ISGylation substrates identified are

themselves IFN-induced proteins, such as MxA (myxovirus

resistance A) and RIG-I (which senses viral RNA). However,

even for proteins whose ISGylation can be confirmed, it has

been difficult to determine whether this modification exerts

a functional consequence, in part because only a very small frac-

tion of any cellular protein is modified by ISG15. In principle,

ISGylation could lead to a gain of function, loss of function, or

dominant-negative effect. A gain of function or dominant-nega-

tive effect could allow a small fraction of ISGylated proteins to

exert a strong effect. On the other hand, a loss of function of

a small fraction of proteins is unlikely to have a functional conse-

quence, unless ISGylation occurs preferentially on an ‘‘active’’

pool of proteins. In some cases studied so far, ISGylation

appears to impair the function of target proteins. For example,

ISGylation of filamin B impairs its ability to support IFN-induced

Jun N-terminal kinase (JNK) activity and apoptosis (Jeon et al.,

2009).

There are at least two examples in which ISGylation results in

a gain of function of a cellular target protein. 4EHP binds to the

cap structure of mRNA and inhibits translation by competing

with the translation initiation factor eIF4E. ISGylated 4EHP binds

to the mRNA cap with greater affinity than the unmodified

protein. It has been postulated that ISGylation of 4EHP leads

to selective inhibition of viral RNA translation, which may partly

account for the inhibition of viral protein synthesis by IFN

(Okumura et al., 2007).

A recent study has uncovered a role for ISGylation by HERC5

in the regulation of IRF3, a transcription factor that controls the

production of IFN (Shi et al., 2010). HERC5 interacts with IRF3

and promotes its ISGylation. This ISGylation stabilizes IRF3

by inhibiting its interaction with PIN1, a protein that promotes

IRF3 ubiquitination and degradation. Consistent with a gain-of-

function mechanism, HERC5 promotes expression of IRF3-

dependent genes during viral infection and attenuates replica-

tion of several viruses, including VSV.

In addition to cellular ISGylation targets, recent reports

implicate viral proteins as targets of ISG15 modification. These

Figure 1. ISGylation and Its Antiviral Mechanisms(A) ISG15, like ubiquitin, is attached to substrates in a three-step enzymatic

cascade. In the first step, ISG15 is ‘‘activated’’ by UBE1L in an ATP-dependent

process. ISG15 is then transferred to the E2 UBCH8 and subsequently to

a target protein through the E3 HERC5. Like ubiquitin, ISG15 is conjugated

to a lysine on the target protein through a C-terminal glycine-glycine motif.

(B) Type I interferons (IFNs) induce expression of ISG15 and ISGylation

machinery including HERC5. During infection with influenza A, nonstructural

protein 1 (NS1A) protein is ISGylated on lysine 41. ISGylation inhibits the

binding of NS1A to the nuclear import factor importin-a. Mutation of this lysine

largely protects influenza A from the antiviral actions of type I IFN.

(C) HERC5, likely due to its association with ribosomes, broadly targets newly

synthesized proteins for ISGylation. ISGylation of certain viral proteins,

including those that make up the capsid, could have a dominant-negative

effect by interfering with the precise assembly of higher-order structures.

Thus ISG15 can cause a significant impairment in viral infectivity despite

ISGylation of only a small percentage of the target proteins.


studies provide fresh insights into the antiviral mechanisms of

ISG15.

Specific Targeting of Influenza A NS1 Protein

To determine whether targeting of any viral proteins is involved in

ISG15-mediated impairment of influenza A replication, Krug and

colleagues coexpressed influenza A proteins with the ISGylation

machinery and found that the NS1 protein of the H3N2 influenza

A/Udorn/72 (Ud) virus is an ISG15 substrate (Zhao et al., 2010).

ISGylation of NS1A could also be observed following infection

of IFN-b-treated cells with Ud virus. Moreover, NS1A binds

specifically to HERC5 but not the closely related HERC4 and

HERC6. Similarly, Wang and colleagues find that NS1A interacts

with HERC5, and that HERC5 promotes its ISGylation (Tang

et al., 2010). NS1A is a virulence factor that can inhibit host cell

pre-mRNA processing and the IFN-induced 20 to 50 oligo(A)

synthetase/RNase L pathway. Importantly, both groups find

evidence that ISGylation of NS1A impairs influenza replication,

although different conclusions were reached regarding the

mechanism(s) of this impairment.

Through a combination of affinity purification, mass spectrom-

etry, and mutagenesis, Krug and colleagues find that NS1A

Lys41 appears to be themajor ISG15 acceptor site. As this lysine

lies within the region of NS1A responsible for binding to double-

stranded RNA and the nuclear import factor importin-a, the

authors assayed the ability of ISGylation to affect either of these

interactions. Whereas ISGylated NS1A binds as well as non-

ISGylated NS1A to polyI:C, it fails to interact with importin-a,

suggesting that ISGylation of NS1A causes a specific loss of

function. Importantly, K41R mutation significantly enhances

the ability of the virus to replicate in the presence of IFN-b, sug-

gesting that specific targeting of NS1A protein by ISG15 impairs

influenza A replication through a loss-of-function mechanism

(Figure 1B).

By contrast, mutagenesis results from Wang and colleagues

indicate that ISGylation of multiple lysines on NS1A contributes

to the impairment of viral replication. Moreover, ISGylation of

NS1A appears to cause a severe impairment in the binding to

U6 snRNA and dsRNA. In addition, ISGylation also impairs

self-interaction of NS1A.

The reasons for the discrepancies regarding NS1A’s ISGyla-

tion site(s) and the ability of ISGylated NS1A to bind to RNA

are unclear. It is noteworthy that the influenza viruses used by

the two groups differ in origin, so their interactions with the

host cell may be different. In any case, these reports identify

the first viral ISG15 target and suggest that ISGylation of this

target impairs viral replication through a loss-of-function mech-

anism. It is at present not clear how ISGylation of a small

percentage of NS1A leads to such a dramatic impairment in viral

replication.

Broad Targeting of Newly Synthesized Viral Proteins

A recent article inMolecular Cell suggests an intriguing model for

understanding the antiviral activity of ISG15 (Durfee et al., 2010).

Only a minority of constitutively expressed proteins from the

aforementioned proteomics study can be confirmed as ISGyla-

tion substrates at their endogenous levels, even when ISG15

and the ISGylation enzymes are overexpressed. By contrast,

most of these proteins are confirmed as ISGylation substrates

when they are exogenously expressed along with the ISGylation

machinery. In fact, most (but not all) exogenously expressed

proteins, including bacterial proteins and the TAP affinity tag,

are also ISGylated using this method. These results raise doubts

regarding the physiological significance of putative ISGylation

substrates.

Yet Huibregtse and colleagues embraced what could easily

have been dismissed as a technical artifact. Their subsequent

results suggest that a key variable determining whether or not

a protein gets ISGylated is its new synthesis in the presence of

ISG15 and ISGylationmachinery. Proteins that are newly synthe-

sized, for instance those that are expressed from a transfected

plasmid, in the presence of the ISGylation machinery are readily

ISGylated. Moreover, multiple fragments of a protein that are

expressed as deletion mutants appear equally susceptible to

ISGylation, suggesting a lack of rigid specificity determinants

within the protein structure as might have been presumed.

A potential explanation of these results is that newly synthesized

proteins are targets for ISGylation; indeed, fractionation of cyto-

solic extracts reveals that HERC5 is associated with ribosomes.

Thus the authors propose that HERC5 broadly, and at least

somewhat nonspecifically, targets newly synthesized proteins

for ISGylation (Figure 1C).

This idea implies that some viral proteins will be ISGylated

during replication. As some viral structural proteins, such as

those that make up the capsid, must precisely assemble into

higher-order structures, it is possible that ISGylation of a small

fraction of these proteins could have a dominant-negative effect.

Indeed, using the human papillomavirus (HPV) pseudovirus

system, in which the HPV L1 and L2 capsid proteins are able

to package a plasmid expressing green fluorescent protein

and deliver it to new cells, the authors show that ISGylation of

approximately 10% of L1 protein is associated with a 70%

decrease in infectivity. The mechanistic basis of the infectivity

impairment by ISG15 remains to be determined; perhaps entry

of the virus into new cells or release of the nucleic acids into

the infected cells is impaired. In any case, the results suggest

that ISG15 can indeed cause a dominant-negative impairment

of viral protein function, an appealing idea that might explain

how ISGylation of a small fraction of a given protein can have

potent antiviral effects. In addition, as postulated by the authors

of this report, these findings suggest that ISGylation of some,

perhaps most, host proteins could be a by-product of the cell’s

effort to maximize ISGylation of viral proteins.

Perspectives

Although ISG15 is the first UBL known to exist, its biological role

and mechanism of action are less well understood than those of

most of the other UBLs, such as SUMO or NEDD8. This is in part

due to the absence of homologs of ISG15 and its conjugation

machinery (e.g, UBE1L) in experimental organisms such as

yeast, Drosophila, or C. elegans. Nevertheless, significant prog-

ress has been made in the past few years in the identification of

the enzymatic machinery that carries out ISGylation and in the

elucidation of the role of ISGylation in antiviral defense. The

recent findings of the direct antiviral activity of ISG15 through

both specific and broad modification of viral proteins represent

a major advance in understanding the antiviral mechanisms of

ISGylation. Some ISGylated host proteins also appear to


mediate its antiviral effects (for example, ISGylated 4EHP and

IRF3 as mentioned above).

Although upregulating the expression of ISGylation machinery

is a primary means of regulating ISGylation, additional regulatory

mechanisms clearly exist, for example, NS1B’s binding to ISG15

and HERC50s association with ribosomes and specific sub-

strates like NS1A. Biochemical reconstitution of the ISGylation

processwould potentially facilitate the identification of additional

factors that regulate ISGylation.

An emerging theme from the recent mechanistic studies is that

ISGylation alters a protein’s ability to engage in its typical inter-

actions (such as with other proteins or RNA). The basis for this

alteration is as yet unclear. It is likely that the presence of

ISG15 could directly interfere with the normal protein-protein

or protein-RNA interface. It is also feasible that ISGylation could

induce allosteric changes in protein structure, or that ISG15-

binding protein(s) may be present in cells and could modulate

interactions between ISGylated proteins and their typical

partners.

It is noteworthy that mice lacking ISG15 are not as susceptible

to viral infection as IFN receptor knockout mice, indicating

that ISGylation contributes to, but is not solely responsible

for, the antiviral effects of IFN in mice (Lenschow et al., 2007).

Recent work demonstrating marked differences in the interac-

tion between influenza B virus and the ISGylation machinery of

mice and humans suggests that ISG15 might play a more prom-

inent antiviral role in human. Indeed, blocking ISGylation in

human cells severely impairs IFN-induced antiviral activity

against influenza A virus (Hsiang et al., 2009). Future research

could also reveal other functions of ISGylation unrelated to its

antiviral effect. Indeed, the levels of ISG15 and its conjugation

to cellular proteins are elevated in several tumors and tumor-

derived cell lines (Desai et al., 2006).

Understanding the roles and mechanism of action of ISGs,

such as ISG15, in antiviral defense may pave the way to more

effective antiviral therapies. For example, viral proteins that

counter the IFN response by antagonizing ISGylation might

make appealing therapeutic targets.

ACKNOWLEDGMENTS

We thank J. Cabrera for assistance with preparation of the figure.

REFERENCES

Chang, Y.G., Yan, X.Z., Xie, Y.Y., Gao, X.C., Song, A.X., Zhang, D.E., and Hu,

H.Y. (2008). J. Biol. Chem. 283, 13370–13377.

Desai, S.D., Haas, A.L., Wood, L.M., Tsai, Y.C., Pestka, S., Rubin, E.H.,

Saleem, A., Nur, E.K.A., and Liu, L.F. (2006). Cancer Res. 66, 921–928.

Durfee, L.A., Lyon, N., Seo, K., and Huibregtse, J.M. (2010). Mol. Cell 38,

722–732.

Giannakopoulos, N.V., Arutyunova, E., Lai, C., Lenschow, D.J., Haas, A.L., and

Virgin, H.W. (2009). J. Virol. 83, 1602–1610.

Hsiang, T.Y., Zhao, C., and Krug, R.M. (2009). J. Virol. 83, 5971–5977.

Jeon, Y.J., Choi, J.S., Lee, J.Y., Yu, K.R., Kim, S.M., Ka, S.H., Oh, K.H., Kim,

K.I., Zhang, D.E., Bang, O.S., et al. (2009). EMBO Rep. 10, 374–380.

Jeon, Y.J., Yoo, H.M., and Chung, C.H. (2010). Biochim. Biophys. Acta 1802,

485–496.

Kim, K.I., Yan, M., Malakhova, O., Luo, J.K., Shen, M.F., Zou, W., de la Torre,

J.C., and Zhang, D.E. (2006). Mol. Cell. Biol. 26, 472–479.

Lai, C., Struckhoff, J.J., Schneider, J., Martinez-Sobrido, L., Wolff, T., Garcia-

Sastre, A., Zhang, D.E., and Lenschow, D.J. (2009). J. Virol. 83, 1147–1151.

Lenschow, D.J., Giannakopoulos, N.V., Gunn, L.J., Johnston, C., O’Guin, A.K.,

Schmidt, R.E., Levine, B., and Virgin, H.W., 4th. (2005). J Virol 79, 13974–

13983.

Lenschow, D.J., Lai, C., Frias-Staheli, N., Giannakopoulos, N.V., Lutz, A.,

Wolff, T., Osiak, A., Levine, B., Schmidt, R.E., Garcia-Sastre, A., et al.

(2007). Proc. Natl. Acad. Sci. USA 104, 1371–1376.

Okumura, F., Zou, W., and Zhang, D.E. (2007). Genes Dev. 21, 255–260.

Osiak, A., Utermohlen, O., Niendorf, S., Horak, I., and Knobeloch, K.P. (2005).

Mol. Cell. Biol. 25, 6338–6345.

Shi, H.X., Yang, K., Liu, X., Liu, X.Y., Wei, B., Shan, Y.F., Zhu, L.H., and Wang,

C. (2010). Mol. Cell. Biol. 30, 2424–2436.

Sridharan, H., Zhao, C., and Krug, R.M. (2010). J. Biol. Chem. 285, 7852–7856.

Tang, Y., Zhong, G., Zhu, L., Liu, X., Shan, Y., Feng, H., Bu, Z., Chen, H., and

Wang, C. (2010). J. Immunol. 184, 5777–5790.

Versteeg, G.A., Hale, B.G., van Boheemen, S., Wolff, T., Lenschow, D.J., and

Garcia-Sastre, A. (2010). J. Virol. 84, 5423–5430.

Yuan, W., and Krug, R.M. (2001). EMBO J. 20, 362–371.

Zhao, C., Denison, C., Huibregtse, J.M., Gygi, S., and Krug, R.M. (2005). Proc.

Natl. Acad. Sci. USA 102, 10200–10205.

Zhao, C., Hsiang, T.Y., Kuo, R.L., and Krug, R.M. (2010). Proc. Natl. Acad. Sci.

USA 107, 2253–2258.


Leading Edge

Primer

Biological Applications

of Protein Splicing

Miquel Vila-Perello1 and Tom W. Muir1,*1Laboratory of Synthetic Protein Chemistry, The Rockefeller University, 1230 York Avenue, New York, NY 10021, USA


DOI 10.1016/j.cell.2010.09.031

Protein splicing is a naturally occurring process in which a protein editor, called an intein, performs

a molecular disappearing act by cutting itself out of a host protein in a traceless manner. In the two

decades since its discovery, protein splicing has been harnessed for the development of several

protein-engineering methods. Collectively, these technologies help bridge the fields of chemistry

and biology, allowing hitherto impossible manipulations of protein covalent structure. These tools

and their application are the subject of this Primer.

Introduction

Molecular biologists have developed powerful methods to study

the details of protein function. Approaches such as X-ray crystal-

lography and site-directed mutagenesis have furnished count-

less insights, highlighting how even the most byzantine of

problems can yield to the right tools. Nonetheless, there is

always demand for more tools. This is perhaps best illustrated

by considering protein posttranslational modifications (PTMs).

Most, if not all, proteins are modified at some point; it is nature’s

way of imposing functional diversity on a single polypeptide

chain (Walsh et al., 2005). Moreover, many proteins are modified

in manifold ways as exemplified by the histones, where dozens

of discrete PTMs have been identified. Existing tools based on

site-directed mutagenesis offer limited opportunities for deter-

mining what all these PTMs are doing. Although it is straightfor-

ward to mutate a protein in such a way as to prevent a PTM from

being installed, the reverse strategy whereby a mutation is intro-

duced that mimics a PTM is a haphazard business at best. To fill

this and other voids, protein chemists have come up with an

array of approaches for the introduction of countless chemical

modifications into proteins, including all of the major types of

PTM.

The chemical modification of proteins can be accomplished

through a variety of means, including bioconjugation techniques

(Hermanson, 2008), total chemical synthesis (Kent, 2009),

enzyme-mediated reactions (Lin and Wang, 2008), nonsense

suppression mutagenesis (Wang et al., 2006), and a variety of

protein ligation methods (Hackenberger and Schwarzer, 2008).

The latter group of strategies include the protein semisynthesis

methods (defined as those where the protein is manufactured

from premade fragments one or both being recombinant in

origin) expressed protein ligation (EPL) and protein trans-splicing

(PTS) (Muir, 2003; Muralidharan and Muir, 2006; Mootz, 2009).

These are unique technologies in that they combine the power

of biotechnology, which provides accessibility to significant

amounts of large proteins, with the versatility of chemical

synthesis, which allows the site-specific incorporation of almost

any chemical modification into the target protein. In the following

sections we provide an overview of EPL and PTS and illustrate

how these technologies have been used to tackle problems in

molecular biology that have proven refractory to other methods.

Expressed Protein Ligation

Expressed protein ligation (EPL) allows a recombinant protein

and a synthetic peptide to be linked together undermild aqueous

conditions (Muir et al., 1998; Evans et al., 1998). The process

involves a chemoselective reaction that yields a final protein

product with a native peptide bond between its two building

blocks. The synthetic nature of one of the fragments enables

the site-specific introduction of almost any chemical modifica-

tion in the protein of interest, including fluorophores, caging

groups, crosslinkers, PTMs, and their analogs, as well as almost

any imaginable combination of modifications. At the same time,

the recombinant nature of the other fragment conveniently gives

access to large proteins, thereby overcoming the size restriction

associated with total chemical synthesis.

EPL is based on the well-known reaction between a polypep-

tide bearing a C-terminal thioester (a-thioester) and a peptide

possessing an N-terminal cysteine residue. This reaction,

termed native chemical ligation (NCL), originated in the field of

peptide chemistry and has proven extraordinarily powerful for

the total synthesis of small proteins and their analogs (Kent,

2009). However, the generation of large proteins using total

synthesis is still a daunting task for the nonspecialist, largely

due to the technical issues associated with performing the

multiple ligation reactions needed to access polypeptides

greater than 100 amino acids. One solution to this size problem

is to employ recombinant polypeptide building blocks in the

process; indeed, this semisynthetic NCL approach was demon-

strated early on by using a recombinant protein fragment con-

taining an N-terminal cysteine (Erlanson et al., 1996). Nonethe-

less, the full integration of NCL and semisynthesis awaited the

development of a general approach to install an a-thioester

moiety into recombinantly derived proteins. The solution to this

problem came from the discovery of a most unusual PTM,

termed protein splicing (Paulus, 2000).


Protein splicing is an autocatalytic process in which an inter-

vening protein domain (intein) excises itself from the polypeptide

in which it is embedded, concomitantly creating a new peptide

bond between its two flanking regions (exteins) (Figure 1). In

a sense, intein-mediated protein splicing is the protein equiva-

lent of RNA splicing involving self-splicing introns. Several

hundred inteins have been identified in unicellular organisms

from all three phylogenetic domains; all share conserved

sequence motifs and are derived from a common precursor

(for a complete listing, see http://www.neb.com/neb/inteins.

html). Thus, protein splicing is presumed to have an ancient

evolutionary origin. Parenthetically, although intein-mediated

protein splicing is not known to occur in multicellular organisms,

protein automodification processes do occur and involve intein-

like domains, most notably the hedgehog-like proteins that are

essential for embryonic development (Paulus, 2000). A biological

role for protein splicing in unicellular organisms has proven

elusive; modern inteins seem to be parasitic genetic elements

that are inserted into the open reading frames of (usually) essen-

tial genes. This frustration aside, the process has found a multi-

tude of applications in biotechnology (Noren et al., 2000) and

quickly attracted the interest of the peptide chemistry commu-

nity, as a-thioesters were identified as crucial intermediates in

the reaction mechanism (Figure 1). Several engineered inteins

have been developed that allow access to recombinant protein

a-thioester derivatives by thiolysis of the corresponding

C-terminal intein fusions (Figure 2). Moreover, inteins have also

been engineered to allow the introduction of an N-terminal

cysteine (Cys) moiety into recombinant proteins. Simple access

to reactive proteins without any size restriction through molec-

Figure 1. Mechanism of Protein SplicingProtein splicing (A) and its variant protein trans-

splicing (B).

ular biology techniques suddenly enabled

the application of NCL to the modification

of a much larger fraction of the proteome.

Indeed, the approach has been used to

generate semisynthetic derivatives of

members of essentially every major class

of protein including antibodies, integral

membrane proteins, cytoplasmic sig-

naling proteins, metabolic enzymes, and

transcription factors (Muir, 2003; Muralid-

haran and Muir, 2006).

Protein Trans-Splicing

A technology related to EPL, also based

on the use of inteins, is protein trans-

splicing (PTS, Figure 1). In PTS, artificially

or naturally split inteins are used to create

a new peptide bond between their flank-

ing exteins. Split inteins are characterized

by the fact that their primary sequence

is cut into two polypeptides giving an

N-terminal fragment (IntN) and a C-ter-

minal fragment (IntC). Fragment complementation leads to

reconstitution of the canonical intein fold, recovery of protein

splicing activity, and ligation of the exteins. Importantly, several

split inteins have been described in which one of the two frag-

ments is small enough to be obtained by peptide synthesis,

thus allowing splicing reactions to be performed between a

recombinant fragment and a synthetic one (Table 1) (Mootz,

2009). This allows the generation of a semisynthetic protein

derivative upon PTS. Use of these autoprocessing domains to

carry out the ligation reaction precludes the need to isolate

a-thioesters or N-terminal Cys peptides or proteins and,

because the IntN and IntC fragments often have high affinity for

one another, the reaction can be carried out at very low concen-

trations (lowmicromolar) under native conditions. This should be

contrasted with EPL, which being a bimolecular process usually

requires high concentrations of reactants (ideally high micro-

molar range) to be efficient.

Applications of EPL and PTS

The simplest application of EPL or PTS is the modification of the

N- or C-terminal regions of a protein because this can be

achieved in a single ligation step involving a synthetic peptide

fragment, containing the desired chemical probe(s) and a

recombinant protein fragment. Central regions of the protein of

interest can also be labeled, but a three-piece ligation strategy

is then required (Muir, 2003), which is more technically chal-

lenging. It should be noted that EPL and PTS can be used to

link a recombinant protein to a nonpeptidic moiety, provided it

has the necessary reactive handles for ligation. Examples of

this include the attachment of proteins to surfaces, polymers,


and nucleic acids (Cheriyan and Perler, 2009). Ligation of two

fully recombinant protein domains is also possible and has

been used to generate toxic proteins that cannot normally be

expressed (Evans et al., 1998), as well as to label specific

domains within large proteins with isotopes for structural studies

using NMR (nuclear magnetic resonance) spectroscopy (Mura-

lidharan and Muir, 2006).

A key decision when performing EPL and PTS is the selection

of the ligation site. Obviously, this must be chosen such that the

region of interest in the protein corresponds to the synthetic

building block in the semisynthesis scheme. The only sequence

requirement for the standard EPL strategy is the Cys residue at

the ligation site—this makes EPL virtually traceless compared

with protein labeling methods involving the use of reactive

tags (Lin and Wang, 2008). Furthermore, recent developments

in the use of ligation auxiliaries as well as desulfurization

methods have broadened the scope of EPL to include other resi-

dues such as glycine (Gly), alanine (Ala), valine (Val), and phenyl-

alanine (Phe) at the ligation site; these more sophisticated

methods employ a Cys surrogate for the ligation step, which is

later converted into the native residue (Hackenberger and

Schwarzer, 2008). As an alternative to the use of traceless liga-

tion methods, it is also possible to simply mutate in a Cys

residue at a convenient site in the protein. Although a commonly

used strategy, care must be taken to minimize the structural

and functional impact of the mutation on the protein; a serine

(Ser)/Ala/Cys mutation is often a good starting point (Valiya-

veetil et al., 2006a). An additional criterion to be considered

for EPL is the identity of the residue immediately upstream of

the Cys at the ligation site (which will be the residue adjacent

to the a-thioester in the N-terminal building block). Bulky,

b-branched amino acids, such as threonine (Thr), isoleucine

(Ile), and Val, slow-down the rate of the NCL reaction and should

be avoided, if possible.

The sequence requirements associated with PTS are some-

what more nebulous than those for EPL and depend to a great

extent on the exact split intein being used (Table 1) (Mootz,

2009). The mechanism of protein splicing dictates that, at

aminimum, the reaction will result in a Ser/Thr/Cys residue being

placed at the splice junction (Figure 1). However, in many cases,

there will be additional sequence requirements immediately

adjacent to this site. In particular, the commonly used cyanobac-

terial DnaE split inteins prefer to have three native C-extein resi-

dues (Cys-Phe-Asn) for optimal splicing efficiency (Mootz, 2009).

This restriction can be relaxed by using mutant split inteins

evolved to splice at non-native splice junctions, although the

Ser/Thr/Cys at the splice junction is still obligate (Lockless and

Muir, 2009).

A final consideration when choosing a ligation site is its posi-

tion within the secondary and tertiary structure of the protein.

Where possible the protein should be dissected between

modular domains as this will afford fragments that are well

Figure 2. Expressed Protein LigationThe boxed region designates the native chemical ligation (NCL) reaction in which trans-thioesterification of the protein a-thioester by the N-terminal Cys poly-

peptide is followed by an S to N acyl shift to generate a new peptide bond linking the two polypeptides. a-thioesters can be obtained recombinantly, using

engineered inteins, or by chemical synthesis. N-terminal Cys polypeptides can also be produced recombinantly or made using standard solid-phase peptide

synthesis (SPPS) methods.


behaved in terms of solubility and, importantly, preclude the

need for any folding step following the ligation reaction. The

need for well-behaved fragments is especially important when

using PTS because the process must be performed under

native-like conditions. If the protein can be efficiently refolded

then one naturally has more flexibility in choosing the ligation

site. In this case, EPL may be the method of choice given that

the actual ligation step can be performed in the presence of

a variety of additives, including chemical denaturants and deter-

gents (Muralidharan and Muir, 2006). Indeed, use of denaturants

is often beneficial for EPL reactions as it allows high concentra-

tions of reactants to be achieved, thereby improving the effi-

ciency of the bimolecular reaction.

The principal bottleneck of any project involving EPL or PTS is

the generation, by synthetic or recombinant means, of the reac-

tive protein fragments. As is usually the case in protein chem-

istry, each protein target presents its own set of (often unique)

challenges, and so some investment in strategy optimization

will be required for every system. Fortunately, after many years

of methodology development, an extensive array of tools is

now available for the generation of protein reactants for EPL

and PTS. An overview of commonly used approaches is given

in Table 1 and Table S1 (available online). These have allowed

a large number of systems to be interrogated through semisyn-

thesis including proteins that might, at first pass, seem beyond

the reach of organic chemistry, such as integral membrane

proteins.

EPL and PTS have been used to incorporate a variety of modi-

fications into proteins (Figure 3) to answer biological questions

that could not be addressed through more traditional

approaches. In the following sections we discuss examples of

these efforts and the biological insight they have revealed.

Semisynthesis of Proteins Containing Posttranslational

Modifications

The most common application of EPL is in the semisynthesis of

posttranslationally modified proteins. PTMs are used to regulate

the activity of most proteins, and to fully understand how this is

achieved inevitably requires access to these modified proteins

for biochemical or structural studies. As noted earlier, standard

site-directed mutagenesis provides limited possibilities in this

regard. Thus, a clear opportunity exists for using more chemi-

cally driven approaches. EPL, in particular, has helped to fill

this void, aided by the availability of robust methods for the

chemical synthesis of peptides containing PTMs. Indeed, EPL

has been used to generate proteins modified through phosphor-

ylation, glycosylation, lipidation, ubiquitination, acetylation, as

well as several other classes of modification (Muir, 2003; Chat-

terjee and Muir, 2010). Below we focus on specific studies that

highlight important themes.

Phosphorylation

Phosphorylation is one of the most common and extensively

studied PTMs. It should not be surprising then that EPL has

been heavily utilized for the preparation of proteins containing

this modification. Indeed, the first report of EPL described the

semisynthesis of a phosphotyrosine (pTyr) containing analog of

the protein kinase Csk (Muir et al., 1998). Subsequently, EPL

has been used to create several phosphorylated proteins for

detailed functional and structural studies (Schwarzer and Cole,

2005; Muralidharan and Muir, 2006). This is exemplified by

biochemical and crystallographic analyses of semisynthetic

versions of the transcription factors Smad2 and Smad3, which

explain how bis-phosphorylation activates them through

homo- and heterotrimerization (Wu et al., 2001; Chacko et al.,

2004). This system has also served as a useful proving ground

Table 1. Split Inteins Commonly Used for PTS

Sizea

Half-Lifeb (min) CommentsIntN IntC

Naturally Split Inteins

Ssp DnaE 123 36 35–175 One of the better studied and broadly used split inteins. Requires three native extein

residues (Cys-Phe-Asn) at the C-terminal junction. IntC is accessible to SPPSc.

Npu DnaEd 102 36 1 Most efficient split intein described so far. Active with a broad set of residues at the splicing

junction. IntC is accessible to SPPS.

Artificially Split Inteins

Mtu RecA 105 38 60–120 Reconstitution of the active intein requires co-refolding of previously denatured fragments.

Sce VMA 184 55 6 Requires induced fragment complementation by auxiliary dimerization domains. Has been

used to control protein function in conditional protein splicing systems in vivo and in vitro.

Ssp DnaB-S0e 104 47 12 Active under native conditions. Ser-Gly required at the N-terminal junction.

Ssp DnaB-S1e 11 143 280 The short IntN is amenable to SPPS and has been used for the labeling of protein N termini.

Same sequence requirements as DnaB-S0.

Ssp GyrB-S11e 150 6 170 Active under native conditions. Ser-Ala-Asp used at the N-terminal junction.

aNumber of residues.bHalf-lives calculated from reported first-order rate constants.cSolid-phase peptide synthesis.dArtificial variants of the Npu DnaE, with shorter IntC (15 and 6 residues), have been designed by shifting the split site closer to the C terminus.eS0, S1, and S11 indicate the site at which the intein is split. S0 corresponds to the split site of naturally split inteins.

Ssp: Synechotcystis sp.; Npu: Nostoc punctiforme; Sce: Saccharomyces cerevisiae; Mtu:Mycobacterium tuberculosis. (Mootz, 2009) and references

therein.


for several EPL-based technologies, including the incorporation

into proteins of new amino acid crosslinkers (Vila-Perello et al.,

2007) and various photoactivation strategies, including caged

phosphates (Hahn and Muir, 2004). One of the powers of

applying chemistry for the study of proteins is the ability to tweak

the covalent structure of the PTM. Cole and coworkers have

exploited this freedom to introduce various nonhydrolyzable

analogs of Ser/Thr/Tyr phosphorylation (termed phosphonates)

into proteins (Schwarzer and Cole, 2005). This strategy is partic-

ularly powerful in systems where the native phospho-amino acid

species is too short lived to permit detailed mechanistic studies.

For example, a semisynthetic version of the protein tyrosine

phosphatase, SHP-2, was prepared containing a stable tyrosine

phosphonate in place of the native pTyr (Lu et al., 2001). Subse-

quent microinjection of this protein into cells helped define a role

for this phosphorylation event in activation of the mitogen-acti-

vated kinase pathway.

Figure 3. Examples of Proteins Modfiied by

EPL and PTSExpressed protein ligation (EPL) and protein trans-

splicing (PTS) can be used to site-specifically

modify a wide variety of structurally and function-

ally diverse proteins, as the examples given in

the figure illustrate. Modifications range from natu-

rally occurring posttranslational modifications

(PTMs) to unnatural moieties and include the

following: (A) D-amino acids (D-Ala), (B) ester

bonds, (C) acetylated (N-acetyl-Lys) and (D) meth-

ylated amino acids (N-tri-methyl-Lys), (E) phos-

pho-Ser/Thr, (F) ubiquitination, (G) isotopes (PET

emitting 18F), (H) fluorophores (fluorescein), (I)

photo-crosslinkers (photo-Met), (J) Ser-ATP bi-

substrate transition state analogs, (K), b turn

mimics (nipecotic acid), (L) photo-caging groups

(photo-caged phospho-Ser), (M) glycosylated

and (N) prenylated amino acids, (O) nonhydrolyz-

able analogs of AMP, and (P) nonhydrolyzable

phosphomimics (Tyr phosphonate).

Lipidation

In terms of ease of chemical synthesis,

O-phosphorylation is among the lower-

hanging fruit of the PTM tree—this is

equally true for N-acetylation and

N-methylation, which have also been

introduced into semisynthetic proteins

(Chatterjee andMuir, 2010).Certainmodi-

fications such as lipidation, glycosylation,

and ubiquitination, however, present an

altogether different level of synthetic chal-

lenge due to their complexity and/or

physical attributes. Nonetheless, even

these have yielded to the EPL and PTS

approaches in recent years. Accordingly,

a variety of lipid modifications have been

introduced into proteins by EPL/PTS,

including prenyl groups and glycophos-

phatidylinositol (GPI) anchors (Brunsveld

et al., 2006). This is nicely illustrated by

the work of Goody and coworkers, who

have used semisynthesis in conjunction with structural and

functional approaches to study how lipidation regulates the func-

tion ofmembers of the Ras superfamily, including, most recently,

elucidation of the mechanism of membrane targeting of geranyl-

genanylated versions of a Rab GTPase (Wu et al., 2010).

Glycosylation

In terms of shear chemical complexity, glycosylation is arguably

the winner among the PTMs. The attached sugars can be

composed of several different monosaccharide building blocks

linked together in elaborate branched structures whose tailoring

can differ from molecule to molecule (Bertozzi and Kiessling,

2001). Studying the structural and functional consequences of

protein glycosylation is thus complicated by the inability to

isolate well-defined glycosylated proteins from natural sources.

Carbohydrate chemists have amassed an impressive arsenal

for the synthesis of complex oligosaccharides (Lepenies et al.,

2010). Recent years have seen this synthetic know-how


integrated with EPL for the preparation of homogeneous glyco-

proteins (Buskas et al., 2006). An impressive recent example of

this is the work of Unverzagt and coworkers, who synthesized

ribonuclease C, a 15 kDa enzymewith 4 disulfides and a bianten-

naric nonasaccharide, using a three-piece EPL strategy (Piontek

et al., 2009). In the coming years, we expect that the semisynthe-

sis of homogeneous glycoproteins will become more routine,

thereby allowing the role of this modification in the storage and

transfer of biological information to be examined in greater detail

than has hitherto been possible.

Ubiquitination

Ubiquitination is another example of a PTM difficult to study

using proteins isolated from natural sources. Ubiquitin (Ub) is

a 76 amino acid protein that is attached through its C terminus

to the 3-amino group of a lysine residue in a target protein.

Proteins can be monoubiquitinated, multiubiquitinated, or

polyubiquitinated, with the precise nature of this conjugation

dictating the functional consequences of the modification. The

E1–E3 protein ligase family is responsible for the attachment of

Ub to target proteins. Understanding the substrate specificity

and enzymology of these enzymes is an area of very active study.

Nonetheless, the details remain sufficiently obscure to make

in vitro ubiquitination of a target protein impractical (at least on

a preparative scale) in all but a few cases. Protein semisynthesis

provides an alternative source of ubiquitinated proteins. Indeed,

recent years have seen a flurry of reports describing chemical

methods to attach Ub to specific sites on a target protein

(McGinty et al., 2008; Li et al., 2009; Ajish Kumar et al., 2009;

Chatterjee et al., 2010; Chen et al., 2010). All of these strategies

employ inteins at one stage or another and allow the conjugation

of Ub to proteins through both native and non-native linkages.

Armed with these approaches, investigators have studied the

function of ubiquitination in several systems, including the role

of the PTM in regulating the activity of PCNA (involved in trans-

lesion DNA synthesis) (Chen et al., 2010) and histones (McGinty

et al., 2008; Chatterjee et al., 2010). These examples further high-

light a unique power of semisynthesis, namely the ability to

manipulate the structure of the PTM in ways that would be

impossible using an enzymatic approach. In particular, the

chemical approach permits Ub to be substituted for related

proteins (so-called ubiquitin-like proteins, Ubls), thereby allow-

ing structure-activity relationships to be explored. In the histone

example, the generation of a series of Ubl-modified mononu-

cleosomes aided in defining themechanism by which ubiquitina-

tion of histone H2B stimulates methylation of histone H3 by the

methyltransferase hDot1L (Chatterjee et al., 2010). More gener-

ally, the biochemical analysis of histone modifications appears

to be particularly fertile ground for the application of protein

semisynthesis. The majority of PTMs in histones are localized

in the flanking regions, making them readily accessible to EPL

and PTS. Indeed, several insights have already emerged from

the study of semisynthetic histones bearing chemically installed

PTMs (Chatterjee and Muir, 2010). We anticipate that this area

will continue to blossom in the years ahead.

Site-Specific Incorporation of Unnatural Building Blocks

EPL and PTS have been heavily utilized in the site-specific incor-

poration of unnatural amino acids into proteins. The ability to

precisely tune the steric and electronic properties of amino acid

side chains is a powerful way to explore the details of protein

function; nowhere is this truer than for the study of enzymes.

Indeed, analogs of a number of enzymes (and their substrates)

have been prepared byEPL. These studies have furnishedmech-

anistic insights by manipulating various properties of key amino

acid side chains, including redox potential (as in the case of ribo-

nuclease reductase), nucleophilicity (such as the protein tyrosine

kinase Src), and steric bulk (for instance the GyrA intein)

(Schwarzer and Cole, 2005; Frutos et al., 2010). Semisynthesis

has also been used to incorporate transition state analogs into

enzymesand their substrates. This is exemplified by thedevelop-

ment of bi-substrate inhibitors (i.e., simultaneously targeting two

substrate-binding sites) of protein kinasesbasedonATP-peptide

conjugates that mimic the phosphoryl-transfer transition state

(Schwarzer and Cole, 2005). This strategy was recently used

to study the mechanism of autophosphorylation of full-length

protein kinase A (PKA) (Pickin et al., 2008). PKA has two regula-

tory phosphorylation sites: one in its activation loop, installed

by PDK1 (pyruvate dehydrogenase kinase 1), and the other one

at Ser338, thought to be autocatalyzed. Semisynthesis of PKA

with a pSer338-ATP analog was used to investigate whether

the autophosphorylation reaction was intra- or intermolecular.

A combination of biochemical and computational experiments

demonstrates that the pSer338-ATP moiety is docked into

the PKA active site in an intramolecular fashion, arguing that

Ser338 phosphorylation is an intramolecular event.

A related concept has recently been applied to the study of E1

ubiquitin ligases (Lu et al., 2010; Olsen et al., 2010). These

enzymes activate Ub and Ubls through adenylation (AMP) of

their C termini followed by thioesterification of a conserved

Cys residue in the enzyme. To probe the first half-reaction,

EPL was used to generate a reversible inhibitor by incorporating

a nonhydrolyzable analog of AMP, 50-O-sulfamoyladenosine

(AMSN), at the C terminus of Ub and the Ubl, SUMO (Lu et al.,

2010). A similar EPL approach was used to obtain a covalent

inhibitor of the second half-reaction, in this case by incorporating

a vinyl-sulfonamide electrophilic trap into the moiety (AVSN).

These elegant chemical biology studies have been followed up

by an equally impressive structural biology analysis (Olsen

et al., 2010). Specifically, the crystal structures of both SUMO-

AMSN and SUMO-AVSN in complex with the SUMO E1 were

solved, revealing that a major reorganization of the enzyme

active site accompanies the second half-reaction, that is, thioes-

terification of the E1. Examples like this highlight the utility of

semisynthesis in the study of enzymes. Nonetheless, we have

barely scratched the surface in terms of what is possible in this

area. There remain many exciting directions that have been

largely or wholly unexplored, including the notion of creating

new catalysts by integrating EPL/PTS with concepts and tools

emanating from the fields of computational protein design and

synthetic organic chemistry (e.g., novel organic catalysts).

Amino acid side chains account for 50% of the mass of

a typical protein; the remainder is composed of the main chain.

Backbone hydrogen bonding is, of course, critical to stabilizing

the secondary and tertiary structures of proteins and frequently

plays a direct role in enzyme catalysis and the recognition of

ligands. Unfortunately, the protein main chain constitutes a


‘‘blind spot’’ for standard mutagenesis, and consequently the

effects of backbonemodifications on protein structure and func-

tion are relatively unexplored compared to side-chain alter-

ations. Chemistry-driven protein-engineering approaches such

as EPL do allow changes to be made to the backbone of

a protein, and there are several excellent examples of this

(Muralidharan and Muir, 2006; Kent, 2009). For instance, Raines

and coworkers prepared analogs of the enzyme ribonuclease A

in which an entire unit of secondary structure, a b turn, was

replaced with a reverse-turn mimetic called nipecotic acid

(Arnold et al., 2002). The resulting ‘‘prosthetic’’ protein displayed

wild-type enzymatic activity but was thermodynamically more

stable than the native protein. Another approach to stabilizing

a protein involving a backbone change is through head-to-tail

cyclization. This can be achieved using either EPL or PTS, and

there are several examples of cyclic proteins exhibiting in-

creased stability (Muir, 2003).

As noted above, backbone interactions can play a direct role in

protein function, a point that is clearly illustrated by the selectivity

filter of potassium ion channels. This is a narrow, 12 A long pore

lined with backbone carbonyl oxygen atoms that allows K+ ions

to pass through, but not other monovalent cations such as Na+.

Access to this region in a semisynthetic version of the bacterial

channel, KcsA, allowed the electronegativity of these carbonyl

groups to be attenuated through an amide-to-ester substitution

(Valiyaveetil et al., 2006b). Crystallographic and electrophysi-

ology studies on the resulting protein reveal alterations in ion

occupancy and conductance consistent with a model of

concerted ion conduction through the channel. The work on

KcsA provides another nice example of how chemistry can be

used to engineer the backbone of a protein, namely by engi-

neering the chirality of the polypeptide. Specifically, substitution

of a highly conserved Gly in the selectivity filter for a D-Ala shows

that the ability of the amino acid at that position to adopt a left-

handed helical conformation is absolutely required for activity,

and hence the native Gly residue acts as a D-amino acid surro-

gate (Valiyaveetil et al., 2006a). Electrophysiology and crystallo-

graphic studies demonstrate that the D-Ala-containing channel

is locked in an open conformation able to conduct Na+ in the

absence of K+. The work shows that selectivity is due in part to

the ability of the channel to structurally adapt in an ion-specific

manner to K+.

Site-Specific Incorporation of Biophysical Probes

EPL and PTS have proven to be extremely powerful for the site-

specific incorporation of spectroscopic probes into proteins

(Muralidharan and Muir, 2006; Mootz, 2009). After PTMs, the

incorporation of optical probes is the next most common appli-

cation of semisynthesis. In most cases, these semisynthetic

proteins are used to study ligand-binding events or internal

conformational changes in proteins, either by monitoring

changes in the fluorescence of a single strategically placed

probe in the protein or by employing multiple probes and using

fluorescence resonance energy transfer (FRET) between them.

These spectroscopic approaches are nicely showcased by the

work of Lorsch and coworkers, who carried out a series of

detailed thermodynamic and kinetic studies on the association

of fluorescent derivatives of eukaryotic initiation factors with

the ribosome (Maag et al., 2005). The generation of FRET-based

reporter proteins via EPL has also been used for the screening of

small-molecule inhibitors of biomedically important proteins

such as Abl kinase (Hofmann et al., 2001) and histone acetyl-

transferases (Xie et al., 2009). The former example highlights

a key attribute of EPL/PTS, namely the ability to incorporate

multiple noncoded elements, in this case two different fluoro-

phores, into a protein. This capacity is taken a step further by

a study in which five noncoded elements are incorporated into

the protein Smad2, namely, two phosphoserines, a fluorophore,

a fluorescent quencher, and a photocleavable trigger of activity

(Hahn et al., 2007). This protein was designed to be inactive

and nonfluorescent until irradiated with ultraviolet (UV) light

whereupon the protein activates (through trimerization) and

simultaneously becomes fluorescent. Microinjection of this

caged protein into mammalian cells allowed the levels of the bio-

logically active form of the protein to be precisely titrated (as

quantified by fluorescence) by varying the amount of irradiation

(Hahn et al., 2007).

EPL and PTS have also aided in the development of methods

for the structural characterization of proteins in solution using

NMR spectroscopy. NMR is a very powerful tool for the study

of protein structure and dynamics; however, spectral overlap

associated with large proteins limits its application. Both EPL

and PTS have been used to isotopically label specific regions,

or even atoms, of a protein in order to obtain simplified spectra

for detailed structural studies (Muralidharan and Muir, 2006).

In one recent example, which evokes the symbology of the

Uroboros (a mythical serpent that consumes its own tail), inteins

were actually used (via EPL) to make inteins containing site-

specific 15N and 13C isotopes (Frutos et al., 2010). NMR studies

on these proteins reveal that formation of the branched interme-

diate in the splicing reaction drastically alters the dynamic prop-

erties of the scissile amide bond between the intein and the

C-extein, rendering it more susceptible to nucleophilic attack.

In the so-called segmental isotopic labeling strategy (Muralid-

haran andMuir, 2006), the target protein is divided up into appro-

priate fragments, which are then expressed individually, allowing

uniform isotopic labeling of only the domain of interest. EPL or

PTS are then used to put the protein back together again via

one or more ligation steps. This strategy has been applied to

study specific domains (flanking aswell as internal) in the context

of larger proteins as well as to identify intramolecular interactions

or explore enzymatic mechanisms (Muralidharan and Muir,

2006). For example, Allain and coworkers prepared several

segmental labeled versions of the polypyrimidine tract-binding

protein and used these in conjunction with transverse-relaxation

optimized NMR spectroscopy to define domain-domain inter-

faces within the protein required for RNA binding (Vitali et al.,

2006). Themajority of segmental labeling studies have employed

samples generated in vitro using individually expressed building

blocks. This is often a technically demanding undertaking due to

the large amounts of protein needed for NMR studies. To

address this, Iwai and coworkers have demonstrated the feasi-

bility of performing segmental labeling within Escherichia coli

cells (Zuger and Iwai, 2005). This employs a clever combination

of PTS and orthogonal promoter systems to allow the in vivo

reaction of a nonlabeled and labeled fragment of the protein.


This important advance not only promises easier access to

segmental labeled proteins for traditional structural studies but

also could have application in the emerging field of cell-based

protein NMR analysis.

In Vivo Applications of EPL and PTS

As we have already discussed, semisynthetic proteins prepared

in the test tube can be injected into cells for the purposes of

studying cell biological processes. This approach can be

extended to animal studies. For instance, EPL was recently

used to prepare a version of the protein hormone leptin contain-

ing an 18F-probe for PET (positron emission tomography)

imaging (Ceccarini et al., 2009). This molecule was used to study

the systemic biodistribution of the hormone in rodents and

primates, revealing, among other things, high-level uptake in

tissues undergoing hematopoiesis. This strategy aside, there

are many situations where it might be advantageous to perform

the protein chemistry inside the living cell or animal. In this

regard, PTS is especially powerful due to the availability of natu-

rally split cyanobacterial inteins that have no cross-reactivity with

any endogenous proteins in eukaryotic cells. Early work from our

own group demonstrated the potential of PTS for the in vivo

labeling of proteins by using the naturally split Ssp DnaE intein

for the traceless ligation of synthetic probes to heterologously

expressed proteins in mammalian cells (Giriat and Muir, 2003).

The efficiency of this cell-based semisynthesis approach is

sure to be improved by utilizing the recently described Nostoc

punctiforme DnaE intein, which possesses a series of remark-

able properties, including being the current record holder for

splicing kinetics (t1/2 1 min) (Mootz, 2009).

One of the most exciting uses of PTS is in the generation of

cyclic peptides in cells. Peptide cyclization is commonly used

inmedicinal chemistry (and in nature) to improve peptide stability

and bioactivity. The ability to biosynthesize cyclic peptides

in vivo offers the possibility of generating large genetically

encoded libraries for high-throughput screening purposes. This

can be accomplished by nesting the sequence (or library of

sequences) to be cyclized between the IntC and IntN intein frag-

ments. Flipping the order of the intein fragments in the precursor

ensures that PTS spits out a cyclic peptide (Scott et al., 1999).

This nifty technology, often referred to as SICLOPPS (split intein-

mediated circular ligation of peptides and proteins, Figure 4A),

has been used to screen for inhibitors of several processes

(Cheriyan and Perler, 2009), includingmost recently the selection

of cyclic peptide inhibitors of a-synuclein toxicity in a yeast

model of Parkinson’s disease (Kritzer et al., 2009).

PTS results in a full-length active protein being generated from

two inactive split fragments. This functional output can be

harnessed for a variety of purposes. Umezawa and coworkers

have developed several cell-based biosensors based on the

activity of split inteins and used these for a variety of purposes,

including the identification of mitochondrial proteins (Ozawa

et al., 2003) and the monitoring of caspase activity (Kanno

et al., 2007). PTS has also found application in the area of

gene therapy. In a recent study, Li et al. expanded the scope

of adeno-associated virus (AAV) as a delivery vehicle for thera-

peutic genes (Li et al., 2008). AAV has several advantageous

properties as a vector but is handicapped by its limited usable

DNA capacity ( 4 kb). To overcome this, the authors created

two AAV vectors each carrying half of a therapeutic gene fused

in-frame to a split intein coding sequence. Coinfection of target

cells with the two AAV vectors leads to production of the thera-

peutic protein after PTS. As a proof of principle, the authors

demonstrated the production of a therapeutic dystrophin protein

upon codelivery of appropriate AAV vectors in a mouse model of

muscular dystrophy.

There are no known natural regulators of protein splicing.

Rather, the process appears to occur spontaneously after trans-

lation of the precursor protein. The idea of controlling protein

splicing is, nonetheless, attractive as this would provide a way

to trigger the posttranslational synthesis of a target protein. In

principle, such a system would be fast (compared to inducible

genes), tunable (allowing protein levels to be adjusted), and

portable (many inteins are remarkably promiscuous). With this

in mind, several conditional protein splicing systems have been

reported that respond to changes in temperature, light, protease

activity, and the presence of various small molecules (Cheriyan

and Perler, 2009; Mootz, 2009). These have been used to control

the activity of proteins both in cultured cells and in living animals.

Examples include the control of Notch signaling in Drosophila

melanogaster via a temperature-inducible intein mutant (Zeidler

et al., 2004) and the control of hedgehog signaling in mammalian

Figure 4. In Vivo Applications of Protein

Splicing(A) Schematic representation of intein-mediated

peptide or protein cyclization. The target polypep-

tide is expressed flanked by IntC and IntN at the

N and C termini, respectively. Protein trans-

splicing (PTS) results in the formation of a new

peptide bond between the N and C termini of the

target and thus generates a circularized peptide

or protein.

(B) Control of protein splicing through ligand-

induced intein complementation. The splicing

activity of artificially split inteins can be controlled

by fusion to exogenous auxiliary domains (in the

figure, a ligand-binding domain). A triggering event

(in the figure, ligand binding) causes a conforma-

tional change in the auxiliary domain, which

induces intein reconstitution and subsequent

protein splicing.


cells via a tamoxifen-inducible engineered intein (Figure 4B)

(Yuen et al., 2006). It should be stressed, however, that condi-

tional protein splicing does not work in every context due to

the, as yet, poorly understood functional interplay between the

intein and the surrounding extein sequences. Nevertheless, the

big advantage of this approach over mainstream small-molecule

screening initiatives is that different conditional intein constructs

can be easily surveyed using standard molecular biology tech-

niques.

Outlook

EPL and PTS have proven remarkably useful for studying protein

function in vitro and in vivo. The number of proteins studied by

semisynthesis is constantly growing, as is the complexity of

the modifications that can be introduced. In this Primer, our

aim is to provide a broad overview of the techniques and to intro-

duce selected systems that have been instrumental in unlocking

biological puzzles. As with any approach, EPL and PTS have

their strengths and weaknesses. The approaches are unparal-

leled in terms of the range and number of noncoded elements

that can be introduced into large proteins. However, they are

at their most practical when the regions to be modified are within

50 amino acids of the N or C terminus of the protein of interest,

given that this allows a single ligation step to be performed. The

interiors of proteins are far more difficult to access via semisyn-

thesis, requiring the use of technically demanding sequential

ligation reactions. This should be contrasted with the nonsense

suppression mutagenesis method. Although more restricted in

the types of modification that can be introduced, it does provide

general access to any part of the protein primary sequence

(Wang et al., 2006). Thus, EPL/PTS and nonsense suppression

are complementary protein-engineering approaches and the

decision to use one or the other will depend on the question at

hand. Moreover, there is no reason why the two approaches

cannot be used in combination, a tactic that we are now begin-

ning to see (Li et al., 2009) and that will surely bemore common in

the future.

A defined biological role for protein splicing has so far eluded

investigators—we currently know of no intein whose activity is

naturally regulated, something that would point the way to a bio-

logical purpose. Inteins are, however, very ancient proteins and

so such regulation may have been lost during evolution. What

we can say about inteins is that they are an amazingly malleable

platform for technologydevelopment. It is a fair bet that thechem-

ical biology communitywill continue to find newuses for inteins in

both the basic and applied biomedical sciences. Thus, these

remarkable protein deviceswill continue to be a part of the thread

that stitches together the fields of chemistry and biology.

SUPPLEMENTAL INFORMATION

Supplemental Information includes one table and can be found with this article

online at doi:10.1016/j.cell.2010.09.031.

ACKNOWLEDGMENTS

We thank B. Fierz and N. Shah for valuable input. Some of the work discussed

in this review was performed in the Muir laboratory and was supported by

the NIH.

REFERENCES

Ajish Kumar, K.S., Haj-Yahya, M., Olschewski, D., Lashuel, H.A., and Brik, A.

(2009). Highly efficient and chemoselective peptide ubiquitylation. Angew.

Chem. Int. Ed. Engl. 48, 8090–8094.

Arnold, U., Hinderaker, M.P., Nilsson, B.L., Huck, B.R., Gellman, S.H., and

Raines, R.T. (2002). Protein prosthesis: a semisynthetic enzyme with a beta-

peptide reverse turn. J. Am. Chem. Soc. 124, 8522–8523.

Bertozzi, C.R., and Kiessling, L.L. (2001). Chemical glycobiology. Science 291,

2357–2364.

Brunsveld, L., Kuhlmann, J., Alexandrov, K., Wittinghofer, A., Goody, R., and

Waldmann, H. (2006). Lipidated ras and rab peptides and proteins–synthesis,

structure, and function. Angew. Chem. Int. Ed. Engl. 45, 6622–6646.

Buskas, T., Ingale, S., and Boons, G.J. (2006). Glycopeptides as versatile tools

for glycobiology. Glycobiology 16, 113–136.

Ceccarini, G., Flavell, R.R., Butelman, E.R., Synan, M., Willnow, T.E.,

Bar-Dagan, M., Goldsmith, S.J., Kreek, M.J., Kothari, P., Vallabhajosula, S.,

et al. (2009). PET imaging of leptin biodistribution and metabolism in rodents

and primates. Cell Metab. 10, 148–159.

Chacko, B., Qin, B., Tiwari, A., Shi, G., Lam, S., Hayward, L., de Caestecker,

M., and Lin, K. (2004). Structural basis of heteromeric Smad protein assembly

in TGF-b signaling. Mol. Cell 15, 813–823.

Chatterjee, C., and Muir, T.W. (2010). Chemical approaches for studying

histone modifications. J. Biol. Chem. 285, 11045–11050.

Chatterjee, C., McGinty, R.K., Fierz, B., and Muir, T.W. (2010). Disulfide-

directed histone ubiquitylation reveals plasticity in hDot1L activation. Nat.

Chem. Biol. 6, 267–269.

Chen, J., Ai, Y., Wang, J., Haracska, L., and Zhuang, Z. (2010). Chemically

ubiquitylated PCNA as a probe for eukaryotic translesion DNA synthesis.

Nat. Chem. Biol. 6, 270–272.

Cheriyan, M., and Perler, F.B. (2009). Protein splicing: A versatile tool for drug

discovery. Adv. Drug Deliv. Rev. 61, 899–907.

Erlanson, D.A., Chytil, M., and Verdine, G.L. (1996). The leucine zipper domain

controls the orientation of AP-1 in the NFAT.AP-1.DNA complex. Chem. Biol.

3, 981–991.

Evans, T.C., Benner, J., and Xu, M.Q. (1998). Semisynthesis of cytotoxic

proteins using a modified protein splicing element. Protein Sci. 7, 2256–2264.

Frutos, S., Goger, M., Giovani, B., Cowburn, D., and Muir, T.W. (2010).

Branched intermediate formation stimulates peptide bond cleavage in protein

splicing. Nat. Chem. Biol. 6, 527–533.

Giriat, I., and Muir, T.W. (2003). Protein semi-synthesis in living cells. J. Am.

Chem. Soc. 125, 7180–7181.

Hackenberger, C.P., and Schwarzer, D. (2008). Chemoselective ligation and

modification strategies for peptides and proteins. Angew. Chem. Int. Ed.

Engl. 47, 10030–10074.

Hahn, M.E., and Muir, T.W. (2004). Photocontrol of Smad2, a multiphosphory-

lated cell-signaling protein, through caging of activating phosphoserines.

Angew. Chem. Int. Ed. Engl. 43, 5800–5803.

Hahn, M.E., Pellois, J.P., Vila-Perello, M., and Muir, T.W. (2007). Tunable pho-

toactivation of a post-translationally modified signaling protein and its unmod-

ified counterpart in live cells. ChemBioChem 8, 2100–2105.

Hermanson, G.T. (2008). Bioconjugate Techniques, Second Edition (San

Diego, CA: Academic Press).

Hofmann, R.M., Cotton, G.J., Chang, E.J., Vidal, E., Veach, D., Bornmann, W.,

and Muir, T.W. (2001). Fluorescent monitoring of kinase activity in real time:

development of a robust fluorescence-based assay for Abl tyrosine kinase

activity. Bioorg. Med. Chem. Lett. 11, 3091–3094.

Kanno, A., Yamanaka, Y., Hirano, H., Umezawa, Y., and Ozawa, T. (2007).

Cyclic luciferase for real-time sensing of caspase-3 activities in living

mammals. Angew. Chem. Int. Ed. Engl. 46, 7595–7599.

Kent, S.B. (2009). Total chemical synthesis of proteins. Chem. Soc. Rev. 38,

338–351.


Kritzer, J.A., Hamamichi, S., McCaffery, J.M., Santagata, S., Naumann, T.A.,

Caldwell, K.A., Caldwell, G.A., and Lindquist, S. (2009). Rapid selection of

cyclic peptides that reduce alpha-synuclein toxicity in yeast and animal

models. Nat. Chem. Biol. 5, 655–663.

Lepenies, B., Yin, J., and Seeberger, P.H. (2010). Applications of synthetic

carbohydrates to chemical biology. Curr. Opin. Chem. Biol. 14, 404–411.

Li, J., Sun,W.,Wang, B., Xiao, X., and Liu, X.Q. (2008). Protein trans-splicing as

a means for viral vector-mediated in vivo gene therapy. Hum. Gene Ther. 19,

958–964.

Li, X., Fekner, T., Ottesen, J.J., and Chan, M.K. (2009). A pyrrolysine analogue

for site-specific protein ubiquitination. Angew. Chem. Int. Ed. 48, 9184–9187.

Lin, M.Z., and Wang, L. (2008). Selective labeling of proteins with chemical

probes in living cells. Physiology (Bethesda) 23, 131–141.

Lockless, S.W., and Muir, T.W. (2009). Traceless protein splicing utilizing

evolved split inteins. Proc. Natl. Acad. Sci. USA 106, 10999–11004.

Lu, W., Gong, D., Bar-Sagi, D., and Cole, P.A. (2001). Site-specific incorpora-

tion of a phosphotyrosinemimetic reveals a role for tyrosine phosphorylation of

SHP-2 in cell signaling. Mol. Cell 8, 759–769.

Lu, X., Olsen, S.K., Capili, A.D., Cisar, J.S., Lima, C.D., and Tan, D.S. (2010).

Designed semisynthetic protein inhibitors of Ub/Ubl E1 activating enzymes.

J. Am. Chem. Soc. 132, 1748–1749.

Maag, D., Fekete, C.A., Gryczynski, Z., and Lorsch, J.R. (2005). A conforma-

tional change in the eukaryotic translation preinitiation complex and release

of eIF1 signal recognition of the start codon. Mol. Cell 17, 265–275.

McGinty, R.K., Kim, J., Chatterjee, C., Roeder, R.G., and Muir, T.W. (2008).

Chemically ubiquitylated histone H2B stimulates hDot1L-mediated intranu-

cleosomal methylation. Nature 453, 812–816.

Mootz, H.D. (2009). Split inteins as versatile tools for protein semisynthesis.

ChemBioChem 10, 2579–2589.

Muir, T.W. (2003). Semisynthesis of proteins by expressed protein ligation.

Annu. Rev. Biochem. 72, 249–289.

Muir, T.W., Sondhi, D., and Cole, P.A. (1998). Expressed protein ligation:

a general method for protein engineering. Proc. Natl. Acad. Sci. USA 95,

6705–6710.

Muralidharan, V., and Muir, T.W. (2006). Protein ligation: an enabling tech-

nology for the biophysical analysis of proteins. Nat. Methods 3, 429–438.

Noren, C.J., Wang, J., and Perler, F.B. (2000). Dissecting the chemistry of

protein splicing and its applications. Angew. Chem. Int. Ed. 39, 450–466.

Olsen, S.K., Capili, A.D., Lu, X., Tan, D.S., and Lima, C.D. (2010). Active site

remodelling accompanies thioester bond formation in the SUMO E1. Nature

463, 906–912.

Ozawa, T., Sako, Y., Sato, M., Kitamura, T., and Umezawa, Y. (2003). A genetic

approach to identifying mitochondrial proteins. Nat. Biotechnol. 21, 287–293.

Paulus, H. (2000). Protein splicing and related forms of protein autoprocessing.


Pickin, K.A., Chaudhury, S., Dancy, B.C., Gray, J.J., and Cole, P.A. (2008).

Analysis of protein kinase autophosphorylation using expressed protein liga-

tion and computational modeling. J. Am. Chem. Soc. 130, 5667–5669.

Piontek, C., Varon Silva, D., Heinlein, C., Pohner, C., Mezzato, S., Ring, P.,

Martin, A., Schmid, F.X., and Unverzagt, C. (2009). Semisynthesis of a homo-

geneous glycoprotein enzyme: ribonuclease C: part 2. Angew. Chem. Int. Ed.

Engl. 48, 1941–1945.

Schwarzer, D., and Cole, P.A. (2005). Protein semisynthesis and expressed

protein ligation: chasing a protein’s tail. Curr. Opin. Chem. Biol. 9, 561–569.

Scott, C.P., Abel-Santos, E., Wall, M., Wahnon, D.C., and Benkovic, S.J.

(1999). Production of cyclic peptides and proteins in vivo. Proc. Natl. Acad.

Sci. USA 96, 13638–13643.

Valiyaveetil, F.I., Leonetti, M., Muir, T.W., and Mackinnon, R. (2006). Ion selec-

tivity in a semisynthetic K+ channel locked in the conductive conformation.

Science 314, 1004–1007.

Valiyaveetil, F.I., Sekedat, M., Mackinnon, R., andMuir, T.W. (2006). Structural

and functional consequences of an amide-to-ester substitution in the selec-

tivity filter of a potassium channel. J. Am. Chem. Soc. 128, 11591–11599.

Vila-Perello, M., Pratt, M.R., Tulin, F., and Muir, T.W. (2007). Covalent capture

of phospho-dependent protein oligomerization by site-specific incorporation

of a diazirine photo-cross-linker. J. Am. Chem. Soc. 129, 8068–8069.

Vitali, F., Henning, A., Oberstrass, F.C., Hargous, Y., Auweter, S.D., Erat, M.,

and Allain, F.H. (2006). Structure of the two most C-terminal RNA recognition

motifs of PTB using segmental isotope labeling. EMBO J. 25, 150–162.

Walsh, C.T., Garneau-Tsodikova, S., andGatto, G.J. (2005). Protein posttrans-

lational modifications: the chemistry of proteome diversifications. Angew.

Chem. Int. Ed. 44, 7342–7372.

Wang, L., Xie, J., and Schultz, P.G. (2006). Expanding the genetic code. Annu.

Rev. Biophys. Biomol. Struct. 35, 225–249.

Wu, J.W., Hu, M., Chai, J., Seoane, J., Huse, M., Li, C., Rigotti, D.J., Kyin, S.,

Muir, T.W., Fairman, R., et al. (2001). Crystal structure of a phosphorylated

Smad2. Recognition of phosphoserine by the MH2 domain and insights on

Smad function in TGF-beta signaling. Mol. Cell 8, 1277–1289.

Wu, Y.W., Oesterlin, L.K., Tan, K.T., Waldmann, H., Alexandrov, K., and

Goody, R. (2010). Membrane targetingmechanism of RabGTPases elucidated

by semisynthetic protein probes. Nat. Chem. Biol. 6, 534–540.

Xie, N., Elangwe, E.N., Asher, S., and Zheng, Y.G. (2009). A dual-mode fluores-

cence strategy for screening HATmodulators. Bioconjug. Chem. 20, 360–366.

Yuen, C.M., Rodda, S.J., Vokes, S.A., McMahon, A.P., and Liu, D.R. (2006).

Control of transcription factor activity and osteoblast differentiation in

mammalian cells using an evolved small-molecule-dependent intein. J. Am.

Chem. Soc. 128, 8939–8946.

Zeidler, M.P., Tan, C., Bellaiche, Y., Cherry, S., Hader, S., Gayko, U., and

Perrimon, N. (2004). Temperature-sensitive control of protein activity by condi-

tionally splicing inteins. Nat. Biotechnol. 22, 871–876.

Zuger, S., and Iwai, H. (2005). Intein-based biosynthetic incorporation of

unlabeled protein tags into isotopically labeled proteins for NMR studies.

Nat. Biotechnol. 23, 736–740.


Cytohesins Are CytoplasmicErbB Receptor ActivatorsAnke Bill,1,8 Anton Schmitz,1,8 Barbara Albertoni,1 Jin-Na Song,1 Lukas C. Heukamp,2 David Walrafen,3

Franziska Thorwirth,4 Peter J. Verveer,4 Sebastian Zimmer,2 Lisa Meffert,2 Arne Schreiber,3 Sampurna Chatterjee,5

Roman K. Thomas,5,6,7 Roland T. Ullrich,5 Thorsten Lang,3 and Michael Famulok1,*1LIMES Institute, Program Unit Chemical Biology & Medicinal Chemistry, Laboratory of Chemical Biology,

Rheinische Friedrich-Wilhelms-Universitat Bonn, Gerhard-Domagk-Str. 1, 53121 Bonn, Germany2Institute of Pathology, Universitatsklinikum, Rheinische Friedrich-Wilhelms-Universitat Bonn, Sigmund-Freud Strasse 25,

53123 Bonn, Germany3LIMES Institute, Program Unit Membrane Biology & Lipid Biochemistry, Laboratory of Membrane Biochemistry,

Rheinische Friedrich-Wilhelms-Universitat Bonn, Carl-Troll-Straße 31, 53115 Bonn, Germany4Department of Systemic Cell Biology, Max-Planck Institute of Molecular Physiology, Otto-Hahn-Str. 11, 44227 Dortmund, Germany5MaxPlanck Institute for Neurological Researchwith Klaus-Joachim-Zulch Laboratories of theMax Planck Society and theMedical Faculty of

the University of Koln, Gleueler Str. 50, 50931 Koln, Germany6Chemical Genomics Centre of the Max Planck Society, Otto-Hahn Str. 15, 44227 Dortmund, Germany7Center of Integrated Oncology and Department I of Internal Medicine, University of Koln, Kerpener Straße 62, 50937 Koln, Germany8These authors contributed equally to this work


DOI 10.1016/j.cell.2010.09.011

SUMMARY

Signaling by ErbB receptors requires the activation

of their cytoplasmic kinase domains, which is initi-

ated by ligand binding to the receptor ectodomains.

Cytoplasmic factors contributing to the activation

are unknown. Here we identify members of the cyto-

hesin protein family as such factors. Cytohesin inhi-

bition decreased ErbB receptor autophosphorylation

and signaling, whereas cytohesin overexpression

stimulated receptor activation. Monitoring epidermal

growth factor receptor (EGFR) conformation by

anisotropy microscopy together with cell-free recon-

stitution of cytohesin-dependent receptor autophos-

phorylation indicate that cytohesins facilitate confor-

mational rearrangements in the intracellular domains

of dimerized receptors. Consistent with cytohesins

playing a prominent role in ErbB receptor signaling,

we found that cytohesin overexpression correlated

with EGF signaling pathway activation in human

lung adenocarcinomas. Chemical inhibition of cyto-

hesins resulted in reduced proliferation of EGFR-

dependent lung cancer cells in vitro and in vivo.

Our results establish cytohesins as cytoplasmic

conformational activators of ErbB receptors that

are of pathophysiological relevance.

INTRODUCTION

ErbB receptors are key regulators of cell differentiation, survival,

proliferation, and migration, and aberrant ErbB receptor function

is a hallmark of many human cancers (Fischer et al., 2003; Bublil

and Yarden, 2007). The ErbB receptor family is comprised of four

members, the epidermal growth factor receptor (EGFR/ErbB1),

Her2/ErbB2, Her3/ErbB3, and ErbB4. Signaling is initiated by

growth factor binding to the extracellular domains of the ErbB

receptors. The ligand-induced conformational change in the

receptor ectodomains results in the association of the cyto-

plasmic tyrosine kinase domains of two receptor molecules.

This association has been considered to be sufficient for

releasing the default autoinhibited state of the kinase domains

(Ferguson, 2008; Bose and Zhang, 2009). However, the picture

appears to be more complex as only a fraction of the dimerized

ErbB receptors are catalytically active (Gadella and Jovin, 1995;

Moriki et al., 2001; Cui et al., 2002), and because receptor dimer-

ization seems to occur continuously and reversibly even in the

absence of ligand (Chung et al., 2010). Recent crystallographic

studies indicate that catalytic activity may be restricted to dimers

that show a special arrangement of the kinase domains, the so-

called asymmetric dimers (Zhang et al., 2006; Qiu et al., 2008;

Jura et al., 2009; Red Brewer et al., 2009). However, determi-

nants defining the fraction of active dimers that form within the

entire population of dimerized receptors remain elusive. This

fraction may simply depend on the rate of the spontaneous

conversion from the symmetric to the asymmetric dimer. Alter-

natively, the fraction of active dimers may not simply be defined

by receptor-inherent properties alone or by an equilibrium

between the two receptor dimer populations but be modulated

by cytoplasmic activator proteins. Such activators would endow

the cell with the possibility to fine-tune the number of actively

signaling receptors within a given pool of ligand-occupied recep-

tors according to cellular needs. However, cytoplasmic activa-

tors of ErbB receptors have not yet been identified.

Here, we report cytohesins as cytoplasmic ErbB receptor acti-

vators. The cytohesin family consists of four highly homologous

Cell 143, 201–211, October 15, 2010 ª2010 Elsevier Inc. 201

members, including ubiquitously expressed cytohesin-1, cyto-

hesin-2 (ARNO), cytohesin-3 (Grp1), and cytohesin-4, which is

exclusively found in cells of the immune system (Kolanus,

2007). Cytohesins are guanine nucleotide exchange factors

(GEFs) for ADP ribosylation factors (ARFs) that belong to the

family of small Ras-like GTPases. As in the case of other small

GTPases, ARF function critically depends on activation by

GEFs (Bos et al., 2007). Thus, because ARFs are involved in con-

trolling cytoskeletal dynamics, cell migration, vesicular traffic,

and signaling (Casanova, 2007; Kolanus, 2007), cytohesins are

important regulators of these processes.

We show that cytohesins enhance EGFR activation by directly

interacting with the cytoplasmic domains of dimerized receptors

and by facilitating conformational rearrangements of these

domains. Chemical inhibition and knockdown of cytohesins

reduce EGFR activation, whereas cytohesin overexpression

has the opposite effect. Our results strongly suggest that EGF

and cytohesins concertedly determine the degree of EGFR acti-

vation. We propose that whereas EGF exhibits its known func-

tion from the extracellular side, namely to relieve the autoinhibi-

tion of the unliganded receptor, cytohesins function to adjust

EGFR signaling from the cytoplasmic side by increasing the

number of EGFR dimers having the active, catalytically compe-

tent conformation within the reservoir of ligand-bound EGFR

dimers. This model is further supported by the finding that cyto-

hesin expression levels in human tumors correlate with EGFR

activation and signaling and that the chemical inhibition of cyto-

hesins reduces cell proliferation in vitro and tumor growth in

mice. Thus, cytohesins are introduced as intracellular EGFR acti-

vators that are relevant in the pathophysiology of certain

cancers.

RESULTS

Chemical Inhibition and Knockdown of Cytohesins

Reduce ErbB Receptor Signaling

To test whether cytohesins are involved in ErbB receptor

signaling, we used the specific cytohesin antagonist SecinH3

(Hafner et al., 2006; Bi et al., 2008). For this purpose, EGFR-

expressing human lung adenocarcinoma-derived H460 cells

were stimulated with EGF in the presence of SecinH3. Using

autophosphorylation as a readout, we observed that SecinH3-

treated cells showed an about 50% inhibition of EGFR activation

(Figure 1A). The inhibitory effect was also found at the level of the

adaptor proteins IRS1 and Shc and of the downstream kinases

p44/42 (Erk1/Erk2). A control compound (XH1009) that is struc-

turally related to SecinH3 but does neither bind nor inhibit cyto-

hesins (Bi et al., 2008) had no effect on EGFR activation and

signaling (Figure S1A available online). To obtain SecinH3-inde-

pendent evidence, the cytohesin-specific aptamer M69 (Mayer

et al., 2001) or cytohesin-specific siRNAs were used. Inhibition

of EGFR activation was observed in both experiments (Figures

S1B and S1C). The re-expression of cytohesin-2/ARNO in

siRNA-treated cells rescued the effect of ARNO knockdown on

EGFR autophosphorylation (Figure S2A, lanes 4 and 6).

We then analyzed whether cytohesins also affected the

signaling of Her2 and Her3, two other members of the ErbB

receptor family forming a heterodimer. When Her2/Her3-ex-

pressing human breast adenocarcinoma-derived SkBr3 cells

were treated with heregulin, SecinH3 reduced the phosphoryla-

tion of Her3 by about 50% (Figure 1B). This reduction in Her3

activation was mirrored in reduced activation of the adaptor

protein IRS1 and the downstream kinases Akt and p44/42.

Hsc70

pEGFR

pIRS1

EGFR

pShc

EGF

SecinH3

+

+

+

-

-

-

pp44/42

+-

+

+

- -

+-

+

+

- -

+-

- -

+-

+

+

- -

+-

+

+

- -

EGFSecinH3

pEGFR pIRS pShc pp44/42

rela

tive

in

ten

sity

***

*** *

totalEGFR

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Hsc70

pHER3

pIRS1

HER3

pp44/42

pAkt

HRG

SecinH3

+

+

+

-

-

-

pShc

+-

+

+

- -

+-

+

+

- -

+-

+

+

- -

+-

+

+

- -

+-

+

+

- -

+-

+

+

- -

HRGSecinH3

pHER3totalHER3 pIRS pAkt pShc pp44/42

rela

tive

in

ten

sity

** ** **

*

EGF + ++- +

ARNO (µg) - - .2 .4 .6

HRG + ++- +

FLAG

pHER3

Hsc70

Hsc70

FLAG

pEGFR

- - .2 .4 .6ARNO (µg)

HRG + ++- +

- - .2 .4 .6ARNO (µg)

+ ++- +

- - .2 .4 .6ARNO (µg)

EGF

A

B

C

D

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

+

+

***

EG

FR

ph

osp

ho

ryla

tio

nH

er3

ph

osp

ho

ryla

tio

n

0.0

1.0

2.0

3.0

4.0

0.0

1.0

2.0

3.0

Figure 1. Cytohesins Enhance Activation of ErbB Receptors

(A and B) The cytohesin inhibitor SecinH3 reduces ErbB receptor signaling.

Western blot analysis of H460 (A) or SkBr3 (B) cells treated with SecinH3 or

solvent and stimulated with EGF or heregulin (HRG), respectively, is shown.

Phosphorylation of the indicated proteins was determined by immunodetec-

tion using phosphospecific antibodies. Heat shock cognate protein 70

(Hsc70) served as loading control. The diagrams show relative phosphoryla-

tion levels after normalization for Hsc70. The untreated ligand-stimulated cells

were set as 1 (n = 6).

(C and D) Overexpression of the cytohesin ARNO enhances ErbB receptor

autophosphorylation. H460 (C) or SkBr3 (D) cells were transfected with

increasing amounts of FLAG-tagged ARNO and stimulated with ligand.

Receptor autophosphorylation was analyzed as above (n = 3).

Data are represented as mean ± SEM. See Figure S1 for further information.

202 Cell 143, 201–211, October 15, 2010 ª2010 Elsevier Inc.

The control compound XH1009 had no inhibitory effect (Fig-

ure S1D). Again, the involvement of cytohesins in the activation

of Her3 was confirmed by the aptamer M69 and by cytohesin-

specific siRNAs (Figures S1E and S1F).

Overexpression of ARNO Enhances EGFR Activation

Having shown that cytohesin inhibition and knockdown reduce

ErbB signaling, we asked whether overexpression of cytohesins

leads to an enhancement of EGF-stimulated EGFR activation.

For this analysis we have selected ARNO, which shows in both

H460 and SkBr3 cells higher expression than cytohesin-1 and

-3 (data not shown). When ARNO-transfected H460 cells were

stimulated with EGF, an ARNO-dependent increase in receptor

activation could be detected (Figure 1C). The same result was

seen in the Her2/Her3-expressing SkBr3 cells (Figure 1D). These

data show that ARNO, when overexpressed, enhances the

ligand-dependent activation of ErbB family members.

ARNO Enhances EGFR Activation Independently

of Its GEF Activity

The known function of ARNO is to act as a GEF on ARF proteins.

To analyze whether the GEF activity was also required for the

activation of the EGFR we made use of the GEF-inactive

ARNO mutant ARNO-E156K (Cherfils et al., 1998). Unexpect-

edly, overexpressed wild-type ARNO and ARNO-E156K were

equally potent in enhancing EGFR autophosphorylation (Fig-

ure 2A). The ability of ARNO-E156K to enhance EGFR activation

was not due to its overexpression as ARNO-E156K expressed at

endogenous protein level rescued the inhibition of EGFR auto-

phosphorylation induced by knockdown of endogenous ARNO

(Figure S2A, lanes 5 and 7). The mutant also stimulated Her2/

Her3 autophosphorylation (Figure 2B), suggesting that the GEF

activity is not required for the ARNO-mediated activation of

ErbB receptors. To substantiate this observation, we reduced

the expression of ARF1 or ARF6 by RNA interference. Neither

the knockdown of ARF1 nor that of ARF6 had an influence

on the activation of the EGFR (Figure S2B) or Her2/Her3 (Fig-

ure S2C). These results indicate that the cytohesin-mediated

activation of ErbB receptors does not involve these ARF

proteins, nor does it require the GEF function of the Sec7

domain, and thus implicate a hitherto unknown GEF-indepen-

dent function of ARNO.

As SecinH3 targets the Sec7 domain of the cytohesins (Hafner

et al., 2006; Bi et al., 2008), we asked whether this domain was

sufficient for EGFR activation or whether cytohesins’ pleck-

strin-homology (PH) and/or coiled-coil (CC) domains were also

required (Lim et al., 2010). Deletion studies showed that ARNO’s

Sec7 domain stimulated EGFR autophosphorylation as well as

the full-length protein (Figure 2C), attributing the EGFR-acti-

vating capability of the cytohesins to this domain.

ARNO Acts on Dimerized Receptors

Depending on determinants that are as yet incompletely under-

stood, ErbB receptor activation by growth factor ligands may

(Nagy et al., 1999) or may not (Abulrob et al., 2010) be accompa-

nied by receptor clustering. As the enhancement of EGFR activa-

tion by cytohesins could be due to an effect of cytohesins on

EGFR clustering, we examined by superresolution light micros-

copy (Hell and Wichmann, 1994) whether ARNO was involved

in the EGF-dependent EGFR clustering. We found a slight

increase in the measured EGFR cluster size upon EGF stimula-

tion, which was not affected by SecinH3 (Figure 3A and Figures

S3B and S3C), indicating that the reduction of EGFR signaling

observed after cytohesin inhibition is not a result of alterations

in cluster size at the observed 100 nm scale.

Cytohesins are involved in endocytosis (D’Souza-Schorey

and Chavrier, 2006) and thus could augment EGFR activa-

tion indirectly by modulating the endocytosis or degradation

of the EGFR. However, quantification of the EGFR at the

plasma membrane after EGF stimulation revealed no differ-

ence between untreated and SecinH3-treated cells, arguing

against this assumption (Figure 3B and Figure S3A). Generally,

EGFR activation by EGF enhances receptor endocytosis

(Sorkin and Goh, 2008) and thus might lead to the assumption

that the reduced EGFR activation after cytohesin inhibition

would slow down EGFR endocytosis. However, recently, it was

shown that receptor dimerization and not receptor activity is

a prerequisite for endocytosis (Wang et al., 2005). Therefore,

our finding that SecinH3 treatment does not reduce receptor

C

HRG + +-

- +-ARNOwt

+

-

- +- -ARNO E156K

Hsc70

pHER3

FLAG

EGF + +-

- +-ARNOwt

+

-

- +- -ARNO E156K

Hsc70

pEGFR

FLAG

A B

EGF + ++- +

Hsc70

FLAG

pEGFR

+

∆c

c

∆

∆∆

PH

Se

c7

FL

mo

ck

mo

ck

FL

CCPH

Sec7

ARNO

Figure 2. The Sec7 Domain Enhances the Autophosphorylation of

ErbB Receptors Independently of Its GEF Activity

(A and B) GEF-inactive ARNO enhances ErbB receptor autophosphorylation.

Shown is western blot analysis of protein lysates prepared from H460 (A) or

SkBr3 (B) cells transfectedwith FLAG-tagged wild-type ARNO or GEF-inactive

ARNO-E156K. Cells were stimulated with EGF or heregulin (HRG) and receptor

autophosphorylation was analyzed with phosphospecific antibodies.

(C) The Sec7 domain is sufficient for EGFR activation. H460 cells were trans-

fected with full-length ARNO (FL), with ARNO lacking the coiled-coil (DCC) or

the pleckstrin homology (DPH) domain, or with the isolated Sec7 domain

(Sec7). Autophosphorylation of the EGFR was determined as above.

See Figure S2 for further information.


internalization suggests that EGFR dimerization does not

depend on cytohesins.

To analyze the effect of cytohesins on receptor dimerization

more directly, H460 cells were preincubated with SecinH3, stim-

ulated, and treated with crosslinker to trap dimeric receptors.

Cytohesin inhibition did not affect receptor dimerization but

reduced the phosphorylation of the dimerized receptors (Fig-

ure 3C). Consistently, ARNO overexpression led to increased

phosphorylation of EGFR dimers, whereas it had no effect

on receptor dimerization (Figure 3D). The same results were

obtained for Her2/Her3 receptors in SkBr3 cells (Figures 3E

and 3F). These data suggest that ARNO facilitates the activation

of already dimerized ErbB receptors.

To obtain further evidence for this assumption, we analyzed

directly whether ARNO acts on dimeric receptors. A constitu-

tively dimerized EGFR (lz-EGFR; Figure 4A) was constructed

by replacing the extracellular domain of the receptor with a

dimerization module consisting of a leucine zipper and a single

cysteine residue that forms a disulfide bridge upon dimeriza-

tion (Stuhlmann-Laeisz et al., 2006). When the lz-EGFR was

expressed in HEK293 cells it was found exclusively as a dimer

(Figure S4A, upper panel). Consistent with its constitutive dimer-

ization, lz-EGFR was phosphorylated (Figure S4A, lower panel).

To test whether the activation of the lz-EGFR kinase domain

was dependent on the formation of the asymmetric dimer, the

effect of MIG6 on the autophosphorylation of the lz-EGFR was

analyzed. MIG6 inhibits receptor autophosphorylation by pre-

venting the formation of the active asymmetric EGFR dimer

(Zhang et al., 2007). Coexpression of the EGFR-binding domain

of MIG6 (MIG6-EBR), which is sufficient to inhibit EGFR signaling

(Anastasi et al., 2007), reduced lz-EGFR receptor autophosphor-

ylation, suggesting that the activation of the lz-EGFR depends

on the formation of the asymmetric dimer (Figure S4B). Thus,

regarding the allosteric activation of the kinase domains, the

lz-EGFR appears to behave like an authentic EGFR. Therefore,

the lz-EGFR is a suitable model to ask whether ARNO enhances

the activation of the EGFR kinase after its dimerization.

To address this question, ARNO activity was modulated in

lz-EGFR-expressing cells. In the presence of SecinH3, the auto-

phosphorylation of lz-EGFR was reduced (Figure 4B). The

control compound XH1009 had no effect (Figure S4C). Consis-

tently, overexpression of ARNO in these cells led to an increased

autophosphorylation of lz-EGFR (Figure 4C). These data pro-

vide strong evidence for the hypothesis that ARNO enhances

the activation of already dimerized EGFR, possibly by facilitating

conformational rearrangements.

ARNO Facilitates a Conformational Rearrangement

of the Cytoplasmic Domains of the Dimerized EGFR

To visualize conformational changes of the EGFR cytoplasmic

domains in living cells we tagged each molecule in the dimeric

lz-EGFR at the C terminus with the fluorescent protein mCitrine

(lz-EGFR-mCitrine). Like the untagged lz-EGFR, the fusion pro-

tein was constitutively dimerized and autophosphorylated (Fig-

ure S4D) and reached the plasma membrane, as visualized by

fluorescence microscopy on plasma membrane sheets (data

not shown), demonstrating that the mCitrine did not perturb

receptor function. Changes in the positions of the two mCitrine

moieties relative to each other result in changes in the fluores-

cence resonance energy transfer between these proteins (homo-

FRET). The efficiency of homo-FRET, which is exquisitely

HER3

pHER3

Hsc70

ARNO

+

-

+

+

-

-

phosphorylation

of Her3 dimers

0.5

1.0

1.5

2.0

0.0

2.5

***

+-ARNO

HRG

ARNO

phosphorylation

of EGFR dimers

0.5

1.0

1.5

2.0

0.0

2.5 ***

+-ARNO

+

-

+

+

-

-

EGF

ARNO

Hsc70

ARNO

pEGFR

EGFR

100

50

0

EGFR

fluorescence [%]

EGFSecinH3 -

-- +++

A

EGFSecinH3 -

-- +++

cluster size [nm]

B

D

HRG

SecinH3

-

-

+

-

+

+

+-SecinH3

0.4

0.8

0.0

1.2

*

phosphorylation

of Her3 dimersHER3

pHER3

Hsc70

E F

C

EGF

SecinH3

-

-

+

-

+

+

0.4

0.8

0.0

1.2

+-SecinH3

*

phosphorylation

of EGFR dimersEGFR

pEGFR

Hsc70

120

60

0

80

20

40

100**

Figure 3. Cytohesins Enhance the Phosphorylation but Not the

Dimerization of EGFR

(A) Cytohesins do not alter EGFR cluster size at the observed 100 nm scale.

SecinH3-treated or untreated H460 cells were stimulated with EGF, and EGFR

cluster sizes were determined by STED microscopy on plasma membrane

sheets. Each condition in each experiment (n = 3) includes 105–480 clusters

measured from 10–12 membrane sheets. *p < 0.05.

(B) SecinH3 does not affect EGF-triggered internalization of EGFR. SecinH3-

treated or untreated H460 cells were stimulated with EGF and the EGFR

remaining at the plasma membrane was quantified on plasma membrane

sheets by immunofluorescencemicroscopy. Statistical evaluation was of three

independent experiments each comprising the analysis of 26–66 membrane

sheets per condition.

(C–F) Cytohesins enhance phosphorylation of ErbB dimers. H460 (C and D) or

SkBr3 (E and F) cells were either treated with SecinH3 (C and E) or transfected

with ARNO (D and F), stimulated with ligand for 5 min and chemically cross-

linked. Receptor phosphorylation was analyzed by phosphospecific anti-

bodies. Arrows indicate receptor dimers. Diagrams show the phosphorylation

of the crosslinked, i.e., dimeric, receptors only after normalization for total

dimeric receptor (n = 9 for SecinH3 treatment, n = 5 for ARNO overexpression).

Data are represented as mean ± SEM. See Figure S3 for further information.


sensitive to both the distance and the orientation of the fluoro-

phores, can be determined by measuring the steady-state fluo-

rescence anisotropy of the cells (Squire et al., 2004). This tech-

nique has recently been used to monitor conformational

changes in the neurotrophin receptor (Vilar et al., 2009). To test

whether it is also suited to detect conformational changes in

the EGFR cytoplasmic domains, we expressed lz-EGFR-

mCitrine in COS-7 cells either alone, together with MIG6, or

together with Rheb. Whereas MIG6 is expected to change the

steady-state fluorescence anisotropy of lz-EGFR-mCitrine,

Rheb, which is not involved in EGFR signaling, should have no

effect. As expected, coexpression of MIG6-EBR led to a change

in the steady-state fluorescence anisotropy of lz-EGFR-mCitrine

whereas coexpression of Rheb did not (Figure S4E). Thus,

anisotropy measurements are suited to detect differences in

lz-EGFR-mCitrine conformation. To detect ARNO-dependent

conformational changes in the EGFR cytoplasmic domains,

lz-EGFR-mCitrine was expressed together with ARNO. The co-

expression of ARNO led to a decrease in anisotropy as com-

pared to lz-EGFR-mCitrine alone (Figure 4D). As ARNO neither

changed the fluorescence anisotropy of lz-mCitrine (which

does not contain the EGFR cytoplasmic domain) nor the fluores-

cence lifetime of lz-EGFR-mCitrine (data not shown), these

results indicate that ARNO coexpression resulted in an altered

conformation of the cytoplasmic domains of the EGFR dimer.

Although the geometries of the EGFR dimers in the EGFR-

ARNO and EGFR-MIG6 complexes are expected to be different,

we found in both cases a decrease in fluorescence anisotropy.

At first view, these results seem mutually contradictory as it

might intuitively be anticipated that changes in anisotropy

produced by an inhibitor would oppose those of an activator.

It should be noted, however, that anisotropy depends on both

the distance and the relative orientation of the fluorophores.

Therefore, even if the anisotropy is equal in two situations the

underlying geometry can be quite different. Although a specific

conformation thus cannot be deduced from a certain value of

anisotropy, a change in anisotropy is a reliable indicator for

a change in geometry (Vilar et al., 2009). Together with the anal-

ysis of receptor crosslinking and phosphorylation, these results

support the hypothesis that ARNO enhances receptor activation

by facilitating a conformational rearrangement of the cyto-

plasmic domains of the dimerized EGFR.

Cell-free Reconstitution of ARNO-Dependent EGFR

Activation

ARNO’s function as a conformational activator of the EGFR

implies ARNO and the EGFR to physically interact. Immunoflu-

orescence microscopy of plasma membrane sheets showed

that ARNO and the EGFR colocalize in H460 cells (Fig-

ure 5A). Moreover, coimmunoprecipitation of ARNO and the

EGFR indicated complex formation between the two pro-

teins (Figure 5B). To gain evidence for direct interaction of

ARNO and the cytoplasmic domain of the EGFR, a cell-free

C

D

0

-0.001

-0.002

-0.003

-0.004

-0.005

-0.007

-0.006

change in a

nis

otropy

ARNO + ++-

*

***

w/o ARNO + ARNO ++ ARNO

anisotropy0.19 0.26

B 1.2

SecinH3 - +0.0

0.2

0.4

0.6

0.81.0

**

plz-

EG

FR

/lz-E

GF

R

- +

Flag

plz-EGFR

Hsc70

SecinH3

Flag

plz-EGFR

ARNO

- + ARNO - +

***

0.0

1.0

2.0

3.0

plz-

EG

FR

/lz-E

GF

R

4.0

Hsc70

ARNO

A

kinase domain(709-984)

transmembrane segment (646-668)

Flag FlagS-S

leucine zipper

juxtamembrane region (669-709)

C-terminal region(985-1210)

lz-EGFR

Figure 4. Cytohesins Facilitate a Conforma-

tional Rearrangement of the Intracellular

Domains of EGFR Dimers

(A) Schematic of the constitutively dimerized lz-

EGFR. The extracellular domain of EGFR was re-

placed by a Flag-tagged disulfide-bridged leucine

zipper dimerization module.

(B and C) ARNO enhances the autophosphorlya-

tion of lz-EGFR. Shown are western blot analyses

of HEK293 cells transfected with lz-EGFR and

treated with SecinH3 (B) or cotransfected with

ARNO (C). The phosphorylation of lz-EGFR was

analyzed by phosphospecific antibodies (p-lz-

EGFR). Diagrams show receptor phosphorylation

after normalization for total receptor (n = 5). The

double bands in the FLAG blots correspond to un-

phosphorylated (lower) and phosphorylated

(upper) lz-EGFR.

(D) ARNO facilitates a conformational rearrange-

ment of the intracellular domains of constitutively

dimerized EGFR. For fluorescence anisotropy

microscopy, the C termini of both EGFRmolecules

in lz-EGFR were tagged with mCitrine (lz-EGFR-

mCitrine). COS-7 cells were cotransfected with

lz-EGFR-mCitrine and empty vector (left) or

together with increasing amounts of ARNO

(middle and right). Homo-FRET between the two

mCitrinemoieties was determined by steady-state

fluorescence anisotropymicroscopy. The diagram

shows the statistic evaluation of five experiments,

each covering 25 fields of view with 1–4 cells.

Data are represented as mean ± SEM. See

Figure S4 for further information.


reconstitution system was used. The complete cytoplasmic

domain of the EGFR (EGFR-ICD) and ARNO were heterolo-

gously expressed (Figures S5A and S5B), and the interac-

tion of the purified, FITC-labeled proteins was analyzed by fluo-

rescence anisotropy measurements (Figure 5C). Full-length

ARNO, the isolated Sec7 domain, and the GEF-inactive Sec7-

E156K bound to the EGFR-ICD with apparent dissociation

constants around 1 mM. Segment 1 of MIG6-EBR (MIG6-S1),

a known binding partner of the EGFR-ICD (Zhang et al.,

2007), bound with a dissociation constant (KD) around 2 mM.

No binding was observed between lysozyme and EGFR-ICD,

nor did ARNO full-length or ARNO-Sec7 show binding to

MIG6-S1 (Figure 5C), indicating that the observed binding is

specific. EGFR-ICD lacking the C-terminal 188 amino acids

(EGFR-ICD1022) bound to ARNO-Sec7 with the same affinity

as the complete EGFR-ICD confining ARNO’s binding site

to the kinase or juxtamembrane domains of the EGFR.

In agreement with ARNO functioning upstream of EGFR auto-

phosphorylation, the binding of ARNO did not require phos-

phorylation of the EGFR-ICD (Figure S5C).

Due to the presence of the juxtamembrane segment, EGFR-

ICD forms a dimer resembling the intracellular domains of the

ligand-bound EGFR (Jura et al., 2009) and thus can be used

to analyze the autophosphorylation of the EGFR in a cell-free

system. To test whether the conformational requirements for

the activation of the authentic EGFR are preserved in EGFR-

ICD, an autophosphorylation reaction of EGFR-ICD was per-

formed in the presence of MIG6-S1, which inhibits the forma-

tion of the asymmetric dimer of the EGFR (Zhang et al.,

2007). MIG6-S1 reduced the autophosphorylation of EGFR-

ICD (Figure S5D), indicating that the activation of the EGFR-

ICD kinase depends on the formation of the asymmetric dimer.

Addition of GST had no effect (Figure S5D). When ARNO was

added to an autophosphorylation reaction of EGFR-ICD,

increased autophosphorylation was found (Figure 5D). A similar

level of stimulation was seen when the isolated Sec7 domain

5 µm

1 µm

EGFR ARNO / cytohesin-1 overlay

C

0' 3'1'

ARNO-Sec7-

E156K

0' 3'1'

ARNO-FL-

wt

0' 3'1'

ARNO-Sec7-

wt

0' 3'1'

-

pY

EGFR-ICD

ARNO-FL

ARNO-Sec7

D

KD 1,2 ± 0,2 µMLigandFITC

ARNO-Sec7-wt

ARNO-Sec7-wt

ARNO-FL-wt

ARNO-FL-wt

MIG6-S1

MIG6-S1

lysozyme

EGFR-ICD

EGFR-ICD

EGFR-ICD

EGFR-ICD

KD 1,1 ± 0,1 µM KD 2,1 ± 0,2 µM

n.b.

n.b.

n.b.

ARNO-Sec7-E156K

MIG6-S1

EGFR-ICD KD 1,1 ± 0,2 µM

1000 2000 30000

50

100

150

200

ligand [nM]

ch

an

ge

in

an

iso

tro

py

ARNO-Sec7-wt EGFR-ICD1022

lysozyme EGFR-ICD1022

KD 1,2 ± 0,2 µM

n.b.

B

EGFR

ARNO

EG

FR

co

ntr

ol

IP

blo

t

A Figure 5. ARNO Stimulates Autophosphor-

ylation of EGFR by Direct Interaction

(A) ARNO colocalizes with EGFR. Plasma mem-

brane sheets were immunostained for EGFR

(red channel, left panels) and ARNO/cytohesin-1

(green channel, middle panels). Right panels

show corresponding overlays. To quantify coloc-

alization, circles were superimposed concen-

trically on selected spots in the red channel

and transferred to identical pixel locations in

the green channel. Continuous and dashed

circles indicate positive and negative colocaliza-

tion signals, respectively. 62% ± 5% of

the EGFR spots were positive for ARNO

(n = 3).

(B) Coimmunoprecipitation of ARNO with EGFR.

EGFR was immunoprecipitated from H460 cells

with agarose-coupled anti-EGFR. Coprecipitated

ARNO was detected by an ARNO-specific anti-

body. Agarose-coupled normal mouse IgG was

used as control matrix.

(C) ARNO interacts with the intracellular domain

of the EGFR (EGFR-ICD) in vitro. The indicated

protein was labeled with FITC and the unlabeled

ligand was added at increasing concentrations.

Binding was measured by fluorescence anisot-

ropy. KD values were calculated assuming a 1:1

stoichiometry (n=4) and are given as mean ±

SEM. n.b., no binding.

(D) ARNO enhances autophosphorylation of

EGFR-ICD. The indicated ARNO construct and

EGFR-ICD were incubated in vitro. Autophos-

phorylation was initiated by addition of ATP.

Samples were taken at the indicated time points

and analyzed using antiphosphotyrosine antibody

(pY). EGFR-ICD and ARNO constructs were

detected with anti-His-antibody.



or Sec7-E156K was added. Taken together with the data

obtained in the cellular assays, these results strongly argue

for cytohesins acting on the intracellular domains of dimerized

EGFR as conformational activators.

Cytohesin Overexpression Correlates with Enhanced

EGFR Signaling in Human Lung Cancers

Enhanced EGFR signaling is known to be a hallmark in many

cancers. Having shown that ARNO enhances EGFR activation

in H460 cells, we wondered whether ARNO or other cytohesins

might be overexpressed in lung cancer. To address this ques-

tion, we immunostained primary human lung adenocarcinomas

with an antibody detecting ARNO and cytohesin-1. Whereas

normal lung tissue showed only background or weak staining,

82% of the carcinomas showed moderate or strong ARNO/

cytohesin-1 staining (Figure 6A), demonstrating cytohesin upre-

gulation in a large fraction of lung adenocarcinomas. According

to our in vitro data, increased cytohesin expression should

result in enhanced EGFR autophosphorylation in these tumors.

pEGFR

pp42/pp44

ARNO/

cytohesin-1

pAkt

A

B

C

D

0

20

40

60

80

100]%[ l

atot f

o n

oitcarf 1 2 30

cytohesin score

strongmoderateweakbackground

staining

0

20

40

60

80

100]%[ l

a tot f

o n

oitcarf 1 2 30

cytohesin score

0

20

40

60

80

100]%[ l

ato t f

o n

oitca rf 1 2 30

cytohesin score

cytohesin score

frequencies [%]

53

2

29

16stainingscore

3 strong2 moderate1 weak0 background

cytohesin score

0 3

Figure 6. High Expression Levels of ARNO/

Cytohesin-1 Correlate with Increased EGFR

Signaling in Human Lung Adenocarcinomas

Consecutive sections of resected human lung

adenocarcinomas were stained for ARNO/cytohe-

sin-1 (A), pEGFR (B), pAkt (C), pp44/42 (D). Repre-

sentative images of tumors with background (left

column) or strong (right column) ARNO/cytohe-

sin-1 expression are shown. The diagram in (A)

shows the fraction of tumors with background

(score 0), weak (score 1), moderate (score 2), or

strong (score 3) staining for ARNO/cytohesin-1.

The diagrams in (B)–(D) depict the phosphorylation

levels of the respective protein in correlation to the

cytohesin score (p = 0.002 for pEGFR, p = 0.002

for pAkt, p = 0.025 for pp44/42, n = 45).


Indeed, we found a highly significant

(p = 0.002) correlation between the

expression level of ARNO/cytohesin-1

and the level of EGFR autophosphoryla-

tion (Figure 6B) in consecutive sections

of tumor tissue. Immunofluorescence

double-staining of phosphorylated EGFR

and ARNO further supported this correla-

tion (Figure S6). The increased EGFR

phosphorylation was not due to overex-

pression of the receptor because total

EGFR expression was independent of

the ARNO/cytohesin-1 expression (p =

0.581). The phosphorylation of Akt (Fig-

ure 6C) and p44/42 (Erk1/Erk2) (Fig-

ure 6D) was also significantly correlated

with higher ARNO/cytohesin-1 expres-

sion (p = 0.002 and p = 0.025, respec-

tively), suggesting that the enhanced acti-

vation is not restricted to the EGFR itself

but continues along these two major branches of the EGF

signaling pathway.

SecinH3 Reduces Growth of EGFR-Dependent Lung

Tumor Xenografts

The strong expression of ARNO/cytohesin-1 in tumor tissue

raised the question of whether cytohesins may, by enhanced

EGFR signaling, promote the proliferation of the tumor cells. To

test this possibility, the proliferation rate of the EGFR-dependent

lung cancer cell line PC9 was determined in the presence or

absence of SecinH3. Indeed, the inhibition of cytohesins led to

a strong reduction of the proliferation of PC9 cells (Figure 7A).

Because the inhibition of EGFR signaling in EGFR-dependent

cells results in cell-cycle arrest and the induction of apoptosis

(Sharma et al., 2007), we examined SecinH3-treated PC9 cells

for cell-cycle arrest and apoptosis. We found an increase of cells

in the G1 phase of the cell cycle and a concomitant decrease of

cells in S and G2/M phases, indicative of SecinH3 inducing an

arrest in G1 of the cell cycle (Figure 7B). Accordingly, Annexin


V staining showed that SecinH3 treatment led to an increase of

apoptotic cells (Figure 7C). To test whether SecinH3 treatment

reduced tumor growth in vivo, tumor xenografts were generated

by subcutaneous injection of PC9 cells into nude mice. Cell

proliferation in the tumors was followed by [18F]-fluoro-L-thymi-

dine uptake positron emission tomography ([18F]FLT PET)

(Shields et al., 1998). The tumors in the SecinH3-treated mice

showed significantly less uptake of [18F]FLT (Figure 7D), indi-

cating reduced tumor growth. Further, immunohistochemical

staining of the cell proliferation marker Ki-67 (Gerdes et al.,

1983) in resected tumors confirmed reduced cell proliferation

(Figure 7E), and TUNEL staining showed an increase in apoptotic

cells in the tumors of SecinH3-treated animals (Figure 7F). Taken

together, these data demonstrate that the chemical inhibition of

A B

1.2

0

%-i

nj.

do

se

1.2

0

%-i

nj.

do

se

day 0

day 7

C

10

20

30

-10

-20

-30

-40

0

% change in maxFLT uptake untreated

SecinH3

**

SecinH3 untreated

0

1

3

4

***

TUNEL positive cells

per field of view

SecinH3 +-

2

5

SecinH3 +-

0

10

20

30

***

% apoptotic cells

SecinH3 +-

D

G2/M-phase

S-phaseG1-phase

0

20

40

60

80

100

% cells

SecinH3 +-

E

Ki-67

untreated SecinH3F

0,0

0,2

0,4

0,6

0,8

1,0

1,2

relative cell number

***

Figure 7. SecinH3 Inhibits Growth of EGFR-

Dependent Lung Tumor Xenografts

(A) SecinH3 inhibits proliferation of PC9 cells. The

diagram shows the relative cell number (MTT

assay) after 72 hr treatment with SecinH3 or

DMSO. The cell number in the solvent-treated

samples was set to 1. ***p < 0.001, n = 9.

(B) SecinH3 induces G1 arrest in PC9 cells. PC9

cells were treated with SecinH3 or solvent for

24 hr, fixed, stained with TOPRO-3, and analyzed

by flow cytometry. The diagram shows the

percentage of cells in the indicated cell-cycle

phases. ***p < 0.001, n = 6.

(C) SecinH3 induces apoptosis in PC9 cells.

Annexin V FACS was performed after 48 hr treat-

ment with SecinH3 or solvent. The diagram shows

the percentage of apoptotic cells. ***p < 0.001,

n = 3.

(D) [18F]FLT PET indicates response to SecinH3.

Representative [18F]FLT PET images of mice

bearing PC9 xenografts before and 7 days after

treatment with SecinH3 or carrier (DMSO). **p <

0.01, n = 7.

(E) SecinH3 decreases proliferation of PC9 xeno-

grafts. Ki-67 staining of PC9 xenograft tumors in

nude mice after treatment with carrier or SecinH3

for 7 days.

(F) SecinH3 induces apoptosis in PC9 xenografts.

TUNEL assay of PC9 xenograft tumors in nude

mice after treatment with carrier or SecinH3 for

7 days. The diagram shows the number of TUNEL-

positive cells per high power microscopic field.

Per treatment group, 10 representative fields

were counted. ***p < 0.001.

Data are represented as mean ± SEM.

cytohesins reduces the proliferation of

EGFR-dependent tumor cells in vitro

and in vivo.

DISCUSSION

In the present study, we identify cytohe-

sins as ErbB receptor activators that

enhance receptor activation by direct

interaction with the cytoplasmic domain

of the receptor. The importance of this kind of ErbB receptor

activator is underlined by the findings that increased cytohesin

expression correlates with increased EGFR activation and sig-

naling in human lung cancers, and that the chemical inhibition

of cytohesins reduces the proliferation of EGFR-dependent

lung cancer cells in vitro and in mice. Except for Dok-7, cyto-

plasmic activators have not been described for any receptor

tyrosine kinase. Dok-7 enhances the activity of the muscle-

specific receptor kinase MuSK by dimerizing partially autophos-

phorylated and thus partially activated receptor monomers

(Inoue et al., 2009; Bergamin et al., 2010). In contrast, cytohesins

do neither influence receptor dimerization nor require receptor

autophosphorylation for binding but function as conformational

activators of receptor dimers.


From crystallographic, biochemical, and biophysical data it is

becoming increasingly evident that EGFR dimerization and acti-

vation of the kinase domains are distinctly regulated and thor-

oughly balanced processes, but the mechanisms by which this

balance is achieved are largely elusive. The fundamental model

of EGFR activation held that the activation of the EGFR kinase

results from the EGF-dependent dimerization of the receptor

cytoplasmic domains (Yarden and Schlessinger, 1987). This

model had to be extended when it was shown that the mere

dimerization of the EGFR is not sufficient for activation (Gadella

and Jovin, 1995; Moriki et al., 2001; Cui et al., 2002; Chung et al.,

2010). Recent crystallographic studies strongly suggest that

only a subset of the dimers that adopt a distinct conformation

called the asymmetric dimers, where one kinase acts as an allo-

steric activator for the other, are catalytically active (Zhang et al.,

2006; Jura et al., 2009; Red Brewer et al., 2009). Integration of

these data into the prior model led to the currently prevailing

model of EGFR activation according to which the activation of

the EGFR kinase results from the intrinsic ability of the receptor

kinase domains to form active (asymmetric) dimers as soon as

they are released from their default autoinhibited state (Fergu-

son, 2008; Bose and Zhang, 2009). The only activator required

in this model is the ligand EGF, which binds to the ectodomain

of the receptor and thereby induces and/or stabilizes the

structural rearrangements that release the kinase domains

from their autoinhibited state. Our finding that EGFR activation

is enhanced by cytohesins both in cells and in a cell-free recon-

stitution system indicates that EGFR activation is likely not

comprehensively explained by ligand-induced release from

autoinhibition and the subsequent spontaneaus formation of

the asymmetric dimer. The existence of cytoplasmic EGFR acti-

vators like cytohesins does not preclude receptor activation to

occur in their absence as seen for EGFR-ICD in our cell-free au-

tophosphorylation experiments and as seen for near-full length

EGFR in experiments by others (Mi et al., 2008; Qiu et al.,

2009). Our results implicate, however, a further extension of

the current model of EGFR activation to include additional layers

of regulation.

Indeed, in a cellular context, the transition from the inactive

symmetric to the active asymmetric dimer represents a stage

where additional layers of modulation of receptor activation,

inhibitory as well as stimulatory, might come into play. Recently,

MIG6 was identified as an inhibitor of EGFR signaling (Ferby

et al., 2006; Anastasi et al., 2007; Reschke et al., 2009) that

acts by blocking the formation of the asymmetric dimer (Zhang

et al., 2007), indicating that a layer of negative regulation appears

actually implemented. Cytohesins represent an example of

a class of EGFR activators that may form a layer of positive regu-

lation by facilitating the structural rearrangements required to

convert the receptor dimer into its active conformation. It is

important to point out that the existence of cytoplasmic EGFR

activators does not abolish ligand dependency of receptor acti-

vation because the autoinhibition that is imposed by the extra-

cellular domains on the kinase domain (Zhu et al., 2003) still

has to be released by ligand binding. Such activators do,

however, allow the cell to modulate, for a given amount of

ligand-bound receptor, the number of activated receptors

according to cellular needs.

On the other hand, dysregulation of cytoplasmic EGFR activa-

tors like the cytohesin ARNO might result in inappropriately

activated EGFR signaling. Enhanced EGFR signaling is a charac-

teristic feature of several cancers including non-small cell

lung cancers (Gazdar, 2009). Cancer cells that critically depend

on a specific signaling molecule for growth and survival are

addicted to that oncogene (Weinstein, 2002), and those lung

cancers that respond to EGFR tyrosine kinase inhibitor therapy

are addicted to EGFR (Sharma et al., 2007). Themajority of these

tumors have either upregulated or mutant EGFR (Lynch et al.,

2004; Paez et al., 2004; Pao et al., 2004). Nevertheless, a signif-

icant fraction of lung cancers with apparently normal EGFR

status also respond to EGFR inhibitors, reflecting their EGFR

addiction (Sharma and Settleman, 2009). How these tumor cells

maintain a sufficient level of EGFR signaling to satisfy their EGFR

addiction is currently unclear. Our observation that ARNO over-

expression is associated with an activated EGF signaling path-

way in human lung adenocarcinoma provides a possible expla-

nation for the EGFR addiction of these cancer cells that have

neither mutant nor overexpressed EGFR. Our finding that

the proliferation of EGFR-dependent tumor cells is drastically

reduced by inhibition of cytohesins underlines the pathophysio-

logical significance of intracellular ErbB receptor activators like

ARNO and opens up avenues for fighting ErbB receptor-depen-

dent cancers by targeting not the receptors themselves but their

activators.

EXPERIMENTAL PROCEDURES

For detailed protocols allowing reproduction of the experiments, see Extended

Experimental Procedures.

Immunoblotting/Immunoprecipitation

Cells were serum-starved overnight in the presence of SecinH3 or DMSO and

stimulated for 5 min with EGF or heregulin-b1. Proteins were first immunopre-

cipitated or directly analyzed by SDS-PAGE and immunoblotting. Visualization

was done by enhanced chemiluminescence or by fluorescence-labeled

secondary antibodies.

Crosslinking

Cells were starved overnight in the presence of SecinH3 or DMSO. Directly

after stimulation (5 min), proteins were crosslinked by adding BS3 and

analyzed by SDS-PAGE and immunoblotting.

Anisotropy Microscopy

Anisotropy microscopy was done as described (Squire et al., 2004) in COS-7

cells.

STED Microsocopy and Immunofluorescence Microscopy

Membrane sheets were generated essentially as previously described (Lang

et al., 2001) and visualized either by epi-fluorescence or stimulated emission

depletion (STED) microscopy.

Cell-free Fluorescence Anisotropy and Autophosphorylation Assays

Fluorescein-labeled ARNO, ARNO-Sec7-WT/E156K, MIG6-EBR, or lysozyme

was mixed with unlabeled EGFR-ICD or MIG6-EBR at room temperature, and

fluorescence anisotropy was measured in a microplate reader. For the auto-

phosphorylation assays, EGFR-ICD was incubated with the indicated protein

in the presence of ATP at room temperature. After the indicated time, aliquots

were removed, separated by SDS-PAGE, and analyzed by immunoblotting.


Tumor Samples

All tumor samples stem from the CIO Biobank at the Institute of Pathology,

University of Bonn, Germany. All tumors were clinically and pathologically

identified as being the primary and only neoplastic lesion and classified

according to World Health Organization (WHO) guidelines (Brambilla et al.,

2001). Sections were stained and evaluated as previously described (Heu-

kamp et al., 2006; Zimmer et al., 2008). Staining intensities were individually

evaluated by three independent observers using a four-tier scoring system

as described before (Zimmer et al., 2008). Immunofluorescence double-stain-

ing of tumor sections was performed as described (Friedrichs et al., 2007).

Proliferation and Apoptosis Assays

PC9 cells were treated with SecinH3 or solvent in medium containing 1% FCS.

Proliferation was analyzed after 3 days using aMTT assay. Apoptosis and cell-

cycle status were determined after 2 days by Annexin V and TOPRO-3 staining

and fluorescence-activated cell sorting (FACS) analysis.

[18F]FLT PET Imaging of Tumor Xenografts

nu/nu athymic mice that had been subcutaneously injected with PC9 cells

were treated with SecinH3 or DMSO for 7 days. After [18F]FLT (30-deoxy-30-

[F-18]fluorothymidine) administration tumors were visualized using a FOCUS

microPET scanner.

Statistics

Results are given as the mean ± standard error of the mean (SEM). Statistical

analyses were performed with Prism (GraphPad Software) applying the two-

tailed t test or one-way ANOVA, as appropriate. All datasets passed the

Kolmogorov and Smirnov test for Gaussian distribution. For the analysis of

the tumor samples the Spearman nonparametric correlation test was used.

Differences of means were considered significant at a significance level

of 0.05.


Supplemental Information includes Extended Experimental Procedures and

six figures and can be found with this article online at doi:10.1016/j.cell.

2010.09.011.

ACKNOWLEDGMENTS

We thank S. Rose-John for plasmid pMWOS-L-gp130, K. Nishio for PC9 cells,

the Department of Nanobiophotonics, MPI Gottingen for Atto647N coupled

secondary antibodies and access to STED microscopy, Silvio Rizzoli for

providing MatLab routines for image analysis, Philippe I.H. Bastiaens for

advice on the anisotropy measurements, V. Fieberg and Y. Aschenbach-

Paul for technical assistance, J. Hannam, A.M. Hayallah, and X.-H. Bi for the

synthesis of SecinH3 and XH1009, B. Neumaier for the synthesis of [18F]FLT,

and the members of the Famulok laboratory for helpful discussions. This

work was supported by grants from the DFG, the SFBs 645, 704, and 832,

and the GRK804. The CIO Biobank is supported by the Deutsche Krebshilfe.

A.B. and B.A. thank the Fonds der Chemischen Industrie and the Roche

Research Foundation for scholarships. R.K.T. is supported by the Deutsche

Krebshilfe, Fritz-Thyssen-Stiftung, and the BMBF NGFNplus-program. A.S.

and M.F. are co-owners of a patent application for SecinH3.

Received: April 20, 2010

Revised: July 13, 2010

Accepted: August 10, 2010

Published: October 14, 2010

REFERENCES

Abulrob, A., Lu, Z., Baumann, E., Vobornik, D., Taylor, R., Stanimirovic, D., and

Johnston, L.J. (2010). Nanoscale imaging of epidermal growth factor receptor

clustering. J. Biol. Chem. 285, 3145–3156.

Anastasi, S., Baietti, M.F., Frosi, Y., Alema, S., and Segatto, O. (2007). The

evolutionarily conserved EBR module of RALT/MIG6 mediates suppression

of the EGFR catalytic activity. Oncogene 26, 7833–7846.

Bergamin, E., Hallock, P.T., Burden, S.J., and Hubbard, S.R. (2010). The cyto-

plasmic adaptor protein Dok7 activates the receptor tyrosine kinase MuSK via

dimerization. Mol. Cell 39, 100–109.

Bi, X., Schmitz, A., Hayallah, A.M., Song, J.N., and Famulok,M. (2008). Affinity-

based labeling of cytohesins with a bifunctional SecinH3 photoaffinity probe.

Angew. Chem. Int. Ed. Engl. 47, 9565–9568.

Bos, J.L., Rehmann, H., and Wittinghofer, A. (2007). GEFs and GAPs: Critical

elements in the control of small G proteins. Cell 129, 865–877.

Bose, R., and Zhang, X. (2009). The ErbB kinase domain: Structural perspec-

tives into kinase activation and inhibition. Exp. Cell Res. 315, 649–658.

Brambilla, E., Travis, W.D., Colby, T.V., Corrin, B., and Shimosato, Y. (2001).

The newWorld Health Organization classification of lung tumours. Eur. Respir.

J. 18, 1059–1068.

Bublil, E.M., and Yarden, Y. (2007). The EGF receptor family: spearheading

a merger of signaling and therapeutics. Curr. Opin. Cell Biol. 19, 124–134.

Casanova, J.E. (2007). Regulation of Arf activation: the Sec7 family of guanine

nucleotide exchange factors. Traffic 8, 1476–1485.

Cherfils, J., Menetrey, J., Mathieu, M., Le Bras, G., Robineau, S., Beraud-

Dufour, S., Antonny, B., and Chardin, P. (1998). Structure of the Sec7 domain

of the Arf exchange factor ARNO. Nature 392, 101–105.

Chung, I., Akita, R., Vandlen, R., Toomre, D., Schlessinger, J., and Mellman, I.

(2010). Spatial control of EGF receptor activation by reversible dimerization on

living cells. Nature 464, 783–787.

Cui, T.-X., Nakagami, H., Nahmias, C., Shiuchi, T., Takeda-Matsubara, Y., Li,

J.-M., Wu, L., Iwai, M., and Horiuchi, M. (2002). Angiotensin II subtype 2

receptor activation inhibits insulin-induced phosphoinositide 3-kinase and

Akt and induces apoptosis in PC12W cells. Mol. Endocrinol. 16, 2113–2123.

D’Souza-Schorey, C., and Chavrier, P. (2006). ARF proteins: roles in

membrane traffic and beyond. Nat. Rev. Mol. Cell Biol. 7, 347–358.

Ferby, I., Reschke, M., Kudlacek, O., Knyazev, P., Pante, G., Amann, K., Som-

mergruber, W., Kraut, N., Ullrich, A., Fassler, R., and Klein, R. (2006). Mig6 is

a negative regulator of EGF receptor-mediated skin morphogenesis and tumor

formation. Nat. Med. 12, 568–573.

Ferguson, K.M. (2008). Structure-based view of epidermal growth factor

receptor regulation. Annu. Rev. Biophys. 37, 353–373.

Fischer, O.M., Hart, S., Gschwind, A., and Ullrich, A. (2003). EGFR signal trans-

activation in cancer cells. Biochem. Soc. Trans. 31, 1203–1208.

Friedrichs, N., Steiner, S., Buettner, R., and Knoepfle, G. (2007). Immunohisto-

chemical expression patterns of AP2alpha and AP2gamma in the developing

fetal human breast. Histopathology 51, 814–823.

Gadella, T.W., Jr., and Jovin, T.M. (1995). Oligomerization of epidermal growth

factor receptors on A431 cells studied by time-resolved fluorescence imaging

microscopy. A stereochemical model for tyrosine kinase receptor activation.

J. Cell Biol. 129, 1543–1558.

Gazdar, A.F. (2009). Activating and resistance mutations of EGFR in non-

small-cell lung cancer: role in clinical response to EGFR tyrosine kinase inhib-

itors. Oncogene 28, S24–S31.

Gerdes, J., Schwab, U., Lemke, H., andStein, H. (1983). Production of amouse

monoclonal antibody reactive with a human nuclear antigen associated with

cell proliferation. Int. J. Cancer 31, 13–20.

Hafner, M., Schmitz, A., Grune, I., Srivatsan, S.G., Paul, B., Kolanus, W.,

Quast, T., Kremmer, E., Bauer, I., and Famulok, M. (2006). Inhibition of cytohe-

sins by SecinH3 leads to hepatic insulin resistance. Nature 444, 941–944.

Hell, S.W., andWichmann, J. (1994). Breaking the diffraction resolution limit by

stimulated emission: stimulated-emission-depletion fluorescence micros-

copy. Opt. Lett. 19, 780–782.

Heukamp, L.C., Fischer, H.P., Schirmacher, P., Chen, X., Breuhahn, K., Nico-

lay, C., Buttner, R., and Gutgemann, I. (2006). Podocalyxin-like protein 1


expression in primary hepatic tumours and tumour-like lesions. Histopa-

thology 49, 242–247.

Inoue, A., Setoguchi, K., Matsubara, Y., Okada, K., Sato, N., Iwakura, Y., Higu-

chi, O., and Yamanashi, Y. (2009). Dok-7 activates the muscle receptor kinase

MuSK and shapes synapse formation. Sci. Signal. 2, ra7.

Jura, N., Endres, N.F., Engel, K., Deindl, S., Das, R., Lamers, M.H., Wemmer,

D.E., Zhang, X., and Kuriyan, J. (2009). Mechanism for activation of the

EGF receptor catalytic domain by the juxtamembrane segment. Cell 137,

1293–1307.

Kolanus, W. (2007). Guanine nucleotide exchange factors of the cytohesin

family and their roles in signal transduction. Immunol. Rev. 218, 102–113.

Lang, T., Bruns, D., Wenzel, D., Riedel, D., Holroyd, P., Thiele, C., and Jahn, R.

(2001). SNAREs are concentrated in cholesterol-dependent clusters that

define docking and fusion sites for exocytosis. EMBO J. 20, 2202–2213.

Lim, J., Zhou, M., Veenstra, T.D., and Morrison, D.K. (2010). The CNK1 scaf-

fold binds cytohesins and promotes insulin pathway signaling. Genes Dev.

24, 1496–1506.

Lynch, T.J., Bell, D.W., Sordella, R., Gurubhagavatula, S., Okimoto, R.A.,

Brannigan, B.W., Harris, P.L., Haserlat, S.M., Supko, J.G., Haluska, F.G.,

et al. (2004). Activating mutations in the epidermal growth factor receptor

underlying responsiveness of non-small-cell lung cancer to gefitinib. N. Engl.

J. Med. 350, 2129–2139.

Mayer, G., Blind, M., Nagel, W., Bohm, T., Knorr, T., Jackson, C.L., Kolanus,

W., and Famulok, M. (2001). Controlling small guanine-nucleotide-exchange

factor function through cytoplasmic RNA intramers. Proc. Natl. Acad. Sci.

USA 98, 4961–4965.

Mi, L.-Z., Grey, M.J., Nishida, N., Walz, T., Lu, C., and Springer, T.A. (2008).

Functional and structural stability of the epidermal growth factor receptor

in detergent micelles and phospholipid nanodiscs. Biochemistry 47,

10314–10323.

Moriki, T., Maruyama, H., and Maruyama, I.N. (2001). Activation of preformed

EGF receptor dimers by ligand-induced rotation of the transmembrane

domain. J. Mol. Biol. 311, 1011–1026.

Nagy, P., Jenei, A., Kirsch, A.K., Szollosi, J., Damjanovich, S., and Jovin, T.M.

(1999). Activation-dependent clustering of the erbB2 receptor tyrosine

kinase detected by scanning near-field optical microscopy. J. Cell Sci. 112,

1733–1741.

Paez, J.G., Janne, P.A., Lee, J.C., Tracy, S., Greulich, H., Gabriel, S., Herman,

P., Kaye, F.J., Lindeman, N., Boggon, T.J., et al. (2004). EGFR mutations in

lung cancer: Correlation with clinical response to gefitinib therapy. Science

304, 1497–1500.

Pao, W., Miller, V., Zakowski, M., Doherty, J., Politi, K., Sarkaria, I., Singh, B.,

Heelan, R., Rusch, V., Fulton, L., et al. (2004). EGF receptor genemutations are

common in lung cancers from ‘‘never smokers’’ and are associated with sensi-

tivity of tumors to gefitinib and erlotinib. Proc. Natl. Acad. Sci. USA 101,

13306–13311.

Qiu, C., Tarrant, M.K., Choi, S.H., Sathyamurthy, A., Bose, R., Banjade, S., Pal,

A., Bornmann, W.G., Lemmon,M.A., Cole, P.A., and Leahy, D.J. (2008). Mech-

anism of activation and inhibition of the HER4/ErbB4 kinase. Structure 16,

460–467.

Qiu, C., Tarrant, M.K., Boronina, T., Longo, P.A., Kavran, J.M., Cole, R.N.,

Cole, P.A., and Leahy, D.J. (2009). In vitro enzymatic characterization of

near full length EGFR in activated and inhibited states. Biochemistry 48,

6624–6632.

Red Brewer, M., Choi, S.H., Alvarado, D., Moravcevic, K., Pozzi, A., Lemmon,

M.A., and Carpenter, G. (2009). The juxtamembrane region of the EGF

receptor functions as an activation domain. Mol. Cell 34, 641–651.

Reschke, M., Ferby, I., Stepniak, E., Seitzer, N., Horst, D., Wagner, E.F., and

Ullrich, A. (2009). Mitogen-inducible gene-6 is a negative regulator of

epidermal growth factor receptor signaling in hepatocytes and human hepato-

cellular carcinoma. Hepatology 51, 1383–1390.

Sharma, S.V., and Settleman, J. (2009). ErbBs in lung cancer. Exp. Cell Res.

315, 557–571.

Sharma, S.V., Bell, D.W., Settleman, J., and Haber, D.A. (2007). Epidermal

growth factor receptor mutations in lung cancer. Nat. Rev. Mol. Cell Biol. 7,

169–181.

Shields, A.F., Grierson, J.R., Dohmen, B.M., Machulla, H.J., Stayanoff, J.C.,

Lawhorn-Crews, J.M., Obradovich, J.E., Muzik, O., and Mangner, T.J.

(1998). Imaging proliferation in vivo with [F-18]FLT and positron emission

tomography. Nat. Med. 4, 1334–1336.

Sorkin, A., and Goh, L.K. (2008). Endocytosis and intracellular trafficking of

ErbBs. Exp. Cell Res. 315, 683–696.

Squire, A., Verveer, P.J., Rocks, O., and Bastiaens, P.I.H. (2004). Red-edge

anisotropy microscopy enables dynamic imaging of homo-FRET between

green fluorescent proteins in cells. J. Struct. Biol. 147, 62–69.

Stuhlmann-Laeisz, C., Lang, S., Chalaris, A., Krzysztof, P., Enge, S., Eichler, J.,

Klingmuller, U., Samuel, M., Ernst, M., Rose-John, S., and Scheller, J. (2006).

Forced dimerization of gp130 leads to constitutive STAT3 activation, cytokine-

independent growth, and blockade of differentiation of embryonic stem cells.

Mol. Biol. Cell 17, 2986–2995.

Vilar, M., Charalampopoulos, I., Kenchappa, R.S., Simi, A., Karaca, E.,

Reversi, A., Choi, S., Bothwell, M., Mingarro, I., Friedman, W.J., et al. (2009).

Activation of the p75 neurotrophin receptor through conformational rearrange-

ment of disulphide-linked receptor dimers. Neuron 62, 72–83.

Wang, Q., Villeneuve, G., and Wang, Z. (2005). Control of epidermal growth

factor receptor endocytosis by receptor dimerization, rather than receptor

kinase activation. EMBO Rep. 6, 942–948.

Weinstein, I.B. (2002). Addiction to oncogenes–the Achilles heal of cancer.

Science 297, 63–64.

Yarden, Y., and Schlessinger, J. (1987). Self-phosphorylation of epidermal

growth factor receptor: evidence for a model of intermolecular allosteric acti-

vation. Biochemistry 26, 1434–1442.

Zhang, X., Gureasko, J., Shen, K., Cole, P.A., and Kuriyan, J. (2006). An allo-

steric mechanism for activation of the kinase domain of epidermal growth

factor receptor. Cell 125, 1137–1149.

Zhang, X., Pickin, K.A., Bose, R., Jura, N., Cole, P.A., and Kuriyan, J. (2007).

Inhibition of the EGF receptor by binding of MIG6 to an activating kinase

domain interface. Nature 450, 741–744.

Zhu, H.J., Iaria, J., Orchard, S., Walker, F., and Burgess, A.W. (2003).

Epidermal growth factor receptor: association of extracellular domain nega-

tively regulates intracellular kinase activation in the absence of ligand. Growth

Factors 21, 15–30.

Zimmer, S., Kahl, P., Buhl, T.M., Steiner, S., Wardelmann, E., Merkelbach-

Bruse, S., Buettner, R., and Heukamp, L.C. (2008). Epidermal growth factor

receptor mutations in non-small cell lung cancer influence downstream Akt,

MAPK and Stat3 signaling. J. Cancer Res. Clin. Oncol. 135, 723–730.


Systematic Protein LocationMapping Reveals Five PrincipalChromatin Types in Drosophila CellsGuillaume J. Filion,1,5 Joke G. van Bemmel,1,5 Ulrich Braunschweig,1,5 Wendy Talhout,1 Jop Kind,1 Lucas D. Ward,3,4,6

Wim Brugman,2 Ines J. de Castro,1,7 Ron M. Kerkhoven,2 Harmen J. Bussemaker,3,4 and Bas van Steensel1,*1Division of Gene Regulation2Central Microarray Facility

Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands3Department of Biological Sciences, Columbia University, 1212 Amsterdam Avenue, New York, NY 10027, USA4Center for Computational Biology and Bioinformatics, Columbia University, 1130 St. Nicholas Avenue, New York, NY 10032, USA5These authors contributed equally to this work6Present address: Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge,

MA 02139, USA7Present address: Genome Function Group, MRC Clinical Sciences Centre, Imperial College School of Medicine,

Hammersmith Hospital Campus, Du Cane Road, London W12 0NN, UK


DOI 10.1016/j.cell.2010.09.009

SUMMARY

Chromatin is important for the regulation of transcrip-

tionandother functions, yet thediversity of chromatin

composition and the distribution along chromo-

somes are still poorly characterized. By integrative

analysis of genome-wide binding maps of 53 broadly

selected chromatin components in Drosophila cells,

we show that the genome is segmented into five

principal chromatin types that are defined by unique

yet overlapping combinations of proteins and form

domains that can extend over > 100 kb. We identify

a repressive chromatin type that covers about half

of the genome and lacks classic heterochromatin

markers. Furthermore, transcriptionally active eu-

chromatin consists of two types that differ in molec-

ular organization and H3K36 methylation and regu-

late distinct classes of genes. Finally, we provide

evidence that the different chromatin types help to

target DNA-binding factors to specific genomic

regions. These results provide a global view of chro-

matin diversity and domain organization in a meta-

zoan cell.

INTRODUCTION

Chromatin consists of DNA and all associated proteins. The

scaffold of chromatin is formed by nucleosomes, which are

histone octamers in a tight complex with DNA. This scaffold

serves as the docking platform for hundreds of structural and

regulatory proteins. Furthermore, histones carry a variety of

posttranslational modifications that form recognition sites for

specific proteins (Berger, 2007; Rando and Chang, 2009). The

local composition of chromatin is a major determinant of the

transcriptional activity of a gene; some chromatin proteins

enhance transcription, whereas others have repressive effects.

Traditionally, chromatin was divided into heterochromatin and

euchromatin. There is now ample evidence that a finer classifica-

tion is required. For example, in Drosophila, at least two types of

heterochromatin exist that have distinct regulatory functions and

consist of different proteins. The first type is marked by Poly-

comb group (PcG) proteins and methylation of lysine 27 of

histone H3 (H3K27). PcG chromatin forms large continuous

domains; it is a repressive type of chromatin that primarily regu-

lates genes with developmental functions (Sparmann and van

Lohuizen, 2006). The second type is marked by heterochromatin

protein 1 (HP1) and several associated proteins, combined with

methylation of H3K9. This type of heterochromatin can also

cover large genomic segments, particularly around centro-

meres. Reporter genes integrated in or near HP1 heterochro-

matin tend to be repressed, but paradoxically, many genes

that are naturally bound by HP1 are transcriptionally active

(Hediger and Gasser, 2006). Direct comparison of genome-

wide binding maps indicates that PcG and HP1 heterochromatin

are nonoverlapping (de Wit et al., 2007).

HP1 and PcG chromatin illustrate two important principles of

chromatin organization: each type is marked by unique combi-

nations of proteins and can cover long stretches of DNA. But

are there other major types of chromatin that follow these

same principles? For example, is euchromatin also organized

into domains with distinct protein compositions? Are there

additional types of repressive chromatin that have remained

unnoticed?

In order to address these questions, we generated genome-

wide location maps of 53 broadly selected chromatin proteins

and four key histone modifications in Drosophila cells, providing


a rich description of chromatin composition along the genome.

By integrative computational analysis, we identified, aside from

PcG and HP1 chromatin, three additional principal chromatin

types that are defined by unique combinations of proteins. One

of these is a type of repressive chromatin that covers 50% of

the genome. In addition, we identified two types of transcription-

ally active euchromatin that are bound by different proteins and

harbor distinct classes of genes.

RESULTS

Genome-wide Location Maps of 53 Chromatin Proteins

We constructed a database of high-resolution binding profiles of

53 chromatin proteins in the embryonicDrosophila melanogaster

cell line Kc167 (Figure 1A and Figure S1A available online). In

order to obtain a representative cross-section of the chromatin

proteome, we selected proteins from most known chromatin

protein complexes, including a variety of histone-modifying

enzymes, proteins that bind specific histone modifications,

general transcription machinery components, nucleosome re-

modelers, insulator proteins, heterochromatin proteins, struc-

tural components of chromatin, and a selection of DNA-binding

factors (DBFs) (Table S1). For 40 of these proteins, full-genome

high-resolution binding maps have not previously been reported

in any Drosophila cell type or tissue. Though chromatin immuno-

precipitation (ChIP) is widely used to map protein-genome inter-

actions (Collas, 2009), large-scale application of this method is

hampered by the limited availability of highly specific antibodies.

A

C

B

Principal component analysis

Hidden Markov model

53 c

hro

matin

pro

tein

s

16000 16200 16400 16600 16800 17000

Position on chr2L (kb)

PC1

PC2

PC3

type

16000 16200 16400 16600 16800 17000

Position on chr2L (kb)

MRG15SU(VAR)3−7SU(VAR)3−9

HP6HP1LHR

CAF1ASF1

MUS209TOP1

RPII18SIR2

RPD3CDK7DSP1DF31MAX

PCAFASH2HP1cCtBPJRA

BRMECRBCD

MED31SU(VAR)2−10

LOLALGAF

CG31367ACT5C

TIP60MNT

SIN3ATBP

DWGPHOLPROD

BEAF32bSU(HW)

LAMD1H1

SUUREFFIAL

GROPHO

CTCFPC

E(Z)PCLSCE

Genes+

-

−20

−10

010

20

PC

1

−15 −10 −5 0 5 10 15

PC2

−15 −10 −5 0 5 10 15

−15

−10

−5

05

10

PC2P

C3

Figure 1. Overview of Protein Binding Profiles and Derivation of the Five-Type Chromatin Segmentation

(A) Sample plot of all 53 DamID profiles (log2 enrichment over Dam-only control). Positive values are plotted in black and negative values in gray for contrast.

Below the profiles, genes on both strands are depicted as lines with blocks indicating exons.

(B) Two-dimensional projections of the data onto the first three principal components. Colored dots indicate the chromatin type of probed loci as inferred by

a five-state HMM.

(C) Values of the first three principal components along the region shown in (A), with domains of the different chromatin types after segmentation by the five-state

HMM highlighted by the same colors as in (B).

See also Figure S1 and Table S1.


Moreover, at least for some chromatin proteins, ChIP results can

greatly depend on the choice of crosslinking reagents (Wang

et al., 2009) and can be unreliable for proteins with short resi-

dence times (Gelbart et al., 2005; Schmiedeberg et al., 2009).

We therefore used the DamID technology, which does not

require crosslinking or antibodies. With DamID, DNA adenine

methyltransferase (Dam) fused to a chromatin protein of interest

deposits a stable adenine-methylation ‘‘footprint’’ in vivo at the

interaction sites of the chromatin protein so that even transient

interactions may be detected (van Steensel et al., 2001). Note

that the fusion protein is expressed at very low levels, averting

overexpression artifacts. The DamID profiles of all 53 proteins

were generated in duplicate under standardized conditions

and were detected using oligonucleotide microarrays that query

the entire fly genome at 300 bp intervals. Comparisons to pub-

lished and new ChIP data confirm the overall reliability of the

DamID data (Figure S1B), which was also reported in previous

comparative studies (Moorman et al., 2006; Negre et al., 2006).

For reference purposes, we also generated ChIP maps of

histone H3 and the histone marks H3K4me2, H3K9me2,

H3K27me3, and H3K79me3 on the same array platform.

Most of the Fly Genome Interacts with Nonhistone

Chromatin Proteins

Comparison of the DamID profiles for all 53 proteins shows

a variety of binding patterns (Figure 1A). Nevertheless, several

sets of proteins exhibit profiles that are similar. Some similarities

were anticipated, such as for PC, PCL, SCE, and E(Z), which are

all PcG proteins (Sparmann and van Lohuizen, 2006), and for

HP1, SU(VAR)3-9, LHR, and HP6, which are part of classic

HP1-type heterochromatin (Greil et al., 2007). We also observe

extensive colocalization of Lamin (LAM), histone H1 (H1),

Effete (EFF), Suppressor of Underreplication (SUUR), and the

AT-hook protein D1, which have not been linked previously

except for LAM and SUUR (Pindyurin et al., 2007). There is a

prominent overlap in the binding patterns of a large set of 30

proteins, including histone-modifying enzymes (e.g., RPD3 and

SIR2), components of the basal transcription machinery (e.g.,

CDK7 and TBP), and others detailed below.

In order to identify target and nontarget loci for each protein,

we applied a two-state hidden Markov model (HMM) to each

individual binding map (Extended Experimental Procedures).

This method identifies themost likely segmentation into ‘‘bound’’

and ‘‘unbound’’ probed loci. According to the resulting binary

classifications, the genome-wide occupancy by individual

proteins varies broadly, ranging from about 2% (GRO) to 79%

(IAL). Of interest, 99.99% of the probed loci are bound by at least

one protein and 99.6% by at least three proteins. This indicates

that, at least at the resolution of our maps, essentially no part of

the fly genome is permanently in a configuration that consists of

nucleosomes only. Approximately 1% of the genome shows

extremely high protein occupancy, being bound by 36–44 of

the 53 mapped proteins.

Principal Chromatin Types Defined by Combinations

of Proteins

Next, we used a computational classification strategy to identify

themajor types of chromatin, defined as distinct combinations of

proteins that are recurrent throughout the genome. To identify

such combinations, we initially performed principal component

analysis on the 53 quantitative DamID profiles to reduce the

dimensionality of the data. We then focused on the first three

principal components, which together account for 57.7% of

the total variance. By projecting the genomic sites on the prin-

cipal components, we could distinguish five distinct lobes in

the three-dimensional scatter plot (Figure 1B). No additional

distinct lobes could be observed upon further inspection of

higher-level principal components. Importantly, the five groups

were also clearly separated when using the previously defined

binary target definitions (Figure S1C), showing that this result is

robust to different quantification methods.

Having established that classification into five types properly

summarizes the data, we fitted a five-state HMM onto the first

three principal components. Thus, every probed sequence in

the genome was assigned one of five exclusive chromatin types

(Extended Experimental Procedures). To avoid semantic confu-

sion, and in line with the Greek word chroma (color), we labeled

each of the five protein signatures with a color (BLUE, GREEN,

BLACK, RED, and YELLOW). The HMM classification produced

a mosaic pattern of chromosomal domains that vary widely in

length (Figure 1C). We emphasize that this segmentation is

purely data driven, without using any other knowledge besides

the 53 DamID profiles. The segmentation is generally robust:

removal of any of the proteins except for PC still yields a five-

state classification that is, on average, 96.7% identical to the

model obtained with all 53 proteins. A detailed analysis of the

robustness is summarized in Figure S1D.

Domain Organization of Chromatin Types

The five types of chromatin differ substantially in their genome

coverage, numbers of domains, and numbers of genes (Fig-

ure 2A). We identified a total of 8428 domains that typically range

from 1 to 52 kb (5th–95th percentiles) with a median length of

6.5 kb, although the size distribution depends on the chromatin

type (Figure 2B). 441 domains are larger than 50 kb, and 155

are larger than 100 kb, with the largest domain being 737 kb.

Many individual domains include multiple neighboring genes

(Figure 2C), the largest number of which within a single domain

is 139 (for a centromere-proximal GREEN domain). Taken

together, these data indicate that the fly genome is generally

organized into large regions that are covered by specific combi-

nations of proteins.

BLUE and GREEN Chromatin Correspond to Known

Heterochromatin Types

Visualization of the protein occupancy in each of the five chro-

matin types (Figure 3A) shows that most proteins are not

confined to a single chromatin type. Rather, the five chromatin

types are defined by unique combinations of proteins. Impor-

tantly, BLUE and GREEN chromatin closely resemble previously

identified chromatin types. GREEN chromatin corresponds to

classic heterochromatin that is marked by SU(VAR)3-9, HP1,

and the HP1-interacting proteins LHR and HP6. As described

previously (Ebert et al., 2006; Greil et al., 2007), this type of chro-

matin is prominent in pericentric regions and on chromosome 4

(Figure S2A). To further validate this classification, we conducted


genome-wide ChIP of H3K9me2, a histone mark that is predom-

inantly generated by SU(VAR)3-9 and bound by HP1 (Hediger

and Gasser, 2006) . Indeed, H3K9me2 is highly and specifically

enriched in GREEN chromatin (Figure 3B).

BLUE chromatin corresponds to PcG chromatin, as shown by

the extensive binding by the PcG proteins PC, E(Z), PCL, and

SCE. Indeed, well-known PcG target loci such as the Hox

gene clusters are localized in BLUE domains (Figure S2B).

Furthermore, genome-wide ChIP of H3K27me3, the histone

mark that is generated by E(Z) and recognized by PC (Sparmann

and van Lohuizen, 2006), is highly enriched in BLUE chromatin

(Figure 3B). We emphasize that these histone modification

profiles serve as independent validation because they were not

used in the five-state HMMclassification. The fact that twomajor

well-known chromatin types were faithfully recovered indicates

that our chromatin classification strategy is biologically mean-

ingful.

Of interest, we identified several additional proteins that mark

BLUE or GREEN chromatin, or both. For example, moderate

degrees of occupancy of the histone deacetylase (HDAC)

RPD3 occur in both BLUE and GREEN chromatin, in accordance

with known biochemical and genetic interactions of RPD3 with

PcG proteins as well as SU(VAR)3-9 (Czermin et al., 2001; Tie

et al., 2003). The presence of EFF in BLUE chromatin is consis-

tent with a reported role of this protein in PcG-mediated silencing

(Fauvarque et al., 2001).

A

B C D

Genome coverage

117 Mb

Number of domains

8428 domains

All genes

15145 genes

Silent genes

4229 silent genes

Length of domains

04

08

00

200

0200

count

0200

Domain length (kb)

0 10 20 30 40

02

50

>50

Number of genes per domain

02

00

06

00

03

00

count

02

00

Number of genes / domain

05

00

0 1 2 3 4 5 6 7 8 9 10 >10

mRNA expression

020

40

0400

01500

gene c

ount

0400

01

00

−1 0 1 2 3 4

log10(RNA tag count)

no tags

Figure 2. Characteristics of the Five Chromatin Types

(A) Coverage and gene content of chromatin domains of each type. The chromatin type of a gene is defined as the chromatin type at its transcription start site

(TSS). Gray sectors correspond to geneswhose TSSmaps at the transition between two chromatin types. Silent genes have an average RNA tag count less than 1

per million total tags (see D).

(B) Length distribution of chromatin domains, i.e., genomic segments covered contiguously by one chromatin type.

(C) Distribution of the number of genes per chromatin domain. Because some genes overlap with more than one domain, genes are assigned to a chromatin type

based on the type at the transcription start site.

(D) Histogram of mRNA expression determined by RNA tag profiling. Data are represented as log10 (tags per million total tags).

Dashed vertical lines in (B)–(D) indicate medians.


BLACK Chromatin Is the Prevalent Type

of Repressive Chromatin

BLACK chromatin covers 48%of the probed genome and is thus

by far the most abundant type (Figure 2A). With a median size of

17 kb and with 134 domains larger than 100 kb, BLACK chro-

matin domains tend to be longer than domains of the four other

types (Figure 2B). BLACK chromatin is overall relatively gene

poor (Figure 2A; compare genome coverage and number of

genes), but it nevertheless harbors 4162 genes. By mRNA

high-throughput sequencing, we detected no transcriptional

activity (<1 mRNA molecule per 10 million) for 66% of the genes

in BLACK chromatin, whereas the remaining 34% have very low

activity (Figure 2D). This is in agreement with the low coverage of

BLACK chromatin by RPII18, a subunit shared by all three RNA

polymerases (Figure 3A), and a lack of the active histone marks

H3K4me2 and H3K79me3 as detected by ChIP (Figure 3B). We

note that the majority of silent genes in the genome are located

in BLACK chromatin (Figure 2A). Thus, BLACK chromatin is

a distinctively silent type of chromatin that covers a large part

of the genome.

BLACK chromatin is almost universally marked by four of the

53 mapped proteins: histone H1, D1, IAL, and SUUR, whereas

SU(HW), LAM, and EFF are also frequently present (Figure 3A).

Close-up views show that H1, D1, IAL, SUUR, and LAM have

a broad distribution within BLACK domains, whereas SU(HW)

exhibits a distinct, more focal pattern (Figure 4A).

Given that genes in BLACK chromatin are expressed at very

low levels, we asked whether BLACK chromatin actively

represses transcription or merely forms secondary to a lack of

transcription. In the former model, transgenes inserted into

BLACK chromatin may exhibit reduced transcription, whereas,

in the latter model, transgenes should be unaffected. To test

this, we examined a data set of 2852 random P element inser-

tions that carry a mini-white eye color reporter gene. For each

BA

MRG15SIR2

RPD3DF31

RPII18BEAF32B

TOP1SIN3AASH2MAX

ASF1DSP1PCAFCDK7HP1C

JRACTBPCAF1

MUS209TIP60

TBPMNTDWG

PHOLSU(VAR)3−7

PRODACT5C

GROPHO

MED31BCD

SU(VAR)2−10GAF

CG31367LOLAL

ECRBRM

CTCFIALH1D1

SUURLAMEFF

SU(HW)PC

E(Z)PCLSCEHP6LHRHP1

SU(VAR)3−9

0 0.5 1

Fraction of bound loci

−1

01

2

H3K9me2

log

2(H

3K

9m

e2

Ch

IP /

H3

Ch

IP)

−3

−2

−1

01

H3K27me3

log

2(H

3K

27

me

3 C

hIP

/ H

3 C

hIP

)

−2

−1

01

23

H3K79me3

log

2(H

3K

79

me

3 C

hIP

/ H

3 C

hIP

)

−2

−1

01

23

H3K4me2

log

2(H

3K

4m

e2 C

hIP

/ H

3 C

hIP

)

−1

.0−

0.5

0.0

0.5

Histone H3

log

2(H

3 C

hIP

/ in

pu

t)

Figure 3. Chromatin Types Are Characterized by Distinctive Protein Combinations and Histone Modifications

(A) Fraction of all probed genomic loci within each chromatin type that is bound by each protein. Bound loci were determined separately for each protein as

described in the text.

(B) Levels of histone H3 and four histone modifications as determined by genome-wide ChIP. The distribution of values is shown as ‘‘violin plots,’’ which are

symmetrized density plots of binding values per chromatin type: the wider the violin, the more data points are associated to that value. Dashed horizontal lines

indicate the median binding value for each chromatin type. Histone modification ChIP data were normalized to H3 occupancy.

See also Figure S2.


of these insertions, the expression level was previously scored

and the integration site mapped (Babenko et al., 2010). Strik-

ingly, of 307 insertions located in BLACK regions, 36% exhibited

various degrees of w silencing, compared to 13% genome wide

(Figure 4B). Moreover, repression of transgene insertions in

BLACK chromatin ismore pronounced than in BLUE andGREEN

chromatin. This result strongly indicates that BLACK chromatin

has an active role in transcriptional silencing.

Developmental Regulation of Genes in BLACK

Chromatin

Not all genes in BLACK regions are expected to remain silenced

in various tissues. Indeed, a survey of tissue expression profiling

data (Chintapalli et al., 2007) indicates that genes in BLACK

chromatin can become active, although their expression tends

to be restricted to a few tissues only (Figure 4C). This suggests

that BLACK chromatin domains, as defined in Kc167 cells, can

be remodeled into a different chromatin type in some cell types.

Consistent with this dynamic regulation, BLACK chromatin is

particularly rich in highly conserved noncoding elements

(HCNEs) (Figure 4D), which are thought to mediate gene regula-

tion (Engstrom et al., 2007). The density of HCNEs in BLACK

chromatin is comparable to that in BLUE chromatin, which

harbors many developmentally regulated genes (Tolhuis et al.,

2006), and is much higher than in the other three chromatin

types. Together, these data suggest that BLACK chromatin is,

at least in part, under developmental control.

YELLOW and RED Chromatin Are Two Distinct Types

of Euchromatin

In contrast to BLACK and BLUE chromatin, RED and YELLOW

chromatin have hallmarks of transcriptionally active euchro-

matin. Most genes in these two chromatin types produce

substantial amounts of mRNA (Figure 2D), and levels of RNA

polymerase (Figure 3A), H3K4me2, and H3K79me3 are typically

high, whereas levels of H3K9me2 and H3K27me3 are low

(Figure 3B).

RED andYELLOWchromatin share various chromatin proteins

(Figure 3A). Among these are the HDACs RPD3 and SIR2, as well

as the RPD3-interacting protein SIN3A. HDACs have recently

also been found in transcriptionally active chromatin in human

cells (Wang et al., 2009). Other proteins that are highly abundant

in both RED and YELLOW chromatin include DF31, a little-

studied protein that drives chromatin decondensation in vitro

(Crevel et al., 2001); ASH2, a homolog of a subunit of a H3K4

methyltransferase complex in yeast and vertebrate cells (Nagy

et al., 2002); and MAX, a DBF that is part of the MYC network of

regulators of growth and proliferation (Orian et al., 2003).

Aside from these similarities, RED and YELLOW chromatin

display striking differences. RED chromatin is abundantly

A

C

B D

GREEN BLUE BLACK RED YELLOW

HC

NE

s p

er

MB

02

04

06

08

01

00

l og

2(D

am

−H

1/D

am

)

H1

−3

−1

1

log

2(D

am

−S

UU

R/D

am

)

SUUR

−1.5

0.5

log

2(D

am

−D

1/D

am

)

D1

−1.5

0.5

log

2( D

am

−Lam

/Dam

)

LAM

−1

01

log

2(D

am

−E

FF

/Dam

) EFF

−1.0

0.0

log

2(D

am

−IA

L/D

am

)

IAL

−1.0

0.0

log

2(D

am

−S

U(H

W)/

D.)

SU(HW)

−1

12

Genes

Type

16000 16100 16200 16300 16400 16500

Position on chr2R (kb)

GREEN BLUE BLACK RED YELLOW

Fra

ctio

n o

f sile

nce

d tra

nsg

en

es

0.0

0.1

0.2

0.3

0.4

26 302 307 841 1345

weakmediumstrong

Silencing:

S2 cellslarval tubulelarval salivary glandlarval midgutlarval hindgutlarval trachea

larval CNSlarval fat body

larval carcasstubulesalivary glandcropmidguthindgut

fat bodyvirgin spermathecamated spermatheca

heart

headeyebrainthorac. ganglionovarytestismale access. glandscarcasswhole fly

difference from

genome mean

(log10)

−2 0 2

4,086 BLACK genes

Figure 4. Properties of BLACK Chromatin

(A) Sample plots of binding profiles of the six proteins that are the most prevalent in BLACK chromatin. Genes on both strands, as well as chromatin types, are

depicted below the profiles. Gray blocks in the background correspond to BLACK chromatin domains.

(B) Silencing of a white reporter gene in 2852 P element insertions in adult eyes (Babenko et al., 2010) separated by chromatin type in Kc cells. The fraction of

silenced insertions is higher among those overlapping with BLACK regions than in the rest of the genome (p < 2.2*10 16, chi-square test).

(C) Relative expression levels (log10 scale, normalized to genome-wide average) of BLACK genes in various tissues (Chintapalli et al., 2007).

(D) Density of highly conserved noncoding elements (HCNEs) per chromatin type.


marked by several proteins that are mostly absent from the

four other chromatin types (Figure 3A). Among these are the

nucleosome-remodeling ATPase Brahma (BRM); the regulator

of chromosome structure SU(VAR)2-10; the Mediator subunit

MED31; the 55 kDa subunit of CAF1, present in various

histone-modifying complexes (Martınez-Balbas et al., 1998; Tie

et al., 2001); and several DBFs, including the ecdysone receptor

(ECR), GAGA factor (GAF), and Jun-related antigen (JRA).

These differences in protein composition prompted us to

investigate the timing of DNA replication during S phase, which

is known to differ in relation with chromatin marks (Gilbert,

2002). Analysis of a genome-wide replication timing map from

Kc167 cells (Schwaiger et al., 2009) shows that DNA in RED

and YELLOW chromatin is generally replicated early in S phase,

as may be expected for euchromatin. However, RED chromatin

tends to be replicated even earlier than YELLOW chromatin

(Figure 5A). This coincides with a strong enrichment of origin

A B

C

D

−0

.50

.51

.01

.52

.0

−5 0 5

Relative position from 5' end (kb)

log

2(D

am

−M

RG

15

Dam

)

−5 0 5


−1

01

23

4

H3K36me3

−5 0 5


log

2(H

3K

36m

e3

i nput )

−5 0 5


MRG15

−1

01

23

4

ORC binding

OR

C e

nri

ch

me

nt

−2

0−

10

01

02

0

Replication timing

log

2 (

Ea

rly /

La

te)

Figure 5. RED and YELLOW Are Two Distinct Types of

Euchromatin

(A) Violin plots of replication timing (Schwaiger et al., 2009) per

chromatin type.

(B) Violin plots of origin of replication complex 2 (ORC2) binding

(MacAlpine et al., 2010) per chromatin type.

(C) Average binding of MRG15 around 50 and 30 ends of genes in

RED and YELLOW chromatin. (Left) Alignment to transcript 50

ends. (Right) Alignment to 30 ends. Only genes that are entirely

within one chromatin type are depicted.

(D) Average enrichment of H3K36me3 (Bell et al., 2010), plotted as

in (C).

recognition complex (ORC) binding in RED chromatin,

as mapped by ChIP (MacAlpine et al., 2010) (Fig-

ure 5B), suggesting that DNA replication is often initi-

ated in RED chromatin. These observations further

underscore that RED and YELLOW chromatin are

distinct types of euchromatin.

Active Genes in YELLOW, but Not RED,

Chromatin Carry H3K36me3

Only one protein of the data set is abundant in

YELLOW, but not in RED, chromatin: MRG15, which

is a chromodomain-containing protein. Because

human MRG15 has previously been reported to bind

H3K36me3 (Zhang et al., 2006), we compared the

fine distribution of MRG15 and H3K36me3 along

genes within the two chromatin types (Bell et al.,

2010). Indeed, both are highly enriched along genes

in YELLOW chromatin but are nearly absent from

RED chromatin (Figures 5C and 5D). These data are

consistent with binding of MRG15 to H3K36me3

in vivo. Of interest, H3K36me3 was previously thought

to be a universal marker of elongating transcription

units (Lee and Shilatifard, 2007; Rando and Chang,

2009). Our analysis reveals that, at least in Drosophila

Kc167 cells, this histone mark is mostly absent from

genes lying in RED chromatin, even though these

genes are expressed at similar levels as genes in YELLOW

chromatin (Figure 2D).

RED and YELLOW Chromatin Mark Different Types

of Genes

The substantial differences between RED and YELLOW chro-

matin suggested that the genes that they harbor may be

regulated by two globally distinct pathways. We therefore inves-

tigated whether genes located in RED and YELLOW chromatin

have different characteristics. We began by comparing the

embryonic tissue expression patterns of genes in the two chro-

matin types. Strikingly, genes with a broad expression pattern

over many embryonic stages and tissues (Tomancak et al.,

2007) are highly enriched in YELLOW chromatin, whereas genes

with more restricted expression patterns are depleted

(Figure 6A). Consistent with this, gene ontology (GO) analysis

revealed that universal cellular functions such as ‘‘ribosome,’’


‘‘DNA repair,’’ and ‘‘nucleic acid metabolic process’’ are almost

exclusively found in YELLOW chromatin (Figure 6B), whereas

genes in RED chromatin are linked to more specific processes

such as ‘‘receptor binding,’’ ‘‘defense response,’’ ‘‘transcription

factor activity,’’ and ‘‘signal transduction’’ (Figure 6C). Such

specific functions and expression patterns require complex

mechanisms of gene regulation. Indeed, intergenic regions in

RED domains contain about 2-fold more HCNEs than YELLOW

chromatin (Figure 4D), although not as much as BLACK and

BLUE chromatin. Furthermore, genome-wide formaldehyde-as-

sisted identification of regulatory elements (FAIRE) (Braunsch-

weig et al., 2009; Giresi et al., 2007) points to a high density of

regulatory chromatin complexes in RED chromatin (Figure 6D).

Motif Binding by DBFs Is Guided by Chromatin Types

Chromatin can affect the ability of DBFs to bind to their cognate

binding sequences, which is thought to explain why, in vivo,

most DBFs bind to only a small subset of their recognition motifs

in the genome (Beato and Eisfeld, 1997). We investigated how

the five chromatin types might modulate DBF-DNA interactions.

We focused on five DBFs in our data set (JRA, MNT, GAF, CTCF,

and SU(HW)) for which the sequence-specificity is well charac-

terized. We first calculated the expected genomic binding

pattern of each DBF based on the occurrence of sequence

motifs that match the known DBF recognition motif. The exact-

ness of these matches is taken into account, yielding for each

DamID-probed locus a predicted relative affinity for the DBF

(Foat et al., 2006). Genome-wide comparison of this sequence-

based predicted affinity and actual protein occupancy indicated

only weak to moderate correlations (Spearman’s rho ranging

from 0.04 to 0.35; dashed gray curves in Figure 7A; Figure S4).

This suggests that chromatin indeed has substantial modulating

effects on DBF-motif interactions.

We then repeated this correlation analysis by chromatin type.

Surprisingly, this revealed that each DBF has its own depen-

dence on chromatin context (Figure 7A and Figure S4). GAF

and JRA both bind to their respective motif variants over a range

of affinities in RED chromatin, but not in the other chromatin

types; MNT binds to its motifs only in RED and YELLOW;

CTCF preferentially binds its motifs in RED and BLUE chromatin;

SU(HW) recognizes its motifs most efficiently in BLACK, BLUE,

and RED chromatin. Thus, each of the five chromatin types is

conducive to DNA binding by specific subsets of DBFs. Some

chromatin types may also weakly bind certain DBFs indepen-

dently of DNA interactions, as suggested by the varying DamID

baseline levels in loci that lack high-affinity motifs (e.g., for SU

(HW) and CTCF; Figure 7A).

Four out of five DBFs exhibit a preference for their motif in RED

chromatin. We wondered whether RED chromatin might have an

intrinsic property suchas ‘‘openness’’ or nucleosome remodeling

activity thatwouldgenerally facilitateDBFaccess. To test this,we

generated a DamID profile for the DNA-binding domain (DBD) of

yeast Gal4. This foreign DBD is not expected to have specific

protein-protein interactions with Drosophila chromatin, and its

recognition motif occurs randomly throughout the fly genome.

We observed similar interactions of Gal4-DBD with its cognate

motifs in all five chromatin types (Figure 7A, bottom-right). This

indicates that RED chromatin does not have a general positive

effect on protein-DNA interactions and that high DBF occupancy

in this chromatin type is more likely due to specific targeting

mechanisms for each DBF. In summary, these results indicate

that the five chromatin types together act as guides that help to

target DBFs to specific regions of the genome even though the

cognate binding motifs are broadly distributed (Figure 7B).

DISCUSSION

By systematic integration of 53 protein location maps, we found

that the Drosophila genome is packaged into a mosaic of five

principal chromatin types, each defined by a unique combination

proteinaceous extracellular matrixextracellular regionreceptor bindingcellular component movementtranscription factor activitybehaviorplasma membranemulticellular organismal development

intracellularnucleusnucleic acid metabolic processstructural molecule activityDNA metabolic processstructural constituent of ribosomeribosomeDNA repair

Fraction of Genes

0.0 0.2 0.4 0.6 0.8 1.0

18

120

74

173

157

163

285

934

2926

1098

726

220

177

155

158

80

all genes broad tissue−specific

Fra

ctio

n o

f g

en

es

0.0

0.2

0.4

0.6

0.8

1.0

A B

C

D

−1

01

23

FAIRE

log

2 (

FA

IRE

/ in

pu

t)

Figure 6. Genes in RED and YELLOW Differ in Regulation and Function

(A) Distribution of genes having ‘‘broad’’ and ‘‘tissue-specific’’ expression patterns (defined in Tomancak et al., 2007) over the five chromatin types. Left bar shows

distribution of all genes for comparison.

(B–C) GO slim categories that are significantly enriched (B) or depleted (C) in RED compared to YELLOW genes. Bars indicate the fraction of RED and YELLOW

genes for the given category (BLACK, GREEN, and BLUE are not considered here). Vertical dotted line represents the distribution expected by random chance.

The total numbers of RED and YELLOW genes within each category are indicated on the left.

(D) Violin plots of the log2 FAIRE signal per chromatin type (Braunschweig et al., 2009).

See also Figure S3.


of proteins. Extensive evidence demonstrates that the five types

differ in a wide range of characteristics aside from protein

composition, such as biochemical properties, transcriptional

activity, histone modifications, replication timing, and DBF tar-

geting, as well as sequence properties and functions of the

embedded genes. This validates our classification by indepen-

dent means and provides important insights into the functional

properties of the five chromatin types.

The Number of Chromatin States

Identifying five chromatin states out of the binding profiles of 53

proteins comes out as a surprisingly low number (one can form

1016 subsets of 53 elements). We emphasize that the five chro-

matin types should be regarded as the major types. Some may

be further divided into subtypes, depending on how fine-grained

one wishes the classification to be. For example, within each of

the transcriptionally active chromatin types, promoters and 30

ends of genes exhibit (mostly quantitative) differences in their

protein composition (data not shown) and thus could be re-

garded as distinct subtypes. However, these local differences

are minor relative to the differences between the five principal

types that we describe here. We cannot exclude that the accu-

mulation of binding profiles of additional proteins would reveal

other novel chromatin types. We also anticipate that the pattern

of chromatin types along the genome will vary between cell

types. For example, many genes that are embedded in BLACK

0 20 40 60 80 100

0.0

0.5

1.0

1.5GAF

0 20 40 60 80 100

−0.2

0.0

0.2

0.4

0.6

0.8

1.0 MNT

70 75 80 85 90 95 100

0.0

0.5

1.0

CTCF

70 75 80 85 90 95 100

−0.5

0.0

0.5

1.0

1.5

2.0 SU(HW)

0 20 40 60 80 100

0.0

0.5

1.0

1.5JRA

0 20 40 60 80 100

−0.4

−0.2

0.0

0.2

0.4

0.6 GAL4

A

B

Sequence affinity (rank %) Sequence affinity (rank %) Sequence affinity (rank %)

DamID score

DamID score

DamID score

Figure 7. Binding of DBFs to Their Cognate Motifs Is Differentially Guided by Chromatin Types

(A) Correlations between predicted DNA affinity and actual binding detected by DamID, genome-wide (gray dashed lines), or for each chromatin type (solid lines)

for six DBFs as indicated. Curves are loess-fitted lines; raw data are shown in Figure S4.

(B) Cartoon model depicting the specific guidance of DBFs to their cognate motifs in only certain chromatin types, illustrated for CTCF and MNT. DBF binding to

its cognate motif (gray box) is guided by protein-protein interactions. The presence of specific interactors (colored shapes) only in some chromatin types may

account for targeting.

See also Figure S4.


chromatin (defined in Kc167 cells) are activated in some other

cell types (Figure 4C). Thus, the chromatin of these genes is likely

to switch to an active type.

Whereas the integration of data for 53 proteins provides

substantial robustness to the classification of chromatin along

the genome, a subset of only five marker proteins (histone H1,

PC, HP1, MRG15, and BRM), which together occupy 97.6% of

the genome, can recapitulate this classification with 85.5%

agreement (Figure S1E). Assuming that no unknown additional

principal chromatin types exist in some cell types, DamID or

ChIP of this small set of markers may thus provide an efficient

means to examine the distribution of the five chromatin types

in various cells and tissues, with acceptable accuracy.

BLACK Chromatin: A Distinct Type of Repressive

Chromatin

Previous work on the expression of integrated reporter genes

(Handler and Harrell, 1999; Kelley and Kuroda, 2003; Markstein

et al., 2008) had suggested that most of the fly genome is tran-

scriptionally repressed, contrasting with the low coverage of

PcG and HP1-marked chromatin. BLACK chromatin, which

consists of a previously unknown combination of proteins and

covers about half of the genome, may account for these obser-

vations. Essentially all genes in BLACK chromatin exhibit

extremely low expression levels, and transgenes inserted in

BLACK chromatin are frequently silenced, indicating that BLACK

chromatin constitutes a strongly repressive environment. Impor-

tantly, BLACK chromatin is depleted of PcG proteins, HP1, SU

(VAR)3-9, and associated proteins and is also the latest to

replicate, underscoring that it is different from previously charac-

terized types of heterochromatin (here identified as BLUE and

GREEN chromatin).

The proteins that mark BLACK domains provide important

clues to the molecular biology of this type of chromatin. Loss

of LAM, EFF, or histone H1 causes lethality during Drosophila

development (Cenci et al., 1997; Lenz-Bohme et al., 1997; Lu

et al., 2009). Extensive in vitro and in vivo evidence has sug-

gested a role for H1 in gene repression, most likely through

stabilization of nucleosome positions (Laybourn and Kadonaga,

1991; Wolffe and Hayes, 1999; Woodcock et al., 2006). The

enrichment of LAM points to a role of the nuclear lamina in

gene regulation in BLACK chromatin (Pickersgill et al., 2006),

consistent with the long-standing notion that peripheral chro-

matin is silent (Towbin et al., 2009). Depletion of LAM causes

derepression of several LAM-associated genes (Shevelyov

et al., 2009), whereas artificial targeting of genes to the nuclear

lamina can reduce their expression (Finlan et al., 2008; Reddy

et al., 2008), suggesting a direct repressive contribution of the

nuclear lamina in BLACK chromatin. D1 is a little-studied protein

with 11 AT-hook domains. Overexpression of D1 causes ectopic

pairing of intercalary heterochromatin (Smith and Weiler, 2010),

suggesting a role in the regulation of higher-order chromatin

structure. SUUR specifically regulates late replication on poly-

tene chromosomes (Zhimulev et al., 2003), which is of interest

because BLACK chromatin is particularly late replicating. EFF

is highly similar to the yeast and mammalian ubiquitin ligase

Ubc4 that mediates ubiquitination of histone H3 (Liu et al.,

2005; Singh et al., 2009), raising the possibility that nucleosomes

in BLACK chromatin may carry specific ubiquitin marks. These

insights suggest that BLACK chromatin is important for chromo-

some architecture as well as gene repression and provide

important leads for further study of this previously unknown yet

prevalent type of chromatin.

RED and YELLOW: Distinct Types of Euchromatin

In RED and YELLOW chromatin, most genes are active, and

the overall expression levels are similar between these two

chromatin types. However, RED and YELLOW chromatin differ

in many respects. One of the conspicuous distinctions is the

disparate levels of H3K36me3 at active transcription units. This

histone mark is thought to be laid down in the course of tran-

scription elongation and may block the activity of cryptic

promoters inside of the transcription unit (Li et al., 2007). Why

active genes in RED chromatin lack H3K36me3 remains to be

elucidated.

The remarkably high protein occupancy in RED chromatin

suggests that RED domains are ‘‘hubs’’ of regulatory activity.

This may be related to the predominantly tissue-specific expres-

sion of genes in RED chromatin, which presumably requires

many regulatory proteins. We note that our DamID assay inte-

grates protein binding events over nearly 24 hr, so it is likely

that not all proteins bind simultaneously; some proteins may

bind only during a specific stage of the cell cycle. It is highly

unlikely that the high protein occupancy in RED chromatin

originates from an artifact of DamID, e.g., caused by a high

accessibility of RED chromatin. First, all DamID data are cor-

rected for accessibility using parallel Dam-only measurements.

Second, several proteins, such as EFF, SU(VAR)3-9, and histone

H1, exhibit lower occupancies in RED than in any other chro-

matin type. Third, ORC also shows a specific enrichment in

RED chromatin even though it was mapped by ChIP, by another

laboratory, and on another detection platform (MacAlpine et al.,

2010). Fourth, DamID of Gal4-DBD does not show any enrich-

ment in RED chromatin.

RED chromatin resembles DBF binding hot spots that were

previously discovered in a smaller-scale study inDrosophila cells

(Moorman et al., 2006). Discrete genomic regions targeted by

many DBFs have recently also been found in mouse ES cells

(Chen et al., 2008); hence, it is tempting to speculate that an

equivalent of RED chromatin may also exist in mammalian cells.

Housekeeping and dynamically regulated genes in budding

yeast also exhibit a dichotomy in chromatin organization (Tirosh

and Barkai, 2008), which may be related to our distinction

between YELLOW and RED chromatin. The observations that

RED chromatin is generally the earliest to replicate and is

strongly enriched in ORC binding suggest that this chromatin

type may be not only involved in transcriptional regulation, but

also in the control of DNA replication.

Chromatin Types as Guides for DBF Targeting

Our analysis of DBF binding indicates that the five chromatin

types together act as a guidance system to target DBFs to

specific genomic regions. This system directs DBFs to certain

genomic domains even though the DBF recognition motifs are

more widely distributed. We propose that targeting specificity

is, at least in part, achieved through interactions of DBFs with


particular partner proteins that are present in some of the five

chromatin types, but not in others (Figure 7B). The observation

that yeast Gal4-DBD binds its motifs with nearly equal efficiency

in all five chromatin types suggests that differences in compac-

tion among the chromatin types represent overall a minor factor

in the targeting of DBFs. Although additional studies will be

needed to further investigate the molecular mechanisms of

DBF guidance, the identification of five principal types of chro-

matin provides a firm basis for future dissection of the roles of

chromatin organization in global gene regulation.


Constructs

DamID constructs used for this study are listed in Table S1. New constructs

were cloned by TOPO cloning and GATEWAY recombination as described

(Braunschweig et al., 2009) or by Cre-mediated recombination. For the latter,

we generated an acceptor vector containing the Hsp70 promoter upstream of

myc-epitope tagged Dam, using the Creator Acceptor Vector Construction Kit

(Clontech, 631618). Chromatin protein open reading frames from pDNR-Dual

donor vectors (Drosophila Genomics Resource Center, Bloomington) were

cloned into the acceptor vector using the Creator DNA Cloning Kit (Clontech

PT3460-1). Nuclear localization was checked for all Dam-fusion proteins by

immunofluorescence microscopy with the 9E10 anti-Myc antibody (Santa

Cruz Biotechnology) after heat shock-induced expression as described (Greil

et al., 2007). Only MNT, GRO, and IAL gave weak nuclear signals but were not

discarded because MNT and GRO were successfully mapped by DamID in

previous studies (Bianchi-Frias et al., 2004; Orian et al., 2003) and IAL binds

metaphase chromosomes (Giet and Glover, 2001).

DamID, ChIP, and Microarrays

DamID assays were carried out under standardized conditions as described

previously (Moorman et al., 2006), with a minor modification: proteins were

grouped in sets sharing the same Dam-only controls for hybridization

purposes. For each group, three to five DamID assays on Dam alone were

carried out in parallel, the product of which was pooled before labeling.

ChIP and subsequent linear amplification reactions were done as described

(Kind et al., 2008) using anti-H3K27me3 (07-449) and anti-H3K4me2

(07-030) from Upstate Biotechnology; anti-H3K9me2 (1220) and anti-H3

(1791) from Abcam; affinity-purified anti-H1 serum (Braunschweig et al.,

2009); and anti-H3K79me3 (Schubeler et al., 2004) kindly provided by Fred

van Leeuwen. Fluorescent labeling of DamID and ChIP samples and two-color

hybridizations on custom-designed 385k NimbleGen arrays (Braunschweig

et al., 2009) were performed according to NimbleGen’s array users guide,

version 4.0. Arrays were scanned at 5 mm resolution, and raw data were

extracted using NimbleScan software. The identity of the hybridized material

was tracked by the presence of unique oligonucleotide spikes in each sample.

Furthermore, because the Dam-fusion expression vectors are produced in

Dam-positive bacteria, small amounts of the transfected plasmids are coam-

plified in the methylation-specific amplification protocol. This leads to a strong

signal in the open reading frame of the mapped protein, which allows us to

verify the identity of the used vector from the microarray data alone. This

open reading frame was masked before further data analysis.

Digital Gene Expression

Total RNA was isolated from growing Kc cells using TriZOL (Invitrogen), and

remaining DNA was degraded by shearing and DNaseI digestion. Poly(A)

RNA tag sequencing was carried out on an Illumina Solexa GAII using the

tag profiling kit with DpnII. Two RNA samples yielded 7.4 and 9.0 million reads.

Tags were mapped by BLAST, requiring at most two mismatches and eleven

consecutively matching bases. Only the tags mapping to the last GATC of

a transcript (FlyBase release 5.8) were counted and represented 70.3% and

69.4% of the total number of reads, respectively. Counts were normalized to

the total number of reads, and replicates were averaged.

Data Availability and Analysis

Computational methods are described in the Extended Experimental Proce-

dures.

ACCESSION NUMBERS

DamID, ChIP, and expression data, as well as binarized DamID data and a list

of the coordinates of all identified chromatin domains are available from

NCBI’s Gene Expression Omnibus, accession number GSE22069.


Supplemental Information includes Extended Experimental Procedures,

five figures, and one table and can be found with this article online at

doi:10.1016/j.cell.2010.09.009.

ACKNOWLEDGMENTS

We thank Francesco Russo for help with vector cloning; Marja Nieuwland and

Arno Velds for help with RNA tag sequencing; Dirk Schubeler’s laboratory for

sharing H3K36 methylation data prior to publication; and Reuven Agami, Fred

van Leeuwen, Wouter Meuleman, Ludo Pagie, and Aleksey Pindyurin for help-

ful suggestions. Supported by an EMBO long-term fellowship to J.K.; National

Institutes of Health grants T32GM082797, R01HG003008, and U54CA121852

to L.D.W. and H.J.B.; and grants from the Netherlands Genomics Initiative,

NWO-ALW VICI, and an EURYI Award to B.v.S.

Received: June 10, 2010

Revised: August 2, 2010


Published online: September 30, 2010

REFERENCES

Babenko, V.N., Makunin, I.V., Brusentsova, I.V., Belyaeva, E.S., Maksimov,

D.A., Belyakin, S.N., Maroy, P., Vasil’eva, L.A., and Zhimulev, I.F. (2010).

Paucity and preferential suppression of transgenes in late replication domains

of the D. melanogaster genome. BMC Genomics 11, 318.

Beato, M., and Eisfeld, K. (1997). Transcription factor access to chromatin.

Nucleic Acids Res. 25, 3559–3563.

Bell, O., Schwaiger, M., Oakeley, E.J., Lienert, F., Beisel, C., Stadler, M.B., and

Schubeler, D. (2010). Accessibility of the Drosophila genome discriminates

PcG repression, H4K16 acetylation and replication timing. Nat. Struct. Mol.

Biol. 17, 894–900.

Berger, S.L. (2007). The complex language of chromatin regulation during

transcription. Nature 447, 407–412.

Bianchi-Frias, D., Orian, A., Delrow, J.J., Vazquez, J., Rosales-Nieves, A.E.,

and Parkhurst, S.M. (2004). Hairy transcriptional repression targets and

cofactor recruitment in Drosophila. PLoS Biol. 2, E178.

Braunschweig, U., Hogan, G.J., Pagie, L., and van Steensel, B. (2009). Histone

H1 binding is inhibited by histone variant H3.3. EMBO J. 28, 3635–3645.

Cenci, G., Rawson, R.B., Belloni, G., Castrillon, D.H., Tudor, M., Petrucci, R.,

Goldberg, M.L., Wasserman, S.A., and Gatti, M. (1997). UbcD1, a Drosophila

ubiquitin-conjugating enzyme required for proper telomere behavior. Genes

Dev. 11, 863–875.

Chen, X., Xu, H., Yuan, P., Fang, F., Huss,M., Vega, V.B.,Wong, E., Orlov, Y.L.,

Zhang, W., Jiang, J., et al. (2008). Integration of external signaling pathways

with the core transcriptional network in embryonic stem cells. Cell 133,

1106–1117.

Chintapalli, V.R., Wang, J., and Dow, J.A. (2007). Using FlyAtlas to identify

better Drosophila melanogaster models of human disease. Nat. Genet. 39,

715–720.

Collas, P. (2009). The state-of-the-art of chromatin immunoprecipitation.

Methods Mol. Biol. 567, 1–25.


Crevel, G., Huikeshoven, H., and Cotterill, S. (2001). Df31 is a novel nuclear

protein involved in chromatin structure in Drosophila melanogaster. J. Cell

Sci. 114, 37–47.

Czermin, B., Schotta, G., Hulsmann, B.B., Brehm, A., Becker, P.B., Reuter, G.,

and Imhof, A. (2001). Physical and functional association of SU(VAR)3-9 and

HDAC1 in Drosophila. EMBO Rep. 2, 915–919.

de Wit, E., Greil, F., and van Steensel, B. (2007). High-resolution mapping

reveals links of HP1 with active and inactive chromatin components. PLoS

Genet. 3, e38.

Ebert, A., Lein, S., Schotta, G., and Reuter, G. (2006). Histonemodification and

the control of heterochromatic gene silencing in Drosophila. Chromosome

Res. 14, 377–392.

Engstrom, P.G., Ho Sui, S.J., Drivenes, O., Becker, T.S., and Lenhard, B.

(2007). Genomic regulatory blocks underlie extensive microsynteny conserva-

tion in insects. Genome Res. 17, 1898–1908.

Fauvarque, M.O., Laurenti, P., Boivin, A., Bloyer, S., Griffin-Shea, R., Bourbon,

H.M., and Dura, J.M. (2001). Dominant modifiers of the polyhomeotic extra-

sex-combs phenotype induced by marked P element insertional mutagenesis

in Drosophila. Genet. Res. 78, 137–148.

Finlan, L.E., Sproul, D., Thomson, I., Boyle, S., Kerr, E., Perry, P., Ylstra, B.,

Chubb, J.R., and Bickmore, W.A. (2008). Recruitment to the nuclear periphery

can alter expression of genes in human cells. PLoS Genet. 4, e1000039.

Foat, B.C., Morozov, A.V., and Bussemaker, H.J. (2006). Statistical mechan-

ical modeling of genome-wide transcription factor occupancy data by

MatrixREDUCE. Bioinformatics 22, e141–e149.

Gelbart, M.E., Bachman, N., Delrow, J., Boeke, J.D., and Tsukiyama, T. (2005).

Genome-wide identification of Isw2 chromatin-remodeling targets by localiza-

tion of a catalytically inactive mutant. Genes Dev. 19, 942–954.

Giet, R., and Glover, D.M. (2001). Drosophila aurora B kinase is required for

histone H3 phosphorylation and condensin recruitment during chromosome

condensation and to organize the central spindle during cytokinesis. J. Cell

Biol. 152, 669–682.

Gilbert, D.M. (2002). Replication timing and transcriptional control: beyond

cause and effect. Curr. Opin. Cell Biol. 14, 377–383.

Giresi, P.G., Kim, J., McDaniell, R.M., Iyer, V.R., and Lieb, J.D. (2007). FAIRE

(Formaldehyde-Assisted Isolation of Regulatory Elements) isolates active

regulatory elements from human chromatin. Genome Res. 17, 877–885.

Greil, F., de Wit, E., Bussemaker, H.J., and van Steensel, B. (2007). HP1

controls genomic targeting of four novel heterochromatin proteins in

Drosophila. EMBO J. 26, 741–751.

Handler, A.M., and Harrell, R.A., 2nd. (1999). Germline transformation of

Drosophila melanogaster with the piggyBac transposon vector. Insect Mol.

Biol. 8, 449–457.

Hediger, F., and Gasser, S.M. (2006). Heterochromatin protein 1: don’t judge

the book by its cover! Curr. Opin. Genet. Dev. 16, 143–150.

Kelley, R.L., and Kuroda, M.I. (2003). The Drosophila roX1 RNA gene can

overcome silent chromatin by recruiting the male-specific lethal dosage

compensation complex. Genetics 164, 565–574.

Kind, J., Vaquerizas, J.M., Gebhardt, P., Gentzel, M., Luscombe, N.M.,

Bertone, P., and Akhtar, A. (2008). Genome-wide analysis reveals MOF as

a key regulator of dosage compensation and gene expression in Drosophila.

Cell 133, 813–828.

Laybourn, P.J., and Kadonaga, J.T. (1991). Role of nucleosomal cores and

histone H1 in regulation of transcription by RNA polymerase II. Science 254,

238–245.

Lee, J.S., and Shilatifard, A. (2007). A site to remember: H3K36 methylation

a mark for histone deacetylation. Mutat. Res. 618, 130–134.

Lenz-Bohme, B., Wismar, J., Fuchs, S., Reifegerste, R., Buchner, E., Betz, H.,

and Schmitt, B. (1997). Insertional mutation of the Drosophila nuclear

lamin Dm0 gene results in defective nuclear envelopes, clustering of nuclear

pore complexes, and accumulation of annulate lamellae. J. Cell Biol. 137,

1001–1016.

Li, B., Gogol, M., Carey, M., Pattenden, S.G., Seidel, C., and Workman, J.L.

(2007). Infrequently transcribed long genes depend on the Set2/Rpd3S

pathway for accurate transcription. Genes Dev. 21, 1422–1430.

Liu, Z., Oughtred, R., and Wing, S.S. (2005). Characterization of E3Histone,

a novel testis ubiquitin protein ligase which ubiquitinates histones. Mol. Cell.

Biol. 25, 2819–2831.

Lu, X., Wontakal, S.N., Emelyanov, A.V., Morcillo, P., Konev, A.Y., Fyodorov,

D.V., and Skoultchi, A.I. (2009). Linker histone H1 is essential for Drosophila

development, the establishment of pericentric heterochromatin, and a normal

polytene chromosome structure. Genes Dev. 23, 452–465.

MacAlpine, H.K., Gordan, R., Powell, S.K., Hartemink, A.J., and MacAlpine,

D.M. (2010). Drosophila ORC localizes to open chromatin and marks sites of

cohesin complex loading. Genome Res. 20, 201–211.

Markstein, M., Pitsouli, C., Villalta, C., Celniker, S.E., and Perrimon, N. (2008).

Exploiting position effects and the gypsy retrovirus insulator to engineer

precisely expressed transgenes. Nat. Genet. 40, 476–483.

Martınez-Balbas, M.A., Tsukiyama, T., Gdula, D., and Wu, C. (1998).

Drosophila NURF-55, a WD repeat protein involved in histone metabolism.

Proc. Natl. Acad. Sci. USA 95, 132–137.

Moorman, C., Sun, L.V., Wang, J., de Wit, E., Talhout, W., Ward, L.D., Greil, F.,

Lu, X.J., White, K.P., Bussemaker, H.J., and van Steensel, B. (2006). Hotspots

of transcription factor colocalization in the genome of Drosophila mela-

nogaster. Proc. Natl. Acad. Sci. USA 103, 12027–12032.

Nagy, P.L., Griesenbeck, J., Kornberg, R.D., and Cleary, M.L. (2002). A tri-

thorax-group complex purified from Saccharomyces cerevisiae is required

for methylation of histone H3. Proc. Natl. Acad. Sci. USA 99, 90–94.

Negre, N., Hennetin, J., Sun, L.V., Lavrov, S., Bellis, M., White, K.P., and

Cavalli, G. (2006). Chromosomal distribution of PcG proteins during

Drosophila development. PLoS Biol. 4, e170.

Orian, A., van Steensel, B., Delrow, J., Bussemaker, H.J., Li, L., Sawado, T.,

Williams, E., Loo, L.W., Cowley, S.M., Yost, C., et al. (2003). Genomic binding

by the Drosophila Myc, Max, Mad/Mnt transcription factor network. Genes

Dev. 17, 1101–1114.

Pickersgill, H., Kalverda, B., de Wit, E., Talhout, W., Fornerod, M., and

van Steensel, B. (2006). Characterization of the Drosophila melanogaster

genome at the nuclear lamina. Nat. Genet. 38, 1005–1014.

Pindyurin, A.V., Moorman, C., de Wit, E., Belyakin, S.N., Belyaeva, E.S.,

Christophides, G.K., Kafatos, F.C., van Steensel, B., and Zhimulev, I.F.

(2007). SUUR joins separate subsets of PcG, HP1 and B-type lamin targets

in Drosophila. J. Cell Sci. 120, 2344–2351.

Rando, O.J., and Chang, H.Y. (2009). Genome-wide views of chromatin

structure. Annu. Rev. Biochem. 78, 245–271.

Reddy, K.L., Zullo, J.M., Bertolino, E., and Singh, H. (2008). Transcriptional

repression mediated by repositioning of genes to the nuclear lamina. Nature

452, 243–247.

Schmiedeberg, L., Skene, P., Deaton, A., and Bird, A. (2009). A temporal

threshold for formaldehyde crosslinking and fixation. PLoS ONE 4, e4636.

Schubeler, D., MacAlpine, D.M., Scalzo, D., Wirbelauer, C., Kooperberg, C.,

van Leeuwen, F., Gottschling, D.E., O’Neill, L.P., Turner, B.M., Delrow, J.,

et al. (2004). The histone modification pattern of active genes revealed

through genome-wide chromatin analysis of a higher eukaryote. Genes Dev.

18, 1263–1271.

Schwaiger, M., Stadler, M.B., Bell, O., Kohler, H., Oakeley, E.J., and

Schubeler, D. (2009). Chromatin state marks cell-type- and gender-specific

replication of the Drosophila genome. Genes Dev. 23, 589–601.

Shevelyov, Y.Y., Lavrov, S.A., Mikhaylova, L.M., Nurminsky, I.D., Kulathinal,

R.J., Egorova, K.S., Rozovsky, Y.M., and Nurminsky, D.I. (2009). The B-type

lamin is required for somatic repression of testis-specific gene clusters.

Proc. Natl. Acad. Sci. USA 106, 3282–3287.

Singh, R.K., Kabbaj, M.H., Paik, J., and Gunjan, A. (2009). Histone levels are

regulated by phosphorylation and ubiquitylation-dependent proteolysis. Nat.

Cell Biol. 11, 925–933.


Smith, M.B., and Weiler, K.S. (2010). Drosophila D1 overexpression induces

ectopic pairing of polytene chromosomes and is deleterious to development.

Chromosoma 119, 287–309.

Sparmann, A., and van Lohuizen, M. (2006). Polycomb silencers control cell

fate, development and cancer. Nat. Rev. Cancer 6, 846–856.

Tie, F., Furuyama, T., Prasad-Sinha, J., Jane, E., and Harte, P.J. (2001).

The Drosophila Polycomb Group proteins ESC and E(Z) are present in

a complex containing the histone-binding protein p55 and the histone

deacetylase RPD3. Development 128, 275–286.

Tie, F., Prasad-Sinha, J., Birve, A., Rasmuson-Lestander, A., and Harte, P.J.

(2003). A 1-megadalton ESC/E(Z) complex from Drosophila that contains

polycomblike and RPD3. Mol. Cell. Biol. 23, 3352–3362.

Tirosh, I., and Barkai, N. (2008). Two strategies for gene regulation by promoter

nucleosomes. Genome Res. 18, 1084–1091.

Tolhuis, B., deWit, E., Muijrers, I., Teunissen, H., Talhout, W., van Steensel, B.,

and van Lohuizen, M. (2006). Genome-wide profiling of PRC1 and PRC2

Polycomb chromatin binding in Drosophila melanogaster. Nat. Genet. 38,

694–699.

Tomancak, P., Berman, B.P., Beaton, A., Weiszmann, R., Kwan, E.,

Hartenstein, V., Celniker, S.E., and Rubin, G.M. (2007). Global analysis of

patterns of gene expression during Drosophila embryogenesis. Genome

Biol. 8, R145.

Towbin, B.D., Meister, P., and Gasser, S.M. (2009). The nuclear envelope—

a scaffold for silencing? Curr. Opin. Genet. Dev. 19, 180–186.

van Steensel, B., Delrow, J., and Henikoff, S. (2001). Chromatin profiling using

targeted DNA adenine methyltransferase. Nat. Genet. 27, 304–308.

Wang, Z., Zang, C., Cui, K., Schones, D.E., Barski, A., Peng, W., and Zhao, K.

(2009). Genome-wide mapping of HATs and HDACs reveals distinct functions

in active and inactive genes. Cell 138, 1019–1031.

Wolffe, A.P., and Hayes, J.J. (1999). Chromatin disruption and modification.

Nucleic Acids Res. 27, 711–720.

Woodcock, C.L., Skoultchi, A.I., and Fan, Y. (2006). Role of linker histone in

chromatin structure and function: H1 stoichiometry and nucleosome repeat

length. Chromosome Res. 14, 17–25.

Zhang, P., Du, J., Sun, B., Dong, X., Xu, G., Zhou, J., Huang, Q., Liu, Q., Hao,

Q., and Ding, J. (2006). Structure of human MRG15 chromo domain and its

binding to Lys36-methylated histone H3. Nucleic Acids Res. 34, 6621–6628.

Zhimulev, I.F., Belyaeva, E.S., Makunin, I.V., Pirrotta, V., Volkova, E.I.,

Alekseyenko, A.A., Andreyeva, E.N., Makarevich, G.F., Boldyreva, L.V.,

Nanayev, R.A., and Demakova, O.V. (2003). Influence of the SuUR gene on

intercalary heterochromatin in Drosophila melanogaster polytene chromo-

somes. Chromosoma 111, 377–398.


The Solution Structure of the ADAR2dsRBM-RNA Complex Reveals a Sequence-Specific Readout of the Minor GrooveRichard Stefl,1,4,6 Florian C. Oberstrass,1,6,7 Jennifer L. Hood,3 Muriel Jourdan,1,8 Michal Zimmermann,5

Lenka Skrisovska,1 Christophe Maris,1 Li Peng,2 Ctirad Hofr,5 Ronald B. Emeson,2 and Frederic H.-T. Allain1,*1Institute of Molecular Biology and Biophysics, ETH Zurich, CH-8093 Zurich, Switzerland2Department of Pharmacology, Vanderbilt University, Nashville, TN 37232, USA3Neuroscience Graduate Program, Vanderbilt University, Nashville, TN 37232, USA4National Centre for Biomolecular Research, Faculty of Science, Masaryk University, CZ-62500 Brno, Czechia5Department of Functional Genomics and Proteomics, Institute of Experimental Biology, Faculty of Science,

Masaryk University, CZ-62500 Brno, Czechia6These authors contributed equally to this work7Present address: Department of Bioengineering, Stanford University, 318 Campus Drive, Stanford, CA 94305, USA8Present address: Departement de Chimie Moleculaire, 38041 Grenoble Cedex09, France


DOI 10.1016/j.cell.2010.09.026

SUMMARY

Sequence-dependent recognition of dsDNA-binding

proteins is well understood, yet sequence-specific

recognition of dsRNA by proteins remains largely

unknown, despite their importance in RNA matura-

tion pathways. Adenosine deaminases that act on

RNA (ADARs) recode genomic information by the

site-selective deamination of adenosine. Here, we

report the solution structure of the ADAR2 double-

stranded RNA-binding motifs (dsRBMs) bound to

a stem-loop pre-mRNA encoding the R/G editing

site of GluR-2. The structure provides a molecular

basis for how dsRBMs recognize the shape, and

also more surprisingly, the sequence of the dsRNA.

The unexpected direct readout of the RNA primary

sequence by dsRBMs is achieved via the minor

groove of the dsRNA and this recognition is critical

for both editing and binding affinity at the R/G site

of GluR-2. More generally, our findings suggest

a solution to the sequence-specific paradox faced

by many dsRBM-containing proteins that are

involved in post-transcriptional regulation of gene

expression.

INTRODUCTION

ADARs convert adenosine-to-inosine (A-to-I) by hydrolytic

deamination in numerous mRNA and pre-mRNA transcripts

(Bass, 2002; Nishikura, 2006). Due to the similar base-pairing

properties of both nucleosides, inosine is interpreted as guano-

sine by cellular machineries during the processes of translation

and splicing. In this way, editing-mediated alterations in

sequence can alter codon identity or base-pairing interactions

within higher-order RNA structures (Bass, 2002; Nishikura,

2006). As a result, ADARs can create protein isoforms or regulate

gene expression at the RNA level (Bass, 2002; Nishikura, 2006;

Valente and Nishikura, 2005). ADARs are widely expressed in

most cell types, yet their expression and activity in neuronal

tissues has been shown to be important for proper nervous

system function (Higuchi et al., 2000; Palladino et al., 2000).

Recent high-throughput sequencing analysis of A-to-I editing

identified over 55 editing sites within the coding regions of

mRNAs, with 38 of these sites involving a codon change that

specifies an alternative amino acid. Many of these changes

involve RNA transcripts encoding proteins that are critical for

nervous system function (Li et al., 2009).

ADARs from all characterized species have a modular domain

organization consisting of one-to-three dsRBMs followed by

a conserved C-terminal catalytic adenosine deaminase domain.

The structures of the two dsRBMs and of the isolated catalytic

domain of ADAR2 have been determined in their free states

(Macbeth et al., 2005; Stefl et al., 2006). Among the best-studied

ADAR substrates are pre-mRNAs encoding subunits of the

a-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate (AMPA)-

subtype of ionotropic glutamate receptor (GluR-2, GluR-3 and

GluR-4; Higuchi et al., 2000, 1993; Melcher et al., 1996) that

contain one or both of two highly edited and functionally relevant

sites, namely the R/G and Q/R editing sites (Aruscavage and

Bass, 2000; Lomeli et al., 1994; Melcher et al., 1996).

ADARs can edit RNA substrates either specifically or nonspe-

cifically depending upon the structures of the RNA substrates

(Bass, 2002). In vitro studies have shown editing of up to 50%

of the adenosine residues in both strands using synthetic

dsRNAs that are perfectly complementary (Cho et al., 2003;

Lehmann and Bass, 2000). Such nonspecific editing can be ex-

plained by the presence of dsRBMs which are thought to bind

dsRNA in a sequence-independent manner (Tian et al., 2004),

yet it remains unclear how certain RNA substrates are edited in

a site-specific fashion. Several studies have suggested that the


presence of noncanonical elements in these dsRNAs–such as

mismatches, bulges, and loops–could be important for site-

selective A-to-I conversion (Bass, 2002; Stefl et al., 2006; Tian

et al., 2004).

The dsRBMs of ADARs are not only essential for editing (Stefl

et al., 2006; Valente and Nishikura, 2007), but the dsRBM also

represents the second most abundant family of RNA recognition

motifs. In addition to RNA editing, dsRBMs are involved in

numerous post-transcriptional regulatory processes and most

prominently in micro RNA (miRNA) biogenesis and function

and RNA export (Dreyfuss et al., 2002; Tian et al., 2004). The

few solved structures of dsRBM-containing proteins bound to

short, synthetic RNA duplexes have suggested that dsRBMs

recognize the A-form helix of dsRNA in a sequence-independent

manner, since the majority of dsRBM-RNA interactions involve

direct contact with the 20-hydroxyl groups of the ribose sugars

and direct or water-mediated contacts with nonbridging oxygen

residues of the phosphodiester backbone (Gan et al., 2006;

Ramos et al., 2000; Ryter and Schultz, 1998; Wu et al., 2004),

and that a subclass of dsRBMs prefer stem-loops over A-form

helices (Ramos et al., 2000; Wu et al., 2004).

We previously determined that each of the two dsRBMs of

ADAR2 bind to a distinct location on the GluR-2 RNA encom-

passing the R/G editing site and that the interdomain linker

(amino acids 147-231) is unstructured both in the free protein

and in the complex (Stefl et al., 2006). To better understand

RNA substrate recognition by ADAR2, we have determined the

solution structure of the RNA helix surrounding the editing site

and the solution structure of the two dsRBMs of ADAR2 bound

to the GluR-2 R/G site.

RESULTS

Structure of the GluR-2 R/G RNA Helix Surrounding the

Editing Site

The GluR-2 R/G site (A8) is embedded within a 71 nt RNA stem-

loop containing three base-pair mismatches and capped by a 50-

GCUAA-30 pentaloop (Figure 1A). We previously determined the

structure of the apical part of the stem-loop and showed that the

pentaloop is structured and adopts a fold reminiscent of

a UNCG-type family of tetraloops (Stefl and Allain, 2005). Here,

we have investigated the structure of the RNA helix surrounding

the editing site that contains two A-C mismatches, one at the

editing site (A8) and a second one ten base-pairs downstream

(A18, Figure 1B). Monitoring adenine C2 chemical shifts (a sensi-

tive probe to monitor the protonation state of N1) during a pH

titration, we observed that A8 and A18 are fully protonated below

pH 6.5, partially protonated between pH 6.5–8.5, and unproto-

nated above pH 8.5 (Figures 1H and 1I). The pKa for the adeno-

sines N1 can be estimated between 7 and 7.5 at 310 K, which is

3.3 units higher than the value determined for an isolated AMP

(pKa of 4.0; Legault and Pardi, 1994). Using 863 nOe-derived

distance restraints, we solved the structure of the free RNA in

the protonated state (pH 6.2). The structure is well defined,

even for the A-C mismatches (Figure 1E and Table 1) that are

stacked inside the stem. Therefore, at pH 6.2, the R/G site

has a regular A-form helix structure (Figure 1D) containing two

A+-C base-pairs adopting a wobble conformation, stabilized by

two hydrogen bonds each (Figures 1F and 1G).

Structure of ADAR2 dsRBMs Bound to Their Respective

RNA Targets

Considering the distinct RNA binding location found previously

for each dsRBM (Stefl et al., 2006) and the high molecular weight

(over 50 kDa) of the complex formed between the two dsRBMs of

ADAR2 and the GluR-2 R/G substrate (Figure 1A), we adopted

a modular approach to solve the structure of this complex in

solution. To this end, we first solved the structure of dsRBM1

in complex with a modified GluR-2 upper stem–loop (USL, Fig-

ure 1C, and Figure S1 available online) and then the structure

of dsRBM2 bound to the GluR-2 lower stem-loop that contains

the editing site (LSL, Figure 1B, and Figure S2). The use of

a GluR-2 R/G USL mutant to determine the structure of dsRBM1

in complex with RNA was dictated by the poor data quality that

we obtained with the wild-type (WT) sequence. In changing the

loop sequence to that found in the GluR-3 USL (Aruscavage

and Bass, 2000), we obtained a smaller and more stable RNA

which provided NMR data of higher quality.

A total of 1707 and 1929 nOe-derived distance restraints

(including 36 intermolecular ones for each complex) for ADAR2

dsRBM1–GluR-2 R/G USL mutant and ADAR2 dsRBM2–GluR-

2 R/G LSL complexes, respectively, were used to obtain well-

defined structures (Figure 2 and Table 1). The two dsRBM-RNA

complexes are stabilized by a combination of hydrophobic inter-

actions, hydrogen bonding and electrostatic contacts. In both

dsRBM–RNA complexes, the dsRBMs adopt the expected

abbba topology in which the two a helices are packed along

the three-stranded antiparallel b sheet. The entire interaction

surface spans 12-14 base-pairs covering two minor grooves

and a major groove (Figure 2). In both complexes, three distinct

regions of the dsRBMs are involved in interaction with RNA. The

first region is the helix a1, which interacts with the first minor

groove of the RNA. The second region is a well-conserved

KKNAK-motif, located at the amino-terminal tip of helix a2 and

the preceding loop, that contact the RNA with nonsequence

specific contacts between lysine side-chains and the phosphate

oxygens across the major groove of the RNA (Lys127, 128, and

131 for dsRBM1 and Lys281, 282, 285 for dsRBM2, Figure 2). In

addition, the dipole moment of helices a2 creates a positive

charge in the N-terminal tip of these helices that interacts with

the negatively charged phosphate backbone. This second set

of interactions is mediated by the main-chain amides of K127

and K281, which are hydrogen bonded with the phosphates

oxygen of A24 and U11, respectively (Figure 2). The third region

of contact is the b1-b2 loop which interacts with the second

minor groove of the RNA. The overall architecture of these two

complexes resembles other previously determined dsRBM–

RNA structures (Blaszczyk et al., 2004; Gan et al., 2008; Gan

et al., 2006; Ramos et al., 2000; Ryter and Schultz, 1998; Stefl

et al., 2005a; Wu et al., 2004). However, a detailed inspection

of the interaction regions revealed striking differences between

the two complexes and other dsRBM-RNA complexes, particu-

larly in the first and the third regions where both dsRBMs present

unexpected sequence-specific contacts to the RNA minor

grooves (Figure 2).


Sequence-Specific Recognition by ADAR2 dsRBM1

In the ADAR2 dsRBM1–RNA complex, contacts from helix a1 are

centered at the A32-U40 base-pair below the UCCG tetraloop

(Figures 2A and 2C). Met84 makes a sequence-specific hydro-

phobic contact with H2 of A32 and Asn87 contacts the

20-hydroxyl and O2 of U40. The O3 of Glu88 is hydrogen bonded

to the amino group of the first cytosine of the tetraloop. In addi-

tion, Leu83 makes hydrophobic contacts with the sugar of G41.

The entire helix a1 is tightly inserted in the minor groove created

by the UCCG tetraloop and two adjacent base-pairs (Figure 2A).

The b1-b2 loop of dsRBM1 binds the following minor groove of

the RNA. This minor groove is widened as it has to accommo-

date base-pairing of two guanosines that make an N1 symmet-

rical G22-G50 mismatch (Figures 2A and 2D) that are the center

of this interaction. Val104 side-chain contacts the H8 of G50 (that

adopts a syn conformation) and a sequence-specific hydrogen

D

F

G

E

H

I

A

3’

5’

3’

5’

A8

A18

A8

A18

C64

C54

C64

C54

A8

C64

A18

C54

H+

H+

3’

5’

8.5 8.0 7.5 7.0 6.5

δ 1H (ppm)

152

150

148

146

144

δ3

1)

mp

p( C

A7C2-H2

A61C2-H2A57C2-H2

A52C2-H2

A18C2-H2A8C2-H2

A19C2-H2

A7 C2A8 C2

A18 C2 A19 C2

A52 C2

A57 C2

A61 C2

1

3

5

7

9

3 5 7 9 11

pH

∆

mp

p 3

1C

δ

U C

U G

C-G

C-G

U-A

A-U18A C 54

G-C

G-C

U-A

G-C

G-C

G.UU-A

G-C

G-C8A C 64

A-U

U.GG-C

G-C

UC A

G A

U-A

A-U

U.GA-U

A-U

C-G

A-U

A-U

U-A

A-U

U-A

G G

A-U

U-A

A-U

A C

G-C

G-C

U-A

G-C

G-C

G U

U-A

G-C

G-C

A C

A-U

U G

U-A

A-U

C-G5’ 3’

.

.

GluR-2 R/G

lower stem-loop (LSL) C C

U G

A-U

A-U

A-U

C-G

A-U

A U

U-A

A-U

U-A22G G 50

A-U

U A

G-C

G-C5’ 3’

upper stem-loop (USL)

-

-

U.G

5’ 3’

G-C

G-C

8

18

22

32 40

50

54

64

32 40

B C

Figure 1. Secondary Structures of the RNAs and Solution Structure of GluR-2 R/G LSL RNA

(A) Secondary structure of GluR-2 R/G RNA. The indicated binding regions for the dsRBMs were proposed previously (Stefl et al., 2006).

(B) Secondary structure of the GluR-2 R/G lower stem-loop (LSL).

(C) Secondary structure of the GluR-2 R/G upper stem-loop (USL).

(D) Stereo view of the most representative structure of GluR-2 R/G LSL RNA. The A+-C wobble base-pairs are highlighted in bold sticks.

(E–G) (E) Overlay of the 20 lowest energy structures of GluR-2 R/G LSL. The A+-C wobble base-pairs A18-C54 (F) and A8-C64 (G) are shown.

(H) H2-C2 region of adenines in the 13C-1H-HSQC spectra of the GluR-B R/G LSL is shown at pH 4.7 (green peaks), 6.6 (blue peaks), 7.9 (orange peaks) and 8.9

(red peaks). The two adenines involved in the A+-C wobble base-pair showed drastic perturbation.

(I) Diagram showing the pH-dependence of 13C chemical shift changes of adenine C20s.


Table 1. NMR and Refinement Statistics for the GluR-2 R/G Upper Stem-Loop RNA Bound to ADAR2 dsRBM1, the Free GluR-2 R/G

Lower Stem-Loop RNA, and Its Complex with ADAR2 dsRBM2, and the RDC-Reconstructed Complex of the Full-Length GluR-2 R/G

Stem-Loop RNA Bound to ADAR2 dsRBM12

USL RNA – dsRBM1 Complex LSL RNA LSL – dsRBM2 Complex SL RNA – dsRBM12 complex

USL RNA dsRBM1 LSL RNA dsRBM2 SL RNA dsRBM12

NMR Distance and Dihedral Constraints and RDCs

Distance restraints

Total NOE 645 927 781 702 1054 1252 1981

Intraresidue 309 201 389 365 216 620 417

Interresidue 336 726 392 337 838 631 1564

Sequential (ji-jj = 1) 270 252 352 306 241 555 493

Nonsequential (ji-jj > 1) 66 474 40 31 597 76 1071

Hydrogen bonds 35 64 81a 75 62 132 126

Protein–RNA intermolecular 36 36 72

Total dihedral angle restraints 180 252 267

RNA

Sugar pucker 34 84 84

Backboneb 146 168 183

RDC restraints 45d

Structure Statisticsc

Violations (mean and SD)

Number of distance restraint

violations > 0.2 A

8.45 ± 2.50 0 1.10 ± 1.25 14.31 ± 3.86

Number of dihedral angle

restraint violations > 5!

0.7 ± 0.47 0 0

5.30 ± 3.32

Max. dihedral angle restraint

violation (!)

5.82 ± 1.22 3.28 ± 0.77 2.69 ± 1.12 15.51 ± 2.36

Max. distance constraint

violation (A)

0.29 ± 0.03 0.16 ± 0.01 0.23 ± 0.06 0.32 ± 0.05

Deviations from idealized

geometryd

Bond lengths (A) 0.0042 ± 0.00007 0.0046 ± 0.00005 0.0041 ± 0.00005 0.0048 ± 0.00005

Bond angles (!) 1.989 ± 0.011 2.137 ± 0.017 1.903 ± 0.011 1.995 ± 0.008

RDCs violations

Absolute RDC violations (Hz) 1.12 ± 0.82

Average pairwise r.m.s.d (A)c

Protein (79-142) for dsRBM1;

(221-282) for dsRBM2

Heavy atoms 1.11 ± 0.17 1.01 ± 0.12 1.60 ± 0.36

Backbone atoms 0.59 ± 0.14 0.37 ± 0.08 1.22 ± 0.42

RNA

All RNA heavy atoms 0.60 ± 0.16 1.15 ± 0.35 1.48 ± 0.51 1.30 ± 0.40

Complex

All complex heavy atoms 1.01 ± 0.15 1.49 ± 0.39 1.75 ± 0.31a In the final structure calculations of the free RNA, H-bond restraints were applied in the two A-Cmismatches. This is based on initial structures and on

the protonation state of A8/A18. For the structures of the RNA in complex no H-bond restraints for the two A-C mismatches have been applied.bBased on A-form geometry derived from high-resolution crystal structures: a(270!–330!), b(150!–210!), g(30!–90!), d(50!–110!), 3(180!–240!), and z

(260!–320!). These restraints were used only for the double-helical region. No angle restraints were imposed on the two A-Cmismatches and the loops.cCalculated for an ensemble of the 20 lowest energy structures.d 16 RDCs of dsRBM1 and 29 RDCs of dsRBM2.


Figure 2. RNA Recognition by ADAR2 dsRBM1 and dsRBM2

(A) Stereo view of the most representative structure of dsRBM1 bound to USL RNA. The RNA is represented as a yellow stick model and the protein is shown as

a ribbon model with residues that contact the RNA shown in green. Helix a1 and the b1-b2 loop that mediate the sequence-specific contacts are colored in red.

Hydrogen bonds are indicated by magenta dotted lines. (B) Scheme showing contacts between dsRBM1 and the USL RNA. Protein residues that form hydrogen

bonds to the RNA are shown in blue and the one having hydrophobic interactions are in yellow. Close-up view of minor groove sequence-specific recognitions

mediated by helix a1 (C) and the b1-b2 loop (D) of dsRBM1. (E) Overlay of the 20 lowest energy structures of the dsRBM1-USL complex. (F) Stereoview of themost

representative structure of the dsRBM2 bound to LSL RNA. Helix a1 and the b1-b2 loop that mediate the sequence-specific contacts are colored in blue. (G)

Scheme showing contacts between dsRBM2 and the LSL RNA. Close-up view of the minor groove sequence-specific recognitions mediated by helix a1 (H)

and the b1-b2 loop (I). (J) Overlay of the 20 lowest energy structures of the dsRBM2-LSL complex. For NMR data of these two complexes, see also

Figure S1 and Figure S2.


bond is formed between the main-chain carbonyl of V104 and

the amino group of G22. The widened minor groove accommo-

dates additional interactions between three side-chains

(Phe109, Pro107, His105) and the sugars of the base-pairs

above and below. Altogether, dsRBM1 binds the RNA stem-

loop at a single register via two sequence-specific contacts at

two consecutive RNA minor grooves: a hydrogen bond to the

amino group of the G22 in the GG mismatch via the b1-b2 loop

and an hydrophobic contact to the adenine H2 of A32 via

Met84 in helix a1.

Sequence-Specific Recognition by ADAR2 dsRBM2

The dsRBM2of ADAR2 is adjacent to the deaminase domain and

is essential for A-to-I editing at the R/G site (Stefl et al., 2006; Xu

et al., 2006). In the ADAR2 dsRBM2–GluR-2 R/G LSL complex,

Asn241, Glu242, Met238, Val 237 of helix a1 contact the minor

groove region centered at the A18-C54 mismatch (Figures 2F

and 2H). At pH 7.6, where the protein-RNA complex has been

determined, this mismatch is unprotonated and Met238 makes

a sequence-specific hydrophobic contact with A18H2. Contacts

to the base-pair above and below by Asn241 and Glu242, and by

Val 237, respectively, further stabilize the interaction of helix a1 in

this region (Figure 2H). The b1 b2 loop of dsRBM2 interacts with

the second minor groove. The contacts are centered at the

G9-C63 Watson-Crick base-pair located above the A8-C64

mismatch containing the editing site. A sequence-specific

hydrogen bond is formed between the main-chain carbonyl of

Ser258 and the amino of G9 (Figures 2F and 2I). Additionally,

nonsequence specific contacts between the side-chains of Ser

258, His 259 and Phe 263 and the G9-C63 base-pair and the

base-pairs above and below increase the stability of the interac-

tion with the RNA minor groove (Figure 2G). In the vicinity of the

editing site, dsRBM2 contacts C63, while A8 is not contacted by

any residue from the b1-b2 loop therefore making A8 accessible

to the deaminase domain. Altogether, dsRBM2 similar to

dsRBM1, recognizes the RNA helix via two sequence-specific

contacts at two consecutive RNA minor grooves: a hydrogen

bond to the amino group of the G9 at the GC 30 to the editing

site via the b1-b2 loop and a hydrophobic contact to the adenine

H2 of A18 via Met238 in helix a1. In the NMR spectra (data not

shown), we could observe intermolecular nOes corresponding

to dsRBM2 being positioned at a second binding register one

base-pair above (although with only 20% occupancy). In this

case the b1-b2 loop contact G10 and Met 238 contact A19.

Although two consecutive binding sites for dsRBM2 are

observed here, they both confirm the sequence-specific nature

of the dsRBM2-RNA interaction.

Structure of ADAR2 dsRBM12 in Complex

with GluR-2 R/G RNA

Next, we determined the structure of ADAR2 dsRBM12 in

complex with GluR-2 R/G RNA (Figures 3A and 3B). To calculate

an atomic model of this complex, we used the distance

constraints measured in the two sub-complexes described

above (Figure 3C). This strategy could be used considering (1)

the distinct RNA binding location for each dsRBMs, with no

mutual interactions (Stefl et al., 2006), (2) the flexible unstruc-

tured linker connecting dsRBM1 and dsRBM2 in the complex

Figure 3. Structure of ADAR2 dsRBM12 Bound to GluR-2 R/G

(A) Stereo view of the most representative RDC-reconstructed structure of the ADAR2 dsRBM12 bound to GluR-2 R/G. The RNA is represented as a stick model

(in gray; the edited adenosine is highlighted in pink) and the protein is shown as a ribbon model (dsRBM1 in red; dsRBM2 in blue; linker in yellow). (B) Top view of

the complex. Overlay of the 20 lowest energy structures calculated without (C) and with RDCs (D), superimposed on dsRBM1.


(Stefl et al., 2006) and (3) an overlap in the RNA sequence of the

joint region of the subcomplexes (Figure 1). Long-range struc-

tural constraints for this elongated complex were derived from

residual dipolar couplings (RDCs) measured with a deuterated

protein on the full-length complex (dsRBM12 bound to GluR-2

R/G RNA, Figure 1A). The pentaloop which is not contacted by

dsRBM1 was modeled using the structure that was determined

previously (Stefl and Allain, 2005). With this strategy, we could

then determine a precise solution structure of this 50 kDa

complex using 45 15N-1HRDCs (Figure 3D, Table 1). In the struc-

ture, the two dsRBMs bind one face of the RNA covering approx-

imately 120 degrees of the space around the RNA helix

(Figure 3B). This suggests that the binding of an additional mole-

cule of ADAR2 would be sterically possible, consistent with

studies indicating that ADAR2 dimerization is necessary for

RNA editing (Chilibeck et al., 2006; Cho et al., 2003; Gallo

et al., 2003; Valente and Nishikura, 2007).

Sequence-Specific Contacts of ADAR2 dsRBMs

Are Important for Binding Affinity

To confirm the ADAR2 dsRBMs sequence-specific preference in

a quantitative solution binding assay, we performed fluores-

cence anisotropy (FA) experiments by titrating dsRBM1 and

dsRBM2 against labeled USL and LSL RNAs, respectively.

Unlabeled wild-type and mutant RNAs (Figure S3) were used

for competition experiments as described in Experimental

procedures. The equilibrium dissociation constants were calcu-

lated from the displacement of the binding curves (Figure 4). We

designed two sets of mutations, one set was designed to change

the recognition sequence of USL and LSL RNAs (Figures 4A and

330 (± 30)

ref. Flc-USL

Kd [nM]

USL wt

USL G22A/G50U/G41A

USL A32G

500 (± 40)

>1000

ref. Flc-USL

Kd [nM]

330 (± 30)USL wt

USL C34U 450 (± 60)

USL G50C 370 (± 40)

ADAR2-dsRBM1 Change of recognition sequence ADAR2-dsRBM1 Change of shape and in the loop

Norm

alis

ed flu

ore

scence a

nis

otr

opy

Norm

alis

ed flu

ore

scence

anis

otr

opy

ADAR2-dsRBM1 [nM] ADAR2-dsRBM1 [nM]

ADAR2-dsRBM2 Change of recognition sequence ADAR2-dsRBM2 Change of shape

Norm

alis

ed

flu

ore

scence a

nis

otr

opy

Norm

alis

ed flu

ore

scence a

nis

otr

opy

ADAR2-dsRBM2 [nM] ADAR2-dsRBM2 [nM]

ref. Flc-LSL

Kd [nM]

370 (± 30)LSL wt

>2000LSLG9A/G10A/C62U/C63U

1300 (± 200)LSL A18G/A19G/U53C

ref. Flc-LSL

Kd [nM]

370 (± 30)LSL wt

700 (± 70)LSL C54U/C64U

A B

C D

Figure 4. ADAR2 dsRBMs Bind Preferentially to RNAs that Contains Their Sequence-Specific Recognition Motifs

(A) ADAR2 dsRBM1 was titrated with fluorescently labeled USL and binding was measured by fluorescence anisotropy (black circles; fluorescein labeled refer-

ence, Flc-USL). The same experiment was then carried out in the presence of competing unlabeled USL wt (B), USL G22A/G50U/G41A mutant (;), and USL

A32G mutant (6). Equilibrium dissociation constants (Kd) were calculated from the best fit to the data as described in Experimental Procedures.

(B) The same assay as shown in (A) but for USL C34U mutant (;) and USL G50C mutant (6).

(C) ADAR2 dsRBM2 was titrated with fluorescently labeled LSL and binding was measured by fluorescence anisotropy (C; fluorescein labeled reference, Flc-

LSL). The same experiment was then carried out in the presence of competing unlabeled LSL wt (B), LSL G9A/G10A/C62U/C63U mutant (;), and LSL A18G/

A19G/U53C mutant (6).

(D) The same assay as shown in (C) but for LSL C54U/C64U mutant (;).Wild-type and mutant sequences are shown in Figure S3.


4C and Figure S3) and a second set was designed to maintain

the recognition sequence, but change the RNA shape via

mismatches of USL and LSL RNAs intoWatson-Crick base-pairs

(Figures 4B and 4D and Figure S3) to measure their effect on

overall binding affinity. In mutating any of the bases that are

recognized in a sequence-specific manner by dsRBM1 in USL

(G22, A32 or C34), the apparent affinity is reduced compared

to the wild-type (Figures 4A and 4B). However when the G22-

G50 mismatch is replaced by a Watson-Crick G22-C50 pair,

the affinity is almost identical to wild-type RNA, confirming that

dsRBM1 recognizes the sequence rather than the shape of the

RNA helix (note that G41 was mutated in the first RNA mutant

to prevent the sequence-specific recognition of G41 by

dsRBM1). Similarly for the LSL, when G9 or A18 are mutated,

dsRBM2 binding is reduced more than five-fold (Figure 4C),

yet when the two AC mismatches are replaced by Watson-Crick

AU pairs, the affinity is only reduced by two-fold (Figure 4D). In

this latter context, the sequence-specific contacts are the

same for the WT and mutant RNAs, but the presence of

a more deformable A18-C54 base-pair in the WT structure could

explain the higher affinity of dsRBM2 to the WT RNA (note that

additional mutations were introduced in the first two RNA

mutants of LSL to abolish the two binding registers found in

the wild-type LSL). Altogether, the FA data strongly support

the idea that the sequence-specific interactions observed in

the structures of ADAR2 dsRBMs-dsRNA are important for the

affinity of both dsRBMs and that they finely tune the preferential

binding to these recognition motifs.

Sequence-Specific Contacts of ADAR2 dsRBMs Are

Important for Editing

To test the functional importance of the four sequence-specific

contacts identified in the ADAR2 dsRBM12-GluR-2 R/G RNA

complex, single amino acid mutants in helix a1 (M84 or M238)

were mutated to alanine or double mutants in the b1 b2 loop

in either dsRBM1 or dsRBM2 were evaluated for their ability to

edit the wild-type GluR-2 R/G site (Figure 5A). It was necessary

to generate double mutants around the carbonyls of V104 in

dsRBM1 and S258 in dsRBM2 to change the structure of the

main-chain of this loop. All four mutants showed a significant

decrease in RNA editing ranging from a near ablation of editing

(S258A,H259A in the b1 b2 loop of dsRBM2), to 20% editing

(V104A,H105A in the b1 b2 loop of dsRBM1 and M84A in helix

a1 of dsRBM1), to 30% editing (M238A in helix a1 of dsRBM2) of

that demonstrated by the wild-type protein. These data clearly

show that the loss of the sequence-specific contacts of any of

the two dsRBMs strongly decreases editing at the R/G site

with the contact mediated by the b1 b2 loop of dsRBM2 more

strongly affecting editing than the other contacts. In agreement

with deletion studies of ADAR2 (Macbeth et al., 2004; Stefl

et al., 2006), the S258A,H259A mutations have a stronger effect,

likely due to the binding of the b1 b2 loop of dsRBM2 near the

editing site.

Converse experiments in which mutations in the sequence-

specific recognitionmotifs of dsRBM2 (mut1 andmut2), dsRBM1

(mut4) or both (mut3) within the GluR-2 RNA (Figure S4) were as-

sessed for their ability to affect R/G editing by wild-type ADAR2

revealed a significant decrease in maximal editing rates (Vmax)

for all RNA mutants tested (Figure 5B) providing further support

for the functional significance of these contacts. Best-fit kinetic

curves for wild-type and mutant RNAs corresponded to a model

of substrate inhibition, consistent with previously observed

kinetic models for ADARs in which the formation of a ternary

complex containing an ADAR dimer and RNA substrate is

required for efficient adenosine deamination.

DISCUSSION

In solving the structure of ADAR2 dsRBMs bound to the GluR-2

R/G site, we demonstrated that despite forty-four possible

Figure 5. Sequence-Specific Contacts of ADAR dsRBMs Are Impor-

tant for Editing Activity

(A) Quantitative analysis of in vitro editing efficiency for ADAR2 dsRBM double

mutants; all mutants were assayed in duplicate for in vitro editing activity at the

GluR-2 R/G site using three independent nuclear extracts (mean ± SEM;

*p < 0.05, **p < 0.005; ***p < 0.001).

(B) Kinetic analysis of wild-type ADAR2 editing with GluR-2 R/G mutants.

Increasing concentrations of GluR-2 RNAs (see Figure S4; wild-type C; mut

1 -, mut 2 B, mut 3 :, mut 4 ,) were incubated with wild-type rat

ADAR2 protein as described above; all mutant RNAswere assayed in triplicate

for determination of in vitro reaction velocity (mean ± SEM). Nonlinear fitting of

kinetic curves corresponded to a model of substrate inhibition (R2 = 0.91-0.98

for all RNAs) with Vmax values corresponding to 3.92, 3.84, 2.08, 1.20, and

1.29 fmol/min for wild-type, mut1, mut2, mut3, and mut4, respectively. Wild-

type and mutant sequences are shown in Figure S4.


binding sites on the GluR-2 R/G RNA stem-loop (considering

a 32 base-pair stem, a 10 base-pair register between the two

sequence-specific contacts and two possible orientations for

the dsRBM), each dsRBM binds at a very specific register on

this large RNA molecule. This binding is achieved by a direct

readout of the RNA sequence in the minor groove of the

A-form helix. The two dsRBMs of ADAR2 use helix a1 and the

b1-b2 loop as molecular rulers to find their binding register in

the RNA minor groove of the GluR-2 R/G RNA. Through the

b1 b2 loop, the carbonyl oxygens of Val104 in dsRBM1 and

Ser258 in dsRBM2 contact the amino groups of base-paired

guanines, G22 and G9 respectively. The same type of

sequence-specific RNA recognition of GC or GU base-pairs in

the minor groove of RNA helices have been observed in several

ribosomal proteins of the large subunit (Klein et al., 2004) and in

some tRNA synthetases bound to RNA (Rould et al., 1989)

although the fold of these proteins and the overall binding

mode are different from a dsRBM. Through helix a1, the side-

chain methyl groups of Met84 in dsRBM1 and of Met238 in

dsRBM2 are in contact with the H2s of A32 and A18, respec-

tively. Recognition of these two anchoring points in the minor-

groove, separated by 9 and 8 base-pairs for dsRBM1 and

dsRBM2, respectively, illustrates how the two dsRBMs find their

sequence-specific binding registers, demonstrating that these

dsRBMs have more sequence-specificity than previously

thought. Interestingly, in each complex, one of the two anchoring

points involves a mismatched base-pair (the G22-G50 base-pair

for dsRBM1 and the A18-C54 base-pair for dsRBM2). It is there-

fore possible that the highly exposed amino or C2H2 groups of

thesemismatches in theminor groove further assist the dsRBMs

of ADAR2 to find their binding register, supporting earlier findings

that these two mismatches are important for positioning ADAR2

at the R/G site (Ohman et al., 2000). In addition to sequence-

specific interactions between ADAR2 dsRBMs and its GluR-2

target, K127 (dsRBM1) and K281 (dsRBM2) make contacts

with phosphate oxygens across the major groove of the RNA

(Figure 2). These basic amino acid moieties are conserved in

the loop between the b3 and a2 regions for all dsRBMs (Tian

Figure 6. RNA Recognition Code of Various

dsRBMs

(A) and (B) Overlay of the ADAR2 dsRBM1 (in blue),

ADAR2 dsRBM2 (in red), and Aquifex aeolicus

RNaseIII dsRBM (in gray) structures highlights

the variability of helix a1 within the dsRBM fold

and its importance for the determination of the

register length between the two specific contacts

on the RNA helix (C). For Aquifex aeolicus RNaseIII

dsRBM–dsRNA interactions, see also Figure S5

and for sequence alignments of different dsRBMs,

see also Figure S6.

et al., 2004) and mutation of these resi-

dues in PKR and Staufen have been

shown to ablate dsRNA-binding activity

(McMillan et al., 1995; Ramos et al.,

2000), indicating the importance of both

sequence-specific and sequence–inde-

pendent recognition of the RNA substrate for site-specific aden-

osine deamination.

Prior to this work, the structures of only four dsRBM-contain-

ing proteins in complex with RNA had been determined by

X-ray crystallography (XlrbpA and Aquifex aeolicus (Aa) RNa-

seIII) or NMR spectroscopy (Staufen and Rnt1p; Gan et al.,

2006; Ramos et al., 2000; Ryter and Schultz, 1998; Wu et al.,

2004). In the two solution structures, the dsRBMs appear to

recognize primarily the loop of the RNA while in the two crystal

structures the dsRBMs are found bound across the junction

between coaxially stacked helices. Lack of clear sequence-

specific contacts led to the general opinion that dsRBMs are

shape-specific rather than sequence-specific RNA binding

domains (Stefl et al., 2005a). The two dsRBM-RNA complexes

of ADAR2 reported here have revealed that dsRBMs recognize

not only the shape of the RNA (a stem-loop for dsRBM1 and an

A-form helix for dsRBM2), but also more surprisingly the

sequence of the RNA. Interestingly, in a recent crystal structure

of an Aa RNaseIII dsRBM bound to a stem-loop, sequence-

specific contacts in the minor groove via helix a1 and the

b1 b2 loop have been observed (Gan et al., 2008). The helix

a1 in Aa RNaseIII is elongated by one turn compared to the

helix a1 of the dsRBMs of ADAR2 and a Gln side-chain recog-

nizes a guanine by two sequence-specific hydrogen bonds

(Figure S5). The contact mediated by the b1 b2 loop in Aa

RNaseIII are similar to the dsRBMs in ADAR2. The b1 b2

loop has the same length (six amino acids) and the main-chain

carbonyl of the third residue of the loop is hydrogen bonded to

a guanine amino of a GU base-pair. Despite similarities in the

mode of binding, the three dsRBMs recognize different

sequences and different register lengths. The dsRBM of Aa

RNaseIII preferentially recognizes an RNA helix containing

a G-X10-G sequence while the dsRBM1 and dsRBM2 of

ADAR2 preferentially recognize G-X9-A and G-X8-A sequences,

respectively (Figure 6). The length and the positioning of helix

a1 relative to the dsRBM fold appear to be the key structural

elements that determine the register length of the different

dsRBMs (Figure 6C).


Our findings regarding the RNA binding specificity of dsRBMs

have important implications for the sequence-specificity

paradox of ADAR2, but also of many other dsRBM-containing

proteins that continue to puzzle investigators (Tian et al., 2004).

Apparent differences in the sequences of dsRBMs between

mammalian ADAR2 and ADAR1 (Figure S6), where ADAR1

dsRBMs appears to have a longer helix a1 and lack the

ADAR2 equivalent of Met 84 and Met 238, could explain why

ADAR1 and ADAR2 have different substrate specificities (Bass,

2002; Lehmann and Bass, 2000). Furthermore, our structure

shows how dsRBM2 of ADAR2 binds the GluR-2 R/G site near

the editing site in recognizing the amino group of the guanosine

30 to the edited A. This would explain the strong preference for

a guanosine moiety 30 to the edited adenosine that is found in

a great majority of substrates selectively edited by ADAR2

(Bass, 2002; Lehmann and Bass, 2000; Li et al., 2009; Riedmann

et al., 2008) andmore recently in long double-stranded RNA (Eg-

gington and Bass, personal communication). This sequence

preference disappears when the dsRBMs are deleted from

ADAR2 (Eggington and Bass, personal communication) further

supporting that this sequence requirement is due to dsRBM

binding. Finally, in interacting with the guanosine 30 to the edited

adenosine and to the nucleotide that base-pairs with the editing

site, dsRBM2 not only brings the deaminase domain in close

proximity to the editing site, but also does not prevent access

of the adenosine to the deaminase domain. When this precise

positioning is impaired, specific editing is nearly abolished (see

the effect of the S258A, H259A mutant) which emphasizes the

functional importance of sequence-specific recognition of RNA

by dsRBMs for A-to-I editing.

The sequence-specific contacts that we observed with the

dsRBMs ADAR2 are interesting when comparing sequence

alignments of several dsRBM structures that have been deter-

mined (Figure S6). This alignment reveals a surprisingly high vari-

ability in the length and amino acid sequence composition of the

two regions of the dsRBMs mediating the sequence-specific

interactions with the RNA, namely the helix a1 and the b1 b2

loop. This strongly suggests that dsRBMs are likely to have

different binding specificity in agreement with reports indicating

that dsRBMs from different proteins are not functionally inter-

changeable (Liu et al., 2000; Parker et al., 2008). Similar to

ADAR2, many dsRBM-containing proteins involved in miRNA

and siRNA processing and function are likely to bind RNA in

a sequence-specific manner, that would modulate their target

selection and mechanism of action. For example, DICER was

shown to compete with ADARs for the same RNA substrates

(Kawahara et al., 2007; Yang et al., 2006). Interestingly, ADARs

modulate the processing of miRNA precursors not only by

A-to-I modifications that alter the secondary structure of pri-

miRNA (Kawahara et al., 2007; Tonkin and Bass, 2003; Yang

et al., 2006), but also simply by RNA-binding alone to pri-miR-

NAs, as recently shown with catalytically inactive ADARs (Heale

et al., 2009). This latter function for ADARs, as regulators of pri-

miRNA processing, closely resemble that found for single-

stranded sequence-specific RNA-binding proteins such as

Lin28, hnRNP A1 or KSRP (Guil and Caceres, 2007; Heo et al.,

2008, 2009; Michlewski et al., 2008; Newman et al., 2008; Tra-

bucchi et al., 2009). Furthermore, RNAi activity has been shown

to coincide with siRNA sequence motifs (Katoh and Suzuki,

2007). Altogether it is becoming clear that sequence-specific

recognition mediated by dsRBMs is functionally important for

dsRBM containing proteins. We have demonstrated here with

ADAR2 how such sequence-specific recognition is mediated in

dsRBMs and how this is relevant for RNA editing. Future work

will be required to elucidate the variations in dsRNA-binding

specificity and their functional relevance for numerous other

members of the dsRBM-containing protein family.


Preparation of Proteins

Details on cloning, expression and purification of the ADAR2 dsRBM1, ADAR2

dsRBM2, and ADAR2 dsRBM12 constructs have been described previously

(Stefl et al., 2005b, 2006).

NMR Spectroscopy

All NMR spectra were acquired at 310 K. Spectra were recorded at 500, 600,

and 900 MHz Bruker spectrometers. All spectra were processed with

XWINNMR or Topspin1.3/2.0 (Bruker BioSpin) and analyzed with Sparky 3.0

(Goddard T.G. and Kellner D.G., University of California, San Francisco). The1H, 13C and 15N chemical shifts of the protein in complex, were assigned by

standard methods (Sattler et al., 1999). The 1H-15N HSQC and 1H-13C HSQC

spectra of dsRBM1 and dsRBM2 in free and bound forms are shown in

Figure S1 and Figure S2. All distance restraints were derived from 3D15N,13C-edited NOESYs and 2D 1H-1H NOESY (tm = 150 ms) collected at

900 MHz. RNA exchangeable proton resonances were assigned using 1H-1H

NOESY spectrum (tm = 200 ms) at 278 K. Nonexchangeable proton reso-

nances were assigned using 1H-1H, NOESY, 1H-1H TOCSY, 1H-13C HSQC,

3D 13C-edited NOESY, 2D 1H-1H double-half-filtered NOESY (tm = 150 ms)

(Peterson et al., 2004) and 3D 13C F1-edited, F3-filtered NOESY-HSQC spec-

trum (tm = 150 ms) (Zwahlen et al., 1997) in 99.99% 2H2O (v/v). The NOEs were

semiquantitatively classified based on their intensities in the 2D and 3D NO-

ESY spectra. Hydrogen bond distance restraints were used for base-pairs,

when the imino-protons were observed experimentally. The assignments of

intermolecular NOEs were based on 3D 13C F1-edited, F3-filtered NOESY-

HSQC spectrum (tm = 150ms), 2D 1H-1H F1-13C-filtered F2-

13C-edited NOESY

(tm = 150 ms) on the protein-RNA complexes with either the protein or the RNA13C-15N labeled. In case of dsRBM2–GluR-2 R/G LSL RNA complex, we

observed an extra set of five weaker intermolecular nOes, which were dis-

carded from structure calculation. These intermolecular restraints cannot be

explained with the presented structure of dsRBM2-GluR-2 R/G LSL RNA

complex. They originate from a minor conformation in which the protein is

shifted up by one base pair toward the UUCG tetraloop.

Structure Calculation and Refinement

Distance constraints for the proteins bound to RNA where generated by the

ATNOS/CANDID package (Herrmann et al., 2002). The accuracy of the list of

automatically generated distance constraints was manually checked.

Distance constraints for the free and bound RNAs aswell as for the intermolec-

ular NOEs were assigned manually. Preliminary structures of the free RNA and

the protein-RNA complexes were obtained by a simulated annealing protocol

in CYANA (Guntert et al., 1997; Herrmann et al., 2002). To impose better

convergence of the ensemble, an artificial torsion angles for the canonical

dsRNA regions were used as described previously (Oberstrass et al., 2006).

Additional angle restraints to maintain proper local geometries were used

(Tsui et al., 2000). The final refinement of all structures was performed using

a 20 ps simulated annealing protocol in AMBER (Case et al., 2002) as

described in the Supplemental Information. From 40 refined structures, the

twenty conformers with the lowest AMBER energy were selected to form the

final ensemble of structures. Structural quality was assessed using PRO-

CHECK (Laskowski et al., 1996). Figures were prepared withMOLMOL (Koradi

et al., 1996) and Pymol (DeLano, 2002).


Fluorescence Anisotropy

Fluorescence anisotropy wasmeasured on a FluoroMax-4 spectrofluorometer

(Horiba Jobin-Yvon, USA) equipped with a thermostated cell holder and an

automatic titrator. All measurements were conducted in 50mM sodium phos-

phate buffer (pH 7.0) and 100mMNaCl at 10 C. To avoid any effects caused by

50-end labeling of RNAs, the experiments were designed as a competition

assay. At first, a reference measurement was carried out in which 1400 ml of

10nM fluorescein labeled wild-type RNA was titrated by the protein. Then,

the same titration experiment was repeated in the presence of 500nM unla-

beled RNA (either wild-type or mutants; Vasiljeva et al., 2008). Total volume

of protein added to each reaction was 33 ml. The fitting was performed using

DynaFit software (Kuzmic, 1996, 2006). Initially, the Kd for the reference

protein–labeled RNA complex was determined. The obtained Kd value was

then used as a fixed parameter when fitting the competition data. A 1:1 binding

stoichiometry was assumed in all cases. The data were normalized for visual-

ization purposes.

Quantitative Analysis of In Vitro RNA Editing

For in vitro editing reactions, a 116 nt RNA encoding a portion of the mouse

GluR-2 pre-mRNAwith the complete R/G duplex was transcribed in vitro (Stefl

et al., 2006) and incubated with wild-type or mutant ADAR2 proteins derived

from nuclear extracts obtained from transiently transfected HEK293 cells

(Sansam et al., 2003). Equivalent amounts of wild-type and mutant ADAR2

protein, as determined by Western blotting, were incubated with 40 ng of

the R/G transcript at 30 C for 20 min. These incubation conditions were deter-

mined empirically by performing time-course analyses with wild-type ADAR2

protein to ensure that the assay was in the linear range (data not shown).

The reaction was stopped and the R/G transcript isolated by direct addition

of TRI Reagent (Molecular Research Center) at the end of the incubation

period. For quantification of RNA editing, the in vitro reaction product was

reverse transcribed using AMV Reverse Transcriptase (Promega) and an anti-

sense primer (50-CGGCCAATCGTACGTACCTCCGGCCGAATTCTACAAACC

GTTAAGAGTCTTA-30) with a unique 50-extension (underlined). The resulting

amplicon was diluted 1:1000 and 1 ml was subsequently amplified by PCR

using sense (50-CCGGGAGCTCATCGCCACACCTAAAGGATCC-30) and anti-

sense (50-CGGCCAATCGTACGTACCTCC-30) primers corresponding to

GluR-2 and the unique 50-extension sequences, respectively. PCR amplicons

were purified using the Wizard SV PCR and Gel Cleanup System (Promega)

and digested with Mse I (New England Biolabs) to generate 100 and 70 bp

products representing edited and nonedited transcripts, respectively. The

resulting digestion products were resolved on a 4% Agarose gel and editing

efficiency was quantified by phosphorimager analysis (GE Healthcare).

In vitro editing reactions using GluR-2 R/G mutant RNAs were performed as

described above with equivalent amounts of wild-type ADAR2 protein derived

from nuclear extracts obtained from transiently transfected HEK293 cells

(Sansam et al., 2003). Wild-type and mutant transcripts were trace labeled

with [a-32P]-UTP and concentrations of in vitro transcribed RNAs were deter-

mined using a Perkin-Elmer Tri-Carb 2800TR scintillation spectrometer based

upon the calculated specific activity for each transcript.


Supplemental Information includes Extended Experimental Procedures and

six figures and can be found with this article online at doi:10.1016/j.cell.

2010.09.026.

ACKNOWLEDGMENTS

This work was supported by the Swiss National Science Foundation (Nr.

3100A0-118118) and the SNF-NCCR structural biology to F.H.T.A and the

National Institutes of Health (R01 NS33323) to R.B.E. R.S. is supported by

the Ministry of Education of the Czech Republic (MSM0021622413, Ingo

LA08008), GACR (204/08/1212, 305/10/1490), GAAV (IAA401630903),

HHMI/EMBO start-up grant, and HFSP Career Development Award. M.Z.

and C.H. are supported by GACR (204/08/H054) and by the Ministry of Educa-

tion of the Czech Republic (MSM0021622415). M.Z. is in receipt of a Brno City

Scholarship for Talented Ph.D. Students. The coordinates of the structures of

GluR-2 LSL RNA, ADAR2 dsRBM1 bound to GluR-2 USL RNA, ADAR2

dsRBM2 bound to GluR-2 LSL RNA and ADAR2 dsRBM12 bound to GluR-2

have been deposited in the Protein Data Bank with accession codes 2l2j,

2I3c, 2l2k, and 2I3j, respectively.

Received: September 21, 2009

Revised: May 26, 2010



REFERENCES

Aruscavage, P.J., and Bass, B.L. (2000). A phylogenetic analysis reveals an

unusual sequence conservation within introns involved in RNA editing. RNA

6, 257–269.

Bass, B.L. (2002). RNA editing by adenosine deaminases that act on RNA.


Blaszczyk, J., Gan, J., Tropea, J.E., Court, D.L., Waugh, D.S., and Ji, X. (2004).

Noncatalytic assembly of ribonuclease III with double-stranded RNA.

Structure 12, 457–466.

Case, D.A., Pearlman, D.A., Caldwell, J.W., Cheatham, T.E., III, Wang, J.,

Ross, W.S., Simmerling, C.L., Darden, T.A., Merz, K.M., Stanton, R.V., et al.

(2002). AMBER 7 (San Francisco: University of California).

Chilibeck, K.A., Wu, T., Liang, C., Schellenberg, M.J., Gesner, E.M., Lynch,

J.M., and MacMillan, A.M. (2006). FRET analysis of in vivo dimerization by

RNA-editing enzymes. J. Biol. Chem. 281, 16530–16535.

Cho, D.S., Yang, W., Lee, J.T., Shiekhattar, R., Murray, J.M., and Nishikura, K.

(2003). Requirement of dimerization for RNA editing activity of adenosine

deaminases acting on RNA. J. Biol. Chem. 278, 17093–17102.

DeLano, W.L. (2002). The PyMOL Molecular Graphics System (Palo Alto, CA,

USA: DeLano Scientific).

Dreyfuss, G., Kim, V.N., and Kataoka, N. (2002). Messenger-RNA-binding

proteins and the messages they carry. Nat. Rev. Mol. Cell Biol. 3, 195–205.

Gallo, A., Keegan, L.P., Ring, G.M., and O’Connell, M.A. (2003). An ADAR that

edits transcripts encoding ion channel subunits functions as a dimer. EMBO J.

22, 3421–3430.

Gan, J., Shaw, G., Tropea, J.E., Waugh, D.S., Court, D.L., and Ji, X. (2008). A

stepwise model for double-stranded RNA processing by ribonuclease III. Mol.

Microbiol. 67, 143–154.

Gan, J., Tropea, J.E., Austin, B.P., Court, D.L., Waugh, D.S., and Ji, X. (2006).

Structural insight into the mechanism of double-stranded RNA processing by

ribonuclease III. Cell 124, 355–366.

Guil, S., and Caceres, J.F. (2007). The multifunctional RNA-binding protein

hnRNP A1 is required for processing of miR-18a. Nat. Struct. Mol. Biol. 14,

591–596.

Guntert, P., Mumenthaler, C., andWuthrich, K. (1997). Torsion angle dynamics

for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273,

283–298.

Heale, B.S., Keegan, L.P., McGurk, L., Michlewski, G., Brindle, J., Stanton,

C.M., Caceres, J.F., and O’Connell, M.A. (2009). Editing independent effects

of ADARs on the miRNA/siRNA pathways (EMBO J.).

Heo, I., Joo, C., Cho, J., Ha, M., Han, J., and Kim, V.N. (2008). Lin28 mediates

the terminal uridylation of let-7 precursor MicroRNA. Mol. Cell 32, 276–284.

Heo, I., Joo, C., Kim, Y.K., Ha, M., Yoon, M.J., Cho, J., Yeom, K.H., Han, J.,

and Kim, V.N. (2009). TUT4 in concert with Lin28 suppresses microRNA

biogenesis through pre-microRNA uridylation. Cell 138, 696–708.

Herrmann, T., Guntert, P., and Wuthrich, K. (2002). Protein NMR structure

determination with automated NOE assignment using the new software

CANDID and the torsion angle dynamics algorithm DYANA. J. Mol. Biol. 319,

209–227.

Higuchi, M., Maas, S., Single, F.N., Hartner, J., Rozov, A., Burnashev, N., Feld-

meyer, D., Sprengel, R., and Seeburg, P.H. (2000). Point mutation in an AMPA


receptor gene rescues lethality in mice deficient in the RNA-editing enzyme

ADAR2. Nature 406, 78–81.

Higuchi, M., Single, F.N., Kohler, M., Sommer, B., Sprengel, R., and Seeburg,

P.H. (1993). RNA editing of AMPA receptor subunit GluR-B: a base-paired

intron-exon structure determines position and efficiency. Cell 75, 1361–1370.

Katoh, T., and Suzuki, T. (2007). Specific residues at every third position of

siRNA shape its efficient RNAi activity. Nucleic Acids Res. 35, e27.

Kawahara, Y., Zinshteyn, B., Sethupathy, P., Iizasa, H., Hatzigeorgiou, A.G.,

and Nishikura, K. (2007). Redirection of silencing targets by adenosine-to-ino-

sine editing of miRNAs. Science 315, 1137–1140.

Klein, D.J., Moore, P.B., and Steitz, T.A. (2004). The roles of ribosomal proteins

in the structure assembly, and evolution of the large ribosomal subunit. J. Mol.

Biol. 340, 141–177.

Koradi, R., Billeter, M., and Wuthrich, K. (1996). MOLMOL: a program for

display and analysis of macromolecular structures. J Mol Graph 14, 51–55,

29-32.

Kuzmic, P. (1996). Program DYNAFIT for the analysis of enzyme kinetic data:

application to HIV proteinase. Anal. Biochem. 237, 260–273.

Kuzmic, P. (2006). A generalized numerical approach to rapid-equilibrium

enzyme kinetics: application to 17beta-HSD. Mol. Cell. Endocrinol. 248,

172–181.

Laskowski, R.A., Rullmannn, J.A., MacArthur, M.W., Kaptein, R., and Thorn-

ton, J.M. (1996). AQUA and PROCHECK-NMR: programs for checking the

quality of protein structures solved by NMR. J. Biomol. NMR 8, 477–486.

Legault, P., and Pardi, A. (1994). In-Situ Probing of Adenine Protonation in Rna

by C-13 Nmr. J. Am. Chem. Soc. 116, 8390–8391.

Lehmann, K.A., and Bass, B.L. (2000). Double-stranded RNA adenosine

deaminases ADAR1 and ADAR2 have overlapping specificities. Biochemistry

39, 12875–12884.

Li, J.B., Levanon, E.Y., Yoon, J.K., Aach, J., Xie, B., Leproust, E., Zhang, K.,

Gao, Y., and Church, G.M. (2009). Genome-wide identification of human

RNA editing sites by parallel DNA capturing and sequencing. Science 324,

1210–1213.

Liu, Y., Lei, M., and Samuel, C.E. (2000). Chimeric double-stranded RNA-

specific adenosine deaminase ADAR1 proteins reveal functional selectivity

of double-stranded RNA-binding domains from ADAR1 and protein kinase

PKR. Proc. Natl. Acad. Sci. USA 97, 12541–12546.

Lomeli, H., Mosbacher, J., Melcher, T., Hoger, T., Geiger, J.R., Kuner, T., Mon-

yer, H., Higuchi, M., Bach, A., and Seeburg, P.H. (1994). Control of kinetic

properties of AMPA receptor channels by nuclear RNA editing. Science 266,

1709–1713.

Macbeth, M.R., Lingam, A.T., and Bass, B.L. (2004). Evidence for auto-inhibi-

tion by the N terminus of hADAR2 and activation by dsRNA binding. RNA 10,

1563–1571.

Macbeth, M.R., Schubert, H.L., Vandemark, A.P., Lingam, A.T., Hill, C.P., and

Bass, B.L. (2005). Inositol hexakisphosphate is bound in the ADAR2 core and

required for RNA editing. Science 309, 1534–1539.

McMillan, N.A., Carpick, B.W., Hollis, B., Toone, W.M., Zamanian-Daryoush,

M., and Williams, B.R. (1995). Mutational analysis of the double-stranded

RNA (dsRNA) binding domain of the dsRNA-activated protein kinase, PKR.

J. Biol. Chem. 270, 2601–2606.

Melcher, T., Maas, S., Herb, A., Sprengel, R., Seeburg, P.H., and Higuchi, M.

(1996). A mammalian RNA editing enzyme. Nature 379, 460–464.

Michlewski, G., Guil, S., Semple, C.A., and Caceres, J.F. (2008). Posttranscrip-

tional regulation of miRNAs harboring conserved terminal loops. Mol. Cell 32,

383–393.

Newman,M.A., Thomson, J.M., and Hammond, S.M. (2008). Lin-28 interaction

with the Let-7 precursor loop mediates regulated microRNA processing. RNA

14, 1539–1549.

Nishikura, K. (2006). Editor meets silencer: crosstalk between RNA editing and

RNA interference. Nat. Rev. Mol. Cell Biol. 7, 919–931.

Oberstrass, F.C., Lee, A., Stefl, R., Janis, M., Chanfreau, G., and Allain, F.H.

(2006). Shape-specific recognition in the structure of the Vts1p SAM domain

with RNA. Nat. Struct. Mol. Biol. 13, 160–167.

Ohman, M., Kallman, A.M., and Bass, B.L. (2000). In vitro analysis of the

binding of ADAR2 to the pre-mRNA encoding the GluR-B R/G site. RNA 6,

687–697.

Palladino, M.J., Keegan, L.P., O’Connell, M.A., and Reenan, R.A. (2000). A-to-I

pre-mRNA editing in Drosophila is primarily involved in adult nervous system

function and integrity. Cell 102, 437–449.

Parker, G.S., Maity, T.S., and Bass, B.L. (2008). dsRNA binding properties of

RDE-4 and TRBP reflect their distinct roles in RNAi. J. Mol. Biol. 384, 967–979.

Peterson, R.D., Theimer, C.A., Wu, H., and Feigon, J. (2004). New applications

of 2D filtered/edited NOESY for assignment and structure elucidation of RNA

and RNA-protein complexes. J. Biomol. NMR 28, 59–67.

Ramos, A., Grunert, S., Adams, J., Micklem, D.R., Proctor, M.R., Freund, S.,

Bycroft, M., St Johnston, D., and Varani, G. (2000). RNA recognition by a Stau-

fen double-stranded RNA-binding domain. EMBO J. 19, 997–1009.

Riedmann, E.M., Schopoff, S., Hartner, J.C., and Jantsch, M.F. (2008). Spec-

ificity of ADAR-mediated RNA editing in newly identified targets. RNA 14,

1110–1118.

Rould, M.A., Perona, J.J., Soll, D., and Steitz, T.A. (1989). Structure of E. coli

glutaminyl-tRNA synthetase complexed with tRNA(Gln) and ATP at 2.8 A reso-

lution. Science 246, 1135–1142.

Ryter, J.M., and Schultz, S.C. (1998). Molecular basis of double-stranded

RNA-protein interactions: structure of a dsRNA-binding domain complexed

with dsRNA. EMBO J. 17, 7505–7513.

Sansam, C.L., Wells, K.S., and Emeson, R.B. (2003). Modulation of RNA edit-

ing by functional nucleolar sequestration of ADAR2. Proc. Natl. Acad. Sci. USA

100, 14018–14023.

Sattler, M., Schleucher, J., and Griesinger, C. (1999). Heteronuclear multidi-

mensional NMR experiments for the structure determination of proteins in

solution employing pulsed field gradients. Prog. Nucl. Magn. Reson. Spec-

trosc. 34, 93–158.

Stefl, R., and Allain, F.H. (2005). A novel RNA pentaloop fold involved in target-

ing ADAR2. RNA 11, 592–597.

Stefl, R., Skrisovska, L., and Allain, F.H. (2005a). RNA sequence- and shape-

dependent recognition by proteins in the ribonucleoprotein particle. EMBO

Rep. 6, 33–38.

Stefl, R., Skrisovska, L., Xu, M., Emeson, R.B., and Allain, F.H. (2005b). Reso-

nance assignments of the double-stranded RNA-binding domains of adeno-

sine deaminase acting on RNA 2 (ADAR2). J. Biomol. NMR 31, 71–72.

Stefl, R., Xu, M., Skrisovska, L., Emeson, R.B., and Allain, F.H. (2006). Struc-

ture and specific RNA binding of ADAR2 double-stranded RNA binding motifs.

Structure 14, 345–355.

Tian, B., Bevilacqua, P.C., Diegelman-Parente, A., and Mathews, M.B. (2004).

The double-stranded-RNA-binding motif: interference and much more. Nat.

Rev. Mol. Cell Biol. 5, 1013–1023.

Tonkin, L.A., and Bass, B.L. (2003). Mutations in RNAi rescue aberrant chemo-

taxis of ADAR mutants. Science 302, 1725.

Trabucchi, M., Briata, P., Garcia-Mayoral, M., Haase, A.D., Filipowicz, W.,

Ramos, A., Gherzi, R., and Rosenfeld, M.G. (2009). The RNA-binding protein

KSRP promotes the biogenesis of a subset of microRNAs. Nature 459,

1010–1014.

Tsui, V., Zhu, L., Huang, T.H., Wright, P.E., and Case, D.A. (2000). Assessment

of zinc finger orientations by residual dipolar coupling constants. J. Biomol.

NMR 16, 9–21.

Valente, L., and Nishikura, K. (2005). ADAR gene family and A-to-I RNA editing:

diverse roles in posttranscriptional gene regulation. Prog. Nucleic Acid Res.

Mol. Biol. 79, 299–338.

Valente, L., and Nishikura, K. (2007). RNA binding-independent dimerization of

adenosine deaminases acting on RNA and dominant negative effects of

nonfunctional subunits on dimer functions. J. Biol. Chem. 282, 16054–16061.


Vasiljeva, L., Kim, M., Mutschler, H., Buratowski, S., and Meinhart, A. (2008).

The Nrd1-Nab3-Sen1 termination complex interacts with the Ser5-phosphory-

latedRNApolymerase IIC-terminal domain.Nat. Struct.Mol. Biol.15, 795–804.

Wu, H., Henras, A., Chanfreau, G., and Feigon, J. (2004). Structural basis for

recognition of the AGNN tetraloop RNA fold by the double-stranded RNA-

binding domain of Rnt1pRNase III. Proc. Natl. Acad. Sci. USA 101, 8307–8312.

Xu, M., Wells, K.S., and Emeson, R.B. (2006). Substrate-dependent contribu-

tion of double-stranded RNA-binding motifs to ADAR2 function. Mol. Biol. Cell

17, 3211–3220.

Yang,W., Chendrimada, T.P., Wang, Q., Higuchi, M., Seeburg, P.H., Shiekhat-

tar, R., and Nishikura, K. (2006). Modulation of microRNA processing and

expression through RNA editing by ADAR deaminases. Nat. Struct. Mol.

Biol. 13, 13–21.

Zwahlen, C., Legault, P., Vincent, S.J.F., Greenblatt, J., Konrat, R., and Kay,

L.E. (1997). Methods for measurement of intermolecular NOEs by multinuclear

NMR spectroscopy: Application to a bacteriophage lambda N-peptide/boxB

RNA complex. J. Am. Chem. Soc. 119, 6711–6721.


Exon Junction Complex Subunits AreRequired to Splice Drosophila MAPKinase, a Large Heterochromatic GeneJean-Yves Roignant1 and Jessica E. Treisman1,*1Kimmel Center for Biology and Medicine of the Skirball Institute, NYU School of Medicine, Department of Cell Biology,

540 First Avenue, New York, NY 10016, USA


DOI 10.1016/j.cell.2010.09.036

SUMMARY

The exon junction complex (EJC) is assembled on

spliced mRNAs upstream of exon-exon junctions

and can regulate their subsequent translation, local-

ization, or degradation. We isolated mutations in

Drosophila mago nashi (mago), which encodes

a core EJC subunit, based on their unexpectedly

specific effects on photoreceptor differentiation.

Loss of Mago prevents epidermal growth factor

receptor signaling, due to a large reduction in

MAPK mRNA levels. MAPK expression also requires

the EJC subunits Y14 and eIF4AIII and EJC-associ-

ated splicing factors. Mago depletion does not affect

the transcription or stability of MAPK mRNA but

alters its splicing pattern. MAPK expression from

an exogenous promoter requires Mago only when

the template includes introns. MAPK is the primary

functional target of mago in eye development; in

cultured cells, Mago knockdown disproportionately

affects other large genes located in heterochromatin.

These data support a nuclear role for EJC compo-

nents in splicing a specific subset of introns.

INTRODUCTION

The exon junction complex (EJC) plays an important role in

coupling nuclear and cytoplasmic events in gene expression;

its recruitment allows nuclear pre-mRNA splicing to influence

the subsequent fate of the spliced mRNAs (Tange et al., 2004).

The EJC is assembled ontomRNAs during splicing, 20–24 bases

upstream of each exon junction (Gehring et al., 2009a).

The DEAD box RNA helicase eIF4AIII is the first subunit to asso-

ciate with pre-mRNA through interactions with the intron-binding

protein IBP160 (Gehring et al., 2009a; Ideue et al., 2007). eIF4AIII

then recruits Magoh (known as Mago in Drosophila) (Boswell

et al., 1991; Kataoka et al., 2001) and Y14 (Hachet and Ephrussi,

2001; Kataoka et al., 2000; Le Hir et al., 2000; Mohr et al., 2001),

which stabilize eIF4AIII binding by inhibiting its ATPase activity

(Andersen et al., 2006; Ballut et al., 2005; Bono et al., 2006).

These three subunits constitute the pre-EJC; the fourth core

subunit, MLN51 (Barentsz [Btz] in Drosophila) (Degot et al.,

2004; van Eeden et al., 2001), is added after export of spliced

mRNA to the cytoplasm (Gehring et al., 2009a; Herold et al.,

2009). Many accessory proteins transiently interact with this

core complex and modulate its function (Tange et al., 2004).

The EJC remains bound to cytoplasmic mRNA until it is

displaced by the ribosome-associated disassembly factor Pym

during the first round of translation (Dostie and Dreyfuss, 2002;

Gehring et al., 2009b).

The EJC has been shown to regulate posttranscriptional

events that include mRNA localization, translation, and degrada-

tion. In vertebrate cells, the presence of the EJC on spliced

mRNAs increases their translational yield (Nott et al., 2004;

Wiegand et al., 2003), in part by recruiting S6 kinase 1

(Ma et al., 2008). The EJC is best known for its role in

nonsense-mediated decay (NMD), a surveillance mechanism

that degrades mRNAs containing premature termination codons

(PTCs) (Chang et al., 2007). In mammals, NMD is greatly

enhanced by the presence of a spliceable intron downstream

of a PTC and is mediated by the EJC and accessory factors

that include three up-frameshift (UPF) proteins (Cheng et al.,

1994; Stalder and Muhlemann, 2008; Thermann et al., 1998).

However, NMD can occur independently of splicing or the EJC

in lower organisms such as Drosophila (Gatfield et al., 2003). In

Drosophila, the EJC has a role in mRNA localization; all four

core EJC components are required to localize oskar mRNA to

the posterior pole of the oocyte (Hachet and Ephrussi, 2001;

Mohr et al., 2001; Newmark and Boswell, 1994; Palacios et al.,

2004; van Eeden et al., 2001).

We isolated mutant alleles of mago based on their specific

defects in epidermal growth factor receptor (EGFR)-dependent

processes in eye development. Phosphorylation of mitogen-

activated protein kinase (MAPK) is a critical step in signal trans-

duction downstream of the EGFR and other receptor tyrosine

kinases (Katz et al., 2007). Loss of mago strongly reduces the

total level of the mRNA encoding Rolled (Rl), the Drosophila

extracellular signal-regulated kinase (ERK)-related MAPK. Y14

and eIF4AIII, the other two subunits of the pre-EJC, also

positively regulate MAPK transcript levels, but Btz does not.

An intronless MAPK cDNA is independent of mago and can

rescue photoreceptor differentiation in mago mutant clones;

inclusion of the introns renders it Mago dependent. Mago

does not affect MAPK transcription or mRNA stability but


alters its splicing pattern.MAPK is a large gene located in hetero-

chromatin; a genome-wide survey of Mago-regulated genes

found that genes that shared these features were overrepre-

sented. Based on these observations, we propose that the

pre-EJC is essential to splice a specific set of transcripts that

includes the critical signal transduction component MAPK.

RESULTS

mago Is Required for EGFR Signaling in Eye and Wing

Development

EGFR signaling plays a critical role in Drosophila eye develop-

ment. Differentiation of regularly spaced clusters, each contain-

ing eight photoreceptor cells, progresses from posterior to ante-

rior across the third instar larval eye imaginal disc, led by an

indentation known as the morphogenetic furrow (MF). R8, the

first photoreceptor to form in each developing cluster, induces

EGFR activation in surrounding cells to promote their differenti-

ation into R1–R7 photoreceptors (Figure 1A) (Roignant and

Treisman, 2009). In a genetic screen for mutations affecting

photoreceptor differentiation (Janody et al., 2004), we isolated

three alleles of mago nashi (mago) (Figure 1B). In large clones

ofmagomutant cells in the eye disc, R8 differentiation, visualized

using the marker Senseless (Sens), initiated correctly immedi-

ately posterior to the MF; however, few other photoreceptors

were recruited (Figures 1E and 1F). This phenotype resembles

those reported for mutations in components of the EGFR

pathway (Halfar et al., 2001; Yang and Baker, 2003).

Loss of EGFR signaling also leads to apoptosis in the eye disc

(Halfar et al., 2001; Yang and Baker, 2003).magomutant clones

strongly accumulated activated caspases, indicative of

apoptosis (Figures 1I and 1J). To test whether the lack of photo-

receptor differentiation in mago mutant clones was simply

a consequence of cell death, we blocked cell death in the eye

disc by expressing the anti-apoptotic peptide p35 (Hay et al.,

1994). This rescued the loss of R8 cells but did not restore their

ability to recruit additional photoreceptors (Figures 1G and 1H).

Like known components of the EGFR pathway,mago thus inde-

pendently controls both photoreceptor differentiation and cell

survival. A third function of EGFR signaling in the eye disc is to

arrest differentiating photoreceptors in the G1 phase of the cell

cycle. In the absence of EGFR signaling, re-entry of these cells

into the cell cycle can be visualized by increased expression of

Cyclin B, a marker of S and G2 phases (Yang and Baker,

2003). mago mutant clones accumulated Cyclin B in extra cells

(Figures 1K and 1L), indicating a failure of G1 arrest.

To further confirm a requirement for mago in EGFR signaling,

we examined the expression of EGFR target genes. Expression

of the transcription factor Pointed P1 (PntP1) is induced by EGFR

signaling as photoreceptors initiate their differentiation just

posterior to the MF; in mago mutant clones, PntP1 expression

was lost (Figures 1M and 1N). During wing development, EGFR

signaling activates expression of the target gene argos in the

wing vein primordia (Figures 1O and 1P) (Golembo et al.,

1996). argos expression was strongly reduced in mago mutant

cells in the wing disc (Figures 1Q and 1R). The requirement for

mago for EGFR signaling in both eye and wing development

suggests that it has a general function in this pathway. Its effect

on the EGFRpathway appears quite specific because the normal

pattern of R8 differentiation would be incompatible with a role for

mago in signaling by Hedgehog, Notch, or Wingless in the devel-

oping eye (Roignant and Treisman, 2009).

Mago Acts Downstream of Raf and Upstream of MAPK

Activation

To determine the point at whichmago acts in the EGFR signaling

pathway, we performed epistasis experiments in the eye disc.

Spitz (Spi) is the primary ligand that induces EGFR activation in

R1–R7; activated EGFR feeds into the Ras/MAPK pathway

common to other receptor tyrosine kinases (Figure 2M). The

GTP-bound form of Ras activates the protein kinase Raf,

initiating a kinase cascade in which Raf phosphorylates.

Downstream of Raf1 (Dsor1 or MEK), which in turn phosphory-

lates MAPK. Phosphorylated MAPK enters the nucleus and

phosphorylates specific transcription factors to regulate target

gene expression. We expressed constitutively active forms of

these components of the pathway specifically within mago

mutant clones. Constitutively secreted Spi (Schweitzer et al.,

1995), activated EGFR (Queenan et al., 1997), activated Ras

(Karim and Rubin, 1998), and activated Raf (Martın-Blanco

et al., 1999) all failed to induce photoreceptor differentiation in

mago mutant clones (Figures 2A–2F, 2I, and 2J), although each

induced ectopic photoreceptors when expressed in wild-type

cells (Figures 2G and 2H) (Miura et al., 2006; Roignant et al.,

2006). However, an activated form of MAPK, RolledSEM (Ciap-

poni et al., 2001), fully rescued the lack of photoreceptors in

mago mutant cells (Figures 2K and 2L). Similar epistasis experi-

ments in the wing disc, using argos-lacZ to monitor pathway

activation, likewise showed that only activated MAPK could

induce argos expression in mago mutant cells (Figure S1 avail-

able online). The activity of mago is thus required downstream

of Raf activation but upstream of MAPK activation.

Mago Is Required to Maintain MAPK Levels Sufficient

for Signaling

Becausemago encodes a subunit of the EJC, we reasoned that it

might control the expression of a component of the EGFR

pathway. Indeed, we found that the levels of MAPK protein

were strongly reduced in mago mutant clones in both the eye

and wing discs (Figures 3A–3D). To determine the extent of the

reduction, we compared protein extracts from wild-type eye

discs expressing GFP in all cells and from eye discs containing

largemagomutant clonesmarked by the absence of GFP. Quan-

tification of MAPK on western blots relative to GFP and Tubulin,

to correct for the proportion of wild-type cells, showed thatmago

mutant cells expressed MAPK at only 16% of the wild-type level

(Figure 3E).

We next examined whether these results could be general-

ized to cultured Drosophila S2 and S2R+ cells, in which mago

is expressed and can be knocked down by RNA interference

(RNAi) (Figures 3G and 3H). Mago depletion in these cells

resulted in a 75% reduction in MAPK protein levels in compar-

ison to Tubulin, visible both on western blots and in immunohis-

tochemical stainings (Figures 3F–3H). This effect was specific

because MEK levels were not significantly reduced

(Figure 3G). As expected, the loss of MAPK protein strongly


reduced the potency of EGFR signaling; MAPK phosphorylation

induced by treatment of an S2 cell line that stably expresses

the EGFR (D2F) (Schweitzer et al., 1995) with media condi-

tioned by cells expressing Spi (Miura et al., 2006) was strongly

reduced in cells treated with mago RNAi (Figure 3G). S2R+ cells

treated with insulin to activate the endogenous insulin receptor

also showed reduced MAPK phosphorylation and MAPK

protein levels when Mago was knocked down (Figure 3H), sup-

porting a general role for Mago in receptor tyrosine kinase

signaling.

Figure 1. mago Is Required for EGFR-Dependent Processes in the Eye and Wing Discs

(A) Photoreceptor differentiation proceeds from posterior (P) to anterior (A) across the eye disc. R8 cells differentiate first, immediately posterior to the MF, and

produce the ligand Spi, which activates the EGFR in surrounding cells and induces the sequential formation of R1–R7 cells.

(B) Sequence comparison of Drosophila Mago and human Magoh. Identical amino acids are in red. The amino acid changes in our three mago alleles are

indicated.

(C–N) Third instar eye imaginal discs with anterior to the left. Photoreceptors are stained with anti-Elav (C, E, and G; blue in D, F, H, J, and L). R8 is stained with

anti-Senseless (red in D, F, and H).

(C and D) Wild-type.

(E and F) Largemago93Dmutant clones generated in aMinute background are marked by the absence of GFP (green in F). R8 cells start to differentiate normally,

but differentiation of other photoreceptors is impaired.

(GandH) Largemago93Dmutant clonesaremarkedby theabsenceofGFP (green inH) indiscs expressingp35 inall cells posterior to theMF.Rescueof apoptosis in

magomutant clonesdoes not restoreR1–R7differentiation. Insets showenlargements of the boxed regions; note the loss of photoreceptors other thanR8 in (E–H).

(I–N) mago93D clones are marked by the absence of GFP (green in J, L, and N) and stained with anti-Caspase 3 (I; red in J), anti-Cyclin B (K; red in L), or anti-

Pointed P1 (M; magenta in N). The mutant clones show increased apoptosis and increased Cyclin B expression (insets show enlargements of boxed regions),

indicating a failure to arrest in G1, and do not express PntP1.

(O–R) Wing discs expressing GFP (green in P) or containingmago93D clones marked by the absence of GFP (green in R). argos-lacZ expression revealed by anti-

b-galactosidase staining (O and Q; magenta in P, R) is absent in mago mutant clones.


The Other Subunits of the Pre-EJC Also Control MAPK

Levels

Mago is a subunit of the core EJC, which also includes three

other proteins, Y14, Btz, and eIF4AIII. To determine whether

the effect of Mago on MAPK was due to its function within the

EJC, we tested whether these other EJC subunits were also

required for EGFR signaling and MAPK expression. Existing

alleles of Y14/tsunagi (tsu), which have P element sequences in-

serted in the 50UTR (Mohr et al., 2001) (Figure 4A), did not affect

photoreceptor differentiation (Figures S2A and S2D). However,

these alleles did not abolish Y14 protein expression (Figures

S2B, S2C, S2E, and S2F). We therefore generated a Y14 null

allele by imprecise excision of the P element tsu1. The tsuD18

deletion removed the entire Y14 coding sequence, resulting in

the complete absence of detectable Y14 protein (Figure 4A

and Figures S2H and S2I). Clones homozygous for tsuD18

strongly resembled mago mutant clones, showing both failure

to differentiate R1–R7 photoreceptors and extensive apoptosis

(Figures 4B–4D and Figure S2G). We prevented cell death in

tsuD18 mutant clones using a mutation in dark, a gene necessary

for apoptosis (Srivastava et al., 2007); as for mago, this

allowed R8 survival but did not rescue recruitment of other

photoreceptors (Figures 4E–4G). Depletion of Y14 by RNAi in

S2R+ cells also reduced MAPK levels (Figure 4K). Mago and

Y14 thus appear to act together to maintain MAPK protein levels

sufficient for EGFR signaling.

We examined btz function using the likely null allele btz2,

a deletion of the N-terminal half of the protein (van Eeden

Figure 2. mago Acts Downstream of Raf but Upstream of Phospho-MAPK in the Eye Disc

(A–L) Eye discs in which photoreceptors are stained with anti-Elav (A, C, E, G, I, and K; magenta in B, D, F, H, J, and L). Clones mutant for mago93D and/or ex-

pressing UAS transgenes are labeled byGFP expression (green in B, D, F, H, J, and L) and are indicated by green arrows in A, C, E, G, I, and K. Expression of UAS-

sSpi (A and B), UAS-EGFRltop (C and D), UAS-RasV12 (E and F), or UAS-Raf179 (I and J) does not rescue photoreceptor differentiation inmagomutant clones. (G

and H) Expression of UAS-Raf179 in wild-type clones leads to excessive photoreceptor differentiation. (K and L) Expression of UAS-RolledSEM induces excessive

photoreceptor differentiation in mago93D clones.

(M) A simplified diagram of the EGFR signaling pathway.

See also Figure S1.


et al., 2001). Surprisingly, we found that photoreceptors

differentiated normally in btz mutant clones (Figures 4H–4J).

Consistently, RNAi directed against btz in S2R+ cells had no

effect on MAPK expression (Figure 4K), although it was effective

at lowering btz levels (Figure S2M). Because eIF4AIII is the

RNA-binding component of the EJC, its absence is predicted

to abolish all EJC functions. No null allele of eIF4AIII is available,

and clones homozygous for the missense allele eIF4AIII19

(Palacios et al., 2004) only weakly affected photoreceptor

differentiation (Figures S2J–S2L). However, we found that

depletion of eIF4AIII by RNAi in S2R+ cells strongly reduced

MAPK levels relative to Tubulin (Figure 4K). The requirement of

Mago, Y14, and eIF4AIII, but not Btz, for MAPK regulation

suggests that this function is performed by the pre-EJC prior

to Btz addition.

MAPK Is the Primary Target of the Pre-EJC

in Photoreceptor Differentiation

If the pre-EJC bound upstream of MAPK exon junctions acts

directly onMAPKmRNA, then expression of aMAPK cDNA lack-

ing introns should be independent of EJC components. We

Figure 3. mago Is Required to Maintain

Normal MAPK Protein Levels

(A–D) Anti-MAPK staining (A and C; magenta in B

and D) of eye discs (A and B) or wing discs

(C and D) containing mago93D clones marked by

the absence of GFP (green in B and D). MAPK

protein levels are strongly reduced in mago

mutant clones.

(E) Western blot using protein extracts derived

either from eye discs expressing GFP in all cells

(WT) or from eye discs containing large mago93D

clones lacking GFP. The ratio of GFP to Tubulin

was used to quantify the amount of remaining

wild-type tissue in mago93D mutant eye discs.

MAPK levels are decreased by 84%when normal-

ized to GFP.

(F) MAPK and Tubulin staining of S2R+ cells

treated with lacZ ormago dsRNA. MAPK is specif-

ically reduced.

(G) D2F cells treated with lacZ or mago dsRNA

were incubated with sSpi conditioned media for

0 or 30 min. Protein lysates were blotted with anti-

bodies to Tubulin, MAPK, diphospho-MAPK,

MEK, and Mago.

(H) S2R+ cells treated with lacZ or mago dsRNA

were incubated with 25 mg/ml insulin for 0 or 10

min. Lysates were blotted with antibodies to

Tubulin, MAPK, and diphospho-MAPK. mago

dsRNA reduced MAPK phosphorylation after

sSpi or insulin treatment due to a decrease in total

MAPK protein.

found that HA-tagged MAPK expressed

from a cDNA template in S2R+ cells

was unaffected by treatment with mago

dsRNA, although endogenous MAPK

levels were strongly reduced (Figures

5A and 5B). We next asked whether this

construct could mediate EGFR signaling

in cells lacking EJC subunits in vivo. Expression of UAS-

MAPK-HA in mago or Y14 mutant clones in the eye disc largely

restored the differentiation of R1–R7 photoreceptors and

enabled thesemutant cells to respond to activated Ras by differ-

entiating extra photoreceptors (Figures 5C–5H and Figure S3).

Reduction of MAPK levels is thus the primary reason for the

photoreceptor defects in mago or Y14 mutant clones.

The Pre-EJC Regulates MAPK Posttranscriptionally

To discriminate whether the pre-EJC acts by affecting the

synthesis, stability, or translation ofMAPKmRNA, we first exam-

ined MAPK mRNA levels by northern blotting in S2R+ cells in

which Mago was knocked down by RNAi. We observed

a dramatic reduction in MAPK transcript levels (Figure 6A), indi-

cating that Mago is required for the production or stabilization of

mature MAPK mRNA. Using quantitative RT-PCR (qRT-PCR)

with primers in the first two exons to measure the levels of

MAPK transcripts, we found that Mago depletion caused

a 70% reduction inMAPKmRNA, whereas other mRNAs exam-

ined were expressed at normal levels (Figures 6B and 6C). This

decrease is similar to the 75% reduction in MAPK protein,


suggesting that Mago does not significantly affect MAPK trans-

lation. Consistent with this interpretation, RNAi directed against

the EJC disassembly factor Pym, which enhances translation of

EJC-bound transcripts in vertebrates (Diem et al., 2007), did not

affect MAPK expression (Figure 7B). In C. elegans and human

stem cells, translation ofMAPK transcripts is inhibited by Pumilio

family (PUF) proteins bound to their 30UTRs (Lee et al., 2007).

However, knocking down Pumilio (Pum), the only Drosophila

PUF protein, did not increase MAPK levels in the presence or

absence of Mago (Figures S4A and S4B), indicating that the

pre-EJC does not act by counteracting Pum-mediated transla-

tional repression.

The primary step regulated by the pre-EJC is thus likely to be

the transcription, processing, or degradation of MAPK mRNA.

To assess MAPK mRNA stability, we treated cells with actino-

mycin D to block transcription and measured the levels of

MAPK transcripts over time. The stability of MAPK mRNA

relative to Ribosomal protein L15 (RpL15) mRNA was only

slightly decreased by Mago RNAi treatment (Figure S4C),

making it unlikely to account for the large reduction in total

MAPK mRNA levels. A reduction in transcriptional initiation

would be expected to reduce all regions of the MAPK mRNA

to the same extent; however, both qRT-PCR and deep

sequencing of mRNA showed that transcripts of 50 exons were

decreased more than 30 exons (Figure 6C and Figure S4D).

To further evaluate MAPK transcription, we examined the

expression level of the pre-mRNA by qRT-PCR, using primers

to amplify intron-exon junctions. We found that, whereas some

regions of theMAPK pre-mRNAwere reduced inMago-depleted

cells, others were increased (Figure 6D), arguing against an

effect of Mago on MAPK transcription. The variability in both

mRNA and pre-mRNA levels over the length of the gene is

suggestive of defects in splicing. Consistent with this model,

we detected an abnormal MAPK splice product in Mago-

depleted, but not control, cells. Sequencing of this product

revealed that it results from splicing of a cryptic 50 splice site

7 bases into exon 4 directly to exon 7, skipping exons 5 and 6

(Figure 6E). Although splicing defects could result in frameshifts

that would lead to NMD, we did not observe further stabilization

ofMAPK pre-mRNAwhen we knocked down both Mago and the

NMD factor Upf1 (Figure S5A); abnormal splice products may

thus be degraded within the nucleus.

MAPK Expression Requires Splicing-Related EJC

Accessory Factors

The core EJC has been reported to associate with accessory

factors involved in splicing, NMD, translation, and mRNA export

(Figure 7A) (Diem et al., 2007; Le Hir et al., 2000, 2001; Li et al.,

2003; Ma et al., 2008). To investigate which of these processes

is involved in regulation of MAPK by the pre-EJC, we knocked

down a representative set of factors in S2R+ cells. Depletion

of the NMD factors Upf1 and Upf2 (Chang et al., 2007) had no

Figure 4. mago, Y14, and eIF4AIII, but Not btz, Are Required for

MAPK Expression and Function

(A) Genomic structure of Y14, showing the coding regions (black), UTRs

(white), and the position of the tsu1 P element and tsu5 allele. The tsuD18 dele-

tion removes the whole Y14 open reading frame without disrupting the adja-

cent gene Mys-45A.

(B–J) Third instar eye discs containing large clones marked by the absence of

GFP (green in D, G, and J) homozygous for the Y14 null allele tsuD18 (B–D); for

tsuD18, darkN28 (E–G); and for btz2 (H–J). Photoreceptors are stained with anti-

Elav (B, E, and H; blue in D, G, and J), and R8 is stained with anti-Sens (C, F,

and I; red in D, G, and J). Arrows in (D) and (G) point to clusters containing only

R8. Likemago, Y14 is required independently for both photoreceptor differen-

tiation and cell survival, but btz is not necessary for either.

(K) Protein lysates from S2R+ cells treated with the indicated dsRNAs were

blotted with antibodies to Tubulin and MAPK. Mago, Y14, and eIF4AIII are

required to maintain MAPK levels, but Btz is not.

See also Figure S2.


effect on MAPK (Figure 7B). However, depletion of the EJC-

associated splicing factors SRm160 and RnpS1 (Reichert

et al., 2002; Trembley et al., 2005) reduced MAPK levels relative

to Tubulin, especially when both were knocked down simulta-

neously (Figure 7B), supporting a role for the EJC in splicing

the MAPK transcript. Our data probably underestimate the

effect of these splicing factors on MAPK because SRm160

was only partially depleted (Figure S5B). Depletion of RnpS1

by transgenic RNAi in vivo in the wing imaginal disc also reduced

MAPK levels (Figures S5C and S5D).

If the pre-EJC is required for MAPK splicing, it should affect

the expression of a MAPK construct that contains the endoge-

nous introns but is expressed from an exogenous promoter.

We used recombineering to place the MAPK genomic region,

including all its introns, downstream of a UAS promoter and

HA tag. Expression of HA-MAPK from this construct driven by

actin-GAL4 in S2R+ cells was strongly reduced by mago RNAi

(Figure 7C and Figure S5E). In contrast, an actin promoter-driven

HA-MAPK cDNA construct containing only the smallest intron,

I5, was not affected by Mago depletion (Figure 7C and

Figure S5E), arguing that one or more of the largerMAPK introns

confers the requirement for the pre-EJC.

Mago Promotes the Expression of Heterochromatic

Genes with Large Introns

The MAPK gene has two unusual features: it contains introns

of up to 25 kb, much longer than the average size of 1.4 kb

for Drosophila (Yu et al., 2002) (Figure 6C), and it is expressed

despite its location within a region of constitutive heterochro-

matin (Eberl et al., 1993). We investigated whether other genes

that shared these features showed a similar requirement for the

pre-EJC. We found that transcript levels of several other genes

with large introns located in heterochromatin were strongly

reduced in cells depleted for Mago, whereas small genes in

heterochromatin or euchromatin were unaffected (Figure 7D).

To extend these results, we carried out a genome-wide survey

of genes affected by Mago depletion by deep sequencing of

mRNA isolated from control ormago dsRNA-treated S2R+ cells.

We found that genes located in heterochromatin were overrepre-

sented among those showing reduced expression after Mago

knockdown. Expression of 18.5% of the heterochromatic genes

detected in these cells (43/232), but only 6.6%of the euchromatic

genes (505/7638),was reducedby at least 1.5-fold in comparison

to cells treated with lacZ RNAi (Figure 7E and Table S1). Among

heterochromatic genes, those with introns larger than 15 kb

were twice as likely to be affected by mago RNAi as genes with

smaller or no introns, but regulation showed no correlation with

intron size for euchromatic genes (Figure 7E and Table S1).

Some large introns are spliced by a recursive mechanism

that relies on cryptic splice sites located within the intron (Bur-

nette et al., 2005); however, no consensus recursive splice site

is present in the MAPK gene (A.-J. Lopez, personal communi-

cation). Of the 84 genes with predicted recursive splice sites

(Burnette et al., 2005) that are expressed in S2R+ cells, only

four (5%) were downregulated at least 1.5-fold by Mago deple-

tion. Because 7% of all expressed genes were downregulated

to this extent, recursively spliced genes are underrepresented

among Mago targets. It is possible that the pre-EJC specifically

facilitates the splicing of large introns that cannot be subdi-

vided by recursive splice sites, which may be more common

in heterochromatic genes (Smith et al., 2007). Alternatively,

other features of the affected introns, such as the strength of

their splice sites, the presence of repetitive sequences (Dimitri

et al., 2003), or the chromatin structure of the DNA template,

may contribute to determining the requirement for the pre-EJC.

Figure 5. A MAPK cDNA Rescues Photoreceptor Differentiation

in mago Mutant Cells

(A) Diagram of the MAPK-HA cDNA construct.

(B) Western blot of protein extracts from S2R+ cells transfected with UAS-

MAPK cDNA, UAS-GFP, and actin-GAL4 and treated with lacZ or mago

dsRNA. Expression of endogenousMAPK (lower band), but not the HA-tagged

MAPK cDNA (upper band), is reduced in the absence of Mago.

(C–H) Eye discs containingmago93D clones alone (C and D) ormago93D clones

expressingMAPK cDNA (E–H), positively marked byGFP expression (H; green

in D and F) and by anti-HA staining (G; blue in F). Photoreceptors are stained

with anti-Elav (C and E; red in D and F). MAPK cDNA restores almost normal

photoreceptor differentiation to mago mutant cells.

See also Figure S3.


DISCUSSION

The EJC is thought to bind to all spliced mRNAs independently

of their sequence (Le Hir et al., 2000), allowing them to be

distinguished from unspliced transcripts in the cytoplasm.

Despite these very general binding properties, we find that loss

of core EJC subunits causes surprisingly specific defects. Our

investigation of the basis for the effect of EJC subunits on one

target gene, MAPK, has revealed a function of the pre-EJC

during the splicing process.

Specificity of EJC Function

Our genome-wide expression analysis found that loss of Mago

reduces the transcript levels of only 7% of the genes

expressed in S2R+ cells by 1.5-fold or more. The number of

genes directly regulated by the pre-EJC is likely to be much

Figure 6. MagoAffectsMAPKmRNALevels

Posttranscriptionally

(A) MAPK and Ribosomal protein L32 (RpL32)

mRNA detected by northern blotting in S2R+ cells

treated with lacZ or mago dsRNA.

(B) Transcript levels in S2R+ cells treated with the

control dsRNA aveugle (ave), an upstream

component of the EGFR pathway, or with mago

dsRNA were measured by qRT-PCR. mago RNAi

reduced MAPK mRNA levels by 70% but did not

affect mRNAs encoding other EGFR pathway

components.

(C) Diagram showing the seven exons and six

introns of the 50 kbMAPK gene. Primers spanning

each exon-exon junction were used to detect

mRNA levels by qRT-PCR in lacZ or mago

dsRNA-treated S2R+ cells.

(D) MAPK pre-mRNA levels in S2R+ cells treated

with lacZ or mago dsRNA were assessed by

qRT-PCR using primers spanning the exon-intron

junctions. Pre-mRNA levels are reduced in some

regions of theMAPK gene and increased in others

in Mago-depleted cells, suggesting that Mago

does not affect MAPK transcription. For (C) and

(D), the mean of five experiments is shown.

b-tubulin (tub), RpL15, and Histone H3 (His3)

were used as controls. Signals detected in the

absence of reverse transcriptase are also plotted

(lacZ-no RT, mago-no RT) but are negligible on

the scale of these graphs.

(E) RT-PCR using primers in exons 3 and 7 ampli-

fied a smaller product in cells treated with mago,

but not lacZ, dsRNA (arrow). The structure of this

product is diagrammed below.

Error bars in (B)–(D) represent standard devia-

tions. See also Figure S4 and Table S2.

smaller because transcript levels were

measured after an extensive period of

RNAi treatment that was necessary to

eliminate the Mago protein. The ability

of MAPK to rescue photoreceptor differ-

entiation in mago mutant clones also

suggests that many genes are downre-

gulated as an indirect consequence of loss of MAPK. Similarly,

many of the defects of mouse neuroepithelial stem cells hetero-

zygous for Magoh are rescued by restoring the expression of

a single gene, Lis1 (Silver et al., 2010). Cytoplasmic functions

of the EJC also show specificity; for instance, the EJC is required

to localize oskarmRNA to the posterior of the oocyte but has no

effect on the subcellular localization of other spliced mRNAs

such as bicoid or gurken (Hachet and Ephrussi, 2001; Newmark

and Boswell, 1994; Newmark et al., 1997). This functional

specificity might indicate that EJC components are, in fact,

assembled on only a subset of spliced transcripts. Indeed, only

the first intron in the oskar transcript contributes to its localization

by the EJC (Hachet and Ephrussi, 2004). However, experiments

in vertebrate andDrosophila cells have found no specific require-

ment for EJC assembly other than an upstream exon at least 20

bases long (Herold et al., 2009; Ideue et al., 2007; Le Hir et al.,


Figure 7. The Pre-EJC May Facilitate Splicing of Large Introns in Heterochromatic Genes

(A) Proteins associated with the core EJC and their known functions.

(B) MAPK and Tubulin levels detected by western blotting of lysates from S2R+ cells treated with the indicated dsRNAs. Quantification of MAPK levels relative to

the lacZ control is shown below the blot. Knocking down the splicing factors SRm160 and RnpS1, especially in combination, reduces MAPK levels.

(C) qRT-PCR was used to measure mRNA transcribed from the UAS-HA-MAPK genomic construct shown below (HA-MAPKg, primers in the HA sequence and

exon 2) and from actin-MAPKi5-HA (primers in exon 6 and the HA sequence) in S2R+ cells treated with lacZ ormago dsRNA. Controls were tub, RpL15, Dbp80,

and endogenous MAPK (primers in exons 2 and 3), and HA-MAPKg levels were normalized to transcripts from cotransfected UAS-GFP, also driven by

actin-GAL4. Mago depletion reduces mRNA expressed from the genomic construct.

(D) Northern blots of RNA from S2R+ cells treated with lacZ ormago dsRNA. Expression of the large heterochromatic (Het) genes light, CG40263, and Dbp80 is

reduced by mago RNAi, but expression of the small genes tub, 14-3-33, and RpL15 is unaffected.

(E) Genes downregulated R 1.5 fold in mago dsRNA-treated S2R+ cells broken down by location in euchromatin or heterochromatin and by the size of their

largest introns are shown as a percentage of the total number of genes in each category that are expressed in control lacZ dsRNA-treated cells. Heterochromatic

genes with introns larger than 15 kb are the most likely to be dependent on Mago.

Error bars in (C) represent standard deviations. See also Figure S5 and Table S1.


2001; Tange et al., 2004). Localization of EJC components to

particular cytoplasmic regions in Drosophila oocytes (Hachet

and Ephrussi, 2001; Mohr et al., 2001) and mammalian neurons

(Giorgi et al., 2007) may simply represent their selective retention

on mRNAs that are translationally repressed (Dostie and Drey-

fuss, 2002; Gehring et al., 2009b).

The importance ofMAPK for receptor tyrosine kinase signaling

has led to the evolution of multiple mechanisms to regulate its

expression as well as its phosphorylation (Lee et al., 2007;

Nykamp et al., 2008). Other vital targets for the pre-EJC may

be found in the ovary. mago and Y14, but not btz, are required

early in oogenesis for germline stem cell differentiation and

oocyte specification (Parma et al., 2007; van Eeden et al.,

2001). Because germline inactivation of the Ras pathway has

no effect on oogenesis (Hou et al., 1995), these functions of

Mago and Y14 may reflect a requirement for the pre-EJC to

splice transcripts other than MAPK.

A Role for the Pre-EJC in Splicing

The EJC has been shown to act on previously spliced mRNAs in

the cytoplasm to increase their translation, direct their subcel-

lular localization, or target them for degradation if they contain

premature stop codons (Tange et al., 2004). However, none of

these mechanisms could explain the strong reduction of MAPK

mRNA levels in the absence of pre-EJC subunits. We have

provided several lines of evidence suggesting that the pre-EJC

facilitates splicing of a specific subset of introns, including at

least one present in the MAPK pre-mRNA. First, MAPK is not

an indirect transcriptional target of the pre-EJC because

MAPK pre-mRNA is not uniformly reduced in the absence of

mago, and Mago is required for the expression of a MAPK

genomic construct driven by a heterologous promoter. Second,

the EJC-associated splicing factors RnpS1 and SRm160

(Reichert et al., 2002) contribute to maintaining normal MAPK

levels, whereas Btz, the only core EJC subunit absent from the

spliceosomal complex (Gehring et al., 2009a), is dispensable

for MAPK expression. Third, an abnormally spliced MAPK

product is detected in Mago-depleted cells. Finally, heterochro-

matic genes with large introns show an increased propensity for

regulation by Mago. Previous experiments did not detect any

positive function for the EJC in splicing; however, they were per-

formed in vitro using short introns with strong splice sites (Zhang

and Krainer, 2007) and would therefore have missed a function

specific to one class of introns.

It will be interesting to determine what features of intronsmake

their splicing dependent on the pre-EJC. Our genome-wide anal-

ysis points to heterochromatic location and intron size as two

characteristics that are likely to be important. Unlike mammalian

genomes, the Drosophila genome contains primarily short

introns (Yu et al., 2002). Large introns are most common in

heterochromatic genes such as MAPK, where they are rich in

repetitive DNA composed of transposons, retrotransposons,

and satellite sequences (Dimitri et al., 2003). Production of

endo-siRNAs from such repetitive elements (Ghildiyal et al.,

2008) or the presence of splice sites within these elements

(Ponicsan et al., 2010) could interfere with the splicing of the

introns they occupy. Chromatin structure might also directly

influence splicing, as suggested by recent studies showing

differences in nucleosome occupancy and histonemodifications

between exons and introns and recruitment of splicing regula-

tors by chromatin-binding proteins (Schwartz and Ast, 2010;

Luco et al., 2010).

Recognition of splice sites over long distances poses a chal-

lenge to the splicing machinery. Splice sites for large introns

are initially identified by an exon definition mechanism

(Fox-Walsh et al., 2005). The pre-EJC, which is assembled

upstream of the 50 splice site during splicing (Gehring et al.,

2009a), might interact with other factors across the exon to facil-

itate recognition of the upstream 30 splice site. Perhaps pre-EJC

complexes deposited upstream of introns that can be easily

detected due to their small size, strong splice sites, or other

features contribute to the subsequent recognition of neighboring

introns. Alternatively, because the pre-EJC is assembled prior to

exon ligation (Gehring et al., 2009a), it might act during its own

recruitment into the spliceosome to promote the second step

of splicing. We cannot distinguish these alternatives at present

because our measurements of 50 and 30 splice junctions in the

MAPK pre-mRNA were made at steady state and thus reflect

the balance between transcription, splicing, and degradation.

The presence of recursive splice sites that allow large introns

to be spliced in multiple steps (Burnette et al., 2005) makes

genes less likely to require the EJC. Of interest, recursive splice

sites are much less common in vertebrate introns than in

Drosophila (Shepard et al., 2009), suggesting that the EJC-

dependent mechanism might be more widely used in higher

organisms. Our data challenge the view that the EJC acts only

as a marker that affects postsplicing events and suggest that

this complex also functions within the nucleus to process

a specific set of transcripts.


Drosophila Genetics

Three lethal alleles of mago were isolated in a mosaic screen for genes

required for photoreceptor differentiation (Janody et al., 2004). The mutations

were mapped to the mago genomic region by meiotic recombination with

P(w+) elements (Zhai et al., 2003). The coding region of mago was amplified

by PCR from homozygous mutant larvae and sequenced. Two alleles altered

conserved amino acids (mago93D, S39F;mago115F, E21K), and one introduced

a premature stop codon (mago69L, R121@). These alleles failed to complement

the previously described strong allele mago3 (Boswell et al., 1991), which also

showed the same eye disc phenotype. To generate a Y14 null allele, we mobi-

lized the P(w+) element P{EP}tsuEP567 (tsu1) (Flybase) by crossing it to the

transposase stock D2–3, CyO; TM3/T(2;3)apterousxa. We generated 89 inde-

pendent excision lines, identified deletions by PCR using primers flanking

the P element, and determined the breakpoints by sequencing the PCR prod-

ucts. tsuD18 is a 711 bp deficiency that removes the entire Y14 coding

sequence.

Immunohistochemistry and Western Blot Analysis

Antibody staining of eye and wing imaginal discs was performed as described

(Roignant et al., 2006). Secondary antibodies were from Jackson Immunore-

search; FITC, TRITC, or Cy5 conjugates were used at 1:200 and Alexa

488 conjugates at 1:1000. Images were captured on a Leica TCS NT or a

Zeiss LSM 510 confocal microscope. Western blots were performed as

described (Miura et al., 2006). To make protein extract from eye-antennal

discs, 100 discs of each genotype were lysed in ice-cold lysis buffer (50 mM

Tris-HCl [pH 8.0], 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail

[Roche], 5 mM EDTA, 5 mMNaF, 1 mMNa3VO4, 0.1% SDS, and 0.5% sodium

deoxycholate).


Cell Culture and RNAi

S2R+ cells were maintained in Schneider’s medium supplemented with 10%

fetal calf serum; D2F cells (Schweitzer et al., 1995) were additionally supple-

mented with 150 mg/ml G418. Double-stranded RNAs (dsRNAs) were gener-

ated using the MEGAscript T7 and T3 kit (Ambion) as described (Roignant

et al., 2006). S2R+ cells (106 cells/well) were treated with 15 mg dsRNA. Cells

were transfected using Effectene (QIAGEN) 3 days after dsRNA incubation and

returned tomedium containing 15 mg dsRNA. Cells were harvested after 7 days

and lysed in ice-cold lysis buffer. MAPK phosphorylation was induced in S2R+

cells by incubation in PBS containing 25 mg/ml bovine insulin (Sigma) for

10 min. In D2F cells, MAPK phosphorylation was induced by incubation in

sSpiCS-conditioned medium for 30 min (Miura et al., 2006), following EGFR

induction for 3 hr with 200 mM Cu2SO4.

Measurement of RNA Levels

Total RNA was extracted from cells using TRIzol (Invitrogen) and treated with

RQ1 RNase-Free DNase (Promega). Reverse transcription was performed

from 2 mg of total RNA using M-MLV Reverse Transcriptase (Promega). The

exponential phases of PCR reactions were determined on 18–23 cycles to

allow semiquantitative comparisons of cDNAs. For qRT-PCR analysis, cDNA

was amplified using Power SYBR green and a real-time PCR ABI 7900HT

Sequence Detection Systems machine (Applied Biosystems). The relative

abundance of transcripts was calculated as described (Carrera et al., 2008).

All experiments were repeated at least three times, and the data are presented

as the mean ± standard deviation. Primer sequences are given in Table S2. For

northern blots, 10 mg of total RNA were denatured for 30 min at 55 C in glyoxal

loading buffer and separated on a 1% agarose gel. RNA was transferred to

a Hybond-XL membrane (Amersham), UV crosslinked, and probed with PCR

fragments (500 bp–1 kb) radioactively labeled with [32P]-dCTP. Membranes

were exposed to X-ray film for 24–72 hr at !80 C.

Deep Sequencing of mRNA

S2R+ cells were treatedwith 15 mg lacZ ormago dsRNAs for 3 days and placed

in fresh medium with 15 mg dsRNA for another 3 days. Total RNA was

extracted using TRIzol, cleaned using RNeasy Mini Kit (QIAGEN), and treated

with RQ1 RNase-Free DNase (Promega). Isolation of Poly(A)+ mRNA, RT

reactions, and purification of the cDNA templates were performed following

the mRNA-Seq Sample Preparation kit protocol from Illumina. Each cDNA

sample was uploaded onto one lane of a flow cell and sequenced in a 54 nucle-

otide single-end run on an Illumina Genome Analyzer II.

Raw images were analyzed by Illumina RTA version 1.6 using Phix control

lane for estimating base calling parameters. Reads were generated and

aligned to the D. melanogaster genome (dm3) and exon-exon splice junction

database (prepared using a UCSC annotation database downloaded on April

23, 2010) by Illumina CASAVA version 1.6 using default filtering parameters.

The first and last two nucleotides were trimmed out. In total, 22,537,621 and

21,130,945 50 bp reads for LacZ and Mago samples, respectively, were

sequenced. 46%of LacZ and 44% ofMago reads had eligible alignments after

filtering for contaminants, repeats, and reads aligned to multiple genes.

We normalized gene counts using the RPKM method (reads per kilobase of

transcript per million mapped sequence reads) (Mortazavi et al., 2008) and

calculated the fold change of RPKM between the LacZ and Mago samples.

A list of genes in heterochromatin was obtained from Madeline Crosby

(Flybase).

ACCESSION NUMBERS

The RNA sequencing data for lacZ and mago RNAi-treated S2R+ cells have

been deposited in NCBI GEO (accession number GSE23997) and in the Gen-

Bank Short Read Archive (accession number SRA022032.3).


Supplemental Information includes Extended Experimental Procedures, five

figures, and two tables and can be found with this article online at doi:10.

1016/j.cell.2010.09.036.

ACKNOWLEDGMENTS

We thank Robert Boswell, Barry Dickson, Anne Ephrussi, Ruth Lehmann,

James Skeath, Daniel St Johnston, Marc Therrien, the Developmental Studies

Hybridoma Bank, the Drosophila Genomics Resource Center, and the Bloo-

mington Drosophila stock center for reagents. We also thank Antonio-Javier

Lopez for searching for recursive splice sites in MAPK pre-mRNA and Made-

line Crosby for providing a list of heterochromatic genes. We are grateful to the

NYU Cancer Institute Genomics Facility and Laura Hogan for assistance with

qRT-PCR and deep sequencing and to Stuart Brown and Zuojian Tang for bio-

informatics analysis. The manuscript was improved by the critical comments

of Sergio Astigarraga, Ines Carrrera, Kerstin Hofmeyer, Kevin Legent, Sylvie

Ozon Rickman, Hyung Don Ryoo, Josie Steinhauer, Andrea Zamparini, and

Jiri Zavadil. This work was supported by the National Institutes of Health (grant

EY13777 to J.E.T.).

Received: December 29, 2009


Accepted: September 2, 2010


REFERENCES

Andersen, C.B., Ballut, L., Johansen, J.S., Chamieh, H., Nielsen, K.H., Oliveira,

C.L., Pedersen, J.S., Seraphin, B., Le Hir, H., and Andersen, G.R. (2006).

Structure of the exon junction core complex with a trapped DEAD-box

ATPase bound to RNA. Science 313, 1968–1972.

Ballut, L., Marchadier, B., Baguet, A., Tomasetto, C., Seraphin, B., and Le Hir,

H. (2005). The exon junction core complex is locked onto RNA by inhibition of

eIF4AIII ATPase activity. Nat. Struct. Mol. Biol. 12, 861–869.

Bono, F., Ebert, J., Lorentzen, E., and Conti, E. (2006). The crystal structure of

the exon junction complex reveals how it maintains a stable grip onmRNA. Cell

126, 713–725.

Boswell, R.E., Prout, M.E., and Steichen, J.C. (1991). Mutations in a newly

identified Drosophila melanogaster gene,mago nashi, disrupt germ cell forma-

tion and result in the formation of mirror-image symmetrical double abdomen

embryos. Development 113, 373–384.

Burnette, J.M., Miyamoto-Sato, E., Schaub, M.A., Conklin, J., and Lopez, A.J.

(2005). Subdivision of large introns in Drosophila by recursive splicing at non-

exonic elements. Genetics 170, 661–674.

Carrera, I., Zavadil, J., and Treisman, J.E. (2008). Two subunits specific to the

PBAP chromatin remodeling complex have distinct and redundant functions

during Drosophila development. Mol. Cell. Biol. 28, 5238–5250.

Chang, Y.F., Imam, J.S., and Wilkinson, M.F. (2007). The nonsense-mediated

decay RNA surveillance pathway. Annu. Rev. Biochem. 76, 51–74.

Cheng, J., Belgrader, P., Zhou, X., and Maquat, L.E. (1994). Introns are cis

effectors of the nonsense-codon-mediated reduction in nuclear mRNA abun-

dance. Mol. Cell. Biol. 14, 6317–6325.

Ciapponi, L., Jackson, D.B., Mlodzik, M., and Bohmann, D. (2001). Drosophila

Fos mediates ERK and JNK signals via distinct phosphorylation sites. Genes

Dev. 15, 1540–1553.

Degot, S., Le Hir, H., Alpy, F., Kedinger, V., Stoll, I., Wendling, C., Seraphin, B.,

Rio, M.C., and Tomasetto, C. (2004). Association of the breast cancer protein

MLN51 with the exon junction complex via its speckle localizer and RNA

binding module. J. Biol. Chem. 279, 33702–33715.

Diem, M.D., Chan, C.C., Younis, I., and Dreyfuss, G. (2007). PYM binds the

cytoplasmic exon-junction complex and ribosomes to enhance translation of

spliced mRNAs. Nat. Struct. Mol. Biol. 14, 1173–1179.

Dimitri, P., Junakovic, N., and Arca, B. (2003). Colonization of heterochromatic

genes by transposable elements in Drosophila. Mol. Biol. Evol. 20, 503–512.

Dostie, J., and Dreyfuss, G. (2002). Translation is required to remove Y14 from

mRNAs in the cytoplasm. Curr. Biol. 12, 1060–1067.


Eberl, D.F., Duyf, B.J., and Hilliker, A.J. (1993). The role of heterochromatin in

the expression of a heterochromatic gene, the rolled locus of Drosophila mel-

anogaster. Genetics 134, 277–292.

Fox-Walsh, K.L., Dou, Y., Lam, B.J., Hung, S.P., Baldi, P.F., and Hertel, K.J.

(2005). The architecture of pre-mRNAs affects mechanisms of splice-site pair-

ing. Proc. Natl. Acad. Sci. USA 102, 16176–16181.

Gatfield, D., Unterholzner, L., Ciccarelli, F.D., Bork, P., and Izaurralde, E.

(2003). Nonsense-mediated mRNA decay in Drosophila: at the intersection

of the yeast and mammalian pathways. EMBO J. 22, 3960–3970.

Gehring, N.H., Lamprinaki, S., Hentze, M.W., and Kulozik, A.E. (2009a). The

hierarchy of exon-junction complex assembly by the spliceosome explains

key features of mammalian nonsense-mediated mRNA decay. PLoS Biol. 7,

e1000120.

Gehring, N.H., Lamprinaki, S., Kulozik, A.E., and Hentze, M.W. (2009b). Disas-

sembly of exon junction complexes by PYM. Cell 137, 536–548.

Ghildiyal, M., Seitz, H., Horwich, M.D., Li, C., Du, T., Lee, S., Xu, J., Kittler, E.L.,

Zapp, M.L., Weng, Z., and Zamore, P.D. (2008). Endogenous siRNAs derived

from transposons andmRNAs inDrosophila somatic cells. Science 320, 1077–

1081.

Giorgi, C., Yeo, G.W., Stone, M.E., Katz, D.B., Burge, C., Turrigiano, G., and

Moore, M.J. (2007). The EJC factor eIF4AIII modulates synaptic strength and

neuronal protein expression. Cell 130, 179–191.

Golembo, M., Schweitzer, R., Freeman, M., and Shilo, B.Z. (1996). Argos tran-

scription is induced by the Drosophila EGF receptor pathway to form an inhib-

itory feedback loop. Development 122, 223–230.

Hachet, O., and Ephrussi, A. (2001). Drosophila Y14 shuttles to the posterior of

the oocyte and is required for oskar mRNA transport. Curr. Biol. 11, 1666–

1674.

Hachet, O., and Ephrussi, A. (2004). Splicing of oskar RNA in the nucleus is

coupled to its cytoplasmic localization. Nature 428, 959–963.

Halfar, K., Rommel, C., Stocker, H., and Hafen, E. (2001). Ras controls growth,

survival and differentiation in theDrosophila eye by different thresholds of MAP

kinase activity. Development 128, 1687–1696.

Hay, B.A., Wolff, T., and Rubin, G.M. (1994). Expression of baculovirus P35

prevents cell death in Drosophila. Development 120, 2121–2129.

Herold, N., Will, C.L., Wolf, E., Kastner, B., Urlaub, H., and Luhrmann, R.

(2009). Conservation of the protein composition and electron microscopy

structure of Drosophila melanogaster and human spliceosomal complexes.

Mol. Cell. Biol. 29, 281–301.

Hou, X.S., Chou, T.B., Melnick, M.B., and Perrimon, N. (1995). The torso

receptor tyrosine kinase can activate Raf in a Ras-independent pathway.

Cell 81, 63–71.

Ideue, T., Sasaki, Y.T., Hagiwara, M., and Hirose, T. (2007). Introns play an

essential role in splicing-dependent formation of the exon junction complex.

Genes Dev. 21, 1993–1998.

Janody, F., Lee, J.D., Jahren, N., Hazelett, D.J., Benlali, A., Miura, G.I.,

Draskovic, I., and Treisman, J.E. (2004). A mosaic genetic screen reveals

distinct roles for trithorax and polycomb group genes in Drosophila eye devel-

opment. Genetics 166, 187–200.

Karim, F.D., and Rubin, G.M. (1998). Ectopic expression of activated Ras1

induces hyperplastic growth and increased cell death in Drosophila imaginal

tissues. Development 125, 1–9.

Kataoka, N., Diem, M.D., Kim, V.N., Yong, J., and Dreyfuss, G. (2001). Magoh,

a human homolog of Drosophila mago nashi protein, is a component of the

splicing-dependent exon-exon junction complex. EMBO J. 20, 6424–6433.

Kataoka, N., Yong, J., Kim, V.N., Velazquez, F., Perkinson, R.A., Wang, F., and

Dreyfuss, G. (2000). Pre-mRNA splicing imprints mRNA in the nucleus with

a novel RNA-binding protein that persists in the cytoplasm. Mol. Cell 6,

673–682.

Katz, M., Amit, I., and Yarden, Y. (2007). Regulation of MAPKs by growth

factors and receptor tyrosine kinases. Biochim. Biophys. Acta 1773, 1161–

1176.

Le Hir, H., Izaurralde, E., Maquat, L.E., and Moore, M.J. (2000). The spliceo-

some deposits multiple proteins 20-24 nucleotides upstream of mRNA

exon-exon junctions. EMBO J. 19, 6860–6869.

Le Hir, H., Gatfield, D., Izaurralde, E., and Moore, M.J. (2001). The exon-exon

junction complex provides a binding platform for factors involved in mRNA

export and nonsense-mediated mRNA decay. EMBO J. 20, 4987–4997.

Lee, M.H., Hook, B., Pan, G., Kershner, A.M., Merritt, C., Seydoux, G., Thom-

son, J.A., Wickens, M., and Kimble, J. (2007). Conserved regulation of MAP

kinase expression by PUF RNA-binding proteins. PLoS Genet. 3, e233.

Li, C., Lin, R.I., Lai, M.C., Ouyang, P., and Tarn, W.Y. (2003). Nuclear Pnn/DRS

protein binds to spliced mRNPs and participates in mRNA processing and

export via interaction with RNPS1. Mol. Cell. Biol. 23, 7363–7376.

Luco, R.F., Pan, Q., Tominaga, K., Blencowe, B.J., Pereira-Smith, O.M., and

Misteli, T. (2010). Regulation of alternative splicing by histone modifications.

Science 327, 996–1000.

Ma, X.M., Yoon, S.O., Richardson, C.J., Julich, K., and Blenis, J. (2008). SKAR

links pre-mRNA splicing to mTOR/S6K1-mediated enhanced translation effi-

ciency of spliced mRNAs. Cell 133, 303–313.

Martın-Blanco, E., Roch, F., Noll, E., Baonza, A., Duffy, J.B., and Perrimon, N.

(1999). A temporal switch in DER signaling controls the specification and differ-

entiation of veins and interveins in the Drosophila wing. Development 126,

5739–5747.

Miura, G.I., Buglino, J., Alvarado, D., Lemmon, M.A., Resh, M.D., and Treis-

man, J.E. (2006). Palmitoylation of the EGFR ligand Spitz by Rasp increases

Spitz activity by restricting its diffusion. Dev. Cell 10, 167–176.

Mohr, S.E., Dillon, S.T., and Boswell, R.E. (2001). The RNA-binding protein

Tsunagi interacts with Mago Nashi to establish polarity and localize oskar

mRNA during Drosophila oogenesis. Genes Dev. 15, 2886–2899.

Mortazavi, A., Williams, B.A., McCue, K., Schaeffer, L., and Wold, B. (2008).

Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat.

Methods 5, 621–628.

Newmark, P.A., and Boswell, R.E. (1994). The mago nashi locus encodes an

essential product required for germ plasm assembly in Drosophila. Develop-

ment 120, 1303–1313.

Newmark, P.A., Mohr, S.E., Gong, L., and Boswell, R.E. (1997). mago nashi

mediates the posterior follicle cell-to-oocyte signal to organize axis formation

in Drosophila. Development 124, 3197–3207.

Nott, A., Le Hir, H., and Moore, M.J. (2004). Splicing enhances translation in

mammalian cells: an additional function of the exon junction complex. Genes

Dev. 18, 210–222.

Nykamp, K., Lee, M.H., and Kimble, J. (2008). C. elegans La-related protein,

LARP-1, localizes to germline P bodies and attenuates Ras-MAPK signaling

during oogenesis. RNA 14, 1378–1389.

Palacios, I.M., Gatfield, D., St Johnston, D., and Izaurralde, E. (2004). An

eIF4AIII-containing complex required for mRNA localization and nonsense-

mediated mRNA decay. Nature 427, 753–757.

Parma, D.H., Bennett, P.E., Jr., and Boswell, R.E. (2007). Mago Nashi and Tsu-

nagi/Y14, respectively, regulate Drosophila germline stem cell differentiation

and oocyte specification. Dev. Biol. 308, 507–519.

Ponicsan, S.L., Kugel, J.F., and Goodrich, J.A. (2010). Genomic gems: SINE

RNAs regulate mRNA production. Curr. Opin. Genet. Dev. 20, 149–155.

Queenan, A.M., Ghabrial, A., and Schupbach, T. (1997). Ectopic activation of

torpedo/Egfr, a Drosophila receptor tyrosine kinase, dorsalizes both the

eggshell and the embryo. Development 124, 3871–3880.

Reichert, V.L., Le Hir, H., Jurica, M.S., and Moore, M.J. (2002). 50 exon interac-

tions within the human spliceosome establish a framework for exon junction

complex structure and assembly. Genes Dev. 16, 2778–2791.

Roignant, J.Y., and Treisman, J.E. (2009). Pattern formation in the Drosophila

eye disc. Int. J. Dev. Biol. 53, 795–804.

Roignant, J.Y., Hamel, S., Janody, F., and Treisman, J.E. (2006). The novel

SAM domain protein Aveugle is required for Raf activation in the Drosophila

EGF receptor signaling pathway. Genes Dev. 20, 795–806.


Schwartz, S., and Ast, G. (2010). Chromatin density and splicing destiny: on

the cross-talk between chromatin structure and splicing. EMBO J. 29, 1629–

1636.

Schweitzer, R., Shaharabany, M., Seger, R., and Shilo, B.Z. (1995). Secreted

Spitz triggers the DER signaling pathway and is a limiting component in embry-

onic ventral ectoderm determination. Genes Dev. 9, 1518–1529.

Shepard, S., McCreary, M., and Fedorov, A. (2009). The peculiarities of large

intron splicing in animals. PLoS ONE 4, e7853.

Silver, D.L., Watkins-Chow, D.E., Schreck, K.C., Pierfelice, T.J., Larson, D.M.,

Burnetti, A.J., Liaw, H.J., Myung, K., Walsh, C.A., Gaiano, N., and Pavan, W.J.

(2010). The exon junction complex component Magoh controls brain size by

regulating neural stem cell division. Nat. Neurosci. 13, 551–558.

Smith, C.D., Shu, S., Mungall, C.J., and Karpen, G.H. (2007). The Release 5.1

annotation of Drosophila melanogaster heterochromatin. Science 316, 1586–

1591.

Srivastava, M., Scherr, H., Lackey, M., Xu, D., Chen, Z., Lu, J., and Bergmann,

A. (2007). ARK, the Apaf-1 related killer in Drosophila, requires diverse

domains for its apoptotic activity. Cell Death Differ. 14, 92–102.

Stalder, L., and Muhlemann, O. (2008). The meaning of nonsense. Trends Cell

Biol. 18, 315–321.

Tange, T.O., Nott, A., and Moore, M.J. (2004). The ever-increasing complex-

ities of the exon junction complex. Curr. Opin. Cell Biol. 16, 279–284.

Thermann, R., Neu-Yilik, G., Deters, A., Frede, U., Wehr, K., Hagemeier, C.,

Hentze, M.W., and Kulozik, A.E. (1998). Binary specification of nonsense

codons by splicing and cytoplasmic translation. EMBO J. 17, 3484–3494.

Trembley, J.H., Tatsumi, S., Sakashita, E., Loyer, P., Slaughter, C.A., Suzuki,

H., Endo, H., Kidd, V.J., and Mayeda, A. (2005). Activation of pre-mRNA

splicing by human RNPS1 is regulated by CK2 phosphorylation. Mol. Cell.

Biol. 25, 1446–1457.

van Eeden, F.J., Palacios, I.M., Petronczki, M., Weston, M.J., and St Johnston,

D. (2001). Barentsz is essential for the posterior localization of oskarmRNAand

colocalizes with it to the posterior pole. J. Cell Biol. 154, 511–523.

Wiegand, H.L., Lu, S., and Cullen, B.R. (2003). Exon junction complexes

mediate the enhancing effect of splicing on mRNA expression. Proc. Natl.

Acad. Sci. USA 100, 11327–11332.

Yang, L., and Baker, N.E. (2003). Cell cycle withdrawal, progression, and cell

survival regulation by EGFR and its effectors in the differentiating Drosophila

eye. Dev. Cell 4, 359–369.

Yu, J., Yang, Z., Kibukawa, M., Paddock, M., Passey, D.A., and Wong, G.K.

(2002). Minimal introns are not ‘‘junk’’. Genome Res. 12, 1185–1189.

Zhai, R.G., Hiesinger, P.R., Koh, T.W., Verstreken, P., Schulze, K.L., Cao, Y.,

Jafar-Nejad, H., Norga, K.K., Pan, H., Bayat, V., et al. (2003). Mapping

Drosophila mutations with molecularly defined P element insertions. Proc.

Natl. Acad. Sci. USA 100, 10860–10865.

Zhang, Z., and Krainer, A.R. (2007). Splicing remodels messenger ribonucleo-

protein architecture via eIF4A3-dependent and -independent recruitment of

exon junction complex components. Proc. Natl. Acad. Sci. USA 104, 11574–

11579.


The Exon Junction Complex Controlsthe Splicing ofmapk and Other LongIntron-Containing Transcripts in DrosophilaDariel Ashton-Beaucage,1,4 Christian M. Udell,1,4 Hugo Lavoie,1,4 Caroline Baril,1 Martin Lefrancois,1 Pierre Chagnon,1

Patrick Gendron,1 Olivier Caron-Lizotte,1 Eric Bonneil,1 Pierre Thibault,1,3 and Marc Therrien1,2,*1Institute for Research in Immunology and Cancer, Laboratory of Intracellular Signaling2Departement de Pathologie et de Biologie Cellulaire3Departement de Chimie

Universite de Montreal, C.P. 6128, Succursale Centre-Ville, Montreal, Quebec, H3C 3J7, Canada4These authors contributed equally to this work


DOI 10.1016/j.cell.2010.09.014

SUMMARY

Signaling pathways are controlled by a vast array of

posttranslational mechanisms. By contrast, little is

known regarding the mechanisms that regulate the

expression of their core components. We conducted

an RNAi screen in Drosophila for factors modulating

RAS/MAPK signaling and identified the Exon Junc-

tion Complex (EJC) as a key element of this pathway.

The EJC binds the exon-exon junctions of mRNAs

and thus far, has been linked exclusively to post-

splicing events. Here, we report that the EJC is

required for proper splicing of mapk transcripts by

a mechanism that apparently controls exon defini-

tion. Moreover, whole transcriptome and RT-PCR

analyses of EJC-depleted cells revealed that the

splicing of long intron-containing genes, which

includes mapk, is sensitive to EJC activity. These

results identify a role for the EJC in the splicing of a

subset of transcripts and suggest that RAS/MAPK

signaling depends on the regulation of MAPK levels

by the EJC.

INTRODUCTION

The RAS/MAPK signaling pathway is linked to a wide range of

cellular processes that include proliferation, differentiation and

survival (Kolch, 2005). Its critical role in oncogenesis and various

developmental disorders has also been recognized early on and

abundantly studied (Schubbert et al., 2007; Zebisch et al., 2007).

The pathway is minimally defined by the small GTPase RAS and

a core of three kinases (RAF, MEK and ERK/MAPK) that transmit

signals incoming mostly from plasma membrane receptors.

Upon its activation by a Guanine nucleotide Exchange Factor

(GEF), GTP-loaded RAS triggers RAF activation, which in turn

phosphorylates MEK. Activated MEK then phosphorylates and

activates MAPK. MAPK then phosphorylates a specific set of

substrates that will elicit distinct cell responses (McKay andMor-

rison, 2007; Turjanski et al., 2007).

Within recent years, it has become apparent that the RAS/

MAPK pathway is not merely a linear route involving only four

proteins as commonly referred to, but in fact corresponds to

a larger protein network that comprises multiple regulatory char-

acteristics such as feedback loops (Dougherty et al., 2005),

compartmentalization (Ebisuya et al., 2005; McKay and Morri-

son, 2007) and crosstalk with other pathways (Hurlbut et al.,

2007). However, many of the features inherent to this network

are still poorly understood and its protein composition remains

incompletely defined. Despite their diversity, these regulatory

mechanisms typically affect the enzymatic activity of the core

components or the accessibility to their substrates (Kolch,

2005; McKay and Morrison, 2007). Although much less docu-

mented, mounting evidence suggests that the pathway is also

influenced by other means that affect the steady-state levels of

core components. For example, the expression of LET-60/RAS

in C. elegans or mammalian RAS proteins has been shown to

be modulated by the let-7 family miRNAs (Johnson et al.,

2005). More recently, PumiliomRNAbinding proteins were found

to restrict MAPK activity by attenuating the expression of the

C. elegans MAPK, Mpk-1, as well as ERK2/MAPK1 and p38a/

MAPK14 in human ES cells, which occurred via the binding of

specific sites in the 30UTR of their respective transcripts (Lee

et al., 2007). Moreover, the LARP-1 RNA-binding protein has

also been shown to control the mRNA abundance of RAS/MAPK

pathway components, including those of MAPK in theC. elegans

germ line (Nykamp et al., 2008). Thus, it appears that the expres-

sion of the core components themselves can be the target of

specific mechanisms that in turn impact signaling output.

In order to identify novel factors that modulate RAS/MAPK

signaling, we conducted a genome-wide RNAi screen in

Drosophila S2 cells. Interestingly, the screen led to the identifica-

tion of the Exon Junction Complex (EJC) as a critical factor

that specifically impacts MAPK protein levels in Drosophila.

The EJC has recently emerged as an important determinant in

many aspects of mRNA regulation (Tange et al., 2004). This

protein complex is deposited on mRNAs 20-24 nucleotides


upstream of exon-exon junctions in a splicing-dependent

manner and serves as a tether for other peripheral factors that

are associated with various activities. Three main functions for

the EJC have been described to date, namely, the nuclear

export/subcellular localization of specific transcripts (Hachet

and Ephrussi, 2004; Le Hir et al., 2001b; Palacios et al., 2004),

translational enhancement (Diem et al., 2007; Nott et al., 2004;

Wiegand et al., 2003) and nonsense mediated mRNA decay

(NMD) (Gatfield et al., 2003; Gehring et al., 2003; Lykke-Ander-

sen et al., 2001; Palacios et al., 2004), which is an mRNA surveil-

lance mechanism that eliminates faulty transcripts harboring

premature stop codons (Chang et al., 2007). A few species-

specific exceptions exist to these roles; for instance, the EJC

does not appear to mediate NMD in Drosophila, C. elegans or

S. pombe (Gatfield et al., 2003; Longman et al., 2007; Wen and

Brogna, 2010). The EJC is comprised of four core components:

EIF4A3, MAGOH, RBM8A/Y14 and MLN51. These are respec-

tively known in flies as EIF4AIII, Mago Nashi (MAGO), Tsunagi

(TSU) and Barentz (BTZ), which wewill refer to herein. In contrast

to the other EJC core factors, BTZ is predominantly cytoplasmic

(Degot, 2004; Macchi et al., 2003; Palacios et al., 2004) and is

thought to be added to the complex upon mRNA export from

the nucleus, thus forming the cytoplasmic EJC. It is this config-

uration of the EJC that is involved in mRNA localization and

NMD (Palacios et al., 2004). The nuclear EJC (EIF4AIII/MAGO/

TSU) can modulate mRNA export, yet it is not required for export

of all spliced transcripts (Gatfield and Izaurralde, 2002). Intrigu-

ingly, despite the fact that the EJC associates with many splicing

cofactors such as RNPS1 and SRM160 and also interacts with

the core spliceosome (Bessonov et al., 2008; Herold et al.,

2009; Merz et al., 2007), it has not been implicated in pre-

mRNA splicing per se and is therefore considered to be exclu-

sively dedicated to postsplicing events.

Here, we investigated the function of the EJC in RAS/MAPK

signaling. We show that the EJC acts downstream of MEK in

the RAS/MAPK cascade and is required for MAPK expression.

Unexpectedly, we found that the EJC controls the splicing of

mapk pre-mRNAs as its disruption leads to exon skipping

events. Interestingly, RNA-Seq and RT-PCR data revealed that

the alteredmRNA expression profile provoked by EJC disruption

was not a general phenomenon, but instead correlated with

intron length, whereby pre-mRNAs bearing very large introns,

such as those of mapk, were more sensitive to EJC depletion.

Given that exon exclusion was observed for several transcripts

in addition to mapk, we propose that the EJC is involved in

exon definition of large intron-containing genes. Together, our

findings reveal a critical factor that controls RAS/MAPK signaling

in Drosophila and unveils a role for the EJC in the splicing regu-

lation of long intron-containing genes.

RESULTS

EJC Components Modulate RAS1 Signaling

Downstream of MEK

Expression of a constitutively activated form of Drosophila RAS1

(RAS1V12) in S2 cells leads to sustained activation of endoge-

nous MAPK that is quantitatively measurable by immunohisto-

chemistry (Figure 1A and see Experimental Procedures). We

(GF

P d

sR

NA

norm

.)

RAS1V12 RAFCT MEKEE

eIF4AIII

mago tsu cnk

mapk/rl

dsRNA

avg.

signa

l / ce

ll(G

FP

dsR

NA

norm

.)

0.0

0.5

1.0

1.5

Insulin (pMAPK)

EGFR spitz (pMAPK)

SEVS11 (pMAPK)

RAC1V12 (pJNK)

eIF4AIII

mago tsu cnk

mapk/rl

dsRNA

A

B

RAS1

RAF

MEK

RTKRAC1

JNKKK

JNKK

MAPK JNK

RAC1V12

C

avg.

pMAP

K sig

nal /

cell

0.0

0.5

1.0

1.5

RAS1V12

RAFCT

MEKEE

Insulin EGFR + spitzSEVS11

Figure 1. EJC Components Functionally Impact RAS1/MAPK

Signaling Downstream of MEK

(A) Schematic representations of the MAPK and JNK pathways. The stimuli

used in the different cell-based functional assays are positioned within these

pathway models (colored text).

(B) Epistasis analysis of the nuclear EJC components. Quantitative micros-

copy-based measurement of the average pMAPK intensity/cell for quadrupli-

cate pMet-Ras1V12 S2 cell samples and duplicate pMet-rafCT and pMet-mekEE

S2 cell line samples treated with the indicated dsRNAs. dsRNA probes target-

ingmapk/rl and cnkwere added as controls of the epistasis strategy; cnk func-

tions between RAS1 and RAF and MAPK functions downstream of MEK.

(C) Values from RTK-based MAPK activation assays (orange) and RAC1V12-

based activation of JNK (purple) are shown for S2 cells treated with the indi-

cated dsRNAs. The average pMAPK intensity/cell is shown for duplicate

samples of insulin treated S2 cells, pMet-Egfr S2 cells treated with Spitz

(EGFR + SPI) and heat-shock induced pHS-SevS11 cells. The average pJNK

intensity/cell is shown for duplicate samples of pMet-Rac1V12 S2 cells.


used this assay to conduct a genome-wide RNAi screen to iden-

tify novel fly genes that modulate signaling through the RAS1/

MAPK pathway. Approximately 100 validated hits were identi-

fied, of which two-thirds of the known components of the

pathway were found including raf/phl, mek/Dsor1 and mapk/rl

(D.A.-B., unpublished data). Interestingly, among the candidate

genes, the three components of the nuclear EJC (eIF4AIII,

mago and tsu) were identified as suppressors considering that

the average phospho-(p)MAPK signal per cell was found to

decrease consistently upon knockdown of their respective tran-

scripts (Figure 1B and Figures S1A and S1B available online).

This indicates that the EJC positively contributes to pathway

activity at a step between RAS1 and MAPK. Intriguingly,

although the peripheral EJC component RnpS1 was also found

as a suppressor (Figures S1A and S1C), btz, the fourth core

component that is part of the cytoplasmic EJC, was not identi-

fied in the RNAi screen. Subsequent experiments confirmed

that knockdown of btz did not reduce pMAPK levels (Fig-

ure S1A), even though the dsRNA targeting btz did efficiently

deplete its transcript levels (Figure S1B) as did the dsRNAs tar-

geting mago, tsu, eIF4AIII and RnpS1 (Figures S1B and S1C).

In addition to suppressing RAS1V12, depletion of the nuclear

EJC also reduced pMAPK induced by upstream RTKs including

EGFR, SEVS11 and the insulin receptor (InR), but did not

suppress RAC1V12-induced JNK signaling (Figures 1A and 1C).

This indicates that the EJC is required for RTK/RAS1/MAPK

signaling, but is not required in the context of the RAC1/JNK

pathway. Next, we assayed the position of the EJC’s effect

with regards to the kinase cascade downstream of RAS1. Epis-

tasis experiments conducted using several cell lines expressing

constitutively active forms of RAF and MEK also showed

a consistent decrease in pMAPK levels upon EJC knockdown

(Figures 1A and 1B and Figure S1D). In contrast, cnk, a regulator

of RAF activation (Douziech et al., 2003), suppressed RAS1V12,

but not active RAF or MEK (Figure 1B). Thus, these results place

the requirement for EJC activity at a step downstream of MEK.

To determine whether the ability of the EJC tomodulate RAS1/

MAPK signaling was not restricted to S2 cells, but also controls

this pathway in vivo, we conducted genetic interaction experi-

ments using eIF4AIII, mago and tsu loss-of-function alleles.

RTK-induced RAS1/MAPK activity is required for neuronal

photoreceptor and cone cell differentiation during Drosophila

eye development (Wassarman et al., 1995). Expression of

Ras1V12 under the control of the eye specific sev promoter/

enhancer regulatory sequences produces extra photoreceptor

cells, which causes a characteristic rough eye phenotype (Fortini

et al., 1992) that has been extensively used in genetic interaction

assays (Karim et al., 1996; Therrien et al., 2000). As shown in

Figure 2A, this rough eye phenotypewas dominantly suppressed

by heterozygous mutations in the eIF4AIII, mago or tsu genes.

Four additional genetic interaction assays also tied EJC activity

to RAS1/MAPK signaling in vivo. First, the lethality associated

with raf/phl12 hemizygous mutant males was enhanced in

eIF4AIII, mago or tsu heterozygous backgrounds (Figure S2A).

Second, wing vein deletions and lethality caused by a hemizy-

gous mutation in csw, which encodes a Shp-2 phosphatase

homolog (Perkins et al., 1992), were enhanced by a mago1

heterozygous allele (Figures 2B and Figures S2B and S2C).

Third, extra wing vein material produced by a constitutively

active Egfr allele was dominantly suppressed by eIF4AIII, mago

or tsu mutant alleles (Figures 2C and Figures S2D–S2F). Finally,

the rough eye andwing vein deletion phenotypes of homozygous

mapk/rl1, which corresponds to a hypomorphic allele of the

mapk/rl gene, strongly increased in severity in heterozygous

mago mutant backgrounds (Figures 2D and 2E). Collectively,

these data provide compelling evidence that the EJC is required

for RAS1/MAPK signaling in Drosophila.

The EJC Regulates Drosophila MAPK Protein

Expression

In light of the EJC function in mRNA translation (Tange et al.,

2004), we investigated the effects of its depletion on the protein

levels of selected RAS1/MAPK pathway components. Strikingly,

western blot analysis revealed that MAPK levels were signifi-

cantly reduced upon the knockdown of EJC components

(Figure 3). This effect appeared to be specific toMAPK as disrup-

tion of the EJC did not have an appreciable effect on the protein

levels of other pathway components, such as RAS1, RAF, MEK,

or CNK, on other kinases such as AKT and JNK, or on Actin

(Figure 3). Moreover, silver-stained protein lysates from EJC-

depleted S2 cells did not show any significant difference

compared to control cells, thus providing additional evidence

that EJC depletion does not lead to a global reduction of protein

levels in S2 cells (Figure S3A and see below). Interestingly, we

noticed that the levels of ectopically-produced MAPK from

amapk cDNA were insensitive to EJC depletion, which indicates

that the effect of the EJC on MAPK does not occur following its

translation (Figure S3B). Together, these findings provide an

explanation as to why EJC depletion consistently reduced

endogenous pMAPK levels in the various RAS1/MAPK signaling

assays presented above and suggest that the EJC is specifically

involved in MAPK protein expression at a step that precedes its

translation or that occurs concomitantly.

To verify whether the EJC controls MAPK levels in vivo as it

does in S2 cells, we generated mago homozygous mutant

clones during eye development and stained third instar eye-

antennal discs using an anti-MAPK antibody. In agreement

with the cell culture data, both MAPK and pMAPK levels were

strongly reduced in mago mutant clones, whereas CNK was

unaffected (Figures 4A and 4D and Figure S4A). Similar results

were obtained when mago, tsu or mapk itself were clonally

depleted by RNAi (Figures 4B and 4C and Figure S4B). Consis-

tent with reduced MAPK levels, EJC depletion during eye devel-

opment significantly impeded photoreceptor and cone cell

differentiation (Figures 4E and 4F). Together, these results indi-

cate that the EJC plays a critical role in establishing proper

MAPK protein levels in vivo.

Loss of EJC Activity Induces Exon Skipping Events

during Mapk Pre-mRNA Splicing

Given the role that the EJC plays in several aspects of mRNA

maturation, we investigated the effect of EJC disruption on

mapk transcript levels. An initial RT-qPCR evaluation of mapk

transcript levels showed a reproducible 1.5- to 2-fold decrease

upon EIF4AIII and MAGO depletion, whereas, consistent with

the results from the functional assays, mapk transcript levels


did not decrease upon knockdown of btz (Figure 5A). While the

reduction in the amount of mapk mRNA was reproducible, it

did not appear sufficient to account for the decrease in protein

expression. Therefore, we examinedwhethermapkmRNA trans-

lation might be affected upon EJC depletion by comparing its

incorporation into polysomes between control and MAGO-

depleted cells. While the expected 1.5- to 2-fold decrease

in mapk mRNA levels in MAGO-depleted cells was observed,

no difference in polysome incorporation efficiency could be

detected (Figures S5A–S5C). This suggests that the EJC does

not directly impact mapk translation, nor significantly influence

processes that would lead to a decrease in translation efficiency.

Also, the fact that btz was not involved in this function further

L5

L4

L1L2

L3acv

pcv

WT

A

B

C

E

D

sev-Ras1V12/rlS-352WT sev-Ras1V12/+

sev-Ras1V12/mago3

sev-Ras1V12/+; eIF4AIII19/+

sev-Ras1V12/mago1 sev-Ras1V12/mago2 sev-Ras1V12/tsu1

rl1, mago1 / rl1rl1 / rl1 rl1, mago3 / rl1

EgfrElp / + EgfrElp / rl1 EgfrElp / mago1

cswlf wscY / lf / Y ; mago1 / +

rl1 / rl1WT rl1, mago1 / rl1 rl1, mago3 / rl1

Figure 2. Components of the EJC Geneti-

cally Interact with RAS1/MAPK Pathway

Components

(A) Ras1V12 rough eye phenotype is dominantly

suppressed by heterozygous mutations in EJC

components. Fly eyes of the indicated genotypes

were scored for severity of the rough eye pheno-

type (n > 30) and classified as strong, medium

or weak. All of the EJC alleles scored as medium

to strong suppressors. Representative ESEM

microscopy images are presented here. The

mapk/rlS-352 allele is used as a positive control.

Anterior is to the right.

(B) Enhancement of cswlfwing vein phenotypes by

mago1 loss-of-function allele. A wild-type wing is

shown as a reference (left panel). The character-

istic deletion of the distal end of the L5 (and some-

times L4 and L2) wing veins typically observed in

cswlf hemizygous males is shown (central panel)

as well as the enhanced phenotype observed in

mago1 heterozygous background (right panel).

Quantification of the enhancement is presented

in Figure S2C.

(C) Suppression of Egfr gain of function wing vein

phenotype. Typical extra wing vein material

produced by the EgfrElp gain of function allele

near the distal end of the L2 vein is shown in the

left panel (arrow). Suppression of the phenotype

by the heterozygous mapk/rl1 allele (positive

control; central panel) or by heterozygous mago1

(right panel) is also shown and quantification of

the effect is presented in Figure S2E.

(D and E) Enhancement of mapk/rl1 homozygous

rough eye and wing vein phenotypes by two

distinct mago alleles. Vein deletions (L4 mainly,

arrow) and smaller/rougher eyes are observed in

heterozygous mutant backgrounds for mago1

and mago3.

implied that the effect occurred at an

earlier step in mapk expression involving

the nuclear EJC heterotrimer.

In order to confirm the RT-qPCR

results, we surveyed mapk mRNA by

northern blot and RT-PCR analyses. In

both cases, a reduction in the amount

of the expected mapk transcripts was

apparent upon EJC depletion, but

surprisingly, truncated mapk transcripts were also visible in

these samples (Figures 5B and 5C). Specifically, transcript size

distribution, as assayed by northern blot, showed that the two

previously described mapk transcripts of 1.7 and 2.7 kb

(Berghella and Dimitri, 1996) were detected in control S2 cells

(Figure 5B, lane 1). Both transcripts were significantly reduced

upon EJC depletion and at least two novel transcripts of 1.1

and 2.1 kb could be detected (Figure 5B, lanes 3–4). Notably,

the reduction in total mapk mRNA levels observed in EJC-

depleted samples was comparable to those observed by RT-

qPCR. Also, RT-PCR with primers in the 50 and 30 UTRs of

mapk mRNA, which were expected to amplify the predicted

RB, RD, RE and RF isoforms, revealed a significant decrease


in the amount of the expected transcripts following EJC deple-

tion (Figure 5C, compare lane 1 to lanes 3–6). Similar to the

northern blot, several truncated species could be observed in

EJC-depleted samples by RT-PCR. In agreement with our

previous observations, this effect on the mapk transcripts was

observed for eIF4AIII, mago, tsu and RnpS1 RNAi (Figure 5C,

lanes 3–6), but not in btz dsRNA-treated cells (Figure 5C, lane 7).

Finally, in support for the specificity of this effect, jnk/bsk full-

length transcripts did not show any discernable change in abun-

dance nor size when amplified by RT-PCR from EJC-depleted

samples compared to control (Figure 5C, compare lane 8 to

lanes 10–13).

In order to ascertain the nature of these truncated transcripts,

we cloned and sequenced several RT-PCR products from both

control cells and cells depleted of eIF4AIII. Remarkably, this

analysis revealed that the shorter transcripts could be accounted

for by a series of exon skipping events (Figure 5D). The vast

majority of these events involved the precise skipping of multiple

consecutive exons with the first exon being joined to different

downstream exons at their expected 30 splice acceptor site. In

agreement with the RT-PCR data, the most commonly observed

transcripts in the eIF4AIII-depleted samples were those lacking

exons II to III, II to IV, and II to V (Figures 5C and 5D). A compila-

tion of all skipping events revealed that exon II was the most

frequently skipped exon (96% of the cases), followed by exon

III (89%) (Figure S5E). As exons II and III include the start codon

and a substantial portion of the kinase domain of MAPK, their

absence in most mapk transcripts from EJC-depleted cells

explains why EJC depletion leads to severe reduction in MAPK

protein levels. Surprisingly, exon skipping was also observed,

but to a limited extent, in control S2 cells (Figures 5C and 5D)

and could also be observed in RNA samples prepared from flies

at all developmental stages (Figure S5D). This indicates that

exon skipping during the splicing of themapk pre-mRNA occurs

dsRNA - Ras1 eIF4AIII

raf/phl

mek/Dsor

1

mapk/rl

mago tsu EJC

poo

l

α-RAS1

α-CNK

α-RAF

α-MEK

α-MAPK

α-JNK

α-AKT

α-ACTIN

Figure 3. Depletion of EJCComponents Specifically ReducesMAPK

Levels in S2 Cells

S2 cells were treated with the indicated dsRNAs. EJC pool refers to a pool of

three dsRNAs targeting eIF4AIII, mago, and tsu. Protein lysates were sub-

jected to SDS-PAGE and immunoblotted with antibodies indicated at the right

of each panel.

mergeGFP MAPK

GFP merge

GFPELAV

GFPCUT

E

F

E'

F'

MAPK

GFP mergepMAPK

A A' A''

C

D

C' C''

D' D''ma

go3

clone

stsu

dsRN

A

clone

sma

gods

RNA c

lones

MAPKGFP merge

mago

dsRN

A

clone

s

B B' B''

Figure 4. Depletion of EJC Components Reduces MAPK Levels

In Vivo and Impacts Photoreceptor Cell Differentiation

(A–C) MAPK staining of third instar larval eye discs. (A) MAPK levels are under

detection levels inmago3 mutant clones, which are marked by the absence of

GFP fluorescence. MAPK is also absent in GFP-positive clones expressing

mago (B) and tsu (C) RNAi.

(D–F) mago dsRNA-expressing clones marked by GFP fluorescence impede

typical RTK-dependent signaling markers in third instar larval eye discs. (D)

Phospho-MAPK (pMAPK) signal that is normally induced by EGFR signaling

in cell clusters along the morphogenetic furrow (arrow) and in a limited set of

cells posterior to it, is abrogated in mago dsRNA-expressing cells. (E and F)

mago dsRNA-expressing cells are impaired in their ability to differentiate as

neurons or cone cells as expression of the neuronal marker ELAV (E) or the

cone marker CUT (F) is eliminated or significantly reduced in those cells. In

all panels, anterior is to the right.


naturally at a certain rate. Taken as a whole, these findings

suggest that the nuclear EJC plays a critical role in exon inclusion

during mapk pre-mRNA splicing.

Intron Length Determines the EJC’s Effect on Splicing

Given that the EJC is thought to generally bind to intron-contain-

ing mRNAs (Le Hir et al., 2000), we deemed unlikely that the

splicing defect we observed was unique to the mapk mRNA.

However, the observation that EJC depletion had no effect on

the expression of several other proteins encoded by intron-con-

taining genes indicated that splicing in general does not require

the EJC. Thus, in order to evaluate the full spectrum of the EJC’s

effects on the transcriptome of S2 cells, we conducted a tran-

scriptional profiling experiment using massively parallel RNA

sequencing (RNA-Seq) technology on a SOLiD platform (Applied

Biosystems) that yielded highly reproducible transcript coverage

across samples (Figure S6A). The abundance of 6760 expressed

transcripts (53 coverage threshold) assayed in this manner

revealed that mapk was among the 50 genes whose expression

was most reduced upon EJC depletion (Figure 6, Figure S6B,

dsRNA

2.7 kb

1.7 kb

*

* mapk/rl

B C

1 2 3 4 5

jnk/bsk1.7 kb

RDRB / RE∆II∆II-III

∆II-IV∆ II-V

A

0.0

0.5

1.0

1.5

Mapk tra

nscript

levels

( GFP

dsR

NA

norm

alis

ed)

dsRNA

GFP mapk/rleIF4AIII

magobtz

GFPmapk/rl

eIF4AIIImago btz

1kb GFPmapk/rleIF4AIII

magotsu btz

dsRNA

1kbRnpS11kb GFPmapk/rl

eIF4AIIImago

tsu btz

dsRNA

1kbRnpS1

3569

3569

III IV

III IV

D

5 ∆II-VII (531)

RD (1,800)

RB (1,569)

RE (1,550)

∆II (1,378)

∆II-III (1,251)

occurences

5

10

5

3

1

ol5'

ol3'

GFPdsRNA

(ntotal

= 24)

6eIF4AIIIdsRNA

(ntotal

= 55)

∆II-III (1,251)

2 ∆ II (1,378)

9 ∆II-IV (1,140)

3 ∆II-IV' (1,101)

RE (1,550)1

6 ∆II-V (1,005)

9986456127504525914502

222

998645612750452222

I II V VI VII

I II

V VI VII VIII

VIII4195

ksb/knjlr/kpam4131211101987654321

observed mapk/rl species

Figure 5. The EJC Is Required for Faithful Splicing of mapk/rl Pre-mRNA

Total RNA was prepared from S2 cells treated with the indicated dsRNA. (A) mapk/rl transcript levels were assayed by RT-qPCR as described in the methods.

Results are the average of three independent experiments (±SD). Statistical significance was evaluated using a two-tailed Student’s t test (asterisk denotes

p value < 0.01).

(B) Poly(A)+ mRNA was subjected to northern blot with probes for mapk/rl or jnk/bsk mRNA (asterisk indicate shifted mapk/rl species).

(C) RT-PCR with primers in the 50 and 30 UTR ofmapk/rl or jnk/bskmRNA. The positions of species depicted in (D) are indicated at the right of themapk/rl panel.

(D)mapk/rl RT-PCR products were cloned and sequenced. The schematic represents the most abundant species observed (for example, DII-III indicates a tran-

script lacking exons II and III). Red bars indicate the position of primers used for RT-PCR. The coding sequence is colored in blue and the exons are numbered I

through VIII based on the RE transcript. The size of each species is shown in parentheses and lengths of the introns are indicated. Long introns (>250 bp) are not

drawn to scale and are represented by dashed lines. The predicted RF isoform was not observed and is therefore not represented.


and Table S1). Therefore, the mapk transcript is highly sensitive

to perturbation of the EJC in comparison with most other genes

expressed in S2 cells.

We next sought to determine what characteristics of mapk

determined this sensitivity to the EJC.We noted three character-

istics that distinguish mapk from other fly genes. First, mapk is

localized in the pericentromeric constitutive heterochromatin.

Second, the first exon which we found to be aberrantly joined

to other downstream exons has a noncanonical splice donor

site (the first three intronic residues, GUU, do not match the

GUR consensus; Mount et al., 1992). Finally, mapk has introns

of larger than average size, including one 25 kb intron, which

is among the longest introns in the genome; less than 1% of

D. melanogaster introns are larger than 25 kb (Figure S6C). Of

these three characteristics, intron length emerged as the deter-

mining factor of the EJC’s effect. While localization in hetero-

chromatin or the presence of atypical splice sites did somewhat

correlate with sensitivity to EJC depletion, this effect was depen-

dent on the presence of long introns. Indeed, genes with these

characteristics, but with introns shorter than 250 bp were not

found to be sensitive to the EJC’s effect (Table S2). On the other

hand, transcripts with long introns were more likely to be

sensitive to EJC depletion (Table S2, Figures 6A and 6B, and

Figure S6B). Additionally, this sensitivity appeared to increase

as a function of intron length in such a manner that a general

trend could be observed for the entire population of transcripts

after both eIF4AIII and mago knockdowns (Figures 6A and 6B,

and Figure S6B). This relationship between EJC sensitivity and

intron length is highly significant since 81 out of the 114 genes

downregulated by more than 2-fold after eIF4AIII depletion had

an intron larger than 1000 bp (p value = 8.593 10 21 (calculated

with a hypergeometric distribution) (Figure 6A)).

Following this observation, we sought to verify whether the

fold-changes in transcript levels detected by RNA-Seq were

due to splicing defects similar to those observed for the mapk

transcript. Therefore, we performed RT-PCR on a subset of

candidate transcripts which were affected by EJC depletion

and also contained long introns. We found that a similar reduc-

tion in the levels of the full-length transcript could be observed

and, in many cases, truncated products were also detected in

EJC-depleted cells (Figure 7A), suggesting that splicing was

also altered in these cases. Indeed, sequencing of the products

for the PMCA, lt, and Tequila genes showed that exon skipping

was, in large part, responsible for the appearance of the trun-

cated products (Figure S7). In contrast, a set of control tran-

scripts with multiple exons and whose introns were all smaller

than 250 bp did not show additional truncated isoforms or

variation in abundance upon EJC depletion (Figure 7B). These

results demonstrate a clear link between intron length, splicing

and the EJC.

Finally, to complement our comprehensive study of transcript

abundance after EJC depletion, we wanted to determine

whether protein levels were accordingly modulated upon EJC

knockdown in an intron length-dependent manner. For this, we

used a nonbiased, label-free quantitative proteomics approach

to globally assess the abundance of proteins in S2 cells depleted

for the EJC (see Extended Experimental Procedures and

Figure S6D). Of the 6760 expressed transcripts detected by

RNA-Seq, we were able to correlate the abundance of 2410

corresponding proteins. Interestingly, no dramatic change was

observed in the EJC-depleted proteome and no global trend

could be found between protein fold-change and intron length

(Figures S6E and S6F). This observation is consistent with

western blot analysis of several proteins (Figure 3) and with the

Figure 6. The Impact of the EJC on Transcript Levels Correlates with Intron Length

Poly(A)+, rRNA-depleted mRNAs from eIF4AIII, mago and GFP dsRNA-treated S2 cells were subjected to RNA-Seq on the SOLiD sequencing platform.

(A) eIF4AIII knockdown causes a systematic and gradual downregulation of transcripts log2(fold-change) in a manner proportional to the length of the longest

intron of the gene. mapk is among the 50 most downregulated gene detected by transcriptome analysis. The lt, tequila, PMCA, sxc and Dbp80 transcripts

that were subjected to further validation are labeled in red and eIF4AIII, in green. The Venn diagram depicts the overlap between the groups of 2-fold downregu-

lated transcripts and genes containing long introns (>1000 bp).

(B)mago and eIF4AIII depletion yield a reproducible decrease of themapkmRNA abundance together with other long intron-containing transcripts. Transcripts

that were further validated are labeled in red. eIF4AIII and mago mRNAs are depicted in green and served as internal controls.


general aspect of silver-stained protein lysates from EJC-

depleted S2 cells (Figure S3A). Nonetheless, MAPK, PMCA

and SXC were among the most significantly downregulated

protein products in both eIF4AIII and mago-depleted samples

compared to GFP dsRNA-treated cells (Figures S6E–S6G).

This absence of strong correlation between transcript and

protein abundance has been known for some time (Gygi et al.,

1999) and is attributed to the contrasting stability of mRNAs

and proteins, especially in the context of short-term depletion

of the EJC by RNAi. This global analysis of transcript and protein

levels shows that most of the effect of EJC depletion is mani-

fested at the level of transcript abundance and splicing and

that mapk is one of the most sensitive targets of the EJC.

DISCUSSION

Previous work has chiefly associated the EJC to the control of

postsplicing events. Here, we describe a function for the nuclear

EJC in the regulation of splicing. This function is not universal,

because in the context of RAS1/MAPK signaling it was limited

to mapk. Our observation that intron length determines sensi-

tivity to EJC depletion provides an explanation for this. Impor-

tantly, this also provides an important insight into the splicing

of transcripts with long introns, which suggests that the EJC is

required for exon definition in this context.

Splicing of short introns (<200 bp) occurs via the recognition of

50 and 30 splice sites across the intron (intron definition). The

process differs in long introns (>250 bp), where bordering exons

require a priori recognition of their respective splice sites across

the exon before splicing can occur (exon definition) (Fox-Walsh

et al., 2005; Sterner et al., 1996). Exon definition is less robust

than intron definition, and is thus thought to be more permissive

to regulation (Fox-Walsh et al., 2005; Sterner et al., 1996).

Consistent with this, long genes with multiple exons tend to

present more alternative splice variants (Budagyan and Loraine,

2004) and exons bordered by large introns are much more

likely to be excluded (Fox-Walsh et al., 2005; Kim et al., 2007;

McGuire et al., 2008; Roy et al., 2008). This correlation is more

pronounced in lower eukaryotes, where large introns are

comparatively rare and seem to act as major determinants of

alternative splicing (Fox-Walsh et al., 2005). In vertebrates, the

presence of additional modes of regulation is thought to explain

the less predominant impact of intron size. Still, it remains

unclear whether the effect of long introns on alternative splicing

GFPeIF4AIII

magoRnpS1

btz

Dbp80(RB)

lt(RA / RB / RC)

PMCA(RJ / RM)

sxc(RA / RB / RC)

Teq(RA / RE)

CG9149(RA)

CG10417(RA / RB)

CG8042(RA)

Dbp45A(RA)

sip3(RA)

GFPeIF4AIII

magoRnpS1

btz

RNAi

RNAi

A

B

GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz

GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz GFPeIF4AIII

magoRnpS1

btz

Figure 7. Truncated Transcripts Are Observed for Other EJC targets

Total RNA was prepared from S2 cells treated with the indicated dsRNA. RT-PCR products for EJC-sensitive transcripts (A) and EJC-insensitive transcripts (B)

were resolved on agarose/EtBr gels. Specific transcript isoforms targeted by RT-PCR are indicated in parentheses. Each of the transcripts assayed in (A) has at

least one large intron (>1000 bp), while the transcripts in (B) have at least five exons, but no introns longer than 200 bp.


involves specific protein factors, although the binding of multiple

hnRNPs to long introns has been proposed as one explanation

for this effect (Fox-Walsh et al., 2005).

Our finding that the EJC is associated with exon definition in

transcripts containing long introns suggests that the nuclear

EJC is one of the factors determining alternative splicing in this

context. Importantly, among the different types of alternative

splicing events, exon exclusion – often of multiple consecutive

exons – was the salient feature of EJC disruption. This is consis-

tent with previous reports that exon exclusion is predominantly

associated to long introns (Fox-Walsh et al., 2005; Kim et al.,

2007; McGuire et al., 2008). However, we also observed some

intron retention and alternative donor and acceptor site usage

in the transcripts we sequenced (Figure 5 and Figure S7). In addi-

tion to these splicing changes, EJC disruption also caused

a reduction in transcript levels of genes with long introns (Figures

6 and 7). This may be due to a lower stability of the new transcript

variants or to other splicing defects culminating in the degrada-

tion of the RNA product.

As introns tend to be larger in higher eukaryotes compared to

Drosophila and other lower eukaryotes (Fox-Walsh et al., 2005;

Kim et al., 2007; McGuire et al., 2008), it will be interesting to

verify the extent to which the EJC is involved in splicing in higher

eukaryotes. With respect to the MAPK genes, it is intriguing

to note that human MAPK1/ERK2 bears an intron just under

60 kb, which is significantly above the average human intron

length (Lander et al., 2001). In contrast,MAPK3/ERK1 has signif-

icantly smaller introns, raising the possibility that these two

MAPK genes bear different sensitivities to regulation by the EJC.

Despite the fact that the core components of the nuclear EJC

have not been previously linked to splicing, they have been

shown to associate with the spliceosome during splicing and

before actual deposition of the EJC on the mRNA (Gehring

et al., 2009; Merz et al., 2007). Furthermore, deposition of

EIF4AIII upstream of the exon junction site has been found to

occur during the second step of the splicing reaction, before

cleavage of the 30 splice site (Gehring et al., 2009). Thus, these

previous observations do not preclude an involvement of the

nuclear EJC in the splicing process.

The cytoplasmic EJC, which includes BTZ, is required for

oskar mRNA localization and translation in Drosophila (Hachet

and Ephrussi, 2004; Mohr et al., 2001; van Eeden et al., 2001).

However, btz depletion did not influence MAPK activity, nor

did it cause reduction in mapk mRNA or protein levels. Though

BTZ is sometimes described as a core EJC component, EJC

complexes devoid of BTZ have been found to associate with

the spliceosome (Bessonov et al., 2008; Herold et al., 2009;

Merz et al., 2007), and human EIF4A3, RBM8A/Y14 andMAGOH

have been shown to form a trimeric complex in the absence of

MLN51/BTZ (Ballut et al., 2005; Gehring et al., 2009). Further-

more, BTZ deposition on the mRNA has been found to occur

after completion of the splicing reaction (Gehring et al., 2009).

Also, whereas MAGO, TSU and EIF4AIII have been described

as being mainly located in the nucleus and in nuclear speckles

(Le Hir et al., 2001a; Palacios et al., 2004), BTZ bears a nuclear

export signal and is predominantly cytoplasmic at steady state

(Degot, 2004; Macchi et al., 2003; Palacios et al., 2004). Thus,

BTZ is likely not involved in nuclear events regulated by the

EJC, which is consistent with our observation that btz depletion

does not impact the splicing of mapk or of other pre-mRNAs

(Figure 5C and Figure 7A).

While the EJC has not been previously implicated in splicing,

this is not the case for some EJC-associated factors such as

the SR (serine/arginine-rich) factors RNPS1 and SRM160. SR

factors are important determinants of constitutive and alternative

splicing (Long and Caceres, 2009). RNPS1 was initially charac-

terized as a splicing factor (Mayeda et al., 1999), and has since

been shown to regulate alternative splicing through alternate

exon usage (Sakashita et al., 2004) and to enhance spliceosomal

activity (Trembley et al., 2005). The fact that we also identified

RNPS1 as a factor linked to the EJC’s splicing function is consis-

tent with this and further suggests that the recruitment of RNPS1

or other SR factors could provide an additional level of specificity

to the EJC’s effect. In support of this idea, our RT-PCR data

showed that RnpS1 did not impact the splicing of all EJC target

genes (Figure 7). Importantly, RNPS1 takes part in more than

one aspect of the EJC’s functions as it is also linked to NMD

(Lykke-Andersen et al., 2001) and translational enhancement

(Nott et al., 2004; Wiegand et al., 2003). SRM160, another

EJC-associated SR factor, has also been shown to function as

a splicing coactivator which binds to exonic splicing enhancers

via other SR factors (Blencowe et al., 1998; Eldridge et al., 1999).

However, SRm160 knockdown had modest effects on MAPK

expression and no splicing defects in mapk could be observed

by RT-PCR (data not shown). Interestingly, SRM160 and

RNPS1 also promote pre-mRNA 30 end cleavage, but only

SRM160 can function independently of the EJC in this context

(McCracken et al., 2002; McCracken et al., 2003), indicating

that its activity can be uncoupled from that of the EJC. Still,

it is possible that this component acts in concert with the EJC

in the splicing of other transcripts.

One important issue raised by our findings is whether the

EJC’s effect on splicing is regulated, or if it is part of a constitutive

process involved in exon definition. EJC activity is known to be

regulated in the context of translational enhancement where

mTOR signaling modulates this activity via the EJC cofactor,

SKAR (Ma et al., 2008). A similar mechanism involving EJC

cofactors could also be responsible for bridging different inputs

with the regulation of splicing. Accordingly, the CK2 kinase has

been found to phosphorylate RNPS1 and regulate its splicing

activity (Trembley et al., 2005). More generally, the control of

alternative splicing through SR and hnRNP factors is regulated

by different signaling events (House and Lynch, 2008; Stamm,

2008). For example, the RAS/MAPK dependent regulation of

CD44 splicing occurs via the SRM160 and SAM68 SR factors

(Cheng and Sharp, 2006). Thus, SR factors could provide

further specificity and signal-integration properties to the

EJC’s function in splicing. Further investigation of the splicing

changes brought about by cofactors such as RNPS1 will help

to understand their contribution to EJC-regulated splicing.

Indeed, one possibility is that the EJC acts as an adaptor

platform for SR factors which are required for exon definition.

In this model, the individual SR factors would be the effectors

involved in providing specificity to splice site selection. Alterna-

tively, it is possible that the EJC also directly impacts splice

site selection (and exon definition) by masking either binding


sites for other splicing factors or RNA motifs directly involved in

splice site selection (Yu et al., 2008). A third possibility is that the

EJC stabilizes the interaction of spliceosome complexes with

splice sites and that this is of particular importance for exon

definition. Investigation of these possibilities will be important

in order to further understand the EJC’s role in splicing at the

mechanistic level.

The example of CD44 is of particular interest in the context of

this study as themodulation of CD44 also constitutes a feedback

mechanism that regulates RAS activity (Cheng and Sharp, 2006;

Cheng et al., 2006). Since we initially identified the EJC as specif-

ically regulating mapk expression in the context of the RAS1/

MAPK pathway, and asmapkwas among themost highly modu-

lated EJC targets, both at the mRNA and protein level, the

control of mapk splicing, like CD44, may also be a key element

in modulating signal flow. The identification of the inputs regu-

lating the splicing of mapk via the EJC will provide additional

insights into how signal modulation is achieved through this

route. Also, although many signaling genes are known to encode

alternatively spliced transcripts with different functions, the

extent to which regulation of signaling processes is achieved

by splicing is still largely underappreciated.


Quantitative Immunofluorescence Microscopy

S2 cell lines were distributed in 96-well clear plates (Corning) containing a final

concentration of 200 ng/mL dsRNA. Cells were then fixed, blocked and incu-

bated overnight with a primary antibody (anti-pMAPK 1/2000, Sigma #M8159

or an anti-pJNK 1/500, NEB #9251S), then stained with a secondary antibody

(anti-mouse Alexa Fluor 555-conjugated 1/1000, Invitrogen #A-21424) and

with DAPI. Cells were mounted in Mowiol (9.6% PVA, Fluka). An automated

fluorescence microscopy system (Zeiss Axiovert) was employed for plate

imaging. Autofocus, image acquisition and analysis were conducted using

MetaMorph (MolecularDevices) software. Thecell-scoringapplication inMeta-

Morph was used for quantitative image analysis. Information on cell lines and

cell culture conditions is included in the Extended Experimental Procedures.

Fly Genetics, Immunohistochemistry, and ESEM

Fly husbandry was conducted according to standard procedures. Crosses to

raf/phl12 flies were performed at 18 C. All other crosses were performed at

25 C. The sev-RasV12 and raf/phl12 (formerly referred to as rafHM7) lines have

been described previously (Karim et al., 1996; Melnick et al., 1993). EgfrElp

was described in (Baker and Rubin, 1989). The cswlf (Perkins et al., 1996)

was kindly provided by L. Perkins. Themago alleles were described in (Boswell

et al., 1991). RNAi fly lines were obtained from the VDRC (Dietzl et al., 2007).

All other mutant lines described herein were obtained from the Bloomington

stock center.

Homozygous mago mutant clones were generated using the flp-FRT tech-

nique (Xu and Rubin, 1993). Third instar eye discs were fixed and stained

with anti-MAPK (1/1000, Cell Signaling #4695), anti-CUT (1/200, DSHB) and

anti-ELAV (1/20, DSHB) antibodies. Eye disc images were acquired on a Zeiss

LSM 510 laser scanning confocal microscope. Fly eyes were imaged using an

environmental scanning electron microscope (Quanta 200 FEG) or stereomi-

croscope (Leica MZ FL III). Permount-mounted wings were imaged using

a Nanozoomer (Hamamatsu).

RT-qPCR, RT-PCR, and Northern Blot

S2 cells were cultured in dsRNA (15 mg/ml) for seven days and total RNA was

prepared using TRIzol reagent (Invitrogen).

For RT-qPCR, 2 mg of total RNA was reverse transcribed using the High

Capacity cDNA Archive Kit with random primers (Applied Biosystems). PCR

reactions for 384-well plate formats were performed using 2 ml of cDNA, 5 ml

of the TaqMan fast Universal PCR Master Mix (Applied Biosystems), 2 mM of

each primer, and 1 mM of the Universal TaqMan probe in a total volume of

10 ml. The ABI PRISM 7900HT Sequence Detection System (Applied Biosys-

tems) was used to detect the amplification level.

For RT-PCR, 1 mg of total RNA was primed with oligo(dT)18 followed by

reverse transcription (RT) with SuperScript II Reverse Transcriptase (Invitro-

gen). 1/20 of the RT reaction was used as template for PCR.

For northern blot analyses, poly(A)+ mRNA was isolated from total S2 cell

RNA using oligotex resin (QIAGEN). mRNA samples (1.5 mg) were separated

on a 5% formaldehyde-1% agarose gel and transferred to a nylon membrane

(Hybond-N+; GE Healthcare). Hybridizations were conducted in 0.125M

Na2HPO4 (pH 7.4), 4 mM EDTA, 7% SDS. 32P-labeled probes were synthe-

sized by random priming using mapk/rl or jnk/bsk cDNA fragments.

Membranes were washed three times in 13 SSC/0.1% SDS and once in

0.13 SSC/0.1% SDS for 20 min at 65 C, and then exposed for 3 days at

!80 C using an intensifying screen.

For additional information regarding primer sequences see the Extended

Experimental Procedures and Table S3.

Whole Transcriptome Sequencing and Analysis

Total RNA from S2 cells subjected to treatment with eIF4AIII, mago and GFP

dsRNA in duplicate and was prepared using TRIzol reagent (Invitrogen) and

Poly(A)+ mRNAs were enriched using oligo(dT) selection with the Oligotex

mRNA Midi kit (QIAGEN). The resulting mRNA was then depleted of rRNA

molecules using the Ribominus Eukaryote kit (Invitrogen). High-throughput

sequencing libraries were prepared according to the SOLiD whole transcrip-

tome library preparation protocol (Applied Biosystems). Whole transcriptome

reads were aligned using the Bioscope software package (Applied Biosys-

tems) using the UCSC Drosophila genome release 4.2 as a reference (http://

genome.ucsc.edu/). 953, 837 and 524 million reads mapped uniquely to the

reference genome and yielded an average exon coverage of 154X, 134X and

77X for the GFP, eIF4AIII and mago knocked down samples respectively.

Sequencing data have been deposited in GEO under accession number

GSE24012.


Supplemental Information includes Extended Experimental Procedures, seven

figures, and three tables and can be found with this article online at doi:

10.1016/j.cell.2010.09.014.

ACKNOWLEDGMENTS

We are grateful to R. Boswell, E. Perkins, N. Perrimon, and D. St-Johnston, as

well as the Bloomington and VDRC stock centers for fly stocks and cell lines.

We thank Christian Charbonneau and Monica Nelea for assistance with

microscopy; and Philippe Roux, Pierre Zindy, and Katherine Borden for their

help with the polysome fractionation experiments. We also extend our grati-

tude to the IRIC HTS platform for use of the automated fluorescence micro-

scope; to Michael Kubal (ABI) for help with analysis of the RNA-Seq data;

and to Raphaelle Lambert for assistance with RT-qPCR and sequencing.

D.A.B. is a recipient of Frederick Banting and Charles Best Canada Doctoral

Scholarship. HL is a recipient of a Cancer Research Society postdoctoral

fellowship. P.T. is recipient of a Tier I Canada Research Chair in Proteomics

and Bioanalytical Spectrometry. M.T. is recipient of a Tier II Canada Research

Chair in Intracellular Signaling. This work was supported by the Canadian

Cancer Society and by the Canadian Institutes for Health Research.

Received: December 29, 2009




REFERENCES

Baker, N.E., and Rubin, G.M. (1989). Effect on eye development of dominant

mutations in Drosophila homologue of the EGF receptor. Nature 340, 150–153.


Ballut, L., Marchadier, B., Baguet, A., Tomasetto, C., Seraphin, B., and Le Hir,

H. (2005). The exon junction core complex is locked onto RNA by inhibition of

eIF4AIII ATPase activity. Nat. Struct. Mol. Biol. 12, 861–869.

Berghella, L., and Dimitri, P. (1996). The heterochromatic rolled gene of

Drosophila melanogaster is extensively polytenized and transcriptionally

active in the salivary gland chromocenter. Genetics 144, 117–125.

Bessonov, S., Anokhina, M., Will, C.L., Urlaub, H., and Luhrmann, R. (2008).

Isolation of an active step I spliceosome and composition of its RNP core.

Nature 452, 846–850.

Blencowe, B.J., Issner, R., Nickerson, J.A., and Sharp, P.A. (1998). A coactiva-

tor of pre-mRNAcsplicing. Genes Dev. 12, 996–1009.

Boswell, R.E., Prout, M.E., and Steichen, J.C. (1991). Mutations in a newly

identified Drosophila melanogaster gene,mago nashi, disrupt germ cell forma-

tion and result in the formation of mirror-image symmetrical double abdomen

embryos. Development 113, 373–384.

Budagyan, B., and Loraine, A. (2004). Gene length and alternative transcription

in fruit fly. Computational Systems Bioinformatics Conference, 2004: 2004,

Stanford, CA, USA: Proceedings, 515 - 516.

Chang, Y.F., Imam, J.S., and Wilkinson, M.F. (2007). The nonsense-mediated

decay RNA surveillance pathway. Annu. Rev. Biochem. 76, 51–74.

Cheng, C., and Sharp, P.A. (2006). Regulation of CD44 alternative splicing

by SRm160 and its potential role in tumor cell invasion. Mol. Cell. Biol. 26,

362–370.

Cheng, C., Yaffe, M.B., and Sharp, P.A. (2006). A positive feedback

loop couples Ras activation and CD44 alternative splicing. Genes Dev. 20,

1715–1720.

Degot, S. (2004). Association of the breast cancer proteinMLN51with the exon

junction complex via its speckle localizer and RNA binding module. J. Biol.

Chem. 279, 33702–33715.

Diem, M.D., Chan, C.C., Younis, I., and Dreyfuss, G. (2007). PYM binds the

cytoplasmic exon-junction complex and ribosomes to enhance translation of

spliced mRNAs. Nat. Struct. Mol. Biol. 14, 1173–1179.

Dietzl, G., Chen, D., Schnorrer, F., Su, K.C., Barinova, Y., Fellner, M., Gasser,

B., Kinsey, K., Oppel, S., Scheiblauer, S., et al. (2007). A genome-wide trans-

genic RNAi library for conditional gene inactivation in Drosophila. Nature 448,

151–156.

Dougherty, M.K., Muller, J., Ritt, D.A., Zhou, M., Zhou, X.Z., Copeland, T.D.,

Conrads, T.P., Veenstra, T.D., Lu, K.P., and Morrison, D.K. (2005). Regulation

of Raf-1 by direct feedback phosphorylation. Mol. Cell 17, 215–224.

Douziech, M., Roy, F., Laberge, G., Lefrancois, M., Armengod, A.V., and

Therrien, M. (2003). Bimodal regulation of RAF by CNK in Drosophila.

EMBO J. 22, 5068–5078.

Ebisuya, M., Kondoh, K., and Nishida, E. (2005). The duration, magnitude and

compartmentalization of ERK MAP kinase activity: mechanisms for providing

signaling specificity. J. Cell Sci. 118, 2997–3002.

Eldridge, A.G., Li, Y., Sharp, P.A., and Blencowe, B.J. (1999). The SRm160/300

splicing coactivator is required for exon-enhancer function. Proc. Natl. Acad.

Sci. USA 96, 6125–6130.

Fortini, M.E., Simon, M.A., and Rubin, G.M. (1992). Signalling by the sevenless

protein tyrosine kinase is mimicked by Ras1 activation. Nature 355, 559–561.

Fox-Walsh, K.L., Dou, Y., Lam, B.J., Hung, S.P., Baldi, P.F., and Hertel, K.J.

(2005). The architecture of pre-mRNAs affects mechanisms of splice-site

pairing. Proc. Natl. Acad. Sci. USA 102, 16176–16181.

Gatfield, D., and Izaurralde, E. (2002). REF1/Aly and the additional exon junc-

tion complex proteins are dispensable for nuclear mRNA export. J. Cell Biol.

159, 579–588.

Gatfield, D., Unterholzner, L., Ciccarelli, F.D., Bork, P., and Izaurralde, E.

(2003). Nonsense-mediated mRNA decay in Drosophila: at the intersection

of the yeast and mammalian pathways. EMBO J. 22, 3960–3970.

Gehring, N.H., Lamprinaki, S., Hentze, M.W., and Kulozik, A.E. (2009). The

hierarchy of exon-junction complex assembly by the spliceosome explains

key features of mammalian nonsense-mediated mRNA decay. PLoS Biol. 7,

e1000120.

Gehring, N.H., Neu-Yilik, G., Schell, T., Hentze, M.W., and Kulozik, A.E. (2003).

Y14 and hUpf3b Form an NMD-Activating Complex. Mol. Cell 11, 939–949.

Gygi, S.P., Rochon, Y., Franza, B.R., and Aebersold, R. (1999). Correlation

between protein and mRNA abundance in yeast. Mol. Cell. Biol. 19,

1720–1730.

Hachet, O., and Ephrussi, A. (2004). Splicing of oskar RNA in the nucleus is

coupled to its cytoplasmic localization. Nature 428, 959–963.

Herold, N., Will, C.L., Wolf, E., Kastner, B., Urlaub, H., and Luhrmann, R.

(2009). Conservation of the protein composition and electron microscopy

structure of Drosophila melanogaster and human spliceosomal complexes.

Mol. Cell. Biol. 29, 281–301.

House, A.E., and Lynch, K.W. (2008). Regulation of alternative splicing: more

than just the ABCs. J. Biol. Chem. 283, 1217–1221.

Hurlbut, G.D., Kankel, M.W., Lake, R.J., and Artavanis-Tsakonas, S. (2007).

Crossing paths with Notch in the hyper-network. Curr. Opin. Cell Biol. 19,

166–175.

Johnson, S.M., Grosshans, H., Shingara, J., Byrom, M., Jarvis, R., Cheng, A.,

Labourier, E., Reinert, K.L., Brown, D., and Slack, F.J. (2005). RAS is regulated

by the let-7 microRNA family. Cell 120, 635–647.

Karim, F.D., Chang, H.C., Therrien, M., Wassarman, D.A., Laverty, T., and

Rubin, G.M. (1996). A screen for genes that function downstream of Ras1

during Drosophila eye development. Genetics 143, 315–329.

Kim, E., Magen, A., and Ast, G. (2007). Different levels of alternative splicing

among eukaryotes. Nucleic Acids Res. 35, 125–131.

Kolch, W. (2005). Coordinating ERK/MAPK signalling through scaffolds and

inhibitors. Nat. Rev. Mol. Cell Biol. 6, 827–837.

Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C., Baldwin, J.,

Devon, K., Dewar, K., Doyle, M., FitzHugh, W., et al. (2001). Initial sequencing

and analysis of the human genome. Nature 409, 860–921.

Le Hir, H., Gatfield, D., Braun, I.C., Forler, D., and Izaurralde, E. (2001a). The

protein Mago provides a link between splicing and mRNA localization.

EMBO Rep. 2, 1119–1124.

Le Hir, H., Gatfield, D., Izaurralde, E., and Moore, M.J. (2001b). The exon-exon

junction complex provides a binding platform for factors involved in mRNA

export and nonsense-mediated mRNA decay. EMBO J. 20, 4987–4997.

Le Hir, H., Izaurralde, E., Maquat, L.E., and Moore, M.J. (2000). The spliceo-

some deposits multiple proteins 20[ndash]24 nucleotides upstream of

mRNA exon-exon junctions. EMBO J. 19, 6860–6869.

Lee, M.H., Hook, B., Pan, G., Kershner, A.M., Merritt, C., Seydoux, G.,

Thomson, J.A., Wickens, M., and Kimble, J. (2007). Conserved regulation of

MAP kinase expression by PUF RNA-binding proteins. PLoS Genet. 3, e233.

Long, J.C., and Caceres, J.F. (2009). The SR protein family of splicing factors:

master regulators of gene expression. Biochem. J. 417, 15–27.

Longman, D., Plasterk, R.H.A., Johnstone, I.L., and Caceres, J.F. (2007).

Mechanistic insights and identification of two novel factors in the C. elegans

NMD pathway. Genes Dev. 21, 1075–1085.

Lykke-Andersen, J., Shu, M.D., and Steitz, J.A. (2001). Communication of the

position of exon-exon junctions to the mRNA surveillance machinery by the

protein RNPS1. Science 293, 1836–1839.

Ma, X.M., Yoon, S.O., Richardson, C.J., Julich, K., and Blenis, J. (2008). SKAR

links pre-mRNA splicing to mTOR/S6K1-mediated enhanced translation effi-

ciency of spliced mRNAs. Cell 133, 303–313.

Macchi, P., Kroening, S., Palacios, I.M., Baldassa, S., Grunewald, B., Ambro-

sino, C., Goetze, B., Lupas, A., St Johnston, D., and Kiebler, M. (2003).

Barentsz, a new component of the Staufen-containing ribonucleoprotein parti-

cles in mammalian cells, interacts with Staufen in an RNA-dependent manner.

J. Neurosci. 23, 5778–5788.

Mayeda, A., Badolato, J., Kobayashi, R., Zhang, M.Q., Gardiner, E.M., and

Krainer, A.R. (1999). Purification and characterization of human RNPS1:

a general activator of pre-mRNA splicing. EMBO J. 18, 4560–4570.


McCracken, S., Lambermon, M., and Blencowe, B.J. (2002). SRm160 splicing

coactivator promotes transcript 30-end cleavage. Mol. Cell. Biol. 22, 148–160.

McCracken, S., Longman, D., Johnstone, I.L., Caceres, J.F., and Blencowe,

B.J. (2003). An Evolutionarily Conserved Role for SRm160 in 30-End Process-

ing That Functions Independently of Exon Junction Complex Formation.

J. Biol. Chem. 278, 44153–44160.

McGuire, A., Pearson,M., Neafsey, D., and Galagan, J. (2008). Cross-kingdom

patterns of alternative splicing and splice recognition. Genome Biol. 9, R50.

McKay, M.M., and Morrison, D.K. (2007). Integrating signals from RTKs to

ERK/MAPK. Oncogene 26, 3113–3121.

Melnick, M.B., Perkins, L.A., Lee, M., Ambrosio, L., and Perrimon, N. (1993).

Developmental and molecular characterization of mutations in the

Drosophila-raf serine/threonine protein kinase. Development 118, 127–138.

Merz, C., Urlaub, H., Will, C.L., and Luhrmann, R. (2007). Protein composition

of human mRNPs spliced in vitro and differential requirements for mRNP

protein recruitment. RNA 13, 116–128.

Mohr, S.E., Dillon, S.T., and Boswell, R.E. (2001). The RNA-binding protein

Tsunagi interacts with Mago Nashi to establish polarity and localize oskar

mRNA during Drosophila oogenesis. Genes Dev. 15, 2886–2899.

Mount, S.M., Burks, C., Hertz, G., Stormo, G.D., White, O., and Fields, C.

(1992). Splicing signals in Drosophila: intron size, information content, and

consensus sequences. Nucleic Acids Res. 20, 4255–4262.

Nott, A., Le Hir, H., and Moore, M.J. (2004). Splicing enhances translation in

mammalian cells: an additional function of the exon junction complex. Genes

Dev. 18, 210–222.

Nykamp, K., Lee, M.H., and Kimble, J. (2008). C. elegans La-related protein,

LARP-1, localizes to germline P bodies and attenuates Ras-MAPK signaling

during oogenesis. RNA 14, 1378–1389.

Palacios, I.M., Gatfield, D., St Johnston, D., and Izaurralde, E. (2004). An

eIF4AIII-containing complex required for mRNA localization and nonsense-

mediated mRNA decay. Nature 427, 753–757.

Perkins, L.A., Johnson, M.R., Melnick, M.B., and Perrimon, N. (1996). The non-

receptor protein tyrosine phosphatase corkscrew functions in multiple

receptor tyrosine kinase pathways in Drosophila. Dev. Biol. 180, 63–81.

Perkins, L.A., Larsen, I., andPerrimon, N. (1992). corkscrewencodes aputative

protein tyrosine phosphatase that functions to transduce the terminal signal

from the receptor tyrosine kinase torso. Cell 70, 225–236.

Roy, M., Kim, N., Xing, Y., and Lee, C. (2008). The effect of intron length on

exon creation ratios during the evolution of Mamm. Genomes. RNA 14, 1–13.

Sakashita, E., Tatsumi, S., Werner, D., Endo, H., and Mayeda, A. (2004).

Human RNPS1 and its associated factors: a versatile alternative pre-mRNA

splicing regulator in vivo. Mol. Cell. Biol. 24, 1174–1187.

Schubbert, S., Shannon, K., and Bollag, G. (2007). Hyperactive Ras in devel-

opmental disorders and cancer. Nat. Rev. Cancer 7, 295–308.

Stamm, S. (2008). Regulation of alternative splicing by reversible protein phos-

phorylation. J. Biol. Chem. 283, 1223–1227.

Sterner, D.A., Carlo, T., and Berget, S.M. (1996). Architectural limits on split

genes. Proc. Natl. Acad. Sci. USA 93, 15081–15085.

Tange, T.O., Nott, A., and Moore, M.J. (2004). The ever-increasing complex-

ities of the exon junction complex. Curr. Opin. Cell Biol. 16, 279–284.

Therrien, M., Morrison, D.K., Wong, A.M., and Rubin, G.M. (2000). A genetic

screen for modifiers of a kinase suppressor of Ras-dependent rough eye

phenotype in Drosophila. Genetics 156, 1231–1242.

Trembley, J.H., Tatsumi, S., Sakashita, E., Loyer, P., Slaughter, C.A., Suzuki,

H., Endo, H., Kidd, V.J., and Mayeda, A. (2005). Activation of pre-mRNA

splicing by human RNPS1 is regulated by CK2 phosphorylation. Mol. Cell.

Biol. 25, 1446–1457.

Turjanski, A.G., Vaque, J.P., and Gutkind, J.S. (2007). MAP kinases and the

control of nuclear events. Oncogene 26, 3240–3253.

van Eeden, F.J., Palacios, I.M., Petronczki, M., Weston, M.J., and St Johnston,

D. (2001). Barentsz is essential for the posterior localization of oskar mRNA

and colocalizes with it to the posterior pole. J. Cell Biol. 154, 511–523.

Wassarman, D.A., Therrien, M., and Rubin, G.M. (1995). The Ras signaling

pathway in Drosophila. Curr. Opin. Genet. Dev. 5, 44–50.

Wen, J., and Brogna, S. (2010). Splicing-dependent NMD does not require the

EJC in Schizosaccharomyces pombe (Embo J.).

Wiegand, H.L., Lu, S., and Cullen, B.R. (2003). Exon junction complexes

mediate the enhancing effect of splicing on mRNA expression. Proc. Natl.

Acad. Sci. USA 100, 11327–11332.

Xu, T., and Rubin, G.M. (1993). Analysis of genetic mosaics in developing and

adult Drosophila tissues. Development 117, 1223–1237.

Yu, Y., Maroney, P.A., Denker, J.A., Zhang, X.H., Dybkov, O., Luhrmann, R.,

Jankowsky, E., Chasin, L.A., and Nilsen, T.W. (2008). Dynamic regulation of

alternative splicing by silencers that modulate 50 splice site competition. Cell

135, 1224–1236.

Zebisch, A., Czernilofsky, A.P., Keri, G., Smigelskaite, J., Sill, H., and

Troppmair, J. (2007). Signaling through RAS-RAF-MEK-ERK: from basics to

bedside. Curr. Med. Chem. 14, 601–623.


Patronin Regulates theMicrotubule Network byProtecting Microtubule Minus EndsSarah S. Goodwin1 and Ronald D. Vale1,*1The Howard Hughes Medical Institute and Department of Cellular and Molecular Pharmacology,

University of California, San Francisco, San Francisco, CA 94158-2200, USA


DOI 10.1016/j.cell.2010.09.022

SUMMARY

Tubulin assembles into microtubule polymers that

have distinct plus and minus ends. Most microtubule

plus ends in living cells are dynamic; the transitions

between growth and shrinkage are regulated by

assembly-promoting and destabilizing proteins. In

contrast, minus ends are generally not dynamic, sug-

gesting their stabilization by some unknown protein.

Here, we have identified Patronin (also known as

ssp4) as a protein that stabilizes microtubule minus

ends in Drosophila S2 cells. In the absence of Pa-

tronin, minus ends lose subunits through the actions

of the Kinesin-13 microtubule depolymerase, leading

to a sparse interphase microtubule array and short,

disorganized mitotic spindles. In vitro, the selective

binding of purified Patronin to microtubule minus

ends is sufficient to protect them against Kinesin-

13-induced depolymerization. We propose that Pa-

tronin caps and stabilizes microtubule minus ends,

an activity that serves a critical role in the

organization of the microtubule cytoskeleton.

INTRODUCTION

Microtubules are the principle scaffold of the mitotic spindle,

serve as tracks for intracellular transport of proteins andmRNAs,

and also participate in signaling functions. The repeating subunit

of the microtubule is the a/b-tubulin heterodimer, which poly-

merizes in a head-to-tail fashion to form protofilaments; typically

13 protofilaments associate laterally to form the microtubules

seen in vivo. Due to the head-to-tail assembly, the microtubule

is a polar filament, with b-tubulin facing the plus end and

a-tubulin at theminus end (Mitchison, 1993). In vitro experiments

using purified tubulin first demonstrated that microtubules

exhibit an unusual property called ‘‘dynamic instability,’’ where-

by microtubules undergo prolonged periods of polymerization

and depolymerization with transitions between the two states

called catastrophe (from polymerization to depolymerization)

and rescue (from depolymerization to polymerization) (Desai

andMitchison, 1997). In vitro, plus andminus ends both undergo

dynamic instability over the same range of tubulin concentra-

tions but display small quantitative differences.

As a result of interactions with specific binding proteins, the

dynamic behavior of microtubules in vivo can differ dramatically

from that described in vitro. Many proteins have been identified

that bind at microtubule plus ends and regulate their dynamics.

For example, MAP215 accelerates tubulin subunit addition at the

plus end, EB1 promotes plus end growth and dynamicity, and

Clip170 increases rescue frequency (Akhmanova and Steinmetz,

2008). Opposing these growth-promoting proteins are the

depolymerizing Kinesin-13 motors, which use ATP hydrolysis

to induce a conformational change at plus ends to promote

catastrophe (Moores and Milligan, 2006). The antagonistic

actions of different +TIP proteins account for the more

pronounced dynamic instability of microtubules in vivo

compared to microtubules composed of pure tubulin in vitro

(Kinoshita et al., 2001).

In contrast to the wealth of information on themicrotubule plus

end, the regulation of the microtubule minus end in vivo is poorly

understood. Inmany cell types, theminus ends are clustered and

anchored at a central microtubule-organizing center (MTOC).

This organization has hindered visualization of their dynamics,

in contrast with plus ends, which are more easily viewed at the

cell periphery by microscopy. Even in organisms and cell types

that lack a central MTOC (e.g., S. pombe, D. melanogaster,

A. thailinia, neurons, epithelial cells, and myotubes), the microtu-

bule minus ends appear to be embedded in poorly characterized

anchoring sites around the cell (Bartolini and Gundersen, 2006;

Rusan and Rogers, 2009).

Occasionally, in animal cells, microtubules are released from

a MTOC or break due to actomyosin forces, thereby allowing

minus ends to be observed free from any nucleating material

(Rodionov and Borisy, 1997; Vorobjev et al., 1999; Yvon and

Wadsworth, 1997; Waterman-Storer and Salmon, 1997; Keating

et al., 1997). The conclusion from these studies is that the vast

majority (80%–90%) of free microtubule minus ends are stable,

neither visibly growing nor shrinking. A similar stability of minus

ends has been observed in cytoplasmic extracts (Rodionov

et al., 1999; Vorobjev et al., 1997). Some minus ends, however,

transition to rapid depolymerization resulting in the disappear-

ance of the microtubule, and a very small percentage of microtu-

bules treadmill through the cytoplasm (caused by simultaneous

minus end shrinkage and plus end growth) (Rodionov and


Borisy, 1997). Microtubule elongation from minus ends has not

been reported in vivo. Thus, in contrast to the pronounced

dynamic instability of plus ends, minus ends are mostly static

and are indeed less dynamic than minus ends composed of

pure tubulin in vitro. These results suggest that microtubule

minus ends might be capped by some unknown protein(s) that

suppresses subunit dynamics.

In a whole-genome RNAi screen for spindle morphology

defects in Drosophila S2 cells, we identified a previously unchar-

acterized protein (short spindle phenotype 4 [ssp4]), whose

depletion caused short spindles in mitosis and microtubule

fragments in interphase (Goshima et al., 2007). Three homologs

exist in humans (Baines et al., 2009), one of which localizes at

microtubule minus ends located close to adherens junctions in

epithelial cells (Meng et al., 2008). In this study, we show that

Drosophila ssp4, which we have renamed Patronin for the Latin

‘‘patronus’’ (protector), protects microtubule minus ends in vivo

against depolymerization by Kinesin-13. In the absence of

Patronin, microtubules release from their nucleating sites and

treadmill through the cytoplasm, a result of unhindered minus

end depolymerization. Purified Patronin selectively binds to

and protects minus ends against Kinesin-13-induced depoly-

merization in vitro, demonstrating that Patronin alone is sufficient

to confer minus end stability. We also show that microtubule

minus end dynamics are regulated by competing actions of de-

stabilizing and stabilizing proteins, as has been shown previously

for the plus end.

RESULTS

Depletion of Patronin Results in Free Microtubules

that Move through the Cytoplasm

Drosophila S2 cells do not have a central MTOC in interphase but

rather generate microtubules from multiple small nucleating

sites, with microtubule plus ends generally visible at the cell

periphery, whereas minus ends lie more centrally (Rogers

et al., 2008; Rusan and Rogers, 2009). In wild-type cells, ‘‘free’’

microtubules (where both the plus and minus ends of the same

microtubule are clearly observed) are rarely found in the

periphery (Figure 1A). In striking contrast, when Patronin was

depleted by RNAi (Figure S1A available online), the interphase

microtubule cytoskeleton became less dense (Figure 1A) (45%

polymer decrease; Figure S1B) and the majority of cells had >5

free microtubules visible at the cell periphery (Figure 1A, Movie

S1). Previously, we speculated that freemicrotubulesmight arise

from increased severing after RNAi of Patronin (Goshima et al.,

2007). However, we did not observemicrotubule severing events

in Patronin RNAi cells, and RNAi knockdown of microtubule-

severing proteins did not suppress the number of free microtu-

bules seen after Patronin RNAi (Figure S1F).

Time-lapse observation of GFP-tubulin in Patronin-depleted

cells provided insight into how Patronin affects microtubules.

Free microtubules appeared to move in a linear manner within

the cytoplasm (Figure 1B, Movie S2). In many cases, we

observed microtubules releasing from sites of nucleation and

moving toward the cell periphery, which might explain the

appearance of free microtubules near the cell boundary (Figures

1A and 1C, Figure S1C, Movie S2). As microtubules are nucle-

ated at their minus ends, these observations indicated that the

free microtubules were ‘‘moving’’ with their plus ends leading

and their minus ends trailing. This conclusion is further sup-

ported by observations of EB1-GFP, which always localized to

the leading end of the translocating microtubule in Patronin

RNAi cells (Figure 1D, Movie S3).

Free Microtubules Move by Treadmilling

in Patronin-Depleted Cells

The movement of microtubules in the cytoplasm of Patronin-

depleted cells could result from either (1) transport by an

anchored minus end-directed motor protein (e.g., cytoplasmic

dynein) or (2) microtubule treadmilling caused by tubulin addition

at the plus end at a similar rate as tubulin loss at the minus end.

To distinguish between these two mechanisms, we photo-

bleached a section of a free GFP-labeled microtubule and

observed how the bleach mark moved relative to the two micro-

tubule ends. If the free microtubule is actively transported, the

bleach mark should remain stationary relative to the plus and

minus ends of the moving microtubule. Conversely, if the micro-

tubule is treadmilling, the bleach mark should appear to move

away from the plus end and get closer to the minus end. In

Patronin-depleted cells, we observed the latter result; all plus

ends moved away from the bleach mark (3.3 ± 0.3 mm/min; n =

20) (mean ± standard deviation [SD]) whereas the minus ends

moved closer (3.2 ± 0.3 mm/min; n = 20) and eventually passed

through the bleached area (Figure 2A). These results indicate

that microtubules move through the cytoplasm by treadmilling.

We next wanted to determine whether microtubule treadmil-

ling occurs for any free microtubule or if this phenomenon

requires the depletion of Patronin. In wild-type cells, it was

possible to find an occasional free microtubule, but these did

not translocate in the cytoplasm. When we photobleached

a free microtubule from a wild-type cell, the bleach mark re-

mained at a constant distance from the minus end (0.01 ±

0.07 mm/min; n = 10), whereas the plus end continued to poly-

merize (3.25 ± 0.24 mm/min; n = 10) (Figure 2A). This finding

suggests that free microtubule minus ends are stable in wild-

type cells, as has been observed in other cell types (Dammer-

mann et al., 2003) and that the minus end depolymerization

that gives rise to microtubule treadmilling requires the depletion

of Patronin. We also examined whether minus end depolymer-

ization occurred after RNAi depletion of g-tubulin and g-TuRC

and g-TuSC components, as the g-TuRC complex has been

shown to bind to microtubule minus ends in vitro (Moritz et al.,

1995; Zheng et al., 1995; Wiese and Zheng, 2000). However, in

these RNAi cells, free microtubules were rare and did not

undergo treadmilling (Figure S1D).

To learn more about microtubule behavior after Patronin

depletion, we measured the plus and minus end dynamics in

wild-type and Patronin-depleted cells. For the microtubule plus

end, the rates of growth and shrinkage and the frequencies of

catastrophe and rescue were similar under Patronin depletion

and wild-type conditions (Table 1). Thus, Patronin appears to

have negligible effects on plus end dynamics. In contrast, minus

ends displayed very different dynamics after Patronin depletion.

In Patronin RNAi cells, minus ends of treadmilling microtubules

often depolymerized at a rate of 3.9 ± 0.9 mm/min (mean ± SD),


which is similar to the plus end polymerization rate of 4.2 ±

1.3 mm/min (Table 1). The similarity in the rates of tubulin addition

at the plus end and dissociation from theminus end explains why

the lengths of treadmilling microtubules often remain relatively

constant, with occasional shortening or lengthening when either

the plus end or minus end pauses (Movie S1 and Movie S2). We

also observed a more rapid minus end depolymerization rate of

10.2 ± 2.2 mm/min, and occasionally individual microtubules

would transition between the slow and fast depolymerization

rates (Figure S1E). Interestingly, minus end depolymerization

often halted when it reached the EB1-enriched microtubule

plus end tip (Figure S2A, 20 of 30 depolymerizing microtubules

paused for an average of 35.8 ± 13.1 s), indicating that +TIP

proteins might help the microtubule resist continued minus-

end depolymerization. After such a pause, the microtubule

would either continue to depolymerize and disappear (11 of 20

microtubules) or resume plus end growth and increase in length

(Figure S2A, 9 of 20 microtubules). In summary, microtubule

minus ends can depolymerize at two rates in vivo: one similar

to plus end growth (resulting in treadmilling) and a second

A Wildtype Patronin RNAi

0

20406080

100

Wildtype Patronin RNAi0 MT 1-5 MT > 5 MT

Cells

with

free

MT

(% of

total

)

30 s 60 sB 0 s

0 sD 21 s 48 s

C 0 s 9 s 15 s 18 s

60 s

Figure 1. Depletion of Patronin Results in Free Microtubules that Move through the Cytoplasm

(A) Time-lapse microscopy of GFP-tubulin wild-type and Patronin-depleted Drosophila S2 cells show that Patronin-depleted cells have numerous ‘‘free’’ micro-

tubules (both the plus and minus ends of the same microtubule are clearly visible, arrows) that are rarely seen in wild-type cells and also have a sparser micro-

tubule network (insert shows a region with several free microtubules). The chart to the right shows the quantitation of free microtubules per cell from two inde-

pendent experiments; colored bars indicate the percentage of cells with the number of indicated freemicrotubules observed (n = 200 cells per experiment; SEM<

6%). Scale bars, 10 mm. See Movie S1.

(B) Time-lapse TIRF microscopy of Patronin-depleted GFP-tubulin cells demonstrates that free microtubules move throughout the cytoplasm (colored arrows

follow the motion of the leading end of three microtubules). Scale bar, 10 mm. See Movie S2.

(C) In Patronin-depleted cells, microtubules (arrows) release and move away from the centrosome (prophase cell). Scale bar, 5 mm. See Movie S2.

(D) In cells coexpressing EB1-GFP (green) and mCherry-tubulin (red), EB1 localizes to the leading end of moving microtubules (arrows), indicating that this is the

microtubule plus end. See Movie S3. Brightness was adjusted in each color channel separately. Scale bar, 5 mm. See also Figure S1 and Figure S2.


more rapid rate that can lead to complete microtubule disap-

pearance and may account for the sparser microtubule network

after Patronin RNAi.

Depletion of the Kinesin-13 Microtubule Depolymerase,

Klp10A, Suppresses the Patronin Phenotype

in Interphase and Mitosis

The above results reveal that Patronin protects the microtubule

minus end against depolymerization in vivo. We next wanted to

determine if the minus end depolymerization in Patronin-

depleted cells was an intrinsic property of the minus end or

whether another protein was actively involved. Kinesin-13s are

microtubule depolymerizers that localize to both plus and minus

ends in vitro, and in vivo they bind to microtubule plus ends

during interphase and promote their depolymerization (Desai

et al., 1999; Hunter et al., 2003; Mennella et al., 2005). Kinesin-

13s also promote the poleward flux of tubulin subunits toward

the spindle pole during mitosis, a process that involves minus

end tubulin turnover (Kwok and Kapoor, 2007; Rogers et al.,

2004). To determine whether a Drosophila Kinesin-13 family

member is involved in depolymerizing the microtubule minus

ends after Patronin depletion, we performed double RNAi of Pa-

tronin with the three Drosophila Kinesin-13s (Klp10A, Klp59C,

Klp59D) and examined the effect on interphase microtubule

dynamics. Strikingly, codepletion of Klp10A rescued the Pa-

tronin RNAi phenotype; the microtubule array was denser and

free microtubules were no longer observed in the majority of

the cells (Figures 2B and 2C). In contrast, double RNAi of either

Klp59C or Klp59Dwith Patronin did not affect the number of free,

treadmilling microtubules (Figure 2C). When a rare, free microtu-

bule was found in a Patronin and Klp10A codepleted cell, the

minus end either remained stationary or appeared to grow,

0 s

210 s

150 s

60 s

Patronin RNAi Patronin + KLP10A RNAi

Klp10A-GFP

mCh-Tubulin

B

D 0 s 6 s

% C

ells w

ith >

5 free

MT

C

pre-bleachWildtype

RNAi:

0 s

3 s

6 s

9 s

3 s E Klp10A-GFP

EB1-mCh

0

20

40

60

80

100

WT Pat Klp10a Klp59C Klp59D Pat +Klp10a Pat +Klp59C Pat +Klp59D

A Patronin RNAi pre-bleach

Figure 2. Free Microtubules Move by Klp10A-Mediated Treadmilling in Patronin-Depleted Cells

(A) Photobleaching amark in the middle of moving microtubules in Patronin RNAi cells reveals that the bleach mark is stationary and the trailing minus endmoves

toward the bleach mark (see arrows) (n = 20). This indicates that the apparent motion of microtubules occurs through simultaneous tubulin polymerization at

the plus end and depolymerization at the minus end. In wild-type cells, the bleach mark in a rare free microtubule remains stationary relative to the minus

end, indicating that it is neither polymerizing nor depolymerizing (n = 10). Scale bars, 5 mm.

(B) Comparison of GFP-tubulin cells depleted of Patronin alone or both Patronin and Klp10A. Cells codepleted of Patronin and Klp10A have a wild-type-like

microtubule network and rarely have free microtubules. Scale bar, 10 mm.

(C) Quantitation of the percentage of cells with >5 free microtubules shows that codepletion of Patronin and Klp10A, but not Klp59C or Klp59D, rescues the

Patronin RNAi phenotype. The mean and SEM are shown from two independent experiments (n = 200 cells per experiment).

(D) In Patronin-depleted cells coexpressing Klp10A-GFP (green) and mCherry-tubulin (red), Klp10A localizes to and tracks along the depolymerizing minus ends

of treadmilling microtubules (arrows). Scale bar, 5 mm. See Movie S3.

(E) In Patronin-depleted cells coexpressing Klp10A-GFP (green) and EB1-mCherry (red), Klp10A localizes to the trailing end (arrows), whereas EB1 localizes to the

leading ends of treadmilling free microtubules (frame from a time-lapse sequence). Scale bar, 5 mm. Brightness was adjusted in each color channel separately.

See also Figure S1 and Figure S2.


resulting in an increase in microtubule length. Interestingly, EB1-

GFP localized to both ends of these growing microtubules,

although it appeared more abundant at the presumed plus end

at the cell periphery (Figure S2B). We also occasionally observed

a transient localization of EB1-GFP at free microtubule minus

ends in cells depleted of Patronin alone, which was accompa-

nied by a pause in minus end depolymerization and a brief

increase in microtubule length (Figure S2B). This, to our knowl-

edge, is the first observation of in vivo minus end polymerization.

We also examined the interphase localization of Klp10A-GFP

in Patronin-depleted cells. Previous studies showed that

Klp10A-GFP localizes to microtubule plus ends prior to their

catastrophe/depolymerization; the loading of Klp10A to growing

plus ends is mediated by an interaction with EB1 (Mennella et al.,

2005). In Patronin-depleted cells, we observed a prominent

puncta of Klp10A-GFP tracking along the depolymerizing minus

ends of treadmilling microtubules (Figure 2D, Movie S3). In cells

coexpressing Klp10A-GFP and EB1-mCherry, we found that

Klp10A is concentrated at the depolymerizing minus end,

whereas EB1 is at the growing plus end (Figure 2E). These local-

ization data support the conclusions of the Klp10A rescue exper-

iments, indicating that Klp10A is actively depolymerizing minus

ends in the absence of Patronin, and suggest that this minus

end localization is not dependent on EB1.

We next examinedwhether Klp10A is involved in producing the

short spindle phenotypeobserved after Patronin depletion (Gosh-

ima et al., 2007). Wild-type spindles have a pole-to-pole length of

10.1 ± 1.7 mm (mean ± SD), which was reduced to 6.1 ± 1.3 mm

after Patronin depletion (Figures 3A and 3B). A similar reduction

was observed in an acentrosomal mitotic spindle produced by

centrosomin (Cnn) RNAi (Li and Kaufman, 1996) (9.6 ± 1.9 mm in

Cnn RNAi cells and 6.7 ± 1.3 mm after Cnn/Patronin double

RNAi [n = 35]), suggesting that Patronin’s function is not limited

to thecentrosome. Interestingly,weobserved twodistinct classes

of short, bipolar spindles after Patronin RNAi: one in which the

spindle had normal morphology with a clearly alignedmetaphase

plate, and another where the spindle appeared ‘‘collapsed’’ and

the bipolar array penetrated across the metaphase plate

(FigureS3B).CodepletionofKlp10AandPatronin restorednormal

morphology (Figures 3A and 3B) and produced longer spindles

(12.4 ± 2.6 mm) than those in wild-type cells, a length comparable

to Klp10A depletion alone (11.2 ± 2.2 mm). Conversely, codeple-

tion of Klp59C or Klp59D and Patronin produced shorter spindles

than wild-type cells (Figure S3A). These results suggest that Pa-

tronin protects microtubule minus ends against Klp10A-induced

depolymerization during mitosis and that the balance of counter-

acting stabilizing and destabilizing forces at the minus ends

governs spindle length (see Discussion).

Poleward flux of tubulin subunits during metaphase has been

associated with minus end depolymerization by Klp10A and

linked to the regulation of spindle length; less poleward flux

results in longer spindle length and vice versa (Rath et al.,

2009). Depletion of Patronin resulted in an increased flux

(2.03 ± 0.06 mm/min) over wild-type (1.44 ± 0.28 mm/min), thus

explaining the shorter spindle. As previously reported, Klp10A

RNAi caused a dramatic reduction in flux (0.68 ± 0.09 mm/min)

(Laycock et al., 2006; Rath et al., 2009). Codepletion of Patronin

and Klp10A produced a flux (0.66 ± 0.03 mm/min) similar to

Klp10A alone (Figure 3C), thus explaining the long spindle

phenotype.

Taken together, our results suggest that Klp10A is actively

depolymerizing free microtubule minus ends in interphase and

mitosis and that the presence of Patronin is able to suppress

this depolymerization activity.

GFP-Patronin Localizes to Microtubule Nucleation

Centers

To learn more about Patronin’s functions, we determined its

intracellular localization. A polyclonal antibody made against

the C-terminal region of Patronin, although having considerable

background staining, showed that endogenous protein localizes

to centrosomes in prophase, the midbody during cytokinesis,

throughout the metaphase spindle, and to punctae in interphase

that often overlap with microtubules (Figure S4C).

A GFP-Patronin fusion protein, which rescued the Patronin

phenotype and thus is functional (Figure S4A), localized in punc-

tae alongmicrotubules in interphase, bundling them at moderate

to high expression levels, and localized throughout the mitotic

Table 1. Quantitation of Dynamic Instability Parameters in Wild-

Type and Patronin-Depleted GFP-Tubulin Cells

Wild-Type Patronin RNAi

Microtubule Plus End

Growth (mm/min) 3.58 ± 1.10 4.22 ± 1.31

Shrinkage (mm/min) 10.21 ± 2.12 10.93 ± 1.56

Catastrophe (min 1) 0.12 ± 0.06 0.11 ± 0.05

Rescue (min 1) 0.16 ± 0.08 0.21 ± 0.08

Microtubule Minus End

Shrinkage I (mm/min) 0.01 ± 0.07* 3.93 ± 0.87

Shrinkage II (mm/min) N.D. 10.20 ± 2.21

Polymerization and depolymerization rates were measured for 25 indi-

vidual microtubules (per type of measurement) from 8–16 cells over three

different experiments. The number reported is the mean and SD from the

25 measurements. Polymerization and depolymerization rates were

measured by kymograph analysis using ImageJ. For Patronin RNAi cells,

‘‘free’’ microtubules were measured (both ends clearly visualized, see

Movie S2). The exception (noted by an *) is the minus end dynamics in

wild-type cells. Because of the high degree of stability and possible

movement of the microtubule in the wild-type cytoplasm over long

measurement times, we measured the microtubule minus end relative

to a photobleach mark as in Figure 2 (n = 10); the value shown is within

the error of our measurement and indicates that the minus end is very

stable. ‘‘N.D.,’’ indicates that a second rate was not detected. A compa-

rable measurement of a minus end relative to a bleach mark in Patronin

RNAi cells yielded two shrinkage rates (3.21 ± 0.31 and 10.81 ± 0.94;

n = 20 for each rate), similar to that observed for tracking the minus

end in microtubules without photobleach marks (shown in the table).

The microtubules scored for this table exhibited a single, constant minus

end shrinkage rate. However, these two different rates of minus end

shrinkage occasionally were observed for individual microtubules

(Figure S1E). Catastrophe and rescue frequencies were calculated for

10 cells per condition. In each cell, 10 microtubules were observed and

the frequency of catastrophe and rescue calculated over the course of

3 min. The number reported is the mean and SD of the frequencies

calculated for each cell.


spindle (Figure 4A and Figure S4B). We also examined the

localization of Patronin’s three major domains: an N-terminal

calponin homology domain (CH), a middle domain containing

three predicted coiled-coils (CC), and a C-terminal microtu-

bule-binding domain (CKK domain; Baines et al., 2009). The

CH domain appeared diffuse throughout the cytoplasm (Fig-

ure S5A), whereas the CKK domain localized along all microtu-

bules as previously reported (Baines et al., 2009) (Figure 4A).

Interestingly, the central CC domain localized to small microtu-

bule-nucleating foci (Figure 4A and Figure S5F) and occasionally

along short stretches of microtubules.

We used the microtubule-depolymerizing drug colcemid to

visualize GFP-Patronin during the depolymerization and refor-

mation of the microtubule cytoskeleton. After complete microtu-

bule depolymerization, small foci containing both GFP-Patronin

and mCherry-tubulin were observed throughout the cytoplasm

(data not shown). Sas-4 and g-tubulin, established markers of

microtubule-nucleating centers, localized to similar foci in

GFP-tubulin cells under the same conditions (data not shown).

In the initial phase of microtubule regrowth, microtubules elon-

gated out from these foci, eventually reforming the interphase

microtubule array (Figure 4B). Therefore Patronin localizes to

sites of new microtubule formation. A connection between Pa-

tronin and microtubule-nucleating centers was also suggested

by coexpression studies of mCherry-Patronin with GFP-Sas-4

(Figures 4C and 4D) or GFP-SAK (Figures S5B and S5C). GFP-

Sas-4 and SAK normally are distributed as discrete cytoplasmic

punctae (Figure 4C and Figure S5B). However, when full-length

Patronin (Figure 4D and Figure S5C) or its CC domain

(Figure S5F) were overexpressed along with Sas-4 and SAK,

these proteins colocalized with Patronin. However, Sas-6, a cen-

triolar protein, did not colocalize with Patronin (Figures S5D and

S5E). Thus, Patronin may directly or indirectly interact with

a subset of proteins associated with microtubule-nucleating

centers.

Purified Patronin Specifically Binds to and Protects

Microtubule Minus Ends against Depolymerization

In Vitro

Our in vivo studies revealed that Patronin stabilizes microtubule

minus ends and protects them against Kinesin-13 depolymeriza-

tion. To determine whether Patronin alone is sufficient for such

protection, we expressed and purified full-length GFP-Pa-

tronin-6xHis (224 kDa) from baculovirus-infected Sf9 cells

(Figure 5A) to test its activity in vitro.

We first wanted to establish how Patronin interacted with

microtubules made from purified tubulin. We attached GFP-

Patronin to a coverslip using a surface-adsorbed anti-GFP anti-

body and then added GMP-CPP-stabilized, rhodamine-labeled

microtubules. Strikingly, the microtubules attached to the cover-

slip by only one end, resulting in filaments that swiveled in

space while anchored at a single point (Figure 5B, Movie S4).

In most cases, a clear spot of GFP-Patronin colocalized with

the anchored end of the microtubule (asterisks, Figure 5B).

Microtubules did not bind to the coverslip surface in the absence

of Patronin and attached along their length when bound by

anti-tubulin antibody or kinesin (data not shown). To determine

if Patronin preferentially bound to the microtubule plus or minus

end, microtubule gliding was induced by introducing kinesin or

dynein to the assay. With kinesin, the Patronin-bound end

became the leading end as kinesin moved the microtubule

across the glass (128 out of 130 preanchored microtubules

exhibited this polarity) (Figure 5C). The leading ends of gliding

microtubules also frequently stopped, presumably due to re-

binding to Patronin, causing the microtubule to buckle due to

the pushing force of kinesin (asterisk in Figure 5C, Movie S5).

Conversely, when dynein was added, the Patronin-bound end

now became the trailing end of the gliding microtubule (138

out of 139 microtubules) (Figure 5C, Movie S5). These results

show that Patronin binds highly selectively to the microtubule

minus end in vitro.

Wildtype Patronin + KLP10A RNAiPatronin RNAiA

B

Spind

le siz

e (µm

)

KLP10A RNAi

RNAi:

Spind

le flu

x (µm

/min)

C

RNAi:0

4

8

12

16

Wildtype Patronin Klp10A Patronin +Klp10A0

0.51

1.52

2.5

Wildtype Patronin Klp10A Patronin +Klp10A

Figure 3. Depletion of Klp10A Suppresses

the Patronin Phenotype in Mitosis

(A) Codepletion of Patronin and Klp10A rescues

the short spindle phenotype observed in Pa-

tronin-depleted cells and results in elongated

spindles similar to those seen in Klp10A-depleted

cells. Scale bar, 10 mm.

(B) The mean pole-to-pole metaphase spindle

length under each condition was quantified for

two independent experiments (n > 60 spindles

per condition; error bar, SEM; p < 0.001 for each

reported condition).

(C) The flux of tubulin toward the spindle poles was

measured by photobleaching an 1 mm stripe in

the GFP-tubulin spindle and tracking its move-

ment. The mean flux rates were quantified under

each condition from two independent experiments

(n = 20 spindles per condition; error bar, SEM;

p < 0.001 for each reported condition except the

pair of Klp10A RNAi and Klp10A/Patronin RNAi

flux [p < 0.9]). Thus poleward flux is increased after

Patronin depletion and decreased belowwild-type

levels when Patronin and Klp10A are codepleted.

See also Figure S3.


To further confirm these conclusions, we sought to visualize

GFP-Patronin bound to the microtubule. In this assay, we first

attached kinesin or dynein to the coverslip and then added

GMP-CPP rhodamine-labeled microtubules along with purified

GFP-Patronin. By TIRF microscopy, GFP-Patronin most often

bound at only one end of the microtubule. With kinesin pushing

the microtubule, GFP-Patronin was on the leading microtubule

end (of 84 microtubules with bound GFP-Patronin, 80 had a

GFP-Patronin spot at the minus end, 1 was at the plus end,

and 3 appeared internal) (Figure 5D, Movie S6). With dynein

transporting the microtubule, GFP-Patronin was at the trailing

end (of 101 microtubules with bound GFP-Patronin, 91 were at

CH CC CKK

BGFP-PatroninmCh-Tubulin

A GFP-Patronin Domain mCh-Tubulin Merge

GFP

1a.a. 1689

CCGFP

535a.a. 1457

CKKGFP

1447a.a. 1689

34 m 49 m 55 m 64 m

C GFP-Sas-4mCh-Patronin MergeGFP-Sas-4 D

Figure 4. GFP-Patronin Localization and Domain Analysis

(A) Coexpression of GFP-fusions of full-length Patronin (TIRF microscopy) or Patronin domains with mCherry-tubulin (merge: GFP-Patronin in green and

mCherry-tubulin in red). Localization patterns are discussed in the text. Scale bars, 10 mm.

(B) Time-lapse microscopy of GFP-Patronin (green) and mCherry-tubulin (red) expressing cells regrowing their microtubule network after washout of the micro-

tubule-depolymerizing drug colcemid (time after washout is indicated). The inserts correspond to the box at 34 min. Patronin and tubulin localize to small foci,

which serve as points of microtubule nucleation during the reformation of the cytoskeleton.

(C) Cells expressing GFP-Sas-4 alone form cytoplasmic foci, but when GFP-Sas-4 is coexpressed with mCherry-Patronin (D), Sas-4 is recruited to sites of

mCherry-Patronin along microtubules. Brightness was adjusted in each color channel separately in the merged images.

See also Figure S4 and Figure S5.


the minus end, 4 were at the plus end, and 6 appeared internal)

(Figure 5E,Movie S6). Thus, by direct observation, GFP-Patronin

binds selectively to the minus end.

We next used a reconstituted assay with purified MCAKmotor

domain from P. falciparum (P.f. MCAK) (homolog of Klp10A,

Moores et al., 2002) to test whether purified Patronin is sufficient

to protect minus ends from Kinesin-13-induced depolymeriza-

tion (Figures 6A and 6B). GMP-CPP polarity-marked microtu-

bules were adhered to the coverslip via anti-rhodamine

antibody, and P.f. MCAK was added in the presence or absence

of Patronin. Without Patronin, both ends of the microtubule

depolymerized (plus end: 2.5 ± 0.4 mm/min, minus end: 1.8 ±

0.7 mm/min; comparable to rates reported previously in vitro)

(Hunter et al., 2003; Desai et al., 1999; Cooper et al., 2009). In

the presence of purified Patronin, however, depolymerization

from the plus end still occurred (2.2 ± 0.3 mm/min) whereas

depolymerization from the minus end was negligible (0.01 ±

0.06 mm/min) (Figures 6A and 6B, Movie S7). We also observed

selective minus end stabilization with Patronin and full-length

hamster MCAK (C.g. MCAK) (Figure 6C). Higher concentrations

of P.f. or C.g. MCAK lead to the depolymerization of someminus

ends, suggesting that there is a competition between Patronin

and MCAK for minus end binding (Figure 6C). The full-length

MCAK competed more effectively than the motor domain, likely

because of its higher association rate (Cooper et al., 2009). We

also performed an alternative assay in which the microtubule

minus end was anchored to surface-adhered Patronin and

a solution of P.f. MCAK was added. Once again, MCAK depoly-

merized the plus end rapidly, whereas the Patronin-anchored

minus end did not shorten at our level of detection (Figures

S6A–S6C). In summary, our in vitro studies reveal that purified

Patronin binds selectively to the microtubule minus end and

this binding confers protection against Kinesin-13-induced

microtubule depolymerization.

DISCUSSION

Microtubule minus end dynamics has remained one of the least

well understood properties of the microtubule cytoskeleton.

Here, through in vivo and in vitro approaches, we have demon-

strated that Patronin binds with high selectivity to microtubule

minus ends and acts as a ‘‘cap,’’ stabilizing these ends and pro-

tecting them against the actions of microtubule depolymerases.

The consequence of losing Patronin-mediated capping in S2

cells is dramatic. During interphase, the microtubule density

decreases and microtubules released from nucleating sites

0 s 15 s 40 s

A

0 s 50 s 115 s

-+

kinesin

dynein

kinesin dyneinC 0 s 0 s

15 s

60 s

405 s

705 s

D

E

- +

B*

*

*

*

*

*

*

250

150

100

75

*

Figure 5. Purified Patronin Selectively

Binds to Microtubule Minus Ends In Vitro

(A) Purified GFP-Patronin-6xHis analyzed by SDS

polyacrylamide gel electrophoresis and stained

with Coomassie blue. Immunoblot analysis

reveals that lower band of the doublet is Patronin

lacking the GFP (not shown).

(B) When GFP-Patronin is attached to a coverslip

with anti-GFP antibody, it binds GMP-CPP-stabi-

lized, rhodamine-labeled microtubules by one

end. See Movie S4. Asterisks indicate the site of

microtubule anchoring, which often overlaps with

a GFP-Patronin spot. Scale bar, 10 mm.

(C) To reveal whichmicrotubule endwas anchored

to GFP-Patronin, kinesin or dynein was added

after microtubule anchoring. Arrows follow a

microtubule that was initially anchored by one

end and then bound along its length to the

motor-covered surface. With kinesin, the formerly

anchored end is leading (until the leading end reat-

taches and the microtubule buckles (asterisk, 60

s); with dynein, the formerly anchored end is trail-

ing. See Movie S5. These assays reveal that

microtubules are anchored to surface-bound Pa-

tronin selectively at their minus ends (see statistics

from three independent experiments in the text).

Scale bar, 5 mm.

Conventional kinesin (D) or dynein (E) microtubule-

gliding assays in the presence of GFP-Patronin

(6 nM; green) demonstrate that GFP-Patronin

binds selectively to the minus end. In the kinesin

assay, GFP-Patronin (green) is most frequently

observed at the leading ends of gliding microtu-

bules, whereas in the dynein assay, it resides at

the trailing ends. The results from three indepen-

dent experiments indicate that GFP-Patronin

binds selectively to the minus end. See Movie

S6. Scale bars, 10 mm. Brightness was adjusted

in each color channel separately.


treadmill through the cytoplasm. During mitosis, the spindle

becomes significantly shorter and in some cases collapses to

a shape that more resembles a monopolar spindle. In addition

to clarifying the role of Patronin, our studies also provide insight

into the regulation of microtubule minus end dynamics. We

demonstrate that minus ends are substrates for capping

(Patronin), destabilizing (Kinesin-13), and possibly growth-pro-

moting or -stabilizing (EB1) activities, as has been demonstrated

for themicrotubule plus end. The behavior of minus ends reflects

a net balance of these actions, which plays an important role in

the overall organization of the microtubule cytoskeleton.

Patronin Mechanism

Patronin binds with high selectivity to the minus end of microtu-

bules (>92% from our in vitro experiments), suggesting that it

recognizes some unique, exposed feature at this end. In the

polar microtubule, a-tubulin faces the minus end whereas

b-tubulin faces the plus end (Mitchison, 1993). Thus we specu-

late that Patronin recognizes features of a-tubulin that are

normally buried at the a/b interface but are exposed at the

end of the microtubule. Consistent with this possibility, an

anti-a-tubulin antibody has been produced that binds selectively

to the microtubule minus end (Fan et al., 1996). Interestingly,

selective minus end binding appears to require the cooperation

of multiple regions of the Patronin protein, as the C-terminal

CKK domain alone binds uniformly along the microtubule

surface (Figure 4A; Baines et al., 2009).

An important functional consequence of Patronin binding to

minus ends is protection against Kinesin-13 depolymerization.

Kinesin-13 destabilizes microtubule ends by bending microtu-

bule protofilaments, causing them to lose lateral interactions

(Moores and Milligan, 2006). Patronin might sterically block

Kinesin-13 binding and/or strengthen the lateral interactions of

protofilaments, rendering minus ends resistant to depolymeriza-

tion. A better understanding of how Patronin caps and protects

minus ends will require higher-resolution structural information

of the Patronin-microtubule minus end complex.

In addition to its cappingandprotecting activity, Patroninmight

act as a scaffolding protein at microtubule nucleation centers in

S2 cells. When full-length Patronin or the central coiled-coil

region is expressed in cells, they localize to foci that nucleate

microtubules. Overexpression of either of these constructs

results in the recruitment of Sas-4 and SAK, two proteins that

are associated with centrioles/centrosomes (Bornens, 2002).

These results raise questions of whether a scaffolding activity

of Patronin might be involved in minus end capping/protection

and possibly microtubule nucleation in vivo. Our in vitro data

showing that purified Patronin can protect the minus end

reveal that Patronin alone is sufficient for this activity, although

A

- +

No Patronin 0 s Patronin 0 s

120 s

160 s

240 s

00.5

11.5

22.5

33.5

Plus end Minus end Plus end Minus end

Depoly

meriza

tion r

ate

(µ

m/m

in)B

- Patronin + Patronin

- +

60 s

0

20

40

60

80

100

2 µMPf MCAK

20 µMPf MCAK

35 nM Cg MCAK

350 nMCg MCAK

2 µMPf MCAK

35 nMCg MCAK

C

% P

rote

cted M

inus

Ends

- Patronin+ 35 nM PatroninFigure 6. GFP-Patronin ProtectsMicrotubuleMinus Ends fromKine-

sin-13-Induced Depolymerization In Vitro

(A) Polarity-marked, GMP-CPP-stabilized rhodamine-labeled microtubules

were attached to the coverslip by an anti-rhodamine antibody. The minus

end is closest to the region of higher fluorescence intensity in the microtubule.

In the absence of Patronin, purified Kinesin-13 motor domain from

P. falciparum (3 mM) depolymerizes both ends of the microtubule. In contrast,

in the presence of GFP-Patronin (30 nM), Kinesin-13 only depolymerizes the

dimmer plus end (white arrows), whereas the minus end (yellow arrows) is

stable. See Movie S7. (Note: the higher concentration of Patronin precludes

imaging of individual Patronins at microtubule ends as in Figure 5.)

Scale bar, 10 mm.

(B) Quantitation of Kinesin-13-induced depolymerization rates at the plus and

minus ends (n = 30 microtubules for each condition; mean and SD). Data are

representative of three independent experiments with different microtubule

preparations.

(C) Patronin was mixed with the indicated concentration of either full-length

Kinesin-13 from hamster (C.g.) or the motor domain from P. falciparum (P.f.)

and added to polarity-marked microtubules. Minus ends were scored as pro-

tected if they showed no detectable depolymerization by the time the plus end

depolymerized by >50% of the microtubule length. Higher concentrations of

the Kinesin-13s are able to compete with Patronin to depolymerize a subset

of minus ends. Percentages are representative of two independent experi-

ments. See also Figure S6.


minus end capping might be more complex and augmented by

additional proteins within the cell. We thus far have not found

that purified Patronin stimulates microtubule nucleation from

purified tubulin in vitro, and the initial regrowth of microtubules

after colcemid washout was similar in wild-type and Patronin

RNAi cells (data not shown). However, the current experiments

cannot exclude some role in nucleation. Thus, possible roles of

Patronin as a scaffolding factor involved in the assembly of other

proteins at microtubule minus ends awaits further investigation.

Regulation of Microtubule Minus End Dynamics In Vivo

A large and still growing number of proteins have been discov-

ered that associate with microtubule plus ends and many exhibit

opposing effects on microtubule dynamics (Akhmanova and

Steinmetz, 2008; Howard and Hyman, 2007), which gives rise

to the high dynamicity of plus ends in vivo and enables cells to

rapidly restructure their microtubule cytoskeleton. The dynamics

of microtubuleminus ends in vivo has not been aswell studied as

that of plus ends, particularly at the level of single microtubules.

In the few studies where minus ends have been observed in

animal cells, they have been reported to bemostly stable (neither

growing or shrinking; Rodionov and Borisy, 1997; Yvon and

Wadsworth, 1997; Waterman-Storer and Salmon, 1997). Minus

end shrinkage and microtubule treadmilling, however, is

common in Arabidopsis (Shaw et al., 2003; Ehrhardt, 2008). In

contrast, microtubule minus ends composed of pure tubulin

grow, shorten, and exhibit dynamic instability (Desai and Mitch-

ison, 1997). The discrepancy between such in vitro dynamicity

and in vivo stability suggests the presence of a minus end

capping factor. g-TuRC interacts directly with the microtubule

minus end (Moritz et al., 1995; Zheng et al., 1995; Wiese and

Zheng, 2000). However, although g-TuRC has a clear role in

microtubule nucleation in vivo, it is uncertain whether it remains

bound to and stabilizes minus ends after the microtubule is

nucleated. Indeed, we feel that this may not be the case, at least

inDrosophila, as g-TuRC RNAi knockdown does not greatly alter

the appearance of the interphase array (Bouissou et al., 2009),

produces elongated rather than short mitotic spindles (Verollet

et al., 2006), and does not generate free, treadmilling microtu-

bules (Figure S1D and Figure S3A). Another protein, ninein, plays

a role in anchoring microtubules toMTOCs and other sites within

cells (Delgehyr et al., 2005); however this interaction appears to

be facilitated by g-TuRC, and ninein has not been shown yet to

interact directly with minus ends. RNAi of other genes that

produced a short spindle phenotype (Goshima et al., 2007) or

centrosomal proteins (Sas-4, SAK, Asp, and Cnn; data not

shown) did not give rise to a microtubule treadmilling phenotype

indicative of minus end instability. Thus, Patronin is the only

protein for which minus end capping activity has been demon-

strated in vivo.

Our experiments also demonstrate that in the absence of

Patronin-mediated capping, microtubule minus ends in vivo

exhibit the range of behaviors seen in vitro (polymerization,

depolymerization, catastrophe, and rescue) and also are acted

upon by previously identified plus end binding proteins. EB1

has been used as a canonical marker of microtubule plus

ends in vivo. Here, we show that EB1-GFP can interact with

microtubule minus ends during episodes of subunit addition

(Figure S2B). Kinesin-13, which binds to plus ends and induces

their catastrophe, has been suggested to depolymerize microtu-

bule minus ends during mitosis based upon its role in spindle

flux (Rogers et al., 2004) but has not been directly visualized at

microtubule minus ends. Here, we show that in the absence of

Patronin, the Kinesin-13 Klp10A-GFP binds to and tracks along

depolymerizing minus ends and is also required for this depoly-

merization (Figure 2). In Patronin-depleted cells, the actions of

Klp10A appear to dominate over any minus end growth-

promoting factors, as most microtubule minus ends undergo

depolymerization and only rarely display brief periods of growth.

In summary, microtubuleminus ends can grow, depolymerize, or

be capped in vivo and the balance of proteins that promote these

activities govern the behavior of microtubule minus ends in cells.

The importance of balancing stabilizing and destabilizing

activities on microtubule ends is illustrated in the mitotic spindle.

Net polymerization occurs at microtubule plus ends near the

kinetochore and net depolymerization occurs at minus ends at

the poles, resulting in a poleward flux of tubulin subunits within

the microtubule lattice (Kwok and Kapoor, 2007; Rogers et al.,

2004). The overall balance of polymerizing and depolymerizing

activities of microtubule-associated proteins governs the size

and shape of the spindle (Goshima et al., 2005; Dumont and

Mitchison, 2009). Studies in several organisms have implicated

Kinesin-13s as major regulators of mitotic microtubule length,

spindle size, and poleward flux (Mitchison et al., 2005; Rath

et al., 2009; Kwok and Kapoor, 2007). Our results suggest that

Patronin provides a ‘‘brake’’ rather than a full block on the minus

end depolymerizing actions of Kinesin-13. In the absence of Pa-

tronin, Kinesin-13 is unchecked, resulting in a higher flux rate and

shorter, sometimes collapsed spindles. With the depletion of

both Patronin and Kinesin-13, flux is low and spindle length is

longer than normal. These results imply that microtubule minus

ends are not completely protected by Patronin but are subject

to competing activities of Patronin and Kinesin-13, as we also

demonstrate in vitro (Figure 6C). Thus, a balance of Patronin

and Kinesin-13 actions on microtubules minus ends governs

the length of the mitotic spindle.

The Patronin Family and Minus End Capping

in Acentrosomal Microtubule Arrays

A single Patronin gene is found in invertebrate genomes and

clear homologs do not exist or are difficult to identify in nonme-

tazoan organisms. After Patronin (then named ssp4) was first

described in Drosophila (Goshima et al., 2007), three vertebrate

homologs with the same domain organization and sequence

identity were reported and have been called the CAMSAP/

ssp4 family of proteins (the three vertebrate branches are

referred to as CAMSAP1, CAMSAP2, and CAMSAP3; Baines

et al., 2009). All Patronin-related genes have a characteristic

domain organization of an N-terminal CH domain, a long central

domain with interspersed predicted coiled-coil regions, and a

C-terminal microtubule-binding domain (termed the CKK

domain), which is the most highly conserved region of the poly-

peptide (Baines et al., 2009). While this work was in progress,

vertebrate CAMSAP1 and a CAMSAP3 member, Nezha, were

reported to interact with microtubules (Baines et al., 2009;

Meng et al., 2008). Meng et. al. (2008) found that Nezha localizes


specifically at microtubuleminus ends located close to adherens

junctions in epithelial cells and bound preferentially to the minus

end in vitro (67% of microtubule-associated Nezha). However,

their study did not explore whether Nezha affected the dynamics

of microtubules or influenced the organization of the microtubule

cytoskeleton. If the vertebrate homologs also are found to

protect microtubule minus ends as shown here for Drosophila

Patronin, we suggest that the currently named CAMSAP/ssp4

family be renamed as the Patronin family, retaining the phyloge-

netic classificationof the threevertebratebranches (Patronin 1, 2,

and 3) (Baines et al., 2009).

Minus end capping has been proposed to be particularly

important for the formation and organization of nonradial, acen-

trosomal interphase microtubule arrays (Dammermann et al.,

2003; Bartolini and Gundersen, 2006). The roles of the three

Patronin family members in vertebrates are not yet defined, but

they may have evolved to interact with distinct partners for local-

izing microtubule minus end capping/anchoring activities to

distinct subcellular regions in epithelial cells (Meng et al.,

2008), neuronal cells (Berglund et al., 2008), and other cells

with acentrosomal arrays. Thus, the three Patronin family

members might provide new molecular tools for probing the

organization and function of microtubules in different vertebrate

cell types.


Cell Culture and RNAi

Drosophila S2 cells (UCSF) were cultured and incubated with dsRNA as previ-

ously described (Goshima and Vale, 2003). Unless noted, cells were treated

with dsRNA for 4 days and when indicated were treated with additional dsRNA

at day 4 and analyzed at day 8. Plasmids and cell lines are described in the

Extended Experimental Procedures.

Live-Cell Imaging

Cells were plated on Con A (Sigma) coated MatTek dishes for 1 hr unless

noted. Live-cell imaging was performed by spinning disk confocal microscopy

or occasionally by TIRF microscopy (noted in the legends). Microscope

equipment is described in the Extended Experimental Procedures. For the

photobleaching experiments, GFP-tubulin cells were imaged on an LSM 510

or 710 (Carl Zeiss, Inc.) (633 1.4 NA objective). Two or three imaging scans

were performed with a 488 nm laser at 1.1% power before a selected area

was bleached. On the LSM 510, bleaching was achieved with a 488 nm Argon

laser at 100% laser power for four iterations, while on the LSM710 a 405 nm

laser at 45% power was used for two iterations. After the photobleach, scans

were taken at 488 nm (1.1% power) every 3 s. The position of a bleach mark

relative to microtubule ends or within a spindle (flux measurements) was

measured over time using ImageJ.

In Vitro Assays

GFP-Patronin with aC terminus 6xHIS tagwas expressed using the BaculoDir-

ect system (Invitrogen). Sf9 cells were infected with P3 virus for 3 days and

harvested. GFP-Patronin-6xHis was purified on a NiNTA column (QIAGEN);

the eluted protein was dialyzed overnight into 50 mM Tris-HCl (pH 8),

150 mM KAcetate, 1 mM DTT and 10% glycerol and stored in LN2.

Flow cells were used for all in vitro assays. For the anchoring assay, anti-

GFP antibody was adhered to the coverslip and 150 nM GFP-Patronin was

added for 5 min. Coverslips were blocked with 1 mg/ml casein solution, after

which a solution of GMP-CPP stabilized rhodamine-microtubules (see

Extended Experimental Procedures), an oxygen scavengingmixture (catalase,

glucose oxidase, and glucose), and 1 mg/ml casein in BRB80 was added

(referred to as the ‘‘microtubule solution’’). To determine the polarity of the

anchored microtubule, the experiment was repeated with the following

changes: a mixture of anti-GFP and anti-GST antibody was adhered to the

coverslip, and after microtubules were anchored by Patronin, K560 kinesin

(Woehlke et al., 1997) or GST-D4.4 dynein (Reck-Peterson et al., 2006), an

oxygen scavenger mix, and 5 mM ATP was added.

For the motility assays, a coverslip with immobilized K560 kinesin or

GST-D4.4 dynein (via anti-GST) was blocked with 1 mg/ml casein and the

microtubule mixture plus 6 nM GFP-Patronin and 5 mM ATP was added.

For the Kinesin-13 depolymerization assay, polarity-marked GMP-CPP

rhodamine microtubules (See Extended Experimental Procedures) were

anchored to the coverslip with an anti-rhodamine antibody. The indicated

concentration of Kinesin-13 (either theMCAKmotor domain fromP. falciparum

[purified as described in Moores et al., 2002] or full-length hamster MCAK

obtained from Linda Wordeman [Cooper et al., 2009]) was added with 5 mM

ATP in BRB80 with an oxygen scavenger mix. Images were taken at 20 s

intervals on the TE2000U Nikon microscope using a 403 1.3 NA objective

and Nikon intensilight. Microtubule lengths were measured using ImageJ

software.


Supplemental Information includes Extended Experimental Procedures,

six figures, and seven movies and can be found with this article online at

doi:10.1016/j.cell.2010.09.022.

ACKNOWLEDGMENTS

We thank E. Griffis, N. Stuurman, and G. Goshima for guidance, discussions,

and advice. G. Goshima,M. Sirajuddin, A. Carter, A. Yildiz, and A. Karunakaran

contributed reagents. L. Wordeman and M. Wagenbach (U. of Washington)

generously provided full-length MCAK, and J. Raff (U. of Cambridge) kindly

provided a Sas-4 antibody. We thank the Physiology Course and the Cell

Division Group at the MBL, Woods Hole for helpful discussions.

Received: March 12, 2010




REFERENCES

Akhmanova, A., and Steinmetz, M. (2008). Tracking the ends: a dynamic

protein network controls the fate of microtubule tips. Nat. Rev. Mol. Cell.

Biol. 9, 309–322.

Baines, A.J., Bignone, P.A., King, M.D., Maggs, A.M., Bennett, P.M., Pinder,

J.C., and Phillips, G.W. (2009). The CKK domain (DUF1781) domain binds

microtubules and defines the CAMSAP/ssp4 family of animal proteins. Mol.

Biol. Evol. 26, 2005–2014.

Bartolini, F., and Gundersen, G.G. (2006). Generation of noncentrosomal

microtubule arrays. J. Cell Sci. 119, 4155–4163.

Berglund, L., Bjorling, E., Oksvold, P., Fagerberg, L., Asplund, A., Al-Khalili

Szigyarto, C., Persson, A., Ottosson, J., Wernerus, H., Nilsson, P., et al.

(2008). A genecentric Human Protein Atlas for expression profiles based on

antibodies. Mol. Cell. Proteomics 7, 2019–2027.

Bornens, M. (2002). Centrosome composition and microtubule anchoring

mechanisms. Curr. Opin. Cell Biol. 14, 25–34.

Bouissou, A., Verollet, C., Sousa, A., Sampaio, P., Wright, M., Sunkel, C.E.,

Merdes, A., and Raynaud-Messina, B. (2009). {gamma}-Tubulin ring com-

plexes regulate microtubule plus end dynamics. J. Cell Biol. 187, 327–334.

Cooper, J.R., Wagenbach, M., Asbury, C.L., and Wordeman, L. (2009).

Catalysis of the microtubule on-rate is the major parameter regulating the

depolymerase activity of MCAK. Nat. Struct. Mol. Biol. 17, 77–82.

Dammermann, A., Desai, A., and Oegema, K. (2003). The minus end in sight.

Curr. Biol. 13, R614–R624.


Delgehyr, N., Sillibourne, J., and Bornens, M. (2005). Microtubule nucleation

and anchoring at the centrosome are independent processes linked by ninein

function. J. Cell Sci. 118, 1565–1575.

Desai, A., and Mitchison, T.J. (1997). Microtubule polymerization dynamics.

Annu. Rev. Cell Dev. Biol. 13, 83–117.

Desai, A., Verma, S., Mitchison, T.J., and Walczak, C.E. (1999). Kin I kinesins

are microtubule-destabilizing enzymes. Cell 96, 69–78.

Dumont, S., and Mitchison, T.J. (2009). Force and length in the mitotic spindle.

Curr. Biol. 19, R749–R761.

Ehrhardt, D.W. (2008). Straighten up and fly right: microtubule dynamics and

organization of non-centrosomal arrays in higher plants. Curr. Opin. Cell

Biol. 20, 107–116.

Fan, J., Griffiths, A.D., Lockhart, A., Cross, R.A., and Amos, L.A. (1996).

Microtubule minus ends can be labelled with a phage display antibody specific

to alpha-tubulin. J. Mol. Biol. 259, 325–330.

Goshima, G., and Vale, R.D. (2003). The roles of microtubule-based motor

proteins in mitosis: comprehensive RNAi analysis in the Drosophila S2 cell

line. J. Cell Biol. 162, 1003–1016.

Goshima, G., Wollman, R., Stuurman, N., Scholey, J.M., and Vale, R.D. (2005).

Length control of the metaphase spindle. Curr. Biol. 15, 1979–1988.

Goshima, G., Wollman, R., Goodwin, S.S., Zhang, N., Scholey, J.M., Vale,

R.D., and Stuurman, N. (2007). Genes required for mitotic spindle assembly

in Drosophila S2 cells. Science 316, 417–421.

Howard, J., and Hyman, A.A. (2007). Microtubule polymerases and depoly-

merases. Curr. Opin. Cell Biol. 19, 31–35.

Hunter, A.W., Caplow, M., Coy, D.L., Hancock, W.O., Diez, S., Wordeman, L.,

and Howard, J. (2003). The kinesin-related protein MCAK is a microtubule

depolymerase that forms an ATP-hydrolyzing complex at microtubule ends.

Mol. Cell 11, 445–457.

Keating, T.J., Peloquin, J.G., Rodionov, V.I., Momcilovic, D., and Borisy, G.G.

(1997). Microtubule release from the centrosome. Proc. Natl. Acad. Sci. USA

94, 5078–5083.

Kinoshita, K., Arnal, I., Desai, A., Drechsel, D.N., and Hyman, A.A. (2001).

Reconstitution of physiological microtubule dynamics using purified

components. Science 294, 1340–1343.

Kwok, B.H., and Kapoor, T.M. (2007). Microtubule flux: drivers wanted. Curr.

Opin. Cell Biol. 19, 36–42.

Laycock, J.E., Savoian, M.S., and Glover, D.M. (2006). Antagonistic activities

of Klp10A and Orbit regulate spindle length, bipolarity and function in vivo.

J. Cell Sci. 119, 2354–2361.

Li, K., and Kaufman, T. (1996). The homeotic target gene centrosomin encodes

an essential centrosomal component. Cell 85, 585–596.

Meng, W., Mushika, Y., Ichii, T., and Takeichi, M. (2008). Anchorage of

microtubule minus ends to adherens junctions regulates epithelial cell-cell

contacts. Cell 135, 948–959.

Mennella, V., Rogers, G.C., Rogers, S.L., Buster, D.W., Vale, R.D., and Sharp,

D.J. (2005). Functionally distinct kinesin-13 family members cooperate to

regulate microtubule dynamics during interphase. Nat. Cell Biol. 7, 235–245.

Mitchison, T.J. (1993). Localization of an exchangeable GTP binding site at the

plus end of microtubules. Science 261, 1044–1047.

Mitchison, T.J., Maddox, P., Gaetz, J., Groen, A., Shirasu, M., Desai, A.,

Salmon, E.D., and Kapoor, T.M. (2005). Roles of polymerization dynamics,

opposed motors, and a tensile element in governing the length of Xenopus

extract meiotic spindles. Mol. Biol. Cell 16, 3064–3076.

Moores, C.A., and Milligan, R.A. (2006). Lucky 13-microtubule depolymerisa-

tion by kinesin-13 motors. J. Cell Sci. 119, 3905–3913.

Moores, C.A., Yu, M., Guo, J., Beraud, C., Sakowicz, R., and Milligan, R.A.

(2002). A mechanism for microtubule depolymerization by KinI kinesins. Mol.

Cell 9, 903–909.

Moritz, M., Braunfeld, M., Sedat, J., Alberts, B., and Agard, D. (1995).

Microtubule nucleation by gamma-tubulin-containing rings in the centrosome.

Nature 378, 638–640.

Rath, U., Rogers, G.C., Tan, D., Gomez-Ferreria, M.A., Buster, D.W., Sosa,

H.J., and Sharp, D.J. (2009). The Drosophila Kinesin-13, KLP59D, impacts

pacman and flux-based chromosome movement. Mol. Biol. Cell 20,

4696–4705.

Reck-Peterson, S.L., Yildiz, A., Carter, A.P., Gennerich, A., Zhang, N., and

Vale, R.D. (2006). Single-molecule analysis of dynein processivity and

stepping behavior. Cell 126, 335–348.

Rodionov, V., Nadezhdina, E., and Borisy, G. (1999). Centrosomal control of

microtubule dynamics. Proc. Natl. Acad. Sci. USA 96, 115–120.

Rodionov, V.I., and Borisy, G.G. (1997). Microtubule treadmilling in vivo.

Science 275, 215–218.

Rogers, G.C., Rogers, S.L., Schwimmer, T.A., Ems-McClung, S.C., Walczak,

C.E., Vale, R.D., Scholey, J.M., and Sharp, D.J. (2004). Two mitotic kinesins

cooperate to drive sister chromatid separation during anaphase. Nature 427,

364–370.

Rogers, G.C., Rusan, N.M., Peifer, M., and Rogers, S.L. (2008). A multicompo-

nent assembly pathway contributes to the formation of acentrosomal microtu-

bule arrays in interphase Drosophila cells. Mol. Biol. Cell 19, 3163–3178.

Rusan, N.M., and Rogers, G.C. (2009). Centrosome function: Sometimes less

is more. Traffic 10, 472–481.

Shaw, S.L., Kamyar, R., and Ehrhardt, D.W. (2003). Sustained microtubule

treadmilling in Arabidopsis cortical arrays. Science 300, 1715–1718.

Verollet, C., Colombie, N., Daubon, T., Bourbon, H.M., Wright, M., and

Raynaud-Messina, B. (2006). Drosophila melanogaster gamma-TuRC is

dispensable for targeting gamma-tubulin to the centrosome and microtubule

nucleation. J. Cell Biol. 172, 517–528.

Vorobjev, I., Svitkina, T., and Borisy, G. (1997). Cytoplasmic assembly of

microtubules in cultured cells. J. Cell Sci. 110, 2635–2645.

Vorobjev, I.A., Rodionov, V.I., Maly, I.V., and Borisy, G.G. (1999). Contribution

of plus and minus end pathways to microtubule turnover. J. Cell Sci. 112,

2277–2289.

Waterman-Storer, C., and Salmon, E. (1997). Actomyosin-based retrograde

flow of microtubules in the lamella of migrating epithelial cells influences

microtubule dynamic instability and turnover and is associated with microtu-

bule breakage and treadmilling. J. Cell Biol. 139, 417–434.

Wiese, C., and Zheng, Y. (2000). A new function for the gamma-tubulin ring

complex as a microtubule minus-end cap. Nat. Cell Biol. 2, 358–364.

Woehlke, G., Ruby, A.K., Hart, C.L., Ly, B., Hom-Booher, N., and Vale, R.D.

(1997). Microtubule interaction site of the kinesin motor. Cell 90, 207–216.

Yvon, A.M., and Wadsworth, P. (1997). Non-centrosomal microtubule

formation andmeasurement of minus endmicrotubule dynamics in A498 cells.

J. Cell Sci. 110, 2391–2401.

Zheng, Y., Wong, M., Alberts, B., and Mitchison, T. (1995). Nucleation of

microtubule assembly by a gamma-tubulin-containing ring complex. Nature

378, 578–583.


Structural Basis for Actin Assembly,Activation of ATP Hydrolysis,and Delayed Phosphate ReleaseKenji Murakami,1,6 Takuo Yasunaga,3 Taro Q.P. Noguchi,4 Yuki Gomibuchi,1 Kien X. Ngo,4 Taro Q.P. Uyeda,4

and Takeyuki Wakabayashi1,2,5,*1Department of Biosciences, School of Science and Engineering2Department of Judo Therapy, Faculty of Medical Technology

Teikyo University, Toyosatodai 1-1, Utsunomiya 320-8551, Japan3Department of Bioscience and Bioinformatics, Faculty of Computer Science and Systems Engineering, Kyushu Institute of Technology,

Ooaza-kawazu 680-4, Lizuka, Fukuoka 820-850, Japan4Biomedical Research Institute, National Institute of Advanced Industrial Science and Technology, AIST Tsukuba Central 4,

1-1-1 Highashi, Tsukuba, Ibaraki 305-8562, Japan5Department of Physics, School of Science, University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113-0033, Japan6Present address: Department of Structural Biology, Stanford University, Stanford, CA 94305, USA


DOI 10.1016/j.cell.2010.09.034

SUMMARY

Assembled actin filaments support cellular signaling,

intracellular trafficking, and cytokinesis. ATP hydro-

lysis triggered by actin assembly provides the struc-

tural cues for filament turnover in vivo. Here, we

present the cryo-electron microscopic (cryo-EM)

structure of filamentous actin (F-actin) in the pres-

ence of phosphate, with the visualization of some

a-helical backbones and large side chains.

A complete atomic model based on the EM map

identified intermolecular interactions mediated by

bound magnesium and phosphate ions. Comparison

of the F-actin model with G-actin monomer crystal

structures reveals a critical role for bending of the

conserved proline-rich loop in triggering phosphate

release following ATP hydrolysis. Crystal structures

of G-actin show that mutations in this loop trap the

catalytic site in two intermediate states of the

ATPase cycle. The combined structural information

allows us to propose a detailed molecular mecha-

nism for the biochemical events, including actin

polymerization and ATPase activation, critical for

actin filament dynamics.

INTRODUCTION

The actin-filament system is required in almost all cytoplasmic

processes, including cell adhesion, motility, cellular signaling,

intracellular trafficking, and cytokinesis. Although stable actin

filaments (F-actin) are necessary during muscle contraction,

the active turnover of filaments is required in many cell functions.

Actin has two major domains separated by a nucleotide-binding

cleft (Kabsch et al., 1990). The outer domain is divided into sub-

domains 1 and 2 and the inner domain into subdomains 3 and 4.

All of the subdomains interact with the bound nucleotide. ATP is

hydrolyzed at the rate of 1/3.3 s 1 following the elongation of fila-

ments at the growing end of filaments (Blanchoin and Pollard,

2002), whereas the phosphate release is 100 times slower

(Carlier and Pantaloni, 1986). As a result, newly polymerized fila-

ments consist of stable ADP-Pi actin (abbreviated as F-ADP-Pi),

whereas the older filaments contain mainly ADP actin (F-ADP),

which disassembles more rapidly (Carlier and Pantaloni, 1986).

Under physiological conditions, inorganic phosphate (Pi) binds

to F-actin and reduces the critical concentration for polymeriza-

tion (Rickard and Sheterline, 1986; Fujiwara et al., 2007). Actin

dynamics also depends on the identity of the bound divalent

cation, physiologically Mg2+, associated with the bound nucleo-

tide (Carlier et al., 1986).

Although a vast amount of biochemical data has been accu-

mulated, the quest for a definitive and detailed molecular mech-

anism of the polymerization of monomeric actin (G-actin) to fila-

mentous actin (F-actin) has been hampered by the inherent

flexibility of actin filament. The flexibility has not allowed an

atomic structure of F-actin to be determined. More than

50 atomic structures of G-actin bound with ATP or ADP have

been determined since 1990 (Kabsch et al., 1990), but F-actin

has been visualized to relatively moderate resolution either by

three-dimensional (3D) image reconstruction from electron

micrographs (Belmont et al., 1999) or modeling based on X-ray

fiber diagrams (Holmes et al., 1990; Lorenz et al., 1993). The

inherent flexibility of actin filaments hampers determination of

atomic structure.

Recently, a new model of F-actin based on improved X-ray

fiber diffraction analysis was reported (Oda et al., 2009). Oda

et al. proposed that outer-domain movement upon assembly

flattens the actin molecule in the polymer, similar to the case of

the bacterial actin homolog MreB (van den Ent et al., 2001),


A B

C F

D

E


and that the DNase I binding loop (DNase I loop) adopts an open

loop conformation. However, the mechanism of ATP hydrolysis

and its coupling with actin assembly remains poorly understood.

Here, we present cryo-electron microscopic (cryo-EM) data in

which single-particle analysis has been applied to short and

relatively straight stretches of filaments, with Pi added in the

millimolar range (similar to the intracellular Pi concentration), to

further minimize filament flexibility (Nonomura et al., 1975). The

quality of the cryo-EM images was further refined as described

in the Experimental Procedures.

The resolution of the final reconstruction was estimated to be

5 A (Fourier shell correlation [FSC] of 0.143 at 4.7 A, a criterion

according to Rosenthal and Henderson [2003]) or 8 A (FSC of

0.5 at 7.8 A, a traditional criterion), and some of the a-helical

backbones and large side chains can be directly observed.

This indicates that the data quality was sufficient to visualize

the structural changes upon polymerization and allowed us to

build a quasi-atomic model of F-actin (F-ADP+Pi). Putative

Mg2+-binding sites and Pi-binding sites of F-actin, which play

an important role in actin assembly, were identified in the EM

map, and the proline (Pro)-rich loop (residues 108–112) was

observed to adopt a more bent configuration that would trigger

a phosphate-releasing pathway. Crystal structures of G-actin

with mutations in this loop, in which the ATPase activity was

increased or decreased, further revealed the region required

for Pi release (the so-called back-door region; Wriggers and

Schulten, 1999) and the atomic details of the mechanism of

ATP hydrolysis. The combined structural information sheds

new light on the coupling mechanism of ATP hydrolysis and

F-actin assembly.

RESULTS

Overall Structure of Actin in a Filament

The 3D cryo-EM structure was reconstructed from segments

containing 26 actin molecules (Figure 1B). Approximately

8000 actin molecules from zero energy-loss cryo-EM images

of actin filaments in the presence of phosphate (Figure 1A)

contributed to the final EM map (Figure 1 and Figure S1 avail-

able online). A quasi-atomic model (Figures 1B and 1D) was

constructed by refining the initial F-actin model consisting of

26 G-ADP actin molecules (Rould et al., 2006) to obtain

a good fit into the EM density map (see Figure S1, Movie S1,

Movie S2, and Figure 1E for FSC and figure-of-merit [FOM]

plots). The resolution of the structure of F-actin appeared to

be nonuniform depending on the regions (Figure 1E). In the

region where three actin molecules interact within the filament,

the quality of EM map was better (Figure 1E) and the backbone

structure of a helices 5 and 6 (h5, residues 183–196; h6, resi-

dues 207–216) and the Thr-rich loop (residues 197–204) could

be clearly resolved, allowing the assignment of some large side

chains such as Lys191, Tyr198, and Arg206 (Figures 1C and

1D). Although no b structure could be directly visualized,

most a helices and loops defined in Figure 2C could be

assigned. The N-terminal segment (residues 1–5), h0 (residues

41–48) in the middle of DNase I loop, and h7 (mobile helix:

residues 226–230) were less clearly resolved but still allowed

main-chain placement except for the h0 segment, which is

disordered. The structural details are shown in Figures

S1F–S1H.

Outer-Domain Rotation and Widening of Hydrophobic

Cleft

In the F-actin structure, the relation between the two major actin

domains is different from that in G-actin. The outer domain is

found rotated in a swing-door manner by 16! relative to the

inner domain (Figure 2A). The pivoting point, Asp154 next to P

loop 2, is located near the bound nucleotide, with two hinges:

Gln137-Ala138 and Lys336-Tyr337 (Figures 2A and 2C). The

axis of the rotation was oriented by 40! relative to the filament

helix axis (Figure 2A) and not vertical to the helix axis (Oda et al.,

2009) (Figure S2). The outer-domain rotation enables the DNase I

loop to fit in the rear half of the hydrophobic cleft (Figure 1F) so

that it could reach Leu110, which can be clearly assigned in

Figure 1. Representative EM Density of Actin Filament

(A) An original zero energy loss cryo-EM image of actin filament on the left and a gallery of classified and averaged images of actin filament containing 26 mole-

cules on the right. Although images were averaged after they were classified into 120 groups with 3! step rotation angles, a gallery shows only 72 projections with

5! step. Scale applies only for a gallery.

(B) Stereo pair of the density map of actin filaments (gray contours). The atomic model is also shown. Each actin molecule is represented by a different color.

Phosphate and magnesium ions are shown in orange and white, respectively. White arrowheads indicate the N terminus.

(C) Density map (gray contours) for the intermolecular interface of actin filament. Orange and gray spheres indicate phosphate andmagnesium ions, respectively.

Residues involved in the intermolecular interactions were well resolved and are shown in ball and stick format.

(D) Stereo pair showing the densities of helix 6 (h6) in gray contours and rotated by 90! with respect to (C). The difference densities (EMmapminus atomicmodel;

red contours, 5s) correspond to two phosphate ions (orange spheres) at the phosphate-binding sites 1 and 2.

(E) Fourier shell correlation (FSC) between two reconstructions after dividing the data into two halves (left). For the total structure (in gray), the FSC curve behaves

less regularly, reflecting the nonuniform resolution depending on the regions. For the functionally important region, where the ternary interaction takes place, the

map quality is better and the FSC curve behaves regularly and is suitable to estimate the resolution (in blue), which is 8 A (7.8 A, FSC = 0.5 a traditional criterion)

or 5 A (4.7 A, FSC = 0.143, a criterion taking account of the effect of halving data; Rosenthal and Henderson, 2003). For its calculation, the region enclosed in

a cosine-edged cylinder with a diameter of 36.4 A and a height of 22.8 A (covering 44% of the whole molecular volume) was used. The estimated resolution is

consistent with the observation that some a-helical backbones and large side chains were directly visible in the EM maps shown in (C) and (D). Figure of merit

(FOM) was calculated from the Fourier transform of the EMdensitymap and the atomicmodel in each resolution shell (Yonekura et al., 2003) (right). A shell FOMof

0.5 was used as a criterion to determine the resolution limit. The shell FOM was above 0.5 out to 6.0–7.0 A resolution shell.

(F) Ribbon diagram of F-actin. Two phosphates (orange) and two Mg2+ (gray) located at intermolecular interface are shown in sphere format. Locations of the

N terminus, hydrophobic cleft, hydrophobic loop, and DNase I loop are indicated. ADP is shown in ball and stick format. In all panels, the minus end (pointed

end) of filament is upward.


the EM map, of the actin molecule on the minus-end side

(Figure 1C and Movie S1). As a result of the outer-domain rota-

tion (Figure 2A), the front half of the hydrophobic cleft is widened,

making the side chain of Tyr143 more solvent exposed and

increasing the distance between Tyr143 and Leu346 on the

hydrophobic helix (h12) (Figure 2B). The importance of the front

half of the hydrophobic cleft for polymerization is further high-

lighted by the fact that the Dictyostelium actin with Tyr143Phe

mutation polymerizes poorly. The Tyr143Ile mutation, however,

has only a small inhibitory effect on assembly (Figure 2D). This

is consistent with the fact that the corresponding residue of the

bacterial actin homolog MreB is Ile (van den Ent et al., 2001).

The front half of the hydrophobic cleft is also a primary site for

G-actin- and/or F-actin-binding proteins, which could regulate

actin assembly by promoting or blocking the widening of the

hydrophobic cleft. Indeed, small marine toxins such as kabiri-

mide C and jaspisamide A (Klenchin et al., 2003), which bind to

the front half of the hydrophobic cleft and sever actin filament,

are sterically compatible with our F-actin structure. This may

partially account for the inhibitory effect of modification of

Cys374 with tetramethylrhodamine (TMR) on polymerizability

(Otterbein et al., 2001).

Structural Changes in the Intermolecular Interface

Accompanying the domain rotation, the main-chain atoms shift

more than 4.8 A mainly in the loop regions (Figure 2C), which

facilitates intermolecular interactions. This results in a buried

surface area between three actin molecules of up to 7646 A2,

which is a substantial increase compared to 2998 A2 for the

crystal structure of G-ADP actin docked into the EM map

without remodeling. At the interface of the three actins within

the filament (Figures 1C and 1D and Figure 3A), the Thr-rich

loop (containing Thr202 and Thr203) was remarkably different

from that in G-actin (shown in gray in Figure 3A). Furthermore,

the N-terminal part of h6 was shifted compared to that in G-actin

(shown in gray in Figure 3A and Figure S1A) and possibly stabi-

lized by a putative salt bridge of Glu205 with Arg290 of the upper

molecule and the hydrophobic interaction of Ala204 and Ile208

with Ile287 of the upper molecule. The disruption of h6 is

observed in the crystal structure of the actin-DNase I complex

(Kabsch et al., 1990). Local reordering of the N-terminal part of

h6 could provide a common binding site for actin-binding

proteins, including actin itself, and could also help bind amagne-

sium ion (site 1 in Figure 3A) and two phosphate ions (Pi1 and

Pi2, Figure 1D).

A

C

B

D

Figure 2. Comparison of F-Actin Structure with G-Actin Structure

(A) The inner domain of the G-actin (G-ADP) (gray; Otterbein et al., 2001) was superimposed onto that of actin filament (F-actin) (green). In the F-actin structure

(F-ADP+Pi), the outer domain is rotated by 16! relative to the inner domain. The rotation angle was determined using DynDom (Hayward and Berendsen, 1998).

The bound ADP is shown in ball and stick format.

(B) The enlarged frontal view of the hydrophobic cleft in F-actin (green) and G-actin (gray). DNase I loop (not shown) fits in the rear half of hydrophobic cleft. The

front half of hydrophobic cleft remains empty and widens. The double-headed arrows show that the distance betweenTyr143 and Leu346 is wider in F-actin

compared with that in G-actin. The arrow indicates the hinge point of the outer-domain rotation.

(C) Root-mean-square deviation (rmsd) per residue between the molecules of F-actin and G-actin. For the calculation, each of the inner and outer domains was

superimposed onto that of G-actin independently. The structure within each domain is essentially the same as that of G-actin. The conformational changes occur

mainly in DNase I loop, Pro-rich loop, Thr-rich loop, mobile helix h7, and Ser-rich loop, all of which are involved in the intermolecular interfaces.

(D) Actin polymerization assay. Actin mutants (2.3 mM) were incubated at 25!C, ultracentrifuged, and analyzed by SDS-PAGE. The data represent mean values

with standard errors of the actin pellets (n = 4).


A

B

E F G

C D

Figure 3. Intermolecular Interfaces of the Actin Filament

(A) Cartoon representation of the interfaces of actin filament. The structure of G-actin in gray (Otterbein et al., 2001) was superimposed onto each molecule of

actin filament by fitting subdomain 3. Each of the actin molecules of F-actin is represented by a different color. The phosphates and Mg2+ are represented as

orange and gray spheres, respectively.

(B) Cartoon representation of the actin filament. Actin subdomains are numbered. The areas enclosed with the dotted and solid rectangles are shown in (A)

and (C), respectively.

(C) The stereo pair for the area around the phosphate-binding sites with the corresponding EMdensities (blue contours) and the differencemap (red contours, 5s).

(D) Surface electrostatic potential in the same area as in (C). Two phosphate ions are surrounded by positive electrostatic potential.

(E) The difference map (EM map minus atomic model; red contours, 5s) showing the two phosphate ions at sites 1 and 2.

(F) Confocal fluorescence micrographs of cells expressing GFP-fused actin. GFP-fused wild-type actin and GFP-fused V287D actin show more pronounced

cortical accumulation, particularly at pseudopodia (arrows), than do GFP-fused R290E actin or GFP-fused V287D/R290E actin. All of the micrographs are shown

without automatic contrast adjustment (exposure time: 2 s for GFP-fused wild-type actin and GFP-fused V287D, 4 s for GFP-fused R290E actin and GFP-fused

V287D/R290E). The extent of polymerization of each actin mutant was assayed by quantifying the ratio of GFP-actin in insoluble fractions of cells treated with

Triton to the total GFP-actin of cells treated with Triton: 0.35, 0.34, 0.21, and 0.13 for the GFP-fused wild-type actin, GFP-fused V287D actin, GFP-fused R290E

actin, and GFP-fused V287D/R290E actin, respectively. Scale bar, 10 mm.

(G) Actin polymerization assay. Actin mutants (2.3 mM) were incubated at 25 C, ultracentrifuged, and analyzed with SDS-PAGE. The pellet (ppt) is 6.3-fold

concentrated relative to supernatant (sup).


This feature of h6 is supported by Dictyostelium mutants

Val287Asp or Arg290Glu, which both polymerized poorly

(Figure 3G), emphasizing the importance of Ile287 (Val287 inDic-

tyostelium actin) and Arg290 in the vertical interaction. In addi-

tion, these mutants exhibit more disperse distribution when

fused to GFP in Dictyostelium cells, with the double mutant dis-

playing a more prominent phenotype (Figure 3F).

Phosphate-Binding Loops and Shifted Sensor Loop

Both of two Pi-binding loops, P loop 1 (residues 13–16) and P

loop 2 (residues 156–159), could be assigned (Figure 4B). They

surrounded the densities that correspond to a- and b-phos-

phates of the bound nucleotide, with no evidence for any g-

phosphate density. The region, where the g-phosphate is

located in the ATP form, is occupied by bulk solvent

(Figure 4B), indicating that F-actin has bound ADP in the

ATPase site (Figure 4B). Similar to the majority of crystal struc-

tures of G-actin with ATP or ADP, the nucleotide-binding cleft

(Nolen and Pollard, 2007) is closed. However, the P loop 1 encir-

cles a low-density region (a bubble in the EM map; Figure S4A),

suggesting that two strands, s1 (residues 8–11) and s2 (residues

16–18), might be dynamically deformed. This could allow

accommodation of the outer-domain rotation. Consistent with

this, NMR spectroscopy showed that residues 1–22, which

include P loop 1, are mobile even in an F-actin state (Heintz

et al., 1996).

The sensor loop (residues 71–77), which reflects the state of

bound nucleotide of G-actin (Rould et al., 2006), adopts a confor-

mation similar to that of G-ADP (Otterbein et al., 2001; Rould

et al., 2006) (Figure 3A). Compared with G-ATP actin, Glu72 on

the sensor loop moves upwards by 2 A and closer to Arg183.

This enables a putative salt bridge between Glu72 and Arg183,

which stabilize F-actin, as Arg183 is also associated with two

phosphate ions.

Magnesium- and Phosphate-Binding Sites of F-Actin

Actin is known to have several binding sites for cations and

phosphates (Rickard and Sheterline, 1986; Carlier et al., 1986).

Also, more than 50 crystal structures of G-actin have been

determined, and magnesium-binding sites have been deduced

(Table 1; Klenchin et al., 2006).

After several cycles of refinement of the atomicmodel, a differ-

ence map (EM density minus the atomic model) showed signifi-

cant peaks, greater than 3s, with > 90% confidence by t test

applied to three independently reconstructed EM maps

(Figure 3E and Figure S3). The three strongest peaks (s-values

5.2, 5.9, and 4.8) in the positively charged region (Figures 3C

and 3D) are assigned as Pi (HPO42 ) because Pi was the only

major anion in the specimens. The phosphate-binding sites 1

and 3 (Table 1) are also observed in the crystal structure of G-

actin, in which the sites are occupied by sulfate ions (SO42 ).

The EM data also supported the five magnesium-binding sites

out of seven found in the crystal structures (Klenchin et al.,

2006). They have significant densities in the EM difference

map, with the peaks being more than 3s (Figure S3) and with >

90% confidence by the t test analysis. The FOM values increase

when the ions are incorporated to the model. The peaks in the

EM difference map were found in the same locations as the

cations in the crystal structures of G-actin. Thus, the coordi-

nating residues in the model are in essentially the same confor-

mation as in the crystal structures of G-actin.

The resultant atomic model of F-actin (F-ADP+Pi) included 5

Mg2+ and 3HPO42 ions per actin protomer (Figure S3). Because

Pi is located outside of the ATPase site, it is not designated as

F-ADP-Pi.

Two phosphate ions and two Mg2+ ions bound to the inner

domain appear to have a crucial role in the intermolecular inter-

actions. Mg2+ ions at site 1 and 2 would mediate longitudinal and

oblique interactions, respectively (Figure 3A and Figure S3A).

This could explain the importance of Mg2+ for actin assembly

(Laki et al., 1962; Carlier et al., 1986). Mg2+ at site 1 reinforces

the vertical interaction through coordination with two acidic resi-

dues, Asp288 and Glu207 on the vertically adjacent molecule. It

also interacts with Gln59 of the actin molecule on the plus-end

side. Consequently, Mg2+ site 1 forms an intramolecular bridge

between the subdomains 2 (Gln59) and 4 (Glu207), which keeps

the nucleotide-binding cleft closed. The configuration of Mg2+

site 2 is similar to that found in crystal structures of G-actin

with a Mg2+ coordination to Gln263 and Ser265 of the hydro-

phobic loop (right panel of Figure 3A). In the F-actin model, its

unique position in the groove of the actin double helix and near

Table 1. Putative Phosphates andMagnesium Ions in F-Actin and

Corresponding Sites in Crystal Structures of G-Actin

Sites in

F-Actin

Peak

Heighta

Coordinating

Residues

in F-Actin

Corresponding

Sites in G-Actin with

the PDB Code

1 Pi

(site 1)

5.2 s R183, D184,

T202b, K284

3CI5 (SO42 ), 2Q36

(SO42 ), 1YAG (SO4

2 ),

1D4X (SO42 ), 1YVN

(SO42 ), 1NLV (SO4

2 ),

1NM1 (SO42 ), 1NMD

(SO42 )

2 Pi

(site 2)

5.9 s R183, R206 3CI5 (SO42 )

3 Pi

(site 3)

4.8 s K238, R254 2A5X (TSAc), 2A42

(glycerol), 3CI5 (SO42 )

4 Mg2+

(site 1)

5.6 s D286, D288,

E207, Q59

1NWK (Ca2+), 2FXU (Ca2+),

2VYP (Ca2+), 2HF3 (Ca2+)

5 Mg2+

(site 2)

3.0 s Q263, S265,

T66

2Q1N (Ca2+), 2Q0U (Ca2+),

2A5X (Ca2+), 3EKS (Ca2+),

2HF3 (Ca2+), 2ASM (Ca2+),

3EKU (Ca2+), 2Q0R (Ca2+),

2HF4 (Ca2+), 3EL2 (Ca2+),

2A5X (Ca2+)

6 Mg2+

(site 3)

3.6 s T202, E205 1YXQ (Mg2+), 1J6Z (Ca2+)

7 Mg2+

(site 4)

6.3 s Q354, E361 1J6Z (Ca2+), 1NWK (Ca2+),

2VYP (Ca2+)

8 Mg2+

(site 5)

5.1 s D222, E316,

E259

2HMP (Sr2+)

aPeak height in the EM difference map (EM density minus the atomic

model).bThe residues in the longitudinally and obliquely located actin molecules

are indicated in underline and italic, respectively.cN,N,N-TRIMETHYL-3-SULFOPROPAN-1-AMINIUM.


Thr66 of subdomain 2 promotes a contribution to the oblique

interaction through the hydrophobic loop (right panel of

Figure 3A).

Among three putative phosphate ions, two major peaks in the

difference map (Pi-1 with 5.2s and Pi-2 with 5.9s; Figure 3E) are

located near the center of the filament (Pi-1, 1 A from the fila-

ment axis; Pi-2, 3 A). The Pi at Pi-site 1 would stabilize the

ternary interaction through interacting with Thr202 in the Thr-

rich loop, Arg183 of the lateral molecule, and Lys284 of the upper

molecule on the minus-end side (Figure 3C). The Pi at Pi site 2 is

associated with Arg183 and Arg206 of the same lateral molecule

(Figure3C). Thus, Arg183contributes to form twoPi sites.Of note,

in the G-ATP state, the Arg183 side chain adopts a conformation

unsuitable for either Pi site formation (Otterbein et al., 2001; Rould

et al., 2006) (left panel of Figure 4E), which would suggest that

polymerization and ATP hydrolysis may be required for the crea-

tion of the Pi-binding sites as seen in the F-actin model.

A Cavity as a Proposed Phosphate-Release Pathway

The phosphate release, as it occurs subsequent to actin

polymerization, would be dependent on intrinsic properties of

the F-actin. One key feature observed in the EM map is a cylin-

drical cavity ( 18 A long, 6 A in diameter) along the interdo-

main groove on the backside of the actin molecule, which

was observed in the EM map (Figures 4A–4C and Figure S4)

flanked by two additional actin chains (blue and purple in

Figure 4E) and the b-phosphate of ADP (Figures 4B–4D; see

also Movie S3). The cavity would be the only obvious route

by which the hydrolyzed g-phosphate could readily access

the external solvent (Figure 4C and Figure S4). The Pi site 1,

which is located near the exit of the cavity (Figure 4C), could

constitute a possible intermediary binding site for the

hydrolyzed g-phosphate in a phosphate-release pathway

(Figure 4E).

His73, which also flanks the cavity and is methylated in

most eukaryotes, putatively changes side-chain configuration

after polymerization. In the G-ATP state, the d-nitrogen of

His73 forms an interdomain H bond with the carbonyl oxygen

of Gly158, thereby bridging two major domains (left panel of

Figure 4E), whereas the H bond is likely to be disrupted in

the F-actin. The imidazolic ring and the methyl group of meth-

ylated His73 are located near the phenolic ring of Tyr198 of

A B

C

E

D

Figure 4. Structure around the Bound ADP

in the Actin Filament

(A) The stereo pair for the structure around the exit

of the phosphate-releasing cavity. The corre-

sponding EM densities are shown in mesh format.

Two inorganic phosphates are shown in orange

spheres. The bound ADP is shown in ball and stick

format.

(B) The stereo pair to show the phosphate-

releasing cavity at the intermolecular interface of

the actin filament. The Mg2+-ADP is shown in ball

and stick format, and magnesium ion is repre-

sented as a gray sphere.

(C) The vertically clipped view of the surface

rendering of the EM map together with the sche-

matic representation (right). The clipped surface

is capped with a gray plane. The inside surface

of the phosphate-releasing cavity can be seen in

the top region, and the other side of the surface

can be seen in the rest of the region. The bound

ADP and phosphate ions (Pi1 and Pi2) are labeled

and shown in solid sphere format. The cavity is

indicated by a bracket. Its inside (white) and

outside surface (gray) can be seen at the upper

and lower part, respectively.

(D) The structure of the G-actin (gray) was super-

imposed onto the molecule of the actin filament

(F-actin) (green) by fitting the P loop 2. In F-actin,

the Pro-rich loop bends downmore and the stack-

ing interaction of Pro109 with His161 is disrupted.

(E) Summary of the conformational changes

accompanied by the actin polymerization and

ATP hydrolysis. Dotted straight lines indicate puta-

tive interactions. In F-actin (colored panel), the

intermolecular cavity is formed along the groove

between the outer and inner domains of the green

molecule and is flanked by the subdomain 4 of the

purple actin molecule. Because the phosphate-

binding site 1 (Pi1) is located near the exit of the

cavity, the hydrolyzed phosphate will be bound

to this site before it escapes to the external

solvent.


A B C

D

E F


the obliquely adjacent molecule (Figure 4A) and could restrict the

exit of the cavity (Figure 4E). The methylated His73 modulates

access of the hydrolyzed g-phosphate to the Pi site 1. Such

amechanismwould explain reports of Pi releaseprior to polymer-

ization when His73 is not methylated and Pi release from Hi-

s73Ala-actin showing almost no lag time after polymerization

(Nyman et al., 2002). Thus, we propose that the hydrolyzed g-

phosphate binds to the Pi-binding site 1 (Pi-site 1) and stabilizes

the F-actin by reinforcing the ternary interaction before leaving

actin filaments and that the methyl group of methylated His73

helps confine the Pi-release pathway.

Downward Bending of the Pro-Rich Loop

The EM map shows that the DNase I loop extends along the

lateral molecule toward the upper molecule and the middle

portion (h0, residues 41–48) of the DNase I loop associated

with the Pro-rich loop (residues 108–112) of the upper molecule

(right panel of Figure 3A). The exact conformation of the h0

segment of the DNase I loop could not be determined, but in

our analysis, we have modeled it as a short a helix (Figure

S1F), similar to that in the crystal structure of G-ADP actin (Otter-

bein et al., 2001). A large upward shift of this segment may

explain why polymerization is inhibited by phosphorylation of

Tyr53 (Liu et al., 2006): the phosphate group of phosphorylated

Tyr53 could form H bonds with Lys61 and thus restricts the

extendability of the DNase I loop (Baek et al., 2008).

It is most likely that residues in themiddle portion of the DNase

I loop are close to the Pro-rich loop; in particular, it forms an inter-

action with Leu110, of which the side-chain density was

apparent in the EM map (Figures 1C). The Pro-rich loop, in

turn, bends down toward the plus end. The stacking interaction

of Pro109 with His161 observed in G-actin, however, is not

possible in the F-actin model (Figures 4D and 4E). This is signif-

icant, as His161 and Gln137 both have been suggested to be

involved in the ATP hydrolysis (Matsuura et al., 2000; Vorobiev

et al., 2003), and the sensor loop and the Pro-rich loop (contain-

ing Pro109 and Pro112) constitute the back-door region, which

has been suggested to control phosphate release (Wriggers

and Schulten, 1999).

The importance of the Pro-rich loop was confirmed through

a focused study of the Pro109 mutants using X-ray crystallog-

raphy and ATPase assay (Figure 5 and Figure S5). Structural

changes due to themutationswere observedmainly in the region

around the bound nucleotide and h6, with no outer-domain rota-

tion being observed.

In the structure of wild-type actin, a water molecule (Wat1203)

is H-bonded to the 3-oxygen of Gln137, which renders it nucleo-

philic. However, a nucleophilic in-line attack on g-phosphate

would not occur (Matsuura et al., 2000; Vorobiev et al., 2003),

as that water is too distant from the g-phosphate (4.0 A). The

d-nitrogen of His161 coordinates with another water molecule

(Wat1259 in Figures 5A–5C), which, in turn, forms an H bond

with the nucleophilic water and thus constrains its position

(Figure 5A). In Pro109Ile-actin, His161 has no partner for stack-

ing (Figure 5B), and the imidazolic ring of His161 is rotated by

12!, which disrupts the H bond network between His161 and

the nucleophilic water (Figure 5B). As a consequence, the nucle-

ophilic water is found closer to the g-phosphate (3.6 A), but ATP

hydrolysis does not occur (Figures 5B and 5E), presumably

because the back door is still closed, likely due to a hydrophobic

interaction of the introduced Ile109 with Val163 and Ile175.

However, in Pro109Ala mutant where His161 still has no partner

for stacking, the Pro-rich loop is found bent further downward by

1.8 A (Figure 5C), with a shift in the main chain of Ala109 and

Leu110, and no density for g-phosphate is observed. The sensor

loop is shifted upward in amanner similar to that in the F-ADP+Pi

model or in G-ADP crystals (Otterbein et al., 2001; Rould et al.,

2006). This indicates that the ATP added to the crystallization

specimens is hydrolyzed with subsequent g-phosphate release

(Figure 5C). The ATPase activity of the Pro109Ala mutant in the

presence of Mg2+ increased by 10-fold under nonpolymeriz-

able conditions compared to control actin (1/424 s"1 versus

1/4510 s"1), whereas the activity of Pro109Ile actin was

increased by 2-fold (1/2159 s"1 [Figure 5E]). Using ethenoATP,

the nucleotide exchange rate of Pro109Ala-actin was shown to

be 40 times faster (1/7 s"1) than the hydrolysis rate (Figure 5F).

Taken together, these data indicate that the downward bending

of the Pro-rich loop is the key prerequisite for ATP hydrolysis

rather than the outer-domain rotation, which assists the bending

of the Pro-rich loop during polymerization (Figure 6).

Mechanism of ATP Hydrolysis Based on X-Ray

Crystallography

In the catalytic site of the G-actin structures, an extensive

network of water molecules can be observed, which is stabilized

through interactions with the P loop 1, the P loop 2, the sensor

Figure 5. Atomic Details of the Water-Mediated Hydrogen-Bonding Network of Crystal Structures of Actin Mutants in the Presence of Mg2+

(A–C) Atomic details of the structure in the presence of Mg2+. The region surrounding the bound nucleotide in wild-type actin and actin mutants is shown together

with their schematic figures. (A) Wild-type actin (PDB ID code 1NM1), (B) P109I actin, and (C) P109A actin. All actin mutants were crystallized as complexes with

gelsolin segment 1 (see Extended Experimental Procedures for details). The numbering of water molecules for wild-type actin is changed for easier comparison

withmutant actins.Water molecules are shown in red spheres. Dotted lines indicate putative H bonds. The labels of the nucleophilic water (WAT1203) and the key

water molecule (WAT1259) are highlighted in yellow. In the crystal structure of P109A actin (C), the Pro-rich loop bends downward more and the back door is

opened (indicated by an arrow). ADP hydrolyzed from added ATP was observed. An omit-annealed Fo " Fc map around the bound nucleotide is shown (blue

contour), confirming the nucleotide state.

(D) Summary of the crystallographic studies. Asterisks indicate structures published previously. Atomic details of the structure in the presence of Ca2+ are shown

in Figure S5.

(E) ATPase activities of actin mutants. E205A/R206A/E207A mutations were added into each of the actin mutants to make them polymerization incompetent

(Noguchi et al., 2007) to exclude the effect of polymerization on ATPase activation. Each of the data points represents a mean value with a standard error

(n = 4).

(F) Exchange kinetics of ethenoATP with G-ATP in the presence of Mg2+. The rate is 1/30 s"1 and 1/7 s"1 for control actin (with E205A/R206A/E207A mutations)

and P109A actin, respectively. Profilin accelerates the rates of both actins to the same extent.


loop, and the Pro-rich loop (Figures 5A–5C). Two water

molecules, Wat1203 andWat1259, which change their positions

depending on the introduced mutation, could be critical in

the function of the ATPase cycle in the presence of Mg2+. We

propose a four-step mechanism for the ATPase cycle

(see Movie S4). First, by polymerization (or Pro109Ala mutation),

the stacking interaction of His161 is lost, and thewater-mediated

(Wat1259) H bond between the nucleophilic water (Wat1203)

and His161 is disrupted, with a back door being half open, allow-

ing a shift in the position of the nucleophilic water. Second, the

nucleophilic water (Wat1203) moves closer to the g-phosphate.

Third, the water attacks the ATP, which then is hydrolyzed to

ADP and Pi. Fourth, the sensor loop shifts upward and facilitates

the Pi release to the Pi-release cavity by fully opening the back-

door. The proposed mechanism is supported by the conserva-

tion of key residues for ATPase activation (Pro109, Leu110,

His161, and residues of two P loops [residues 13–16 and

156–159]) (Figure S5B). Figure 5D summarizes how ATPase

activity correlates with the states of the key water molecules,

the back door, and nucleotide.

In the presence of Ca2+, ATPase activity is low (Figure 5E).

The second step of ATPase cycle is slow because the water

molecule Wat1108 is coordinated with Ca2+ ion keeping the

nucleophilic water Wat1203 away from the g-phosphate (4.4 A)

(see Figure S5A for details). This explains, in part, the differences

between Ca2+-actin andMg2+-actin in their dynamic polymeriza-

tion properties (Carlier et al., 1986).

DISCUSSION

Rotational Flexibility of Outer Domain

One of the apparent structural changes in actin observed in the

F-actin model compared to G-actin is a pronounced rotation of

the outer domain in a swing-door manner. Such a rotation was

A B

Figure 6. Proposed Model for Actin Polymerization

(A) Schematic diagram of actin polymerization. At steady state, actin monomers come on the filament at the plus end as G-ATP-actin (colored in white). After they

are incorporated into the filament, ATP is hydrolyzed, forming an intermediate F-ADP+Pi (colored in gray) and then ADP-actin (colored in dark gray).

(B) Proposedmodel for actin polymerization at a plus end (downward elongation). Each actin molecule, with four subdomains being numbered, is represented by

a different color. The phosphate moieties of the bound nucleotides and the hydrolyzed g-phosphates are shown in orange spheres. The subdomain 4 is repre-

sented by two cylinders corresponding to helices h5 and h6. In this model, polymerization and ATP hydrolysis proceed in four steps as described in the text. This

model implicates that the actin filament has at least three actin molecules in an F-ATP state on the plus end (ATP cap) and that the activation of ATPase does not

occur in a trimer (a proposed nucleus for elongation).


reported recently with the outer-domain rotation around the axis

perpendicular to the actin helix axis, a propeller-like twisting

(Oda et al., 2009) (Figure S2A). The observed apparent rotation

in the F-actin model is, however, different, with the rotation

axis oriented by 40 relative to the filament axis (Figure 2A).

This rotation could be described as two perpendicular rotations,

roughly propeller-like twisting and scissor-like closing. The

apparent difference between the two studies may be due to

the torsional and/or bending flexibility of actin filaments and

absence of added Pi in the sample used to obtain the fiber

diffraction data. Of note, the orientation of the outer domain is

variable even among the crystal structures of G-actin (!10 ),

but this flexibility does not trigger ATP hydrolysis, suggesting

that more substantial changes such as these shown in this

work are required for ATPase activation to promote the down-

ward bending of the Pro-rich loop, a key prerequisite for trig-

gering ATPase activity.

How Polymerization Triggers ATP Hydrolysis

At the heart of the G-actin to F-actin dynamics lies the question

how polymerization triggers ATP hydrolysis. We propose

a detailed model for polymerization based on the F-actin model,

outlined in Figure 6B and Movie S5. A newly incorporated mole-

cule (#2 magenta) enables the outer-domain rotation of the

penultimate actin molecule (#3 green). The other oblique interac-

tion between the penultimate actin and the neighboring actin

molecule (#4 blue) helps to orient the subdomain 2 of the penul-

timate actin so that its DNase I loop (the cylinder tethered to

subdomain 2 in Figure 6B) reaches out upward. The middle

portion of the DNase I loop fits in the rear half of the hydrophobic

cleft, where the middle portion associates with the Pro-rich loop

of the molecule on the minus-end side (#5 in yellowish green).

This interaction induces the downward shift of Pro-rich loop

and triggers ATP hydrolysis. The back door becomes fully

open, and Pi is released to the Pi-release pathway (stage 2 in

Figure 6B). Thus, the interaction of the DNase I loop with the

Pro-rich loop is the key step to couple actin assembly with

ATPase activation. These results explain why actin with a nicked

DNase I loop polymerizes poorly (Khaitlina et al., 1993).

Hydrolyzed g-phosphate binds to the Pi site 1 and reinforces

the ternary interaction. Phalloidin, which binds to the site (Lorenz

et al., 1993) near the Pi site 1 and slows the Pi release to the

external solvent (Dancker and Hess, 1990), appears to mimic

and/or reinforce the function of Pi. Unlike phalloidin, Pi can

reversibly dissociate from the sites at low Pi concentrations,

resulting in weakened intermolecular interactions.

In summary, the EM map and resulting quasi-atomic model of

F-ADP+Pi show intricate intermolecular interfaces and binding

sites for Pi and Mg2+ that allow proposal of a molecular mecha-

nism of ATP hydrolysis upon actin assembly and delayed Pi

release, in which the Pro-rich loop has a central role in coupling

polymerization with ATP hydrolysis.


Electron Microscopy and Image Processing

Actin was prepared from rabbit skeletal muscle (Spudich and Watt, 1971).

Actin filaments were prepared in solution containing 50 mM NaCl, 5 mM

MgCl2, 0.025 mM ATP, 10 mM sodium phosphate [pH 7.4], 0.05%NaN3, and

0.7 mM DTT at room temperature (25 C). Zero energy loss cryo-EM micro-

graphs were recorded at 200 kV (Hitachi EF-2000 with a cold field emission

gun and an in-column energy filter) with a low-noise, high-sensitivity, and

high-resolution CCD camera (Yasunaga andWakabayashi, 2008) with electron

dose of 15–20 e"/A2, a nominal magnification of 100–110 k, and underfocus

values of 1–1.5 mm. High-coherence beam generated by cold field emission

gun was useful to minimize the image blurring due to underfocusing. The

CCD camera with an epitaxially-grown scintillator (Yasunaga and Wakabaya-

shi, 2008) helped collect images with high resolution. All of the images were

analyzed with a pixel size of 2.28 A with EOS software (Yasunaga and Waka-

bayashi, 1996). Images were segmented to contain 26 actin molecules and

classified into 120 groups with 3 step rotation angles and were treated as

single particles (Narita et al., 2001; Narita and Maeda, 2007). In the resultant

3D image, noncrystallographic helical symmetry was determined with cross-

correlation analysis. Finally, the 14 actin molecules on the minus-end part in

the EM map were averaged after fitting each other using SPIDER (Frank

et al., 1996). The noncrystallographic helical symmetry with a 166.48 rotation

and a 27.3 A translation along the filament axis was used.

Model Building and Refinement

The initial model consisting of 26 G-ADP actin molecules (Rould et al., 2006)

was manually fitted to the EM density using O (Jones et al., 1991). The refine-

ment was performed using spatially restricted molecular dynamics (Noda

et al., 2006). EM densities were treated as added pseudo-potential so that

atoms tended to be retained in higher-density regions of the EM map.

Throughout the refinement, the validity of the atomic model was checked

with the real-space R factor with O (Jones et al., 1991) and the program

pdbRhoFit (Yasunaga and Wakabayashi, 1996), which sums up the density

of the EM map where the atoms are located. The degree of fit of the atomic

model to the EMdensity was also accessed by calculating FOM from the Four-

ier transforms of the EM map and the atomic model (Yonekura et al., 2003).

Manual model rebuilding using O (Jones et al., 1991) and refinement using

steered-molecular dynamics were performed iteratively to produce the final

model.

Crystallization and Structural Determination

Dictyostelium actin mutants were expressed and purified, as described previ-

ously (Noguchi et al., 2007), with slight modifications. To minimize the effect of

polymerization on biochemical assays, E205, R206, and E207 weremutated to

Ala so that actin becomes nonpolymerizable (Noguchi et al., 2007). Crystals of

the complex of actin mutant (P109I/E205A/R206A/E207A) with gelsolin

segment 1 in the Ca2+ state were obtained by hanging drops containing

13% PEG3550, 130 mM LiCl, 100 mM MES [pH 7.0], 0.2 mM CaCl2, 0.5 mM

DTT, 0.2 mMATP, and 0.2 mMAMP-PNP. To obtain crystals in theMg2+ state,

10 mM MgCl2 and 0.5 mM EGTA were added before the crystallization.

Crystals of the complex of another actin mutant (P109A/E205A/R206A/

E207A) with gelsolin segment 1 were obtained in a similar manner. Crystals

were cryoprotected with glycerol or ethylene glycol and then flash cooled in

liquid nitrogen. Data were collected at Photon Factory (KEK, Japan). The struc-

tures were solved by molecular replacement (Storoni et al., 2004) using

Dictyostelium wild-type actin (Matsuura et al., 2000) as an initial model. The

models were rebuilt and refined using O (Jones et al., 1991) and Refmac5

(Collaborative Computational Project, Number 4, 1994).

Polymerization Assays

Each actin mutant (2.3 mM) was incubated in 150 mM NaCl, 20 mM imidazole-

HCl (pH 7.0), 3 mM MgCl2, 10 mM CaCl2, and 1 mM ATP for 30 min at 25 C,

ultracentrifuged, and analyzed by SDS-PAGE.

In Vivo Optical Microscopy

In vivo copolymerization of mutant actin with wild-type actin was assayed by

quantifying GFP-actin. In brief, Dictyostelium cells expressing GFP-fused

wild-type or mutant actin were grown in HL5. In the mid log phase of growth,

the culture medium was replaced with lysis buffer containing 20 mM HEPES

(pH 7.4), 50 mM NaCl, 2 mM MgCl2, 1 mM EGTA, 0.5% Triton X-100, and


protease inhibitors. After 1 min, the soluble and insoluble fractions were sepa-

rated and subjected to western blotting analysis using anti-GFP antibodies.

Assay of ATPase Activity and Nucleotide Exchange Rate

The ATPase activity of G-actin was measured as described previously (Saeki

et al., 1996) in 10.7 mM actin mutant, 10 mM Tris-HCl (pH 7.5), 0.2 mM CaCl2,

0.5 mM DTT, and 0.2 mM ATP at 25 C. To measure the activity for the Mg2+

state, 0.2 mMMgCl2 and 0.3 mM EGTA were added before the measurement.

E205A/R206A/E207A mutations were added to make the mutant polymeriza-

tion incompetent (Noguchi et al., 2007) and exclude the effect of polymeriza-

tion on ATPase activation. Nucleotide exchange wasmeasured by fluorometry

of the ethenoATP incorporated to G-actin (340 nm excitation, 410 nm

emission). The experiments were performed with 2 mM actin in the presence

or absence of 0.5 mM Dictyostelium profilin I. The reaction was initiated by

diluting the actin solution (10 mM Tris-HCl [pH 7.5], 0.2 mM CaCl2, 0.5 mM

DTT, 0.2 mM MgCl2, 0.3 mM EGTA, and 1 mM ATP) into a buffer containing

20 mM ethenoATP.

Figure Preparation

Figures were prepared using MOLMOL (Koradi et al., 1996) and Chimera

(Pettersen et al., 2004).

ACCESSION NUMBERS

Atomic coordinates and structure factors have been deposited in the Protein

Data Bank under accession codes 3A5L, 3A5M, 3A5N, and 3A5O. The acces-

sion code for actin filament structure is 3G37.



figures, three tables, and five movies and can be found with this article online

at doi:10.1016/j.cell.2010.09.034.

ACKNOWLEDGMENTS

We thankDr.Murray Stewart (MRCLab.Mol. Biol., Cambridge) for discussions

and advice on crystallography; Dr. Ralph Davis and Dr. Karl-Magnus Larsson

(Stanford University) for critical reading of the manuscript; Dr. Akihiro Narita for

his contribution to this work at the early stages; Dr. Soichi Wakatsuki and

colleagues at Photon Factory, KEK for data collection at the synchrotron

site; Dr. Koji Yonekura for the program to calculate a figure of merit; Dr. Hideo

Higuchi for fluorometry; and Yuji Tuji, Akito Tominaga, Ryuta Mikawa, and Har-

umi Kiuchi for assistance with crystallography and/or biochemical experi-

ments. This work was supported by grants from Human Frontier Science

Program and Grant-in-Aid for Scientific Research on Priority Areas (Bio-supra-

molecule) from the Ministry of Education, Science, Technology, and Sports of

Japan to T.W. and a grant from SENTAN, JST to T.Q.P.U.

Received: January 13, 2010




REFERENCES

Baek, K., Liu, X., Ferron, F., Shu, S., Korn, E.D., and Dominguez, R. (2008).

Modulation of actin structure and function by phosphorylation of Tyr-53 and

profilin binding. Proc. Natl. Acad. Sci. USA 105, 11748–11753.

Belmont, L.D., Orlova, A., Drubin, D.G., and Egelman, E.H. (1999). A change in

actin conformation associated with filament instability after Pi release. Proc.

Natl. Acad. Sci. USA 96, 29–34.

Blanchoin, L., and Pollard, T.D. (2002). Hydrolysis of ATP by polymerized

actin depends on the bound divalent cation but not profilin. Biochemistry

41, 597–602.

Carlier, M.F., and Pantaloni, D. (1986). Direct evidence for ADP-Pi-F-actin as

the major intermediate in ATP-actin polymerization. Rate of dissociation of

Pi from actin filaments. Biochemistry 25, 7789–7792.

Carlier, M.F., Pantaloni, D., and Korn, E.D. (1986). Fluorescence measure-

ments of the binding of cations to high-affinity and low-affinity sites on

ATP-G-actin. J. Biol. Chem. 261, 10778–10784.

Collaborative Computational Project, Number 4. (1994). The CCP4 suite:

programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr.

50, 760–763.

Dancker, P., and Hess, L. (1990). Phalloidin reduces the release of

inorganic phosphate during actin polymerization. Biochim. Biophys. Acta

1035, 197–200.

van den Ent, F., Amos, L.A., and Lowe, J. (2001). Prokaryotic origin of the actin

cytoskeleton. Nature 413, 39–44.

Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M., and Leith,

A. (1996). SPIDER and WEB: processing and visualization of images in 3D

electron microscopy and related fields. J. Struct. Biol. 116, 190–199.

Fujiwara, I., Vavylonis, D., and Pollard, T.D. (2007). Polymerization kinetics of

ADP- and ADP-Pi-actin determined by fluorescence microscopy. Proc. Natl.

Acad. Sci. USA 104, 8827–8832.

Hayward, S., and Berendsen, H.J.C. (1998). Systematic analysis of domain

motions in proteins from conformational change: new results on citrate

synthase and T4 lysozyme. Proteins 30, 144–154.

Heintz, D., Kany, H., and Kalbitzer, H.R. (1996). Mobility of the N-terminal

segment of rabbit skeletal muscle F-actin detected by 1H and 19F nuclear

magnetic resonance spectroscopy. Biochemistry 35, 12686–12693.

Holmes, K.C., Popp, D., Gebhard,W., and Kabsch, W. (1990). Atomicmodel of

the actin filament. Nature 347, 44–49.

Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved

methods for building protein models in electron density maps and the location

of errors in these models. Acta Crystallogr. A 47, 110–119.

Kabsch, W., Mannherz, H.G., Suck, D., Pai, E.F., and Holmes, K.C. (1990).

Atomic structure of the actin:DNase I complex. Nature 347, 37–44.

Khaitlina, S.Y., Moraczewska, J., and Strzelecka-Golaszewska, H. (1993). The

actin/actin interactions involving the N-terminus of the DNase-I-binding loop

are crucial for stabilization of the actin filament. Eur. J. Biochem. 218, 918–920.

Klenchin, V.A., Khaitlina, S.Y., and Rayment, I. (2006). Crystal structure of poly-

merization-competent actin. J. Mol. Biol. 362, 140–150.

Klenchin, V.A., Allingham, J.S., King, R., Tanaka, J., Marriott, G., and Rayment,

I. (2003). Trisoxazole macrolide toxins mimic the binding of actin-capping

proteins to actin. Nat. Struct. Biol. 10, 1058–1063.

Koradi, R., Billeter, M., and Wuthrich, K. (1996). MOLMOL: a program for

display and analysis of macromolecular structures. J. Mol. Graph. 14,

51–55, 29–32.

Laki, K., Maruyama, K., and Kominz, D.R. (1962). Evidence for the interaction

between tropomyosin and actin. Arch. Biochem. Biophys. 98, 323–330.

Liu, X., Shu, S., Hong, M.S., Levine, R.L., and Korn, E.D. (2006). Phosphoryla-

tion of actin Tyr-53 inhibits filament nucleation and elongation and destabilizes

filaments. Proc. Natl. Acad. Sci. USA 103, 13694–13699.

Lorenz, M., Popp, D., and Holmes, K.C. (1993). Refinement of the F-actin

model against X-ray fiber diffraction data by the use of a directed mutation

algorithm. J. Mol. Biol. 234, 826–836.

Matsuura, Y., Stewart, M., Kawamoto, M., Kamiya, N., Saeki, K., Yasunaga, T.,

andWakabayashi, T. (2000). Structural basis for the higher Ca(2+)-activation of

the regulated actin-activated myosin ATPase observed with Dictyostelium/

Tetrahymena actin chimeras. J. Mol. Biol. 296, 579–595.

Narita, A., Yasunaga, T., Ishikawa, T., Mayanagi, K., and Wakabayashi, T.

(2001). Ca(2+)-induced switching of troponin and tropomyosin on actin fila-

ments as revealed by electron cryo-microscopy. J. Mol. Biol. 308, 241–261.

Narita, A., and Maeda, Y. (2007). Molecular determination by electron micros-

copy of the actin filament end structure. J. Mol. Biol. 365, 480–501.


Noda, K., Nakamura, M., Nishida, R., Yoneda, Y., Yamaguchi, Y., Tamura, Y.,

Nakuamura, H., and Yasunaga, T. (2006). Atomic model construction of

protein complexes from electron micrographs and visualization of their 3D

structure using VR system. J. Plasma Phys. 72, 1037–1040.

Noguchi, T.Q., Kanzaki, N., Ueno, H., Hirose, K., and Uyeda, T.Q. (2007). A

novel system for expressing toxic actin mutants in Dictyostelium and purifica-

tion and characterization of a dominant lethal yeast actin mutant. J. Biol.

Chem. 282, 27721–27727.

Nolen, B.J., and Pollard, T.D. (2007). Insights into the influence of nucleotides

on actin family proteins from seven structures of Arp2/3 complex. Mol. Cell 26,

449–457.

Nonomura, Y., Katayama, E., and Ebashi, S. (1975). Effect of phosphates on

the structure of the actin filament. J. Biochem. 78, 1101–1104.

Nyman, T., Schuler, H., Korenbaum, E., Schutt, C.E., Karlsson, R., and Lind-

berg, U. (2002). The role of MeH73 in actin polymerization and ATP hydrolysis.

J. Mol. Biol. 317, 577–589.

Oda, T., Iwasa, M., Aihara, T., Maeda, Y., and Narita, A. (2009). The nature of

the globular- to fibrous-actin transition. Nature 457, 441–445.

Otterbein, L.R., Graceffa, P., and Dominguez, R. (2001). The crystal structure

of uncomplexed actin in the ADP state. Science 293, 708–711.

Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M.,

Meng, E.C., and Ferrin, T.E. (2004). UCSF Chimera—a visualization system

for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612.

Rickard, J.E., and Sheterline, P. (1986). Cytoplasmic concentrations of inor-

ganic phosphate affect the critical concentration for assembly of actin in the

presence of cytochalasin D or ADP. J. Mol. Biol. 191, 273–280.

Rosenthal, P.B., and Henderson, R. (2003). Optimal determination of particle

orientation, absolute hand, and contrast loss in single-particle electron cryomi-

croscopy. J. Mol. Biol. 333, 721–745.

Rould, M.A., Wan, Q., Joel, P.B., Lowey, S., and Trybus, K.M. (2006). Crystal

structures of expressed non-polymerizable monomeric actin in the ADP and

ATP states. J. Biol. Chem. 281, 31909–31919.

Saeki, K., Sutoh, K., and Wakabayashi, T. (1996). Tropomyosin-binding site(s)

on the Dictyostelium actin surface as identified by site-directed mutagenesis.

Biochemistry 35, 14465–14472.

Spudich, J.A., and Watt, S. (1971). The regulation of rabbit skeletal muscle

contraction. I. Biochemical studies of the interaction of the tropomyosin-

troponin complex with actin and the proteolytic fragments of myosin. J. Biol.

Chem. 246, 4866–4871.

Storoni, L.C., McCoy, A.J., and Read, R.J. (2004). Likelihood-enhanced fast

rotation functions. Acta Crystallogr. D Biol. Crystallogr. 60, 432–438.

Vorobiev, S., Strokopytov, B., Drubin, D.G., Frieden, C., Ono, S., Condeelis, J.,

Rubenstein, P.A., and Almo, S.C. (2003). The structure of nonvertebrate actin:

implications for the ATP hydrolytic mechanism. Proc. Natl. Acad. Sci. USA

100, 5760–5765.

Wriggers, W., and Schulten, K. (1999). Investigating a back door mechanism of

actin phosphate release by steered molecular dynamics. Proteins 35,

262–273.

Yasunaga, T., and Wakabayashi, T. (1996). Extensible and object-oriented

system Eos supplies a new environment for image analysis of electron micro-

graphs of macromolecules. J. Struct. Biol. 116, 155–160.

Yasunaga, T., and Wakabayashi, T. (2008). Evaluation of a 2k CCD camera

with an epitaxially grown CsI scintillator for recording energy-filtered electron

cryo-micrographs. J. Electron Microsc. (Tokyo) 57, 101–112.

Yonekura, K., Maki-Yonekura, S., and Namba, K. (2003). Complete atomic

model of the bacterial flagellar filament by electron cryomicroscopy. Nature

424, 643–650.


Nuclear Size Is Regulatedby Importin a and Ntf2 in XenopusDaniel L. Levy1 and Rebecca Heald1,*1Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA 94720-3200, USA


DOI 10.1016/j.cell.2010.09.012

SUMMARY

The size of the nucleus varies among different cell

types, species, and disease states, but mechanisms

of nuclear size regulation are poorly understood. We

investigated nuclear scaling in the pseudotetraploid

frog Xenopus laevis and its smaller diploid relative

Xenopus tropicalis, which contains smaller cells and

nuclei. Nuclear scaling was recapitulated in vitro

using egg extracts, demonstrating that titratable

cytoplasmic factors determine nuclear size to

a greater extent than DNA content. Nuclear import

rates correlated with nuclear size, and varying the

concentrations of two transport factors, importin

a and Ntf2, was sufficient to account for nuclear

scaling between the two species. Both factorsmodu-

lated lamin B3 import, with importin a increasing

overall import rates and Ntf2 reducing import based

on cargo size. Importin a also contributes to nuclear

size changes during early X. laevis development.

Thus, nuclear transport mechanisms are physiolog-

ical regulators of both interspecies and develop-

mental nuclear scaling.

INTRODUCTION

Cell size varies widely among different organisms, as well as

within the same organism in different tissue types and during

development, placing variable metabolic and functional

demands on internal organelles (Hall et al., 2004). A fundamental

question in cell biology is how organelle size is regulated to

accommodate cell size differences. Models proposed to

describe the regulation of organelle size can be broadly divided

into those involving static mechanisms, in which the amount or

size of one structural component determines the organelle’s

size, or dynamicmechanismswhereby feedback from the organ-

elle regulates its own assembly or balances rates of assembly

and disassembly (Marshall, 2002, 2008). Although these models

have been applied to size control of relatively simple linear struc-

tures like flagella (Wilson et al., 2008) and stereocilia (Manor and

Kachar, 2008), mechanisms that regulate the size of organelles

with more complex geometries have been difficult to elucidate.

The nucleus is a particularly important example of an organelle

that exhibits wide variations in size among eukaryotes, with

nuclear surface area spanning over two orders of magnitude

from budding yeast to certain amphibians (Maul and Deaven,

1977). Although correlations between ploidy and the size of the

nucleus are well documented (Cavalier-Smith, 2005; Fank-

hauser, 1939), when genetic and growth conditions were altered

in budding and fission yeasts, nuclear size varied with cell size

and not ploidy (Jorgensen et al., 2007; Neumann and Nurse,

2007). The functional significance of maintaining proper nuclear

morphology is unclear, but defects in nuclear size and shape

are associated with and diagnostic of human disease, including

cancer and other pathologies (Webster et al., 2009; Zink et al.,

2004).

Whereas molecular mechanisms that determine nuclear size

are largely unknown, structural components of the nucleus likely

play a role. Inmetazoans, the nuclear envelope (NE) is composed

of a double lipid bilayer perforated by nuclear pore complexes

(NPCs) that mediate nucleocytoplasmic transport. The outer

NE is continuous with the endoplasmic reticulum (ER), and the

inner NE is lined on the nucleoplasmic side with a meshwork of

lamin intermediate filaments constituting the nuclear lamina.

Lamin depletion reduces nuclear size (Jenkins et al., 1993; New-

port et al., 1990), and disease-causing mutations in lamins and

lamin-associated proteins alter nuclear size and shape (Dechat

et al., 2008). The NE breaks down prior to mitosis in most animal

cells, and upon its reformation, the nucleus expands in a process

that requires protein import (Neumann and Nurse, 2007; New-

port et al., 1990), accompanied by insertion of new NPCs (D’An-

gelo et al., 2006). The classical nuclear import pathway is medi-

ated by a family of importin a transport receptors that bind

nuclear localization signal (NLS)-containing proteins and impor-

tin b, the protein that directs translocation through the NPC.

Generation of Ran-GTP by its guanine exchange factor in the

nucleus ensures unidirectional import, as only Ran in its GTP-

bound state binds importin b, thereby releasing importin a and

NLS cargos within the nucleus. Importin b bound to Ran-GTP

is recycled to the cytoplasm, where nucleotide hydrolysis takes

place, and Ran-GDP is then imported by the dedicated transport

factor Ntf2, promoting another round of Ran-GTP production

and cargo release (Madrid and Weis, 2006; Stewart, 2007).

One approach to studying nuclear size control is to investi-

gate scaling, the phenomenon that nuclear size often correlates

with cell size. Two related frog species exemplify scaling: Xen-

opus laevis animals, cells, and eggs are larger than Xenopus

tropicalis (Horner and Macgregor, 1983). A significant advan-

tage of this system is that cell-free extracts prepared from Xen-

opus eggs reconstitute assembly of subcellular structures and


organelles in vitro, including the nucleus and mitotic spindle

(Maresca and Heald, 2006). Thus, it is possible to examine

intrinsic mechanisms of organelle scaling in a cell-free environ-

ment. By this approach, Xenopus species-specific scaling by

cytoplasmic factors has been demonstrated for the mitotic

spindle (Brown et al., 2007). Evidence for scaling of the nucleus

by cytoplasmic factors comes from experiments in fission yeast

showing that nuclear size correlated with the relative amount of

surrounding cytoplasm (Neumann and Nurse, 2007).

In this study, we demonstrate that nuclear size scales

between X. laevis and X. tropicalis and that titratable cyto-

plasmic factors influence nuclear size to a greater extent than

DNA content. We find that importin a and Ntf2 levels mediate

interspecies nuclear scaling, at least in part, by regulating the

import of lamin B3. Whereas importin a regulates the overall

import rate of NLS cargos, Ntf2 modulates import based on

cargo size. We further demonstrate that nuclear size scales

during early X. laevis development and that, similar to our find-

ings in egg extracts, changes in nuclear import and importin

a levels contribute to these developmentally regulated nuclear

size changes.

RESULTS

Nuclear Size and Import Scale between X. laevis

and X. tropicalis In Vitro

Nuclei were assembled in X. laevis and X. tropicalis egg extracts,

using X. laevis sperm as the chromatin source. At different time

points, nuclei were fixed, visualized by immunofluorescence

with an antibody against the NPC (Figure 1A), and quantified

for NE surface area (Figure 1B). Nuclei assembled within

30–40 min after chromatin addition and were initially similar in

size in both extracts but, over time, grew larger in X. laevis extract

compared to X. tropicalis. Though nuclei in these extracts do not

attain a steady-state size, X. tropicalis nuclei never reach the size

of X. laevis nuclei. Extracts prepared from different batches of

eggs exhibited some variability, but analysis of five extracts for

each species yielded an average NE expansion rate of

70 ± 9 mm2/min in X. laevis and 30 ± 9 mm2/min in X. tropicalis

(mean ± SD, Figure 1B). On average, NE surface area was

2.3-fold greater in X. laevis extract compared to X. tropicalis.

Similar interspecies nuclear growth differences were observed

in live samples by time-lapse fluorescence microscopy visual-

izing nuclear import of green fluorescent protein (GFP) fused to

the classical SV40 NLS (Movie S1 and Figure S1A available on-

line). To address whether continual nuclear expansion was

a peculiarity of the extract system, we measured nuclear size

over time in early cleavage stage X. laevis embryos. Nuclei

expanded in vivo at a rate comparable to that of egg extracts

and failed to reach a steady-state size in arrested embryos

(Figure S1B), demonstrating that extracts faithfully recapitulate

nuclear dynamics in the early embryo, where cell-cycle timing

sets the limit for nuclear growth.

Mixing the two extracts at different ratios produced a graded

effect on nuclear size (Figure 1C), suggesting that neither extract

possesses dominant activating or inhibitory factors. Addition of

extract fractionated by high-speed centrifugation to preassem-

bled nuclei revealed that cytosol had a greater effect on nuclear

size than membrane (data not shown). When nuclei were formed

with reduced DNA content, using X. tropicalis sperm with 55%

the DNA of X. laevis sperm, only an average 12% reduction in

nuclear surface area was observed (Figure 1D). Taken together,

these results demonstrate that, in this system, titratable cyto-

plasmic factors determine nuclear size to a greater extent than

the amount of nuclear DNA. X. laevis sperm nuclei were used

in all subsequent egg extract experiments, and the species

denotes whether nuclei were formed in X. laevis or X. tropicalis

extracts.

Of interest, we observed that GFP-NLS accumulated at

a faster rate and to a greater overall extent in X. laevis nuclei

compared to X. tropicalis in both live and fixed samples

(Figure 1E, Figure S1C and S1D, and Movie S1). To elucidate

this difference in nuclear import capacities between the two

species, we first considered their nuclear pores. During early

NE expansion, the total NPC number was similar in X. laevis

and X. tropicalis nuclei, with a slightly higher density in X. tropica-

lis (Figure S1E and S1F). Because nuclear growth is accompa-

nied by new NPC insertion (D’Angelo et al., 2006), the total

NPC number increased more over time in X. laevis nuclei than

in X. tropicalis nuclei, whereas the NPC densities remained

comparable (Figures S1E and S1F). Whereas NPC number did

not correlate with nuclear size during early NE expansion, there

was a marked difference in their import properties. Large cargos

consisting of streptavidin-conjugated quantum dots (Qdots)

coated with a biotin-labeled domain of snurportin-1 that binds

importin b (Lowe et al., 2010) were efficiently imported into X. lae-

vis nuclei but failed to accumulate in X. tropicalis nuclei over time,

although they localized to the NE (Figure 1F). These 40 ± 9 nm

diameter particles are similar in size to a 20 megadalton macro-

molecule. Thus, X. laevis nuclei are capable of importing larger

cargos than X. tropicalis and have a higher overall import

capacity for NLS-containing proteins.

Importin a2 and Ntf2 Levels Differ between X. laevis

and X. tropicalis

Given the observed nuclear import differences in the two

extracts, we measured the relative amounts of nucleocytoplas-

mic transport proteins by western blot and immunofluorescence

to determine whether concentrations of any of these proteins

correlated with import. Whereas the levels of many transport

factors, including Ran, RanGAP, RanBP1, and Cas, were similar

in the two extracts, the concentration of the predominant impor-

tin a isoform, a member of the importin a2 subfamily, was 3-fold

higher in X. laevis compared to X. tropicalis (Figure 2A and

Figure S2). Levels of importin a1, importin a3, and importin

b were also higher in X. laevis but to a lesser degree (Figure 2A

and Figure S2). Furthermore, X. laevis nuclei stained more

intensely for importin a2 and importin b than X. tropicalis nuclei

(Figures 2B and 2C).

In contrast, Ntf2 showed the opposite trend compared to im-

portin a, with levels almost 4-fold higher in X. tropicalis extract,

and more intense Ntf2 nuclear staining (Figure 2). As Ntf2 is the

nuclear import factor dedicated to recycling Ran-GDP from the

cytoplasm to the nucleus (Smith et al., 1998), these higher Ntf2

levels likely explain why nuclear Ran was greater in X. tropicalis

compared to X. laevis, even though total Ran levels were similar


(Figure 2). The marked differences in importin a2 and Ntf2

concentrations led us to investigate their relevance to nuclear

scaling between the two species.

Importin a2 Increases Nuclear Size and Import

First, we altered importin a levels. Endogenous importin a must

be phosphorylated to diffuse freely in Xenopus cytoplasm while

the unphosphorylated form binds large membrane stores

present in egg extracts (Hachet et al., 2004), possibly rendering

it unable to engage in nucleocytoplasmic transport. We there-

fore tested the effects of a phosphomimetic importin a-E con-

taining six glutamate point mutations (Hachet et al., 2004), as

well as in vitro phosphorylated importin a (Hachet et al.,

2004). When added to nuclei assembled in X. tropicalis egg

extract, both of these proteins increased NE surface area,

whereas unphosphorylated importin a had little effect (Fig-

ure S3A). The maximal change in nuclear size occurred in the

range of 0.8–1 mM added importin a-E, increasing NE surface

area 1.5- to 1.7-fold (Figure 3A). Importin a-E likely affects

nuclear size by modulating import because its addition

increased nuclear accumulation of GFP-NLS and addition of

import-defective importin a-E lacking the N-terminal importin

b-binding (IBB) domain failed to increase nuclear size (Figure 3B

and Figure S3A).

A B

C D

E F

Figure 1. Nuclear Size and Import Scale between X. laevis and X. tropicalis

(A) Nuclei were assembled in X. laevis or X. tropicalis egg extract with X. laevis sperm and visualized by immunofluorescence using mAb414 that recognizes the

NPC. Scale bar, 20 mm.

(B) NE surface area was quantified from images like those in (A) for at least 50 nuclei at each time point. Best-fit linear regression lines are displayed for six X. laevis

and five X. tropicalis egg extracts, and the average difference between the two extracts was statistically significant by Student’s t test (p < 0.001). R2 values range

from 0.96 to 0.99 for X. laevis and 0.94 to 0.98 for X. tropicalis. Error bars represent standard deviation (SD).

(C) X. laevis and X. tropicalis extracts were mixed as indicated, and nuclear size was measured at 90 min. One representative experiment of three is shown, and

error bars represent SD.

(D) Nuclei were assembled using the indicated source of extract and sperm, and nuclear size was measured at 90 min. One representative experiment of three is

shown, and error bars represent SD.

(E) GFP-NLSwas added to nuclei at 30min, and imageswere acquired live at 30 s intervals with the same exposure time. Nuclear GFP-NLS fluorescence intensity

per unit area was measured at each time point, averaged for five nuclei from each extract, and normalized to 1.0 (arbitrary units). Error bars represent SD. Repre-

sentative images are at 70 min. Scale bar, 20 mm.

(F) IBB-coated Qdots were added to nuclei at 30min, and imageswere acquired live at the indicated time points for at least 30 nuclei with the same exposure time.

Nuclear Qdot fluorescence intensity per unit area was calculated, averaged, and normalized to 1.0 (arbitrary units). Error bars represent SD. One representative

experiment of three is shown. Representative images are at 75 min. Scale bar, 20 mm.

See also Figure S1 and Movie S1.


In complementary experiments, importin a2 was partially

immunodepleted from X. laevis extracts. Depending on the

extract, 0.5–1 mM importin a2 was depleted, and no other

proteins were stoichiometrically codepleted (data not shown).

Compared to mock-depleted extracts, lowering the importin

a concentration reduced nuclear size and GFP-NLS import,

and both effects could be rescued by addition of importin a-E

but only if it was import competent with an intact IBB domain

(Figure 3C). Addition of excess importin a-E to X. tropicalis nuclei

(Figure 3A) or X. laevis nuclei (data not shown) slightly reduced

nuclear size while minimally affecting bulk import.

To address the specificity of the importin a effect and to deter-

mine whether other import factors contribute to nuclear sizing,

we added importin a-E, importin b, and Ran alone and in combi-

nation to nuclei assembled in X. tropicalis extract. At 0.8 mM,

importin b negatively affected nuclear size, Ran had no effect,

and no combination with importin a-E increased nuclear size to

a greater extent than importin a-E alone (Figure S3B). At 4 mM,

all three proteins individually reduced nuclear size, and no

combination increased size (Figure S3B). We also investigated

a different Ran-regulated nucleocytoplasmic shuttling pathway

that utilizes the transport receptor transportin. Addition of re-

combinant transportin to X. tropicalis nuclei negatively affected

nuclear size at all concentrations tested, likely by interfering

with other Ran-mediated transport (Figure S3C). Furthermore,

transportin levels were indistinguishable between X. laevis and

X. tropicalis (Figure S2), as was nuclear import of YFP-

M9-CFP, a transportin cargo (Figure S3D). We conclude that

nuclear scaling acts predominantly through the NLS-mediated

import pathway, in particular through importin a.

Ntf2Decreases Nuclear Size and Import of LargeCargos

Although importin a contributes to nuclear scaling, its effect was

insufficient to explain the average 2.3-fold size difference

between X. laevis and X. tropicalis nuclei. Because Ntf2 was

the only other import factor we identified that differed signifi-

cantly between the two extracts (Figure 2), being more abundant

in X. tropicalis, recombinant Ntf2 was titrated into X. laevis

extract. Increasing the Ntf2 concentration increased nuclear

Ran, consistent with functional Ntf2 directing Ran import,

and nuclear size was concomitantly reduced (Figure 3D). When

1.6 mM Ntf2 was added to X. laevis extract to approximate the

total Ntf2 concentration in X. tropicalis, nuclear Ran staining

increased to nearly the X. tropicalis level, but NE surface area

was not fully reduced to that of X. tropicalis. Intriguingly,

GFP-NLS import did not correlate with nuclear size. In fact, addi-

tion of Ntf2 slightly increased nuclear GFP-NLS levels, perhaps

due to the higher nuclear Ran concentration (Figure 3D). This

result suggested that Ntf2 was not affecting nuclear size by

altering the global NLS import rate. Instead, increasing the Ntf2

concentration in X. laevis reduced the amount and rate of Qdot

import (Figures 3D and 3E). Because Ntf2 binds proteins of the

NPC (Clarkson et al., 1996), higher Ntf2 levels may occlude the

pore, potentially impeding import of larger particles like Qdots,

A B

C

Figure 2. Importin a2 and Ntf2 Levels Differ in X. laevis and X. tropicalis

(A) 25 mg of protein from three different X. laevis and X. tropicalis egg extracts was separated by SDS-PAGE, transferred to nitrocellulose, and probed with anti-

bodies against the indicated proteins. Values below each set of three lanes represent relative protein amounts (mean ± SD, n = 3) quantified by infrared fluores-

cence. Absolute concentrations were determined by comparing band intensities to known concentrations of recombinant importin a2 or Ntf2 on the same blot.

Two different antibodies against importin a2 and Ntf2 showed similar differences between the two species.

(B) Nuclei at 80 min were processed for immunofluorescence using the same antibodies as in (A), and representative images are shown. For a given antibody,

images were acquired with the same exposure time and scaled identically. Scale bar, 20 mm.

(C) Quantification of nuclei displayed in (B). Nuclear fluorescence intensity per unit area was calculated for at least 50 nuclei per condition, averaged, and normal-

ized to 1.0 (arbitrary units). Error bars represent SD. Two different antibodies against importin a2 and Ntf2 showed similar differences between the two species.

See also Figure S2.


A D

B E

C F

Figure 3. Importin a2 and Ntf2 Regulate Nuclear Size and Import

(A) Nuclei were assembled in X. tropicalis extract, and at 40 min, importin a-E was added at the indicated concentrations in addition to GFP-NLS. At 80 min,

images for at least 50 nuclei per condition were acquired with the same exposure time, and NE surface area was quantified, averaged, and normalized to the

buffer control. Error bars represent standard error (SE). Scale bar, 20 mm.

(B) Experiments were performed as in (A) with a fixed concentration (0.8 mM) of added importin a-E or a mutant version lacking the importin b-binding domain

(DIBB). Average fold change from the buffer control and SD are shown (n = 4 extracts). TheDIBBmutant did not have a strong dominant-negative effect on import

because it was added at a concentration below the endogenous importin a level.

(C) Nuclei were assembled in X. laevis extract mock and partially immunodepleted of importin a2 (0.5–1 mMdepleted). Kinetics of nuclear assembly were similar in

the two extracts. At 40 min, indicated proteins were added at 1 mM as well as GFP-NLS. At 80 min, images for at least 50 nuclei per condition were acquired with

the same exposure time, and NE surface area and nuclear GFP-NLS fluorescence intensity were quantified. Average fold change from themock depletion and SD

are shown (n = 4 extracts).

(D) Recombinant Ntf2 was titrated into X. laevis extract prior to nuclear assembly. Initial kinetics of nuclear assembly were not altered by supplemental Ntf2.

GFP-NLS or IBB-coated Qdots were added at 30 min. At 80 min, nuclei were processed for immunofluorescence with an antibody against Ran, and images

for at least 50 nuclei per condition were acquired with the same exposure time. NE surface area was quantified from Ran-stained nuclei, averaged, and normal-

ized to the buffer control. Nuclear fluorescence intensities for Qdots, GFP-NLS, and Ran were similarly processed. Error bars represent SE. One representative

experiment of three is shown. For each parameter, the difference between 0 and 1.6 mM added Ntf2 was statistically significant by Student’s t test (p < 0.005).

(E) Experiments similar to (D) were performed with a fixed Ntf2 concentration (1.6 mM) and over time. Nuclear Qdot or GFP-NLS fluorescence intensities for at

least 50 nuclei per time point were averaged and normalized to 1.0 (arbitrary units). Error bars represent SE. At 95 min, the difference in Q dot import between

0 and 1.6 mM added Ntf2 was statistically significant by Student’s t test (p < 0.001).

(F) Nuclei were assembled in X. tropicalis extract supplemented with anti-Ntf2 or IgG antibodies (0.1 mg/ml). At 30 min, nuclear assembly was similar in the two

conditions, and Qdots or GFP-NLS was added. At 80 min, immunofluorescence for Ran was performed, and nuclear parameters were quantified as in (D).

Average fold change from the IgG control and SD are shown (n = 6 extracts).

See also Figure S3.


but not small cargos like GFP-NLS. Consistent with this model,

reducing the effective Ntf2 concentration in X. tropicalis by anti-

body inhibition (Figure 3F) or Ntf2 depletion (Figure S3E)

conferred on these nuclei the ability to import Qdots without

significantly altering GFP-NLS import. Concomitantly, NE

surface area increased 1.4- to 1.5-fold, and nuclear Ran staining

decreased on average 11% (Figure 3F and Figure S3E). Taken

together, these data are consistent with Ntf2 regulating nuclear

size by modulating the import rates of large cargos that presum-

ably contribute to nuclear sizing.

Importin a2 and Ntf2 Scale Nuclear Size through Lamin

B3 Import

Because both addition of importin a and inhibition of Ntf2 inX. tro-

picalis increasednuclear size,we testedwhether combining these

manipulations was sufficient to convert X. tropicalis nuclei to the

size ofX. laevisnuclei. Averagedover five experiments,X. tropica-

lisNE surface area increased 2.2-fold with supplemental importin

a-EandNtf2 inhibition (Figure4A),nearlyequivalent to theaverage

2.3-fold interspecies nuclear size difference (Figure 1B).

Importin a and Ntf2 could control nuclear size by regulating

either bulk import of NLS cargos or import of specific structural

components of the nucleus. To differentiate between these

possibilities, we supplemented X. tropicalis extract with different

NLS cargos to specifically increase their import and assessed

the effect on nuclear size. Addition of nucleoplasmin (Npl) or

GFP-NLS, both importin a cargos, did not significantly alter

nuclear size over a wide range of concentrations (Figure 4B, Fig-

ure S4A, and data not shown). In contrast, recombinant lamin B3

(LB3) titrated into X. tropicalis extract increased NE surface area

1.7-fold when added at an optimal concentration of 0.14 mM

(Figure 4B and Figure S4A). LB3 is the major Xenopus egg lamin

A B

C D

Figure 4. Importin a2 and Ntf2 Are Sufficient to Account for Interspecies Nuclear Scaling by Regulating LB3 Import

(A) X. tropicalis nuclei were assembled in the presence of anti-Ntf2 or IgG control antibodies (0.1mg/ml) and 0.14 mMGFP-LB3 as indicated, and at 40min, 0.8 mM

importin a-E was added to some reactions. LB3 was visualized in nuclei by immunofluorescence at 80 min, and images for at least 50 nuclei per condition were

acquired with the same exposure time. NE surface area and LB3 fluorescence were quantified. Average fold change from the buffer control and SD are shown

(n = 5 extracts). Scale bar, 20 mm.

(B) Wild-type and mutant GFP-LB3 proteins, 13 Npl (nucleoplasmin), and GFP-NLS were added at 0.14 mM to X. tropicalis extract. For 53 Npl, 0.7 mMNpl was

added. Nuclei were visualized at 75 min by immunofluorescence using an antibody against Ran. NE surface area was calculated for at least 50 nuclei. Average

fold change from the buffer control and SD are shown (n = 3 extracts). The K31Q mutant had a dominant-negative effect on the structure of the lamina, as nuclei

were smaller and appeared crumpled, whereas the R385P mutant did not efficiently assemble into the lamina.

(C) Nuclei were visualized by immunofluorescence with an antibody against Xenopus LB3. Images for at least 50 nuclei at each time point were acquired with the

same exposure time. Fluorescence intensity was quantified, averaged, and normalized to 1.0 (arbitrary units). Error bars represent SD. One representative exper-

iment of three is shown. The Western blot was performed as in Figure 2A using an antibody against Xenopus LB3.

(D) Nuclei were assembled in X. tropicalis extract mock- and immunodepleted of LB3 (0.1 mMdepleted). Ntf2 antibodies, importin a-E, and GFP-LB3 were added

to LB3-depleted extract in the same manner as in (A) with the exception that GFP-LB3 was added at 0.2 mM. At 80 min, nuclei were stained for Ran by immu-

nofluorescence, images for at least 50 nuclei per condition were acquired, and NE surface area was quantified. Average fold change from themock depletion and

SD are shown (n = 4 extracts). Scale bar, 20 mm.

See also Figure S4 and Table S1.


that is required for NE growth (Jenkins et al., 1993; Newport

et al., 1990) and contains a classical NLS (Loewinger and

McKeon, 1988). At higher concentrations of LB3, nuclear size

was reduced and LB3 puncta were visible, likely representing

the formation of aggregates unable to assemble into a functional

lamina (Figure S4A). Addition of two LB3 point mutants previ-

ously shown to be defective for lamina assembly (Heald and

McKeon, 1990), or LB3 with a mutated NLS, failed to increase

nuclear size (Figure 4B). These data indicate that the concentra-

tion of the specific cargo LB3 can determine nuclear size, depen-

dent on its import and functional assembly.

Because LB3 concentration can affect nuclear size, we

compared LB3 import and levels in the two Xenopus extracts.

Although the rate of nuclear LB3 accumulation in X. tropicalis

extract was 35% the rate in X. laevis, the LB3 concentration

was 2-fold higher in X. tropicalis (Figure 4C), consistent with

the X. laevis egg containing 2.1-fold more total LB3 in

a 4.3-fold larger volume (Table S1). Nuclear size differences in

these two extracts therefore correlate not with lamin concentra-

tion but, rather, with the rate of lamin import as regulated by

importin a and Ntf2. Consistent with this interpretation, upon

addition of importin a and inhibition of Ntf2 in X. tropicalis that

led to increased LB3 import, supplemental LB3 did not further

increase nuclear size (Figure 4A and Figures S4B and S4C).

Conversely, Ntf2 reduced import of LB3 in a concentration-

dependent manner in X. laevis (Figure S4D). Furthermore,

addition of importin a and/or Ntf2 antibodies to LB3-depleted

X. tropicalis extracts had little effect on nuclear size

(Figure 4D), even though these nuclei were still import competent

for GFP-NLS (data not shown). Taken together, these data argue

that differences in importin a and Ntf2 concentrations can

account for nuclear scaling between X. laevis and X. tropicalis

and that they control nuclear size by regulating import of LB3

and possibly other NLS cargos that require an intact lamina to

function.

Nuclear Scaling during Xenopus Development

Is Also Regulated by Importin a

To investigate whether mechanisms of interspecies nuclear

scaling also operate during development, we turned to X. laevis

embryos. Upon fertilization, the 1 mm diameter egg undergoes

12 rapid, synchronous cell divisions (each 30min) with no over-

all growth, generating about four-thousand 50 mm cells at the

midblastula transition (MBT) or stage 8 (Nieuwkoop and Faber,

1967). After the MBT, zygotic transcription initiates, cells

become motile, and cell divisions slow and lose synchrony. As

the embryo proceeds through gastrulation, further reductions

in cell size occur, reaching 12 mm in the tadpole (Montorzi

et al., 2000). Xenopus embryogenesis therefore offers a robust

model for developmental scaling.

Nuclear sizewas quantified inX. laevis embryos by immunoflu-

orescence (Figure 5A). Because nuclei continually expand in

early embryos (Figure S1B), we compared different stage

embryos arrested for 60 min. Though nuclear sizes were similar

during the first few cell divisions after fertilization, NE surface

area became progressively smaller after stage 5 (16 cell) and

through stage 10 (gastrulation), reaching a relatively constant

size in stage 12 and later embryos. Measurements made in situ

were comparable (Figure S5A). A similar trend in nuclear size

changes was observed in X. tropicalis embryos except NE

surface area was on average 51% less than X. laevis at equiva-

lent developmental stages (Figure S5B). Halving the DNA

content in X. laevis embryonic nuclei only reduced NE surface

area by 10%, demonstrating that, like egg cytoplasm, embryo

cytoplasm determines nuclear size to a greater extent than

ploidy (Figure S5C).

To investigate whether nucleocytoplasmic transport also

regulates nuclear scaling during early X. laevis development,

we examined the levels of transport factors. Strikingly, total

importin a2 levels dropped 47% by stage 8 (MBT) relative to

earlier stages and a further 30% by stage 12 (Figure 5B). In

contrast, importin a1, importin a3, Ran, and Ntf2 concentrations

remained relatively constant (Figure 5B and Figure S5D). At

stage 8, concomitant reductions in GFP-NLS import capacity

and nuclear importin a2 and Ntf2 staining occurred, whereas,

at stage 12, import was reduced further but with no significant

change in nuclear importin a2 and Ntf2 (Figures 5C and 5D).

To determine whether importin a directly modulates nuclear

size during development, fertilized one-cell X. laevis embryos

were injected with mRNA encoding importin a-E and were

allowed to develop to later stages. Exogeneous expression of

importin a-E to 0.6 mM ± 0.2 mM (mean ± SD, n = 5) significantly

increased nuclear size in stage 7 and stage 8 embryos to the

range observed in early stage embryos but had a lesser effect

at stage 9 (Figure 5E). Increasing nuclear size in embryos did

not affect their grossmorphology or viability. Addition of importin

a-E to embryo extracts similarly increased nuclear size

(Figure S5E) and also increased GFP-NLS import (data not

shown), whereas Ntf2 addition had little effect (data not shown).

Of interest, we observed that nuclei in stage 7 and stage 8

embryos reached a steady-state size (Figure 5F), unlike earlier

in development (Figure S1B). Overexpression of importin a-E in

stage 7 embryos led to continuous nuclear expansion similar

to that observed in early stages, whereas nuclei in stage 8

embryos grew larger but attained a new equilibrium size,

suggesting that other factors became limiting at the MBT

(Figure 5F). Taken together, these data demonstrate that impor-

tin a is one factor that mediates nuclear scaling during X. laevis

embryogenesis, affecting both the rate of nuclear expansion in

early embryos and the steady-state nuclear size in later

embryos.

DISCUSSION

We investigated how nuclear size is regulated in two related but

different sized frog species as well as during early frog develop-

ment, two physiological examples of nuclear scaling. Using

Xenopus egg extracts to examine intrinsic mechanisms of

nuclear scaling in the absence of the cell showed that titratable

cytoplasmic factors regulate nuclear size to a greater extent

than DNA content and that differences in the concentrations of

importin a and Ntf2 are sufficient to explain most of the observed

interspecies nuclear scaling by altering nuclear import. Importin

a, but not Ntf2, also plays a role in nuclear scaling during

embryogenesis in X. laevis. Whereas nucleocytoplasmic trans-

port was known to be required for NE growth (D’Angelo et al.,


2006; Neumann and Nurse, 2007; Newport et al., 1990), our data

show that titrating nuclear import concomitantly scales nuclear

size and that this mechanism can account for how the size of

the nucleus is controlled in two frog species and during develop-

ment.

Importin a mediates nuclear scaling by regulating overall

import of NLS cargos, consistent with computer modeling and

cell culture experiments showing that importin a concentration

positively correlates with the rate and steady-state level of

nuclear import (Gorlich et al., 2003; Riddick and Macara, 2005,

2007; Smith et al., 2002). However, our results indicate a more

complex relationship between nuclear import factors and

nuclear size. For example, we observe that increasing importin

a concentration more than 3-fold over normal levels reduces

nuclear size (Figure 3A and Figure S3B), probably because

elevated lamin B3 import that occurs under these conditions

(data not shown) is detrimental to nuclear assembly

(Figure S4A). Ntf2 has also been implicated as a positive

A B

C

D E

F

Figure 5. Importin a2 Regulates X. laevis Developmental Nuclear Scaling

(A) Different stage X. laevis embryos were arrested with cycloheximide for 60 min. Nuclei were isolated from embryo extracts and visualized by immunofluores-

cence using mAb414. Scale bar, 20 mm. For the graph, NE surface area was quantified for at least 50 nuclei from each stage. Error bars represent SD.

(B) 25 mg of protein from different stage embryo extracts was analyzed by western blot, as in Figure 2A.

(C) To assess nuclear import, GFP-NLS (1 mM) was added to embryo extracts, and images of unfixed nuclei were acquired 30 min later with the same exposure

time. Immunofluorescence was performed on fixed embryonic nuclei, and images were acquired with the same exposure time. Scale bar, 20 mm.

(D) Quantification of (C). Nuclear fluorescence intensity per unit area was calculated for at least 50 nuclei per stage, averaged, and normalized to 1.0 (arbitrary

units). Error bars represent SD.

(E) Single-cell fertilized X. laevis embryos were injected with 1 ng importin a-E mRNA or water as control. Nuclei were isolated and quantified as in (A), except that

an antibody against Xenopus LB3was used for immunofluorescence. One representative experiment of two is shown for each stage, and error bars represent SD.

(F) Experiments similar to (E) were performed except that embryos were arrested for different lengths of time in cycloheximide. Error bars represent SE. Repre-

sentative stage 7 nuclei at 120 min are shown for control (bottom) and importin a-E (top) injected embryos. Scale bar, 20 mm.

See also Figure S5.


regulator of both Ran and bulk import (Riddick and Macara,

2005, 2007). Though the Ntf2-Ran relationship holds true in our

experiments, we find that increased Ntf2 slows import of large

cargos such as Qdots, but not smaller proteins like GFP-NLS.

Because it associates with the NPC, Ntf2 could influence import

rates based on cargo size (Clarkson et al., 1996). In fact, studies

of X. laevis oogenesis revealed that late-stage oocytes acquire

the ability to import large nucleoplasmin-coated gold particles

concomitantly with a reduction in Ntf2 levels (Feldherr et al.,

1998). Furthermore, addition of Ntf2 to those oocytes reduced

import of gold particles, similar to our observation that increasing

the Ntf2 concentration in X. laevis reduced Qdot import

(Figure 3D). It is worth noting that supplementing X. laevis extract

with Ntf2 up to the X. tropicalis level slowed but did not block

Qdot import, suggesting that other interspecies NPC differences

may affect cargo size-dependent import.

Nuclear size appears to be determined by import of specific

NLS cargos, not by mass action transport. LB3 was a good

candidate because its import is importin a-mediated, it is

required for NE expansion (Jenkins et al., 1993; Newport et al.,

1990), and its overexpression induces proliferation of nuclear

membrane (Goldberg et al., 2008; Prufert et al., 2004). Addition

of LB3, but not Npl or GFP-NLS, to X. tropicalis egg extract

increased nuclear size, but not to the size of X. laevis, suggesting

that additional NLS proteins are involved. Potential nuclear sizing

cargos include inner nuclear membrane proteins that interact

with the lamina, like the lamin B receptor and LAPs, as well as

SUN and KASH family proteins that span the NE and mediate

interactions between the nucleus and cytoskeleton. Consistent

with this idea, NPC manipulations that increase translocation

of integral membrane proteins to the inner NE correlate with

increased nuclear size (Theerthagiri et al., 2010). The fact that

Qdot import positively correlates with nuclear size indicates

that cargos important for scaling are relatively large. Although

lamin monomers are only 70 kD, they minimally form tetramers

made up of two dimers, each composed of 50 nm elongated

coiled coils (Aebi et al., 1986; Heitlinger et al., 1991). Because

particles as large as 20 megadaltons can transit the X. laevis

NPC, LB3 may be imported as large oligomers.

We discovered some striking similarities between interspe-

cies nuclear size regulation and nuclear scaling during Xenopus

embryogenesis. Reductions in nuclear size during development

were accompanied by diminishing import capacity for NLS

cargos, and scaling of nuclear size at the MBT correlated

with a drop in total and nuclear importin a levels. Increasing

the concentration of importin a in embryos increased nuclear

size without noticeably affecting development, suggesting

that nuclear size per se does not regulate early developmental

transitions. Thus, conserved importin a-mediated transport

mechanisms regulate nuclear size both during development

and between frog species, but distinct and yet uncharacterized

mechanisms also contribute to nuclear scaling in Xenopus

embryogenesis.

Our data suggest two nuclear sizing regimes determined by

either reaction rates or abundance of NE components. The egg

is stockpiled in order to form 4000 MBT nuclei, and therefore

these components are not limiting in egg extracts and early

embryos. In this regime, nuclear size is determined by rates of

NE expansion and nuclear import in conjunction with cell-cycle

timing. In contrast, MBT nuclei reach a steady-state size when

import and NE components like lamins are no longer in excess.

Consistent with this idea, increasing importin a expression in

MBT embryos caused nuclei to reach a new steady-state size

(Figure 5F) at which lamins became limiting because coexpress-

ing importin a and LB3 further augmented nuclear size (data not

shown). Of interest, the amount of LB3 loaded into the eggs of

each species correlates well with the total NE surface area at

the MBT, with X. laevis containing 2.1-fold more total LB3 than

X. tropicalis at the onset of development and 2-fold more NE at

the MBT when transcription starts (Table S1). Because the ratio

of NE surface area to embryo volume at this transition is

2.1-fold higher in the smaller X. tropicalis species (Table S1),

the starting LB3 concentration in the egg is also about 2-fold

higher (Figure 4C). ThusXenopus eggs are loadedwith the proper

amount of LB3, and presumably other nuclear envelop compo-

nents, so that they are not limiting during the rapid divisions of

early development.

Our results are consistent with multiple mutually nonexclusive

models of organelle size control. Considering a static model,

importin a and Ntf2 levels limit nuclear import of LB3, thereby

constraining the rate at which nuclei expand. However, dynamic

processes must balance import-mediated growth. Nuclear size

is a regulated cellular parameter that depends on tissue type, de-

velopmental state as demonstrated during Xenopus embryogen-

esis, and species as shown comparing X. laevis and X. tropicalis,

in which nuclear size differences have evolved by fine-tuning the

expression of nuclear import factors. A fundamental question is

why nuclear size is regulated. Changes in the dimensions and

morphology of the nucleus are associated with pathologies,

including cancer (Webster et al., 2009; Zink et al., 2004), but dis-

secting the cause and effect relationship between nuclear size

and disease state is difficult. Understanding the role that nuclear

import plays in scaling nuclear size and identifying relevant

factors and their mechanisms of action provide an avenue to

directly manipulate nuclear size in the context of normal and

diseased cells in order to examine the functional consequences.


Xenopus Egg Extracts and Nuclear Assembly

X. laevis (Maresca and Heald, 2006) and X. tropicalis (Brown et al., 2007) meta-

phase-arrested egg extracts andXenopus sperm (Murray, 1991) weremade as

previously described. The standard nuclear assembly reaction was 25 ml fresh

extract, 100 mg/ml cycloheximide, 1000 Xenopus sperm per ml, and 0.5 mM

CaCl2. X. laevis sperm was used in all experiments except Figure 1D, in

which X. tropicalis sperm was used, as indicated. Reactions were

incubated at 19!C –22!C and import-competent nuclei generally formedwithin

30–40 min.

To monitor nuclear import, GFP-NLS (1 mM), YFP-M9-CFP (1 mM), or IBB-

Qdots (10 nM) were added to nuclei. IBB-Qdots were prepared by mixing

20 mM biotin-labeled IBB-CFP (a gift from Alan Lowe and Jan Liphardt)

with 1 mM Qdot 605 streptavidin conjugate (Invitrogen) at a 1:1 ratio and

incubating on ice 15 min. We also examined import of three smaller IBB-

Qdots using Qdots 525, 565, and 585 streptavidin conjugates (Invitrogen),

finding that all three were imported into X. laevis and X. tropicalis nuclei

(data not shown).

Immunodepletions and recombinant proteins are detailed in the Extended

Experimental Procedures. Proteins and antibodies were dialyzed into XB


(100mMKCl, 1mMMgCl2, 0.1 mMCaCl2, 50mM sucrose, and 10mMHEPES

[pH7.7]) and added to extracts prior to nuclear assembly, except for importin a,

whichwas dialyzed into 300mMKCl, 10mMMgCl2, and 10mMHEPES (pH7.8)

and added to preformednuclei. Total volume of additionwas less than 10% the

reaction volume, and buffer and IgG controls were performed. Reactions were

allowed to proceed to 75–90 min, as nuclear size at these time points was

similar to the size of nuclei in early stage embryos, thus providing a physiolog-

ically relevant situation for comparing nuclear size changes.

Xenopus Embryos and Extracts

Xenopus embryos were obtained as previously described (Grammer et al.,

2005; Sive et al., 2000), and details on how they were generated and injected

are in the Extended Experimental Procedures. Embryos were arrested in late

interphase with 150 mg/ml cycloheximide for 60min unless indicated otherwise

(Lemaitre et al., 1998), washed several times in ELB (250 mM sucrose, 50 mM

KCl, 2.5 mM MgCl2, and 10 mM HEPES [pH 7.8]) containing LPC (10 mg/ml

each leupeptin, pepstatin, chymostatin), cytochalasin D (100 mg/ml),

and cycloheximide (100 mg/ml), packed in a tabletop centrifuge at 200 g for

1 min, crushed with a pestle, and centrifuged at 10,000 3 g for 10 min at

16 C. The cytoplasmic extract containing endogenous embryonic nuclei

was supplemented with LPC, cytochalasin D (20 mg/ml), cycloheximide

(100 mg/ml), and energy mix (3 mM creatine phosphate, 0.4 mM adenosine

triphosphate, 40 mM EGTA, and 0.4 mM MgCl2).

Immunofluorescence and Microscopy

Nuclei in egg extracts or from embryos were mixed with 20 volumes fix buffer

(ELB, 15% glycerol, 2.6% paraformaldehyde) for 15 min at room temperature,

layered over 5 ml cushion buffer (XB, 200 mM sucrose, 25% glycerol), and

spun onto 12 mm circular coverslips at 1000 3 g for 15 min at 16 C. Nuclei

were postfixed in coldmethanol for 5min and rehydrated in PBS-NP40. Cover-

slips were blocked with PBS-3% BSA overnight at 4 C, incubated at room

temperature for 1 hr each with primary and secondary antibody diluted in

PBS-BSA followed by 5 mg/ml Hoechst, mounted in Vectashield (Vector Labo-

ratories), and sealed with nail polish. Antibodies are described in the Extended

Experimental Procedures.

Images were acquired with an Olympus BX51 fluorescence microscope,

403 objective, and Hamamatsu Orca II cooled CCD camera. Nuclear

cross-sectional areas were measured from thresholded images in Meta-

Morph (Molecular Devices) and multiplied by 4 to estimate total NE surface

area. To validate this method for quantifying nuclear size, imaging was per-

formed using a Marianas Spinning Disk Confocal microscope (Intelligent

Imaging Innovations). For a given nucleus, 100 confocal sections were

acquired, and nuclear circumference for each slice was measured in ImageJ

(NIH). NE surface area was calculated as the sum of these circumferences

multiplied by the slice thickness (0.2 mm), and these values agreed within

2% of estimates from the cross-sectional area (data not shown). We therefore

used the cross-sectional area method to estimate NE surface area because it

facilitated the acquisition of data from large numbers of nuclei. For fluores-

cence intensity measurements, images were acquired with the same expo-

sure time, and a region of representative background fluorescence was

used for background correction. Total integrated intensity and nuclear area

were quantified from thresholded images (Metamorph) and used to calculate

intensity per unit area. Statistical methods are described in the figure

legends.

Western Blots

Egg extract protein concentrations were measured by Bradford assay (Bio-

rad). The average total protein concentration was 56 ± 3 mg/ml in X. laevis

and 52 ± 4 mg/ml in X. tropicalis (mean ± SD, n = 6). 25 mg protein from three

different X. laevis and X. tropicalis extracts was separated by SDS-PAGE and

semi-dry transferred to nitrocellulose (Biorad). Blots were blocked with PBS-

5%milk, probed with primary and secondary antibodies (see Extended Exper-

imental Procedures) diluted in PBST-5% milk, and scanned on an Odyssey

Infrared Imaging System (LI-COR Biosciences). Band intensities were quanti-

fied using the Odyssey software. Western blots on different stage embryo

extracts were similarly performed.



figures, one table, and one movie and can be found with this article online at

doi:10.1016/j.cell.2010.09.012.

ACKNOWLEDGMENTS

We thank Steve Bird, Mary Dasso, Dirk Gorlich, David Halpin, Richard Harland,

Petr Kalab, Jan Liphardt, Alan Lowe, Andreas Merdes, Jon Soderholm, and

Karsten Weis for reagents. We acknowledge Steve Bird and Saori Haigo for

conducting the initial nuclear scaling experiments in egg extracts; Saori Haigo,

Esther Kieserman, and Richard Harland’s laboratory for help with Xenopus

embryos; and Abby Dernburg for use of the Marianas SDC microscope. We

thank members of the Heald lab for helpful advice; Rose Loughlin, Jeff Tang,

Karsten Weis, and David Weisblat for constructive comments on the manu-

script; and Favian Hernandez for artwork. R.H. is supported by the NIH Direc-

tor’s Pioneer Award (DP1 OD000818) and The Miller Institute for Basic

Research in Science. D.L.L. acknowledges support from an American Cancer

Society postdoctoral fellowship (PF-09-041-01-DDC).

Received: March 24, 2010




REFERENCES

Aebi, U., Cohn, J., Buhle, L., and Gerace, L. (1986). The nuclear lamina is

a meshwork of intermediate-type filaments. Nature 323, 560–564.

Brown, K.S., Blower, M.D., Maresca, T.J., Grammer, T.C., Harland, R.M., and

Heald, R. (2007). Xenopus tropicalis egg extracts provide insight into scaling of

the mitotic spindle. J. Cell Biol. 176, 765–770.

Cavalier-Smith, T. (2005). Economy, speed and size matter: evolutionary

forces driving nuclear genome miniaturization and expansion. Ann. Bot. 95,

147–175.

Clarkson, W.D., Kent, H.M., and Stewart, M. (1996). Separate binding sites on

nuclear transport factor 2 (NTF2) for GDP-Ran and the phenylalanine-rich

repeat regions of nucleoporins p62 and Nsp1p. J. Mol. Biol. 263, 517–524.

D’Angelo, M.A., Anderson, D.J., Richard, E., and Hetzer, M.W. (2006). Nuclear

pores form de novo from both sides of the nuclear envelope. Science 312,

440–443.

Dechat, T., Pfleghaar, K., Sengupta, K., Shimi, T., Shumaker, D.K., Solimando,

L., and Goldman, R.D. (2008). Nuclear lamins: major factors in the structural

organization and function of the nucleus and chromatin. Genes Dev. 22,

832–853.

Fankhauser, G. (1939). Polyploidy in the salamander, Eurycea bislineata.

J. Hered. 30, 379–388.

Feldherr, C., Akin, D., and Moore, M.S. (1998). The nuclear import factor p10

regulates the functional size of the nuclear pore complex during oogenesis.

J. Cell Sci. 111, 1889–1896.

Goldberg, M.W., Huttenlauch, I., Hutchison, C.J., and Stick, R. (2008). Fila-

ments made from A- and B-type lamins differ in structure and organization.

J. Cell Sci. 121, 215–225.

Gorlich, D., Seewald, M.J., and Ribbeck, K. (2003). Characterization of Ran-

driven cargo transport and the RanGTPase system by kinetic measurements

and computer simulation. EMBO J. 22, 1088–1100.

Grammer, T.C., Khokha, M.K., Lane, M.A., Lam, K., and Harland, R.M. (2005).

Identificationofmutants in inbredXenopus tropicalis.Mech.Dev.122, 263–272.

Hachet, V., Kocher, T., Wilm, M., and Mattaj, I.W. (2004). Importin alpha asso-

ciates with membranes and participates in nuclear envelope assembly in vitro.

EMBO J. 23, 1526–1535.

Hall, M.N., Raff, M.C., and Thomas, G. (2004). Cell growth: control of cell size

(Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press).


Heald, R., and McKeon, F. (1990). Mutations of phosphorylation sites in lamin

A that prevent nuclear lamina disassembly in mitosis. Cell 61, 579–589.

Heitlinger, E., Peter, M., Haner, M., Lustig, A., Aebi, U., and Nigg, E.A. (1991).

Expression of chicken lamin B2 in Escherichia coli: characterization of its

structure, assembly, and molecular interactions. J. Cell Biol. 113, 485–495.

Horner, H.A., and Macgregor, H.C. (1983). C value and cell volume: their

significance in the evolution and development of amphibians. J. Cell Sci.

63, 135–146.

Jenkins, H., Holman, T., Lyon, C., Lane, B., Stick, R., and Hutchison, C. (1993).

Nuclei that lack a lamina accumulate karyophilic proteins and assemble

a nuclear matrix. J. Cell Sci. 106, 275–285.

Jorgensen, P., Edgington, N.P., Schneider, B.L., Rupes, I., Tyers, M., and

Futcher, B. (2007). The size of the nucleus increases as yeast cells grow.

Mol. Biol. Cell 18, 3523–3532.

Lemaitre, J.M., Geraud, G., and Mechali, M. (1998). Dynamics of the genome

during early Xenopus laevis development: karyomeres as independent units of

replication. J. Cell Biol. 142, 1159–1166.

Loewinger, L., and McKeon, F. (1988). Mutations in the nuclear lamin proteins

resulting in their aberrant assembly in the cytoplasm. EMBO J. 7, 2301–2309.

Lowe, A.R., Siegel, J.J., Kalab, P., Siu, M., Weis, K., and Liphardt, J.T. (2010).

Selectivity mechanism of the nuclear pore complex characterized by single

cargo tracking. Nature, in press. Published online September 1, 2010.

10.1038/nature09285.

Madrid, A.S., and Weis, K. (2006). Nuclear transport is becoming crystal clear.

Chromosoma 115, 98–109.

Manor, U., and Kachar, B. (2008). Dynamic length regulation of sensory stereo-

cilia. Semin. Cell Dev. Biol. 19, 502–510.

Maresca, T.J., and Heald, R. (2006). Methods for studying spindle assembly

and chromosome condensation in Xenopus egg extracts. Methods Mol.

Biol. 322, 459–474.

Marshall, W. (2002). Size control in dynamic organelles. Trends Cell Biol. 12,

414–419.

Marshall, W.F. (2008). Engineering design principles for organelle size control

systems. Semin. Cell Dev. Biol. 19, 520–524.

Maul, G.G., and Deaven, L. (1977). Quantitative determination of nuclear pore

complexes in cycling cells with differing DNA content. J. Cell Biol. 73, 748–760.

Montorzi, M., Burgos, M.H., and Falchuk, K.H. (2000). Xenopus laevis embryo

development: arrest of epidermal cell differentiation by the chelating agent

1,10-phenanthroline. Mol. Reprod. Dev. 55, 75–82.

Murray, A.W. (1991). Cell cycle extracts. Methods Cell Biol. 36, 581–605.

Neumann, F.R., and Nurse, P. (2007). Nuclear size control in fission yeast.

J. Cell Biol. 179, 593–600.

Newport, J.W., Wilson, K.L., and Dunphy, W.G. (1990). A lamin-independent

pathway for nuclear envelope assembly. J. Cell Biol. 111, 2247–2259.

Nieuwkoop, P.D., and Faber, J. (1967). Normal table of Xenopus laevis,

Second Edition (Amsterdam: North-Holland Publishing Company).

Prufert, K., Vogel, A., and Krohne, G. (2004). The lamin CxxM motif promotes

nuclear membrane growth. J. Cell Sci. 117, 6105–6116.

Riddick, G., and Macara, I.G. (2005). A systems analysis of importin-alpha-

beta mediated nuclear protein import. J. Cell Biol. 168, 1027–1038.

Riddick, G., and Macara, I.G. (2007). The adapter importin-alpha provides

flexible control of nuclear import at the expense of efficiency. Mol. Syst.

Biol. 3, 118.

Sive, H.L., Grainger, R.M., and Harland, R.M. (2000). Early development of

Xenopus laevis: a laboratory manual (Cold Spring Harbor, N.Y.: Cold Spring

Harbor Laboratory Press).

Smith, A., Brownawell, A., and Macara, I.G. (1998). Nuclear import of Ran is

mediated by the transport factor NTF2. Curr. Biol. 8, 1403–1406.

Smith, A.E., Slepchenko, B.M., Schaff, J.C., Loew, L.M., and Macara, I.G.

(2002). Systems analysis of Ran transport. Science 295, 488–491.

Stewart, M. (2007). Molecular mechanism of the nuclear protein import cycle.

Nat. Rev. Mol. Cell Biol. 8, 195–208.

Theerthagiri, G., Eisenhardt, N., Schwarz, H., and Antonin, W. (2010). The

nucleoporin Nup188 controls passage of membrane proteins across the

nuclear pore complex. J. Cell Biol. 189, 1129–1142.

Webster, M., Witkin, K.L., and Cohen-Fix, O. (2009). Sizing up the nucleus:

nuclear shape, size and nuclear-envelope assembly. J. Cell Sci. 122,

1477–1486.

Wilson, N.F., Iyer, J.K., Buchheim, J.A., and Meek, W. (2008). Regulation of

flagellar length in Chlamydomonas. Semin. Cell Dev. Biol. 19, 494–501.

Zink, D., Fischer, A.H., and Nickerson, J.A. (2004). Nuclear structure in cancer

cells. Nat. Rev. Cancer 4, 677–687.


TGF-b and Insulin Signaling RegulateReproductive Aging via Oocyte andGermline Quality MaintenanceShijing Luo,1 Gunnar A. Kleemann,1 Jasmine M. Ashraf,1 Wendy M. Shaw,1 and Coleen T. Murphy1,*1Lewis-Sigler Institute for Integrative Genomics and Department of Molecular Biology, Princeton University, Princeton, NJ 08544, USA


DOI 10.1016/j.cell.2010.09.013

SUMMARY

Reproductive cessation is perhaps the earliest aging

phenotype that humans experience. Similarly, repro-

duction of Caenorhabditis elegans ceases in mid-

adulthood. Although somatic aging has been studied

in both worms and humans, mechanisms regulating

reproductive aging are not yet understood. Here,

we show that TGF-b Sma/Mab and Insulin/IGF-1

signaling regulate C. elegans reproductive aging by

modulating multiple aspects of the reproductive

process, including embryo integrity, oocyte fertiliz-

ability, chromosome segregation fidelity, DNA

damage resistance, and oocyte and germline mor-

phology. TGF-b activity regulates reproductive

span and germline/oocyte quality noncell-autono-

mously and is temporally and transcriptionally

separable from its regulation of growth. Chromo-

some segregation, cell cycle, and DNA damage

response genes are upregulated in TGF-b mutant

oocytes, decline in aged mammalian oocytes, and

are critical for oocyte quality maintenance. Our data

suggest that C. elegans and humans share many

aspects of reproductive aging, including the correla-

tion between reproductive aging and declining

oocyte quality and mechanisms determining oocyte

quality.

INTRODUCTION

Many biological functions associated with quality of life decline

with age, but female reproductive aging is one of the earliest

declines humans experience. Although progressive loss of

ovarian follicles leads to menopause between the ages of 45

and 55, the risk of infertility, birth defects, and miscarriage

increase a decade earlier, likely because of age-related declines

in oocyte quality (te Velde and Pearson, 2002). Although aged

mammalian oocytes exhibit increased errors in fertilization, chro-

mosome segregation, and cleavage divisions (Goud et al., 1999;

te Velde and Pearson, 2002), little is known about mechanisms

that regulate oocyte quality maintenance with age.

Caenorhabditis elegans is a useful model for aging studies

because of its short life span and the conservation of longevity

pathways from C. elegans to humans (Kenyon, 2005; Suh et al.,

2008). Recently, C. elegans has also been developed as a model

of reproductive aging (Andux and Ellis, 2008; Hughes et al., 2007;

Luo et al., 2009). These studies established that (1) C. elegans

reproductive aging is independent of sperm contribution; (2)

simply reducing ovulation rate or progeny production do not

extend reproductive span; and (3) reproductive aging is

usage independent (i.e., independent of the magnitude and

timing of oocyte use). That is, in worms as in humans, simply

delaying the reproductive schedule does not delay reproductive

aging.

One argument against using C. elegans as a model of human

reproductive aging is that oocytes are continually produced in

worms, whereas humans’ total oocyte supply is produced at

birth. However, both human and C. elegans females reproduce

for about one-third to one-half of their lives, and thus undergo

significant reproductive aging on proportional time scales,

implying that genetic mechanisms may link reproduction to

longevity in both organisms (Cant and Johnstone, 2008;

Luo et al., 2009). Furthermore, both human and C. elegans

oocytes are cell-cycle arrested at meiotic prophase I, release

from arrest is coordinated with oocyte maturation in both, and

the mechanisms underlying oocyte maturation are highly

conserved between the two organisms (Greenstein, 2005; Mehl-

mann, 2005). Most importantly, human reproductive aging

occurs a decade prior to the exhaustion of the oocyte supply,

suggesting that oocyte quality, rather than quantity, is the limiting

factor for successful reproduction with age. Thus, the critical

question that we address in this study is whether worms’ repro-

duction is similarly limited by oocyte quality, and if so, by what

mechanisms.

Several long-lived C. elegans mutants, including the Insulin/

IGF-1 receptor mutant daf-2, delay reproductive aging (Huang

et al., 2004; Hughes et al., 2007; Luo et al., 2009). daf-2mutants

extend life span, delay distal germline integrity decline, and

extend reproductive span through the activity of the FOXO tran-

scription factor DAF-16 (Garigan et al., 2002; Hughes et al., 2007;

Kenyon et al., 1993; Luo et al., 2009), but the role of daf-2 in

oocyte quality maintenance and the mechanisms by which

daf-2 mutants extend reproductive span are unknown.

We recently found that mutants of the TGF-b Sma/Mab

pathway also significantly extend reproductive span (Luo et al.,


2009), whereas mutants in the TGF-b Dauer pathway extend life

span (Shaw et al., 2007) without greatly extending reproductive

span (Luo et al., 2009). The TGF-b Sma/Mab pathway, which is

highly conserved from worms to humans, consists of extracel-

lular ligands (DBL-1), type I (SMA-6) and type II (DAF-4) recep-

tors, R-Smads (SMA-2 and SMA-3), a co-Smad (SMA-4), and

a transcription co-factor (SMA-9) (Massague, 2000; Savage-

Dunn, 2005). Notably, Sma/Mab regulation of reproductive

span is genetically independent of Insulin/IGF-1 signaling (IIS)

and Dietary Restriction (Luo et al., 2009).

Here we show that C. elegans oocytes, like human oocytes,

degrade functionally and morphologically with age and that

reduction of TGF-b Sma/Mab signaling and IIS delays repro-

ductive aging by maintaining oocyte and germline quality.

Although the TGF-b Sma/Mab pathway acts autonomously in

the hypodermis to regulate body size (Wang et al., 2002),

surprisingly, we find that both insulin and TGF-b Sma/Mab

signaling regulate oocyte and distal germline quality mainte-

nance nonautonomously. TGF-b regulates reproductive aging

separately from the developmental regulation of growth, both

temporally and transcriptionally. We find that TGF-b oocyte

transcriptional targets that are required for C. elegans

embryonic and germline integrity maintenance also change

with age in mammalian oocytes. The conserved nature of these

signaling pathways suggests that the mechanisms underlying

the maintenance of C. elegans reproductive capacity with age

may also influence reproductive capacity decline in higher

organisms.

RESULTS

TGF-b Sma/Mab and Insulin/IGF-1 Signaling Regulate

Embryo Viability and Oocyte Quality Maintenance

Wild-typeC. elegans reproduction declineswith age, but reduced

Insulin/IGF-1 signaling (IIS) delays reproductive cessation

(Hughes et al., 2007). We recently found that reduced TGF-b

Sma/Mab signaling also significantly extends reproductive span

in amanner that is independent of insulin signaling, caloric restric-

tion, sperm contribution, and ovulation rate (Luo et al., 2009). To

identify the molecular mechanisms underlying normal reproduc-

tive aging and its delay in insulin and TGF-b signaling mutants,

we systematically investigated each component of the reproduc-

tive system, from fertilized embryos through the distal germline

(Figure 1A).

To examine fertilized embryo quality, we determined embry-

onic lethality rates. Compared to age-matched daf-2 and

sma-2 animals, older wild-type animals produced significantly

more embryos that fail to hatch, though they all produced

more unhatched embryos with age (Figures 1B and 1E and

Figure S1A available online) and that are susceptible to damage

by bleaching, a test of eggshell integrity (Figures S1C and S1D).

Thus, the reproductive span extension exhibited by daf-2 and

sma-2 mutants is at least partly a manifestation of increased

embryo integrity late in reproduction.

Chromosomal abnormalities, in particular aneuploidies, are

a major cause of mammalian embryonic developmental defects

(Magli et al., 2007; Rubio et al., 2003), and nondisjunction rates

also increase with age in Drosophila (Tokunaga, 1970). Increased

chromosomal abnormalities, particularly autosome loss, could

contribute to C. elegans embryonic lethality. Meiotic X chromo-

some nondisjunction produces males (Hodgkin et al., 1979),

which in combination with embryonic lethality, provides a simple

measure of chromosomal loss (Saito et al., 2009). Strikingly, the

fraction of male progeny produced by wild-type mothers

increased 16-fold from day 1 to day 5 (Figure 1C). By contrast,

the rate of male production by daf-2 and sma-2 mutants was

significantly lower (Figure 1C and Figure S1B). To directly test

disjunction fidelity, we counted DAPI-stained bodies (Saito

et al., 2009) in oocytes of spermless (fem-1) animals. We found

that the number of oocytes with the normal number of stained

bodies (six bivalents) decreased significantly with age in fem-1,

but insignificantly in sma-2;fem-1 and daf-2;fem-1 animals

(Figure 1D and Figure S1E), suggesting an increased frequency

of chromosomal segregation errors in wild-type oocytes with

age. Therefore, worms with reduced Insulin/IGF-1 and TGF-b

Sma/Mab signaling better maintain oocyte chromosome segre-

gation fidelity with age.

Oocyte quality decline is also a cause of human age-related

infertility (Goud et al., 1999). To test fertilizability, we mated

hermaphrodites with young adult (Day 1 or 2) wild-type males

and counted the number of fertilized embryos and unfertilized

oocytes produced each day (Figure 1E and Figure S1F),

excluding mothers that stopped producing cross-progeny

before reproductive cessation. (While fertilized embryos are

ovoid with a distinct eggshell, unfertilized oocytes are fuzzy

and round, as shown in Figure 1E). Aging wild-type animals

produced a significant number of unfertilized oocytes with

age, whereas daf-2 and sma-2 mutants produced almost

exclusively successfully fertilized embryos (Figure 1E and

Figure S1F). Although daf-2 and sma-2 mutants produce fewer

total progeny, such usage-dependent mechanisms as total

progeny number, early progeny production, and ovulation rate

have been previously eliminated as contributing factors in

reproductive aging (Andux and Ellis, 2008; Hughes et al.,

2007; Luo et al., 2009). To ensure that sperm is not limiting in

our mated assays, we examined oocytes of mated worms for

ribonucleoprotein (RNP) foci, which form in sperm-depleted

oocytes (Jud et al., 2007); the Day 8 nonreproductive mated

worms do not form RNP foci (Figure S1G). This suggests that

sufficient sperm are available throughout the reproductive

period in our mating experiments, and the unfertilized oocytes

that the wild-type worms produce in old age are likely due to

lower oocyte quality. Mutations in both the TGF-b Sma/Mab

and IIS pathways delay such decline, rendering the oocytes

fertilizable longer.

TGF-b Sma/Mab and IIS Regulate Oocyte Morphology

Maintenance

To determine whether IIS and TGF-b Sma/Mab signaling regu-

late oocyte morphology maintenance, we examined wild-type

and mutant oocytes with age. On Day 1 of adulthood, wild-

type oocytes are large and closely packed with their neighboring

oocytes (Figure 1F). sma-2 mutants have fewer oocytes aligned

in the gonad because of their short length, but themorphology of

the oocytes in both the daf-2 and sma-2 mutants is similar to

wild-type in early adulthood.


mitotic germlineTZpachytene nuclei

embryosspermoocytes

A

wt daf-2 sma-2 sma-2daf-2wtday 5day 1

% e

mb

ryo

nic

le

tha

lity

% m

ale

pro

ge

ny

3540

3025201510

50

B C

wt mated hatched emb

wt mated unhatched emb

wt mated unfertilized ooc

daf-2 mated hatched emb

daf-2 mated unhatched emb

daf-2 mated unfertilized ooc

sma-2 mated hatched

sma-2 mated unhatched emb

sma-2 mated unfertilized oocnu

mb

er

of

em

bry

o/o

ocyte

0

4

8

12

16

20

day 4 day 5 day 6 day 7 day 8 day 9 day 10 day 11 day 12

E

day 1 day 8 day 8 day 8

wt

mate

dsm

a-2 m

at e

dda

f-2 m

ate

d

F

severe

mild

normal

severe

mild

normal

Repro. Repro. Repro.PR PR PRsmall cavity cluster small cavity cluster

% o

f w

orm

s

% o

f w

orm

s

0

20

40

60

80

100

0

20

40

60

80

100

wt daf-2 sma-2 wt wtdaf-2 daf-2sma-2 sma-2

G H

177 80 28 86 73 24 51 21 45 36 23

*** ** *** ** *** ** * *** ***

wt daf-2 sma-2 sma-2daf-2wtday 5day 1

3.5

3.0

2.5

2.0

1.5

1.0

00.5

100

80

60

40

0

20

***

day 5day 1% o

ocyte

with

six

biv

ale

nts

D

embryo oocyte

fem-1

daf-2;fe

m-1

sma-2;fem

-1fem

-1

daf-2;fe

m-1

sma-2;fem

-1

4.0***

45***

******

oocyte clustercavity

smaller oocytesoocytes

oocytes

oocytes

embryo

embryo

embryo

oocytes

oocytes

oocytes

oocytes

embryos

oocytes

embryos

(severe)

(mild) (mild)

embryo embryo

embryos

uterus vulva

oocyte clustercavity

smaller oocytesoocyt s

oocytes

oocytes

embryo

embryo

embryo

oocytes

oocytes

oocytes

oocytes

embryos

oocytes

embryos

(severe)

(mild) (mild)

embryo embryo

embryos

uterus vulva

Figure 1. TGF-b Sma/Mab and Insulin/IGF-1 Signaling Regulate Embryo Viability, Oocyte Fertilizability, and Oocyte Morphology

(A) Schematic of the C. elegans gonad.

(B) Percentage of embryos that fail to hatch (±SEP).

(C) Percentage of male progeny (±SEP).

(D) Percentage of oocytes with 6 DAPI-stained bodies (±SEP).

(E) Number of hatched embryos (inset, left), unhatched embryos, and unfertilized oocytes (inset, right) produced each day after mating with young wild-type (wt)

males (mean ± SEM); percentages shown in Figure S1F.

(F) Oocyte morphology, with defects in yellow.

(G) Oocyte morphology markers scored in mated wt animals that are either reproductive (Repro) or post-reproductive (PR).

(H) Oocyte morphology markers scored in day 8 mated worms. p-values for wild-type versus daf-2 or sma-2 indicated.

* p < 0.05, **p < 0.01, and ***p < 0.001.


On Day 8, when mated wild-type animals have nearly

ceased reproduction, their oocytes have visibly degraded:

some become much smaller, as previously reported (Andux

and Ellis, 2008); some lose contact with their neighbors, result-

ing in cavities; and others fuse into large clusters packed in the

uterus (Figure 1F and Figures S1H–S1K). The defects were

independent of levamisole paralysis treatment used for micros-

copy (Figures S1L and S1M and Figures S2A and S2B).

To test whether these defects are morphological predictors

of reproductive capacity, we compared reproductive and post-

reproductive wild-type animals; oocytes from the postrepro-

ductive animals were significantly more degraded in terms of

oocyte size, cavities, and cluster formation (Figure 1G). By

contrast, oocytes in aged daf-2 and sma-2 animals were still

day 8day 1

wt

ma

ted

day 8

daf-2

mate

dsm

a-2

mate

d

A

normal

severe

0

20

40

60

80

100

wt sma-2daf-2

% o

f w

orm

s

wt sma-2daf-2 wt sma-2daf-2

cavity graininess cellularization

*** *** * ** *** ***

% d

ecre

ase

in

# o

f

wild type daf-2 sma-20

5

10

15

20

25

30

35mitotic region**

germ

cells w

ith a

ge

B

C

% h

atc

hed e

mbry

o

wild type daf-2 sma-20

20

40

60

80

100D

mated post γ-irradiation

***

***

cavity

graininess

cellularization

(mild)

(mild)

(severe)

cavity

graininess

cellularization

(mild)

(mild)

(severe)

Figure 2. TGF-b Sma/Mab and Insulin/IGF-1

Signaling Regulate DNA Damage Response

and Distal Germline Integrity

(A) Distal germline morphology, with defects in

yellow.

(B) Distal germline morphology scores of day 8

mated worms; p-values compare wt versus

daf-2 or sma-2. *p < 0.05, **p < 0.01, and ***p <

0.001.

(C) Percentage decrease in mitotic germ cell

number with age (raw values in Figure S2I).

(D) daf-2 and sma-2 lay significantly more hatched

embryos than wild-type after g-irradiation

(% ± SEP). Animals were mated with young wt

males after irradiation.

* p < 0.05, **p < 0.01, and ***p < 0.001.

young-looking, with significantly fewer

morphological defects than age-

matched wild-type oocytes (Figure 1F

and 1H and Figures S1N–S1S). Thus,

reduced TGF-b Sma/Mab and IIS

activity both improve oocyte mor-

phology maintenance. Together, our

data suggest that oocyte quality, as

defined by chromosome segregation

fidelity, fertilizability, and morphology,

declines with age in C. elegans, and

that reduced TGF-b Sma/Mab and IIS

signaling delay this decline.

Distal Germline Morphology

and Proliferation Is Maintained

in TGF-b Sma/Mab and IIS Mutants

The distal germline undergoes signifi-

cant morphological decline with age,

but IIS mutations significantly slow this

deterioration (Garigan et al., 2002)

(Figures 2A and 2B and Figures S2C–

S2F). We scored the appearance of cavi-

ties, graininess, and cellularization, the

major morphological markers of germ-

line aging (Garigan et al., 2002), and

found that sma-2 mutations also signifi-

cantly slow germline deterioration (Figures 2A and 2B and

Figures S2G and S2H). Although these may be independent

effects of the pathways, oocyte and distal germline morphology

characteristics in the same population of wild-type worms are

correlated (Figure S2J).

The distal germline contains proliferating germline stem cells

(GSCs) and their mitotic descendents. The number of DAPI-

stained germ cell nuclei in this zone declines significantly with

age in wild-type animals (Figure 2C and Figure S2I), but declines

less in daf-2 and sma-2 animals (Figure 2C and Figure S2I),

possibly because of better maintenance of proliferative ability.

Together, our data suggest that both IIS and TGF-b Sma/Mab

signalingmay regulate themaintenance of distal germline prolifer-

ation and germline quality, as well as oocyte quality.


Physiological and DNA Damage-Induced Apoptosis

Are Not Major Contributors to TGF-b and IIS Regulation

of Reproductive Aging

Prior to cellularization into oocytes, germ cell nuclei undergo

programmed cell death (Gumienny et al., 1999). This ‘‘physiolog-

ical germ cell apoptosis’’ has been proposed to be an important

factor in maintaining oocyte quality via resource reallocation

(Andux and Ellis, 2008). We found that sma-2 and daf-2mutants

have higher levels of physiological apoptosis than wild-type,

but wild-type levels decreased insignificantly with age

(Figure S2K).

C. elegans’ germline also undergoes apoptosis as a response

to DNA damage from ionizing radiation (Gartner et al., 2000). We

examined animals after g-irradiation and found that DNA

damage–induced apoptosis declined significantly with age in

wild-type animals, but the rates in older TGF-b and IIS mutants

were not significantly different from wild-type (Figure S2L).

Although neither Insulin/IGF-1 nor TGF-b signaling appears to

regulate this process, the significant decrease in irradiation-

induced apoptosis with age likely contributes to reproductive

aging in general.

DNA Damage Response Contributes to Reproductive

Maintenance by TGF-b and IIS

A different aspect of the DNA damage response is improved

both by reduced IIS and TGF-b signaling: the number of viable

progeny produced after ionizing radiation treatment increased

significantly in daf-2 and sma-2 mutants compared to wild-

type (Figure 2D). The proportion of arrested larvae is also slightly

increased in the mutants (Figure S2M), suggesting that even

damaged animals are more developmentally competent than

the wild-type progeny. Thus, although the rate of DNA-damage

induced apoptosis is not increased, sma-2 and daf-2 germ cells

may better repair damaged DNA or be better protected against

genotoxic stress, which in turn may be partially responsible for

slowed reproductive aging.

TGF-b and IIS Signaling Regulate Reproductive Aging

Nonautonomously

TGF-b Sma/Mab signals cell-autonomously in the hypodermis to

regulate body growth (Wang et al., 2002). To test the cell

autonomy of TGF-b Sma/Mab signaling in the regulation of

reproductive aging, we performed mosaic analyses. Hypo-

dermal expression of the TGF-b Sma/Mab signal transducer

SMA-3, which forms a transcriptional complex with SMA-2, is

necessary and sufficient for normal body length (Wang et al.,

2002). Like sma-2 mutants, sma-3 mutants extend reproductive

span (Luo et al., 2009) and maintain oocyte and germline

morphology longer with age (Figure 3C and 3D). If reproductive

aging is dependent on cell-autonomous TGF-b Sma/Mab

signaling in the germline, loss of the sma-3 transgene in the

germline alone should recapitulate sma-3 reproductive span

extension. Alternatively, if reproductive aging is dependent on

somatic (nonautonomous) TGF-b signaling, somatic sma-3

expression should be sufficient to suppress the long reproduc-

tive span of sma-3. We screened a synchronized population

of sma-3(wk30);qcEx26[sma-3 gDNA;sur-5::gfp] transgenic

animals (Wang et al., 2002), selecting worms expressing GFP

in most somatic tissues, including hypodermis, but without

germline fluorescence (Figure S3A–S3C) (‘‘germline silent’’

animals). Because the sma-3 transcript could still be present

but undetectable, we also selected somatically fluorescent

animals that produced no fluorescent progeny, indicating that

they had completely lost the transgenic array in the germline

(‘‘germline lost’’). As previously reported, somatic sma-3 activity

rescued body length (Figure 3A). Surprisingly, both the germline-

silent and germline-lost animals had wild-type–like reproductive

spans (Figure 3B), indicating that somatic sma-3 expression is

sufficient to rescue reproductive span regulation. We also found

that the sma-3 germline-silent mosaic animals reduced ovulation

rate and progeny number, but have a normal reproductive span

(Figure 3B and Figures S3D and S3E), underscoring our previous

finding that low ovulation rates and progeny numbers do not

extend reproductive span (Luo et al., 2009). Additionally, the

morphology of day 8 oocytes and distal germlines of somatic

sma-3 animals were more similar to wild-type than to sma-3

(Figures 3C and 3D). Thus, TGF-b signaling regulates reproduc-

tive aging nonautonomously, signaling from somatic tissues to

the germline to maintain quality.

Expression of sma-3 under the vha-7 promoter, which is

primarily hypodermal, rescues the small body size phenotype

of sma-3 mutants (Wang et al., 2002). To determine the tissue-

specificity of nonautonomous TGF-b signaling in reproductive

aging regulation, we selected large Pvha-7::sma-3;sma-3

(wk30) worms (Figures S3F and S3G) and found that the repro-

ductive span extension of sma-3 mutants was also rescued by

hypodermal sma-3 expression (Figure 3E and Figure S3H).

Because we were concerned that the vha-7 promoter might

also express in somatic gonad tissues, we checked the effect

of sma-9 RNAi in a somatic-gonad-only RNAi strain (rrf-3;rde-

1;qyIs103[Pfos-1a::rde-1+Pmyo-2::yfp]) (Hagedorn et al.,

2009). Somatic gonad-specific knockdown of TGF-b signaling

did not recapitulate the reproductive span extension we

observed in whole-animal RNAi (Figure 3F). Together, our results

suggest that TGF-b signaling in the hypodermis acts autono-

mously to regulate body size, but nonautonomously to regulate

oocyte and distal germline quality maintenance and, subse-

quently, reproductive aging.

IIS acts both autonomously (Libina et al., 2003) and nonauton-

omously to regulate life span (Apfeld and Kenyon, 1998; Wolkow

et al., 2000). We found that germline silencing of daf-16 activity

still allows daf-2 mutant-like reproductive span extension

(Figure 4A and Figure S4), suggesting that IIS also acts germ-

line-nonautonomously to regulate reproductive aging. Tissue-

specific expression analysis of the DAF-16 transcription factor

showed that intestinal expression DAF-16, which increases life

span (Libina et al., 2003), also significantly increased reproduc-

tive span and improved oocyte and germline morphology of

daf-16;daf-2 mutants (Figures 4B and 4D–4F). Surprisingly,

muscle-expressed DAF-16, which has no effect on life span

(Libina et al., 2003) also increased reproductive span and

improved germline and oocyte quality significantly (Figure 4C-

F), whereas neuronal DAF-16 had little effect on reproductive

aging (Figures 4C–4F). Thus, IIS acts nonautonomously to regu-

late germline and oocyte aging, acting partially in different

tissues from its nonautonomous regulation of longevity.


TGF-bSma/Mab Signaling Acts in Adulthood to Regulate

Reproductive Aging

The timing of IIS’ effects on reproduction and longevity largely

overlap, acting primarily in adulthood with some contribution to

reproduction in late larval stages (Dillin et al., 2002) (Figures

S5A and S5B). By contrast, TGF-b signaling acts in earlier larval

stages to regulate body size (Liang et al., 2003; Savage-Dunn

et al., 2000). To determine the timing of TGF-b regulation of

reproductive span, we used RNAi to knock down Sma/Mab

signaling in RNAi-sensitive rrf-3 mutants either during the

animals’ whole life or only during adulthood. Whole-life sma-9

(RNAi) treatment both reduced body size (Figure 5A and

Figure S5C) and increased reproductive span (Figure 5B). sma-

9(RNAi) treatment only in adulthood, however, did not reduce

body size (Figure 5A), but increased reproductive span to the

same extent as whole-life sma-9(RNAi) treatment (p = 0.46,

Figure 5B). Thus, the effects of TGF-b signaling on body size

are temporally separable from its effects on reproduction. Addi-

tionally, small body size is not required for extended reproduc-

tive span through TGF-b signaling. Our tissue specificity and

temporal analyses suggest that the downstream effectors that

control body size and reproductive aging may be distinct,

despite the fact that they are both controlled by TGF-b signaling

in the hypodermis.

TGF-b Oocyte Quality Targets Are Shared

with Mammalian Oocyte Aging Genes

To identify the targets of TGF-b Sma/Mab signaling that regulate

reproductive aging, we compared the transcription of unfertilized

oocytes isolated from day 8 spermless fem-1 and sma-2;fem-1

worms (Figure S5D). Gene ontology (GO) analysis of significantly

upregulated and downregulated TGF-b oocyte genes (Figure 5C

0

.2

.4

.6

.8

1

0 2 4 6 8 10 12 14

0

.2

.4

.6

.8

1

0 2 4 6 8 10 120

.2

.4

.6

.8

1

0 2 4 6 8 10 12

body length

(m

m)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

wildty

pe

sma-3

germ

line lo

st

germ

line s

ilent

fraction r

epro

ductive

sma-3sma-3 germline lost

wild type

sma-3 germline silent

day of adulthood

mated*

wt

sma-3G

L(-)0

20

40

60

80

100small cavity cluster

% o

f w

orm

s

wt

sma-3G

L(-) wt

sma-3G

L(-)

severe

mild

normal

wt

sma-3G

L(-)0

20

40

60

80

100cavity graininess cellularization

% o

f w

orm

s

wt

sma-3G

L(-) wt

sma-3G

L(-)

severe

mild

normal

wild type

sma-3;wqEx2-hypodermal sma-3sma-3;wqEx1-hypodermal sma-3sma-3

mated mated

fraction r

epro

ductive

day of adulthood

fraction r

epr o

ductive

day of adulthood

rrf-3;control(i)rrf-3;sma-9(i)

rrf-3;rde-1;qyIs103;sma-9(i)-somatic gonad RNAi

*

rrf-3;rde-1;qyIs103;control(i)-somatic gonad RNAi

A B

C oocyte D distal germline

E F

***

Figure 3. TGF-b Sma/Mab Signaling Regulates Reproductive Aging Nonautonomously in Hypodermis

(A) Body length of wt, sma-3(wk30), sma-3(wk30);qcEx26[sma-3 gDNA; sur-5::gfp] animals that have lost or silenced transgenic sma-3 expression in the germline

(mean ± SEM).

(B) Mated reproductive spans of worms in (A). *High matricide rate due to internal progeny hatching. (All reproductive span statistics are shown in Table S1.)

(C and D) Scoring of oocyte (C) and distal germline (D) morphology markers in day 8 mated wt, sma-3, and sma-3 germline-lost (GL) animals.

(E) Two independent transgenic lines (sma-3(wk30);Pvha-7::gfp::sma-3) expressing sma-3 in the hypodermis have mated reproductive spans similar to wild-type

(Table S1).

(F) sma-9 RNAi significantly extends mated reproductive span of rrf-3worms, but does not extend the mated reproductive span of the somatic-gonad-only RNAi

strain rrf-3;rde-1;qyIs103[Pfos-1a::rde-1+Pmyo-2::yfp].

* p < 0.05, **p < 0.01, and ***p < 0.001.


and Figure S5F; Tables S2A and S2B) identified such categories

as oogenesis, cell cycle, chromosome segregation and organi-

zation, DNA damage response, proteolysis, ATP binding,

signaling, transcription regulation, protein transport, aging,

GTP binding, and oxidoreductases (Figures 5D and 5E

Figure S5G, Table S3, and Table S4). More than 70% of the

sma-2-regulated genes are regulated in the same direction in

young relative to old (day 8) fem-1 oocytes (Figures S5E and

S5F), and similar GO terms are also enriched (Figure 5D), sug-

gesting that these genes are goodmarkers of the ‘‘youthfulness’’

of oocytes.

A striking number of the genes and GO terms identified in our

array analysis of sma-2;fem-1 and fem-1 oocytes that were

associated with ‘‘youthful’’ oocytes are shared with genes

0

.2

.4

.6

.8

1

0 2 4 6 8 10 12 14 16

0

.2

.4

.6

.8

1

0 2 4 6 8 10 12 140

.2

.4

.6

.8

1

0 2 4 6 8 10 12 14

0

.2

.4

.6

.8

1

0 2 4 6 8 10 12 14

mated

fraction r

epro

ductive

day of adulthood

mated

fraction r

epro

ductive

day of adulthood

mated

fraction r

epro

ductive

day of adulthood

mated

fraction r

epro

ductive

day of adulthood

daf-2daf-16 germline silentdaf-16;daf-2 daf-2

daf-16;daf-2;muIs124-intestinal daf-16daf-16;daf-2;muIs105-endogenous daf-16daf-16;daf-2

daf-16;daf-2;muEx212-muscular daf-16daf-16;daf-2;muEx169-neuronal daf-16

daf-16;daf-2daf-16;daf-2;muEx176-endogenous daf-16daf-2

daf-16;daf-2;muEx212-muscular daf-16daf-16;daf-2;muEx169-neuronal daf-16

daf-16;daf-2daf-16;daf-2;muEx176-endogenous daf-16daf-2

daf-16;daf-2;muEx227-intestinal daf-16

0

20

40

60

80

100

daf-16

;daf-2

% o

f w

orm severe

mild

normal

daf-2

endoge

nousinte

stine

muscle

neuron

daf-16

;daf-2 daf

-2

endoge

nousinte

stine

muscle

neuron

daf-16

;daf-2 daf

-2

endoge

nousinte

stine

muscle

neuron

small cavity cluster*** *** ** *** *** *** * ** *** *** ****

0

20

40

60

80

100

daf-16

;daf-2

% o

f w

orm

daf-2

endoge

nousinte

stine

muscle

neuron

daf-16

;daf-2 daf

-2

endoge

nousinte

stine

muscle

neuron

daf-16

;daf-2 daf

-2

endoge

nousinte

stine

muscle

neuron

cavity graininess cellularization** ** * * ** ** *** ** * *

severe

mild

normal

daf-16;daf-2;muEx211-intestinal daf-16

A B

C D

E oocyte

F distal germline

*

Figure 4. Insulin/IGF-1 Signaling Regulates Reproductive Aging Nonautonomously in Intestine and Muscle

(A) daf-16 germline silent worms (daf-16(mu86);daf-2(e1370);muIs105 [Pdaf-16::gfp::daf-16 +rol-6(su1006)], Figure S4) with only somatic daf-16 activity have

a reproductive span similar to daf-2 mutants (statistics in Table S1).

(B–D) daf-16 activity in intestine (B and D) and muscle (C and D) significantly restores reproductive span extension, whereas neuronal daf-16 activity (C and D)

does not.

(E and F) Oocyte (E) and distal germline (F) morphology scores of day 8 mated daf-16;daf-2, daf-2, endogenous-promoter-driven and tissue-specific promoter-

driven daf-16 transgenic animals. *p < 0.05, **p < 0.01, and ***p < 0.001 for daf-16;daf-2 versus other genotypes.

* p < 0.05, **p < 0.01, and ***p < 0.001.


0

.2

.4

.6

.8

1

0 2 4 6 8 10 12 14

sma- 2;fem -1 D8 oocytevs fem-1 D8 oocyte

-2

up in

sma-

2;fem

-1

log2

3

-3

2

1

0

-1

day of adulthood

fra

ctio

n r

ep

rod

uctive

B C

D

mated

spindl

e lo

caliz

ation

mito

sis

ATP met

abol

ic p

roce

ss

cell ad

hesion

chro

mos

ome

orga

niza

tion

cell diffe

rent

iatio

n

cell de

ath

cell-ce

ll sign

aling

resp

. to D

NA d

amag

e stim

.

trans

criptio

n re

gula

tion

ovipos

ition

ooge

nesis

chro

mos

ome

segre

gatio

n

intra

cellu

lar s

ig. c

ascad

e

ATP bindi

ng

prot

ein

trans

port

spindl

e or

gani

zatio

n

prot

eolysis

unch

. pro

tein

fam

. UPF03

76

hedg

ehog

rece

ptor

activity

imm

unoglobu

lin

leuc

ine-

rich

repe

at

drug

met

abolism

resp

onse

to m

ech. s

tim.

syna

ptic tr

ansm

ission

EGF-li

ke

colla

gen

hom

eosta

tic p

roce

ss

cyto

chro

me

P45

0

lipid

met

abol

ic. p

roce

ss

carb

ohydr

ate

met

ab. p

roc.

*

*

* *

**

*

**

Oocyte

L4

*L2 (Liang et al.)

14 11 10 8 6

en

rich

me

nt

fold

0

1

2

3

4

5E

rrf-3;control(i) whole-life

rrf-3;sma-9(i) adult-onlyrrf-3;sma-9(i) whole-life

control(i);

sma-9(i);sma-9(i);

A

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

bo

dy le

ng

th (

mm

)

***

whole-lifewhole-life

adult-only

Gene Ontology category Genecount

Genes from oocyte array studies (homologs)

Worm (up in sma-2) Mouse Hamatani, et al. 2004 Human Steuerwald, et al. 2007Cell cycle

cyb-1(Ccnb2), cyb-3(Ccnb3), cdc-25.2(Cdc25a), cki-1*mitosis H M W 8 Ccnb2, Ccna2, Cdc16 CCNA2, CCNG1, CDK7

Chromosome segregation, org’nchromosome segregation H M W 7 smc-4(Smc4l1), klp-7, frm-5 Smc4l1, Nin, Kif3b, Bub1 Smc3l1, BUB1B, BUB3spindle localization M 4 gad-1, mes-1, par-3 Hook1, Nin, Rnf19spindle organization M 3 mbk-2, sur-6, goa-1 Tuba2, Tubd1, Pcnt2chromosome organization M W 12 spr-5, nurf-1, hpl-1, hil-2 Hdac2, Morf4l2, Rbbp7

DNA damage response and repair MBD4 (interacts with MLH1), ATR, NBS1response to DNA damage stim. H M 4 mlh-1(MLH1), clk-2, pme-5, uev-2* Msh-3, Exo1, Shprh

Proteolytic pathwayubc-1(Ube2a), ubc-2(Ube2d1), ulp-1proteolysis H M 19 Ube2a, Ubc, Usp1 USP1, CTSC, GRP58

Energy pathway, mitochondrial fn.pmr-1(Atp2c1), vha-13(Atp6v1a), tat-5(Atp9b)ATP metabolic process M 7 Atp2c1, Atp6v1d, Atp5b

ATP binding H M W 42 pgp-7(Abcb11), mrp-2 (Abcc3), psa-4(SMARCA5), pdk-1, akt-2 Abcb6, Abcf3, Cct2 ABCC4, SMARCA5, SUV3

Cell signalling and communicationintracellular signaling cascade H M W 11 cdc-42(RHO GTPase), vhp-1, sel-12 Rhoh, Kras2, Mek1 ATF1, CREB1, CLK1cell-cell signaling M W 5 unc-18, ace-1, cab-1 Gja7, Shroom3, Mmp2

Protein transportarf-1.1(Arf1), arl-13 (Arl13b),

rab-6.2 (Rab6)protein transport H M W 11 Arf1, Arl4, Rab1 ARF4, ARF6, RAB11aTranscription regulationH M W 19 hlh-1, efl-1, spt-5 Phtf1, Crsp6, Lhx8 PHTF1, NFE2L2, EIF2AK2Reproductive process

oogenesis M 5 hrp-1, goa-1, fem-3 Nalp5, Padi5, Nalp9aoviposition W 15 unc-84, cki-1, mtm-3

Othercell death M W 7 ced-1, ced-8, crn-4 Tnfaip8, Mdm4, Bcl2l10cell differentiation W 24 par-1, eor-2, lin-28cell adhesion M W 6 epi-1(Lama2), hmr-1(Cdh11), cdh-3(Cdh23) Lama2, Cdh2, Pcdhb17

Gene Ontology category Genecount Worm (up in sma-2) Mouse Hamatani, et al. 2004 Human Steuerwald, et al. 2007

-18

7

4

3 -12

. 4

19

7

42 -

11

5

11

19

5

15

7 -24

6

**


downregulated in aging mouse and human oocytes (Hamatani

et al., 2004; Steuerwald et al., 2007) (Figure 5D), such as mitotic

cell cycle regulation, chromosome segregation, response to

DNA damage, proteolysis, ATP binding, signaling, transcrip-

tional regulation, and protein transport. The condensin SMC is

upregulated in sma-2 oocytes and declines in both mouse and

human oocytes with age, suggesting that chromosome segrega-

tion is a shared key determinant of oocyte quality. Cell cycle

regulators (CYB-1/3) and DNA mismatch repair proteins (e.g.,

MLH-1 and MBD4) are also higher in sma-2 oocytes and decline

with age in mammalian oocytes. Interestingly, several TGF-b

signaling genes are upregulated with age in mouse oocytes

(Hamatani et al., 2004), paralleling our observations on the exten-

sion of reproductive span in C. elegans TGF-b mutants.

In addition to the genes that are shared between sma-2

mutants and age-regulated in mouse and human oocytes, our

analysis has uncovered new genes that are potential regulators

of reproductive aging. lin-28, which is important in reproductive

development regulation (Hartge, 2009) and the reprogramming

of differentiated cells into induced pluripotent stem cells (Nimmo

and Slack, 2009), and clk-2, a telomere length regulator that is

involved in DNA damage response and cell cycle checkpoint,

are also significantly upregulated in sma-2 oocytes (Table

S2A). Several Class 2 longevity genes, including dod-23 and

dod-24 (Murphy et al., 2003), as well as many oxidoreductases

and protein metabolism genes, are significantly downregulated

in sma-2 oocytes (Table S2B), suggesting additional novel mech-

anisms that may contribute to the regulation of oocyte aging.

Finally, expression of the insulin-like peptide genes ins-22 and

ins-23 is significantly upregulated in sma-2 oocytes, whereas

ins-7 (Murphy et al., 2003) is downregulated, possibly indicating

insulin signaling from the oocytes themselves.

TGF-b Somatic and Oocyte Transcriptional Targets

Are Distinct

Sma/Mab L2 transcriptional targets regulate body size and male

tail patterning (Liang et al., 2007). We compared the expression

profiles of Sma/Mab mutant and wild-type early L4 whole

animals, prior to oocyte development; these targets are similar

to the L2 targets of Liang et al. (2007) (Figure 5E). In contrast to

sma-2 oocyte gene expression, the genes upregulated in Sma/

Mab larvae (Table S2C) include the GO terms of hedgehog

signaling, immunoglobulin domain proteins, leucine-rich repeat

proteins, cuticle collagens, and lipid and carbohydrate metabo-

lism genes (Figure 5E). Thus, at both the gene and GO term level,

the targets of Sma/Mab signaling in body size and oocyte quality

regulation are largely nonoverlapping (Figure 5E).

TGF-b Oocyte Targets Are Required for Reproductive

Quality Maintenance

To test candidate genes for their roles in reproduction, we used

RNAi knockdown to screen the top-ranking oocyte target genes

for their effects on sma-2 late embryo hatching, reasoning that

loss of important sma-2-upregulated genes might reduce repro-

ductive success. Of 60 genes tested, 27 reduced sma-2

embryo-hatching rates (Figure 6A and Figure S6A). We then

tested the genes with the strongest hatching effects for their

contributions to reproductive span determination and embryo/

oocyte quality (Figures 6B–6J and Figure S6). Three genes,

smc-4 (condensin, structural maintenance of chromosomes),

cyb-3 (cyclin B, sister chromatid segregation), and E03H4.8

(unknown, predicted vesicle coat complex), shortened sma-2

reproductive span substantially, from sma-2’s mean of 9 days

to < 3 days (Figure 6B). We found that these ‘‘early effect’’

genes also had severe effects on sma-2 embryonic viability,

producing almost exclusively unhatched embryos (Figures

6C–6F and Figure S6B). These genes are critical for oocyte

quality, because knocking them down in wild-type also resulted

in severe effects on embryonic viability (Figures S6D–S6F).

Knockdown of these genes also severely affected germline

and oocyte morphology; oocytes were largely unidentifiable,

distal germline cells were not well defined, and the gonads

themselves were misshapen (Figures 6G–6J and Figure S6C).

The loss of other sma-2-oocyte regulated genes also increased

the rate of unhatched embryos and/or unfertilized oocytes with

age in both sma-2 and wild-type, but later or more mildly

(Figures 7B–7D and Figure S7); these include math-33 (putative

apoptosis gene), F47G4.4 (putative chromosome segregation

gene), F52D10.2 and C06E7.4 (both unknown), and F21F3.3

(methyltransferase).

Because we had observed that progeny survival after DNA

damage was increased in sma-2 mutants (Figure 2D), and

a number of the DNA damage response genes upregulated in

sma-2 oocytes and were required for embryo viability

(Figure 5D and Table S3), we investigated these genes’ effects

on sma-2’s oocyte quality and post-g-irradiation embryonic

lethality. We found that loss of mlh-1, a DNA mismatch repair

homolog of human MLH1, increased the rate of unhatched

embryos and unfertilized oocytes late in sma-2 reproduction

(days 7–10; Figure 7A, Figure 6A, and Figure S6A). Loss of

uev-2 (stress/DNA damage response) and pme-5 (PARP/tankyr-

ase) had milder effects on hatching (Figures 7E and 7F,

Figure 6A, and Figure S6A). However, loss of uev-2 had a signif-

icant effect on post-g-irradiation sma-2 embryonic lethality

(Figure 7G).

Figure 5. TGF-b Sma/Mab Signaling Regulates Oocyte Quality and Body Size through Distinct Sets of Downstream Targets

(A) sma-9 RNAi adult-only treatment reduces body size significantly (p < 0.001, 14% decrease), whereas adult-only treatment does not (mean ± SEM).

(B) Mated reproductive spans of rrf-3 animals treated with control RNAi whole-life, with sma-9 RNAi whole-life, or with sma-9 RNAi in adulthood only (Table S1).

(C) Expression heat map of 386 genes significantly upregulated in sma-2;fem-1 oocytes (FDR = 0%).

(D) EnrichedGO terms for genes in (C). Example genes from this study (worm) and genes upregulated in young versus old mouse (Hamatani et al., 2004) or human

(Steuerwald et al., 2007) oocytes are listed, with highly homologous or important interacting genes in bold. (Expanded gene list is provided in Table S3.) GO terms

also enriched in younger human (H), mouse (W), or worm (W) oocytes are labeled with corresponding superscript letters. Asterisk indicates a gene involved

in corresponding GO function but failed to be recognized by DAVID (not included in gene counts).

(E) GO terms enriched in TGF-b Sma/Mab mutant oocytes are largely distinct from those enriched in Sma/Mab L4 and L2 (Liang et al., 2007) larvae.

* p < 0.05, **p < 0.01, and ***p < 0.001.


3 4 5 6 7 8 9 10 11 12

sma-2;control(RNAi) mated

0

20

40

60

80

100

%

day of adulthood

sma-2;smc-4(RNAi) mated

3 4 5 6 7 8 9 10 11 120

20

40

60

80

100

%

day of adulthood

sma-2;cyb-3(RNAi) mated

3 4 5 6 7 8 9 10 11 120

20

40

60

80

100

day of adulthood

sma-2;E03H4.8(RNAi) mated

3 4 5 6 7 8 9 10 11 120

20

40

60

80

100

day of adulthood

0

.2

.4

.6

.8

1

0 2 4 6 8 10 12

fraction r

epro

ductive

day of adulthood

sma-2;control(RNAi)sma-2;smc-4(RNAi)sma-2;E03H4.8(RNAi)sma-2;cyb-3(RNAi)

matedB

C

D

E

F

G

H

I

J

hatched embunhatched embunfertilized ooc

control

T09B

4.5ztf-

6 ag

s-3

C37

H5.3

mbk-1

F52D

10.2

F52B

11.1

atgr-7col

-64F2

1F3.3

C49

G7.7

B02

85.7

C49

G7.5

W01

C9.4

ins-23ugt

-48

T24A

6.7

F14F

9.4

cam-1

E03

H4.8

F14F

9.3

C33

D9.9

F47G

4.4

ced-5

fbxa-9

7sel

-12mbk-

2zak

-1mlh-1math

-330

1

2

3

4

5

6

7

8

9

10

control

mrp-2cdk

-4

C06

E7.4

nhr-15

4

T08B

2.7

alh-8

F10E

9.3

T26C

12.1

F11C

1.2

hnd-1srv

-7

T23F

2.5

crn-4

K09

F6.9

ins-22

F35E

12.4

zip-2

Y17

D7B

.2

epi-1

C06

C3.5

Y46

G5A

.19

lsy-2cyb

-3

F20C

5.5

F47G4.5

pme-5ulp

-1sm

c-4uev-2clk

-2% u

nh

atc

he

d e

mb

ryo

(d

ay 3

on

)

mate

rnal ste

rile

A

red= p<0.05orange= p<0.1green= >2 fold, p>0.1

79 7131

%%

sma-2;control(RNAi) mated

sma-2;smc-4(RNAi) mated

sma-2;cyb-3(RNAi) mated

sma-2;E03H4.8(RNAi) mated


Together, our expression results show that sma-2 regulates

a distinct set of genes in oocytes from its targets in body size

determination. Furthermore, our functional analyses of sma-2

Figure 6. TGF-b Sma/Mab Signaling Regulates Genes Essential for Embryonic Viability in Oocytes

(A) RNAi knockdown of many sma-2-regulated oocyte targets increase sma-2 mutant’s embryonic lethality (mean ± SEM).

(B) RNAi knockdown of smc-4, cyb-3, and E03H4.8 have early and severe effects on reproductive span (Table S1).

(C–F) RNAi knockdown of smc-4, cyb-3, and E03H4.8 greatly increase the percentage of unhatched embryos (orange) in mated sma-2 mutants (compare D-F

with C). Wild-type treated with RNAis shown in Figures S6D–S6F.

(G-J) smc-4, cyb-3, and E03H4.8 RNAi-treated sma-2 animals exhibit severely degraded germlines at day 8 (compare H–J with G). Contours of gonads shown in

yellow, visible oocytes outlined by dotted lines in (G).

E

A B

C D

G

H

F

Figure 7. TGF-b Sma/Mab Signaling Regu-

lates Genes Important for Age-Associated

Oocyte Quality Maintenance

(A–F) RNAi treatments of TGF-b target genes

accelerate oocyte quality decline, increasing the

percentage of unhatched embryos (orange) and/

or unfertilized oocytes (yellow) earlier in life

(compare with Figure 6C). mlh-1, math-33, and

F47G4.4 RNAis have greater effects (A-C),

whereas F52D10.2, uev-2, and pme-5 have milder

effects (D-F). Wild-type treated with RNAis shown

in Figures S7C–S7F and S7I–S7J).

(G) uev-2 RNAi treatment significantly increases

sma-2’s production of unhatched embryos (% ±

SEP) after g-irradiation, whereas pme-5 and mlh-

1 do not. Animals weremated with young wtmales

after irradiation.

(H) Model of reproductive aging regulation by the

TGF-b Sma/Mab (pink) and insulin/IGF-1 signaling

(red) pathways. Ligands (Insulin-Like Peptides,

TGF-b DBL-1) are secreted neuronally and

mediate signaling to the soma (hypodermis, intes-

tine, and muscle), generating as yet unidentified

secondary signals to regulate reproduction. These

secondary signals block distal germline and

oocyte integrity maintenance with age, resulting

in germline morphology decline, slowed germ

cell proliferation, and a decline in oocyte quality.

Downstream effectors in oocytes include chromo-

some segregation, cell cycle, DNA damage

response/repair genes, and so forth. Declines in

embryonic viability and infertility mark reproduc-

tive cessation. The germline and somatic gonad

regulate somatic aging (Hsin and Kenyon, 1999),

suggesting a bi-directional signaling flow in the

coordination of somatic and germline aging.

(Photo courtesy of Ian Chin-Sang.)

* p < 0.05, **p < 0.01, and ***p < 0.001.

oocyte targets suggest several mecha-

nisms that are required for successful

extended reproduction, and many of

these mechanisms are shared with

mammalian oocytes. Chromosome

segregation and cell cycle genes are

more highly expressed in sma-2 than in

wild-type worms, and their loss causes

severe functional and morphological

germline defects, suppressing reproduc-

tion. DNA damage response and other novel genes, by

contrast, are required later in reproduction to maintain repro-

ductive fidelity.


DISCUSSION

Herewe have systematically examined the processes involved in

reproduction, from embryonic viability through distal germline

morphology, to determine which are most susceptible to aging

and which are altered in mutants with extended reproductive

spans. Our data establish that oocyte and distal germline quality

correlate well with reproductive success and that TGF-b Sma/

Mab and Insulin/IGF-1 signaling regulate reproductive aging

primarily through their control of these aspects of reproduction.

A Model for TGF-b Sma/Mab and IIS Regulation

of Reproductive Aging

Our mosaic and hypodermal rescue data suggest that the TGF-b

pathway regulates reproductive aging through an interaction

between the soma and germline. We previously showed that

TGF-b signaling regulates reproductive aging independently of

such somatically controlled mechanical processes as ovulation

and body growth (Luo et al., 2009), and our mosaic analysis

supports this uncoupling of reproductive span and ovulation

(Figures S3D and S3E). Thus, the interaction between the soma

andgermline to regulate reproductive aging is likely tobemediated

by molecular signals. These secondary signals must originate in

somatic (hypodermal) tissues downstream of TGF-b signaling,

and subsequently act in the germline to control quality

(Figure7H).Similarly, IIS acts in themuscle and intestine to regulate

germline and oocyte maintenance. Although the specific signals

have not yet been identified, insulin-like peptides are regulated by

IIS and coordinate the state of the insulin pathway between tissues

(Murphy et al., 2007), and a nuclear hormone receptor is required

for starvation-induced adult reproductive diapause (Angelo and

Van Gilst, 2009). Together with our data, the observation that

signals from the germline and somatic gonad regulate longevity

(Flatt et al., 2008; Ghazi et al., 2009; Hsin and Kenyon, 1999),

suggests a bidirectional flow of information between somatic and

reproductive tissues normally coordinates their rates of aging.

The Distal Germline and Reproductive Aging

TGF-b Sma/Mab and IIS mutations prevent age-related decline

in the integrity of the distal germline-containing germline stem

cells, and the quality of the distal germline and oocytes are corre-

lated (Figure S2J). Interestingly, germline stem cells protected by

starvation have the capacity to regenerate and reestablish repro-

duction, even after a long period of quiescence (Angelo and Van

Gilst, 2009). Although this is the first report of C. elegans TGF-

b signaling possibly regulating germline stem cell activity in

C. elegans, TGF-b/BMP signaling is known to affect GSC devel-

opment in other organisms, including Drosophila germline and

mammalian muscles (Carlson et al., 2009; Yamashita et al.,

2005; Zhao et al., 2008). The upregulation of LIN-28, a key regu-

lator of stem cell induction, in the TGF-b mutant reproductive

system is particularly intriguing.

TGF-b Sma/Mab Signaling Regulates Reproductive

Aging Distinctly from Body Growth

Although TGF-b Sma/Mab signaling regulates both body growth

and reproductive aging, the downstream molecular mechanisms

of these two processes are distinct. First, Sma/Mab signaling is

required for body size regulation during development, before

gametogenesis (Liang et al., 2003; Savage-Dunn et al., 2000),

whereas Sma/Mab regulation of germ line aging is carried out in

adulthood (Figure 5B). Second, body size and reproductive span

are not correlated (Luo et al., 2009). Furthermore, despite the

fact that Sma/Mab activity in the hypodermis directs both body

growth and oocyte quality, the Sma/Mab pathway has distinct

transcriptional targets in the body and oocytes. Interestingly, we

find that theseoocyte-specific targetscanbeseparated intoearly-

and late-effect genes, with chromosome segregation and cell

cycle genes having early and severe effects on reproductive

tissues, and DNA damage response genes primarily regulating

late effects. The late effects are particularly interesting, as they

are most likely to become increasingly important as oocytes age.

C. elegans as a Model of Reproductive Aging

Although worms and humans have vastly different life spans and

reproductive strategies, the cellular and molecular bases of

reproductive span regulation are strikingly similar. As we have

shown here for C. elegans, oocyte quality decline is the major

reason for human reproductive capacity decline, resulting in

sterility and developmental birth defects. Chromosomal abnor-

malities, in particular aneuploidies, are themajor defect in human

embryos from aging mothers (te Velde and Pearson, 2002).

Worms also increase chromosome nondisjunction rates with

age (Rose and Baillie, 1979; Tang et al., 2010) (Figures 1B–1D),

and we find that mutants with extended reproductive success

significantly reduce chromosomal nondisjunction rates. Oocyte

fertilizability, stress resistance, and morphology are compro-

mised with age in humans (Blondin et al., 1997; Goud et al.,

1999); we found that this is also the case forC. elegans, but these

declines aredelayed in TGF-band IISmutants. Finally, our oocyte

transcriptional and functional analyses show that genes upregu-

lated in TGF-bmutants are strikingly similar tomammalian oocyte

genes that decline with age, suggesting that many of the molec-

ular mechanisms underlying reproductive cessation are shared

between C. elegans and humans. Therefore, C. elegans not

only regulates reproductive aging through oocyte quality control,

as do humans, but also, such control is mediated through the

regulation of similar oocyte quality maintenance mechanisms.

The fact that both Insulin/IGF-1 and TGF-b signaling, two path-

ways that are evolutionarily conserved from worms to humans,

havesignificant roles in regulating the rateof reproductive decline

andutilize similarmechanisms, suggests that thesepathways are

also likely to be important in the regulation of human reproductive

decline. Several recent genome-wide association studies of

human reproductive development and menopause identified

genes that regulate development and longevity in C. elegans

(Ong et al., 2009; Stolk et al., 2009). These genes include

FOXO3a, the human homolog of the DAF-16/FoxO transcription

factor downstream of the Insulin/IGF-1 signaling pathway, and

LIN-28, which we find is regulated by TGF-b signaling in oocytes.

TGF-b signaling has also been implicated in several aspects

of mammalian reproduction and reproductive aging. TGF-b

members are upregulated in mouse oocytes with age (Hamatani

et al., 2004) andmany TGF-b superfamily ligands regulate follicu-

logenesis (Knight and Glister, 2006; Trombly et al., 2009).

Although humans have a more complex TGF-b pathway family


that performs many different functions, it is likely that TGF-b

signaling may be involved in regulation of reproductive cessa-

tion. Therefore, the similarities in the regulation of reproductive

aging in worms and humans may allow us to use worms as

genetic and molecular models to study this important human

problem, enabling the development of therapies to address

maternal age-related birth defects and reproductive decline.


Extended Experimental Procedures are presented in Supplemental Informa-

tion, and include C. elegans strains used and analyses of embryonic lethality,

male progeny production, chromosome bivalents, oocyte fertilizability, RNP

foci, mitotic germ cell number, physiological and irradiation-induced apo-

ptosis, reproductive span, ovulation rate, body length, and temporal RNAi

effects.

Oocyte Morphology Analysis

For each oocyte image, a score was assigned for each of the three signs of

deterioration (cavities, graininess, and cellularization), according to the

severity of the phenotype, with 1 equals normal, 2 equals mild, and 3 equals

severe. Mann-Whitney analysis was used to determine whether there were

significant differences in pairwise comparisons. An individual who was blind

to the genotypes scored the images independently.

Distal Germline Morphology Analysis

For each distal germline image, a score was assigned for each of the three

signs of deterioration (cavities, graininess, and cellularization), according to

the severity of the phenotype, with 1 equals normal, and 5 (or 3 for

Figure 3D and Figure 4F) equals most severe, by four individuals (three were

blind to the genotypes) and averaged. Mann-Whitney (pairwise) analyses

were used as described above.

Immunostaining

Staining with RME-2 antibody, a gift from Dr. Barth Grant, was performed as

described elsewhere (Grant and Hirsh, 1999).

Mosaic Analysis

Developmentally synchronized sma-3(wk30) III;qcEx26 X [pCS29+sur-5::gfp]

worms with somatic GFP expression were picked; green fluorescence in

tissues including hypodermis, intestine, neurons, but not germline, was

verified at high magnification (Figures S3A–S3C). Worms were screened for

large body size beforemating. Animals with no fluorescent progeny are ‘‘germ-

line-lost’’ worms.

Hypodermal Rescue Strain Construction

sma-3(wk30) were injected with pCS227[Pvha-7::sma-3] at 90 ng/ml (strains

and plasmid kindly provided by Dr. Cathy Savage-Dunn) with Pmyo-

2::mCherry (PFC590, Addgene) as a coinjection marker (5 ng/ml). Large F1s

were picked to establish independent lines for follow-up analysis.

Oocyte and L4 Microarrays

Hypochlorite-synchronizedwild-typeand sma-2orsma-4 larvaewerecollected

atmid-L4.Oocyteswere isolated (Miller, 2006) from fem-1 (day3 andday 8) and

sma-2;fem-1 (day 8) adults; RNA was extracted, and cRNA was linearly ampli-

fied, Cy3/Cy5 labeled, hybridized to the Agilent 44kC. elegansmicroarray, and

analyzed as described elsewhere (Shaw et al., 2007). GO analysis was per-

formed using DAVID (Dennis et al., 2003; Huang et al., 2009) on significantly

differentially expressed genes (FDR = 0%, SAM; Tusher et al., 2001).

ACCESSION NUMBERS

The microarray data can be found in the Gene Expression Omnibus (GEO) of

NCBI through accession number GSE23509 or in PUMAdb (http://puma.

princeton.edu).


Supplemental Information includes Extended Experimental Procedures, seven

figures, and four tables and can be foundwith this article online at doi:10.1016/

j.cell.2010.09.013.

ACKNOWLEDGMENTS

We thank Cathy Savage-Dunn (CUNY) for CS122 and pCS227; David Sher-

wood (Duke University) for NK640; Barth Grant (Rutgers) for RT130 and a-

RME-2 antibody; members of the Murphy Laboratory and Z. Gitai for

comments on the manuscript; and March of Dimes Basil O’Connor Starting

Scholar, NIH New Innovator (1DP2OD004402-01), and NIH (P50 GM071508)

awards for funding.

S.L. and C.T.M. planned the experiments and wrote the manuscript; S.L.

performed all the experiments except L4 microarrays (W.M.S.), with assis-

tance from G.A.K. (Figure 1D and Figure S1E, generation of mosaic and

tissue-specific transgenic animals for Figure 3, Figure S3, and Figure 7H),

and J.M.A. (technical assistance).

Received: January 20, 2010




REFERENCES

Andux, S., and Ellis, R.E. (2008). Apoptosis maintains oocyte quality in aging

Caenorhabditis elegans females. PLoS Genet. 4, e1000295.

Angelo, G., and Van Gilst, M.R. (2009). Starvation protects germline stem cells

and extends reproductive longevity in C. elegans. Science 326, 954–958.

Apfeld, J., and Kenyon, C. (1998). Cell nonautonomy of C. elegans daf-2 func-

tion in the regulation of diapause and life span. Cell 95, 199–210.

Blondin, P., Coenen, K., and Sirard, M.A. (1997). The impact of reactive oxygen

species on bovine sperm fertilizing ability and oocytematuration. J. Androl. 18,

454–460.

Cant, M.A., and Johnstone, R.A. (2008). Reproductive conflict and the separa-

tion of reproductive generations in humans. Proc. Natl. Acad. Sci. USA 105,

5332–5336.

Carlson, M.E., Conboy, M.J., Hsu, M., Barchas, L., Jeong, J., Agrawal, A., Mi-

kels, A.J., Agrawal, S., Schaffer, D.V., and Conboy, I.M. (2009). Relative roles

of TGF-beta and Wnt in the systemic regulation and aging of satellite cell

responses. Aging Cell 8, 676–689.

Dennis, G., Jr., Sherman, B.T., Hosack, D.A., Yang, J., Gao, W., Lane, H., and

Lempicki, R.A. (2003). DAVID: database for annotation, visualization, and inte-

grated discovery. Genome Biol. 4, 3.

Dillin, A., Crawford, D.K., and Kenyon, C. (2002). Timing requirements for

insulin/IGF-1 signaling in C. elegans. Science 298, 830–834.

Flatt, T., Min, K.-J., D’Alterio, C., Villa-Cuesta, E., Cumbers, J., Lehmann, R.,

Jones, D.L., and Tatar, M. (2008). Drosophila germ-line modulation of insulin

signaling and lifespan. Proc. Natl. Acad. Sci. USA 105, 6368–6373.

Garigan, D., Hsu, A.-L., Fraser, A.G., Kamath, R.S., Ahringer, J., and Kenyon,

C. (2002). Genetic analysis of tissue aging inCaenorhabditis elegans: a role for

heat-shock factor and bacterial proliferation. Genetics 161, 1101–1112.

Gartner, A., Milstein, S., Ahmed, S., Hodgkin, J., and Hengartner, M.O. (2000).

A conserved checkpoint pathway mediates DNA damage–induced apoptosis

and cell cycle arrest in C. elegans. Mol. Cell 5, 435–443.

Ghazi, A., Henis-Korenblit, S., and Kenyon, C. (2009). A transcription elonga-

tion factor that links signals from the reproductive system to lifespan extension

in Caenorhabditis elegans. PLoS Genet. 5, e1000639.

Goud, P., Goud, A., Van Oostveldt, P., Van der Elst, J., and Dhont, M. (1999).

Fertilization abnormalities and pronucleus size asynchrony after intracytoplas-

micsperm injectionare related tooocytepostmaturity.Fertil.Steril.72, 245–252.


Grant, B., and Hirsh, D. (1999). Receptor-mediated endocytosis in the Caeno-

rhabditis elegans oocyte. Mol. Biol. Cell 10, 4311–4326.

Greenstein, D. (2005). Control of oocyte meiotic maturation and fertilization.

WormBook Dec 28, 1–12.

Gumienny, T.L., Lambie, E., Hartwieg, E., Horvitz, H.R., and Hengartner, M.O.

(1999). Genetic control of programmed cell death in the Caenorhabditis ele-

gans hermaphrodite germline. Development 126, 1011–1022.

Hagedorn, E.J., Yashiro, H., Ziel, J.W., Ihara, S.,Wang, Z., and Sherwood, D.R.

(2009). Integrin acts upstream of netrin signaling to regulate formation of the

anchor cell’s invasive membrane in C. elegans. Dev. Cell 17, 187–198.

Hamatani, T., Falco, G., Carter,M.G., Akutsu,H., Stagg, C.A., Sharov, A.A., Du-

dekula, D.B., VanBuren, V., and Ko, M.S. (2004). Age-associated alteration of

gene expression patterns in mouse oocytes. Hum.Mol. Genet. 13, 2263–2278.

Hartge, P. (2009). Genetics of reproductive lifespan. Nat. Genet. 41, 637–638.

Hodgkin, J., Horvitz, H.R., and Brenner, S. (1979). Nondisjunction mutants of

the nematode Caenorhabditis elegans. Genetics 91, 67–94.

Hsin, H., and Kenyon, C. (1999). Signals from the reproductive system regulate

the lifespan of C. elegans. Nature 399, 362–366.

Huang, C., Xiong, C., and Kornfeld, K. (2004). Measurements of age-related

changes of physiological processes that predict lifespan of Caenorhabditis el-

egans. Proc. Natl. Acad. Sci. USA 101, 8084–8089.

Huang, D.W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and inte-

grative analysis of large gene lists using DAVID bioinformatics resources. Nat.

Protoc. 4, 44–57.

Hughes, S.E., Evason, K., Xiong, C., and Kornfeld, K. (2007). Genetic and phar-

macological factors that influence reproductive aging in nematodes. PLoS

Genet. 3, e25.

Jud, M., Razelun, J., Bickel, J., Czerwinski, M., and Schisa, J. (2007). Conser-

vation of large foci formation in arrested oocytes of Caenorhabditis nema-

todes. Dev. Genes Evol. 217, 221–226.

Kenyon, C. (2005). The plasticity of aging: insights from long-lived mutants.

Cell 120, 449–460.

Kenyon, C., Chang, J., Gensch, E., Rudner, A., and Tabtiang, R. (1993). A C.

elegans mutant that lives twice as long as wild type. Nature 366, 461–464.

Knight, P.G., and Glister, C. (2006). TGF-beta superfamily members and

ovarian follicle development. Reproduction 132, 191–206.

Liang, J., Lints, R., Foehr, M.L., Tokarz, R., Yu, L., Emmons, S.W., Liu, J., and

Savage-Dunn, C. (2003). TheCaenorhabditis elegans schnurri homolog sma-9

mediates stage- and cell type-specific responses to DBL-1 BMP-related

signaling. Development 130, 6453–6464.

Liang, J., Yu, L., Yin, J., and Savage-Dunn, C. (2007). Transcriptional repressor

and activator activities of SMA-9 contribute differentially to BMP-related

signaling outputs. Dev. Biol. 305, 714–725.

Libina, N., Berman, J.R., and Kenyon, C. (2003). Tissue-specific activities ofC.

elegans DAF-16 in the regulation of lifespan. Cell 115, 489–502.

Luo, S., Shaw, W.M., Ashraf, J., and Murphy, C.T. (2009). TGF-beta Sma/Mab

signaling mutations uncouple reproductive aging from somatic aging. PLoS

Genet. 5, e1000789.

Magli, M.C., Gianaroli, L., Ferraretti, A.P., Lappi, M., Ruberti, A., and Farfalli, V.

(2007). Embryo morphology and development are dependent on the chromo-

somal complement. Fertil. Steril. 87, 534–541.

Massague, J. (2000). How cells read TGF-[beta] signals. Nat. Rev. Mol. Cell

Biol. 1, 169–178.

Mehlmann, L.M. (2005). Stops and starts in mammalian oocytes: recent

advances in understanding the regulation of meiotic arrest and oocyte matu-

ration. Reproduction 130, 791–799.

Miller, M.A. (2006). Sperm and oocyte isolation methods for biochemical and

proteomic analysis. Methods Mol. Biol. 351, 193–201.

Murphy, C.T., Lee, S.-J., and Kenyon, C. (2007). Tissue entrainment by feed-

back regulation of insulin gene expression in the endoderm of Caenorhabditis

elegans. Proc. Natl. Acad. Sci. USA 104, 19046–19050.

Murphy, C.T., McCarroll, S.A., Bargmann, C.I., Fraser, A., Kamath, R.S., Ah-

ringer, J., Li, H., and Kenyon, C. (2003). Genes that act downstream of DAF-

16 to influence the lifespan of Caenorhabditis elegans. Nature 424, 277–283.

Nimmo, R., and Slack, F. (2009). An elegant miRror: microRNAs in stem cells,

developmental timing and cancer. Chromosoma 118, 405–418.

Ong, K.K., Elks, C.E., Li, S., Zhao, J.H., Luan, J.A., Andersen, L.B., Bingham,

S.A., Brage, S., Smith, G.D., Ekelund, U., et al. (2009). Genetic variation in

LIN28B is associated with the timing of puberty. Nat. Genet. 41, 729–733.

Rose, A.M., and Baillie, D.L. (1979). The effect of temperature and parental age

on recombination and nondisjunction in Caenorhabditis elegans. Genetics 92,

409–418.

Rubio, C., Simon, C., Vidal, F., Rodrigo, L., Pehlivan, T., Remohi, J., and

Pellicer, A. (2003). Chromosomal abnormalities and embryo development in

recurrent miscarriage couples. Hum. Reprod. 18, 182–188.

Saito, T.T., Youds, J.L., Boulton, S.J., and Colaiacovo, M.P. (2009). Caeno-

rhabditis elegansHIM-18/SLX-4 interacts with SLX-1 and XPF-1 andmaintains

genomic integrity in the germline by processing recombination intermediates.

PLoS Genet. 5, e1000735.

Savage-Dunn, C. (2005). TGF-b signaling. WormBook Sep 9, 1–12.

Savage-Dunn, C., Tokarz, R.,Wang, H., Cohen, S., Giannikas, C., and Padgett,

R.W. (2000). SMA-3 Smad has specific and critical functions in DBL-1/SMA-6

TGFbeta-related signaling. Dev. Biol. 223, 70–76.

Shaw, W.M., Luo, S., Landis, J., Ashraf, J., and Murphy, C.T. (2007). The

C. elegans TGF-beta Dauer pathway regulates longevity via insulin signaling.

Curr. Biol. 17, 1635–1645.

Steuerwald, N.M., Bermudez, M.G., Wells, D., Munne, S., and Cohen, J.

(2007). Maternal age-related differential global expression profiles observed

in human oocytes. Reprod. Biomed. Online 14, 700–708.

Stolk, L., Zhai, G., van Meurs, J.B.J., Verbiest, M.M.P.J., Visser, J.A., Estrada,

K., Rivadeneira, F., Williams, F.M., Cherkas, L., Deloukas, P., et al. (2009). Loci

at chromosomes 13, 19 and 20 influence age at natural menopause. Nat.

Genet. 41, 645–647.

Suh,Y., Atzmon,G.,Cho,M.-O.,Hwang,D., Liu,B., Leahy,D.J., Barzilai, N., and

Cohen, P. (2008). Functionally significant insulin-like growth factor I receptor

mutations in centenarians. Proc. Natl. Acad. Sci. USA 105, 3438–3442.

Tang, L., Machacek, T., Mamnun, Y.M., Penkner, A., Gloggnitzer, J., Wegros-

tek, C., Konrat, R., Jantsch, M., Loidl, J., and Jantsch, V. (2010). Mutations in

Caenorhabditis elegans him-19 show meiotic defects that worsen with age.

Mol. Biol. Cell. 21, 885–896.

te Velde, E.R., and Pearson, P.L. (2002). The variability of female reproductive

ageing. Hum. Reprod. Update 8, 141–154.

Tokunaga, C. (1970). The effects of low temperature and aging on nondisjunc-

tion in Drosophila. Genetics 65, 75–94.

Trombly, D.J., Woodruff, T.K., and Mayo, K.E. (2009). Roles for transforming

growth factor beta superfamily proteins in early folliculogenesis. Semin. Re-

prod. Med. 27, 14–23.

Tusher, V.G., Tibshirani, R., and Chu, G. (2001). Significance analysis of micro-

arrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA

98, 5116–5121.

Wang, J., Tokarz, R., and Savage-Dunn, C. (2002). The expression of TGF-

beta signal transducers in the hypodermis regulates body size in C. elegans.

Development 129, 4989–4998.

Wolkow, C.A., Kimura, K.D., Lee, M.S., and Ruvkun, G. (2000). Regulation of C. ele-

gans life-span by insulinlike signaling in the nervous system. Science 290, 147–150.

Yamashita, Y.M., Fuller, M.T., and Jones, D.L. (2005). Signaling in stem cell

niches: lessons from the Drosophila germline. J. Cell Sci. 118, 665–672.

Zhao, R., Xuan, Y., Li, X., and Xi, R. (2008). Age-related changes of germline

stem cell activity, niche signaling activity and egg production in Drosophila.

Aging Cell 7, 344–354.


A Myc Network Accounts for Similaritiesbetween Embryonic Stem and CancerCell Transcription ProgramsJonghwan Kim,1,2,3 Andrew J. Woo,1,3 Jianlin Chu,1,2,3 Jonathan W. Snow,1,2,3 Yuko Fujiwara,1,2,3,4 Chul Geun Kim,5

Alan B. Cantor,1,3 and Stuart H. Orkin1,2,3,4,*1Department of Pediatric Oncology, Children’s Hospital and Dana Farber Cancer Institute2Harvard Stem Cell Institute3Harvard Medical School4Howard Hughes Medical Institute

Boston, MA 02115, USA5Department of Life Science, Hanyang University, Seoul 133-791, Republic of Korea


DOI 10.1016/j.cell.2010.09.010

SUMMARY

c-Myc (Myc) is an important transcriptional regulator

in embryonic stem (ES) cells, somatic cell reprogram-

ming, and cancer. Here, we identify a Myc-centered

regulatory network in ES cells by combining pro-

tein-protein and protein-DNA interaction studies

and show that Myc interacts with the NuA4 complex,

a regulator of ES cell identity. In combination with

regulatory network information, we define three ES

cell modules (Core, Polycomb, and Myc) and show

that the modules are functionally separable, illus-

trating that the overall ES cell transcription program

is composed of distinct units. With these modules

as an analytical tool, we have reassessed the hypoth-

esis linking an ES cell signature with cancer or cancer

stem cells. We find that the Myc module, indepen-

dent of the Core module, is active in various cancers

and predicts cancer outcome. The apparent simi-

larity of cancer and ES cell signatures reflects, in

large part, the pervasive nature of Myc regulatory

networks.

INTRODUCTION

The pluripotent state of embryonic stem (ES) cells is maintained

through the combinatorial actions of core transcription factors,

including Oct4, Sox2, and Nanog (Boyer et al., 2005; Chen

et al., 2008; Kim et al., 2008; Loh et al., 2006), in addition to other

regulatory mechanisms encompassing epigenetic regulation

(Boyer et al., 2006; Lee et al., 2006), microRNAs (Marson et al.,

2008; Melton et al., 2010), and signaling pathways (Niwa et al.,

1998; Sato et al., 2004). The discovery that cocktails of core

pluripotency factors and selected widely expressed factors,

such as Myc and Lin28, reprogram differentiated cells to an

ES-like state (Park et al., 2008; Takahashi and Yamanaka,

2006; Yu et al., 2007) underscores the central role of transcrip-

tion factors in cell fate decisions (Graf and Enver, 2009). Compre-

hensive protein interaction and target gene assessment of core

pluripotency factors has provided a framework for conceptual-

izing the regulatory network that supports the ES cell state.

Striking among the features of this network is the extent to which

the core factors physically associate within protein complexes,

co-occupy target genes, and cross-regulate each other (Boyer

et al., 2005; Chen et al., 2008; Kim et al., 2008; Loh et al.,

2006; Wang et al., 2006).

Although its expression dramatically enhances induced

pluripotent (iPS) cell formation, Myc is not an integral member

of the core pluripotency network (Chen et al., 2008; Hu et al.,

2009; Kim et al., 2008). Myc occupies considerably more

genomic target genes than the core factors, and Myc targets

are involved predominantly in cellular metabolism, cell cycle,

and protein synthesis pathways, whereas the targets of core

factors relate more toward developmental and transcription-

associated processes (Kim et al., 2008). Interestingly, promoters

occupied by Myc show a strong correlation with a histone H3

lysine 4 trimethylation (H3K4me3) signature and a reverse corre-

lation with histone H3 lysine 27 trimethylation (H3K27me3), sug-

gesting a connection between Myc and epigenetic regulation

(Kim et al., 2008). It is notable that the H3K4me3 signature has

a positive correlation with active genes, and an open chromo-

somal structure, a distinctive feature of ES cells (Meshorer

et al., 2006). Studies in non-ES cells have also revealed that

Myc interacts with histone acetyltransferases (HATs) (Doyon

and Cote, 2004; Frank et al., 2003). Improved iPS cell generation

by addition of histone deacetylase inhibitors implies that global

changes in epigenetic signatures are critical to efficient somatic

cell reprogramming (Huangfu et al., 2008).

Although they remain pluripotent, ES cells are capable of indef-

inite self-renewal. Both blocked differentiation and the capacity

for self-renewal, hallmarks of ES cells and adult stem cells, are

shared in part by cancer cells (Clarke and Fuller, 2006; Reya

et al., 2001). Although contested in the literature, expression of

pluripotency factors, such as Oct4 and Nanog, has been


described in some cancers (Kang et al., 2009; Schoenhals et al.,

2009). The involvement of Myc in many cancers (Cole and

Henriksson, 2006) and its effects in iPS cell generation raise

important issues regarding the relationship between cancer

and embryonic stem cell states. Moreover, renewed focus on

tumor subpopulations that initiate tumor formation on transfer

to a suitable host (cancer stem cells) has contributed to the

comparison of cancers and stemcells and to thepotential resem-

blance of metastatic cancer cells to stem cells.

These relationships have been reinforced by reports of ‘‘stem

cell’’ or ‘‘embryonic stem cell’’ (ESC)–like signatures in human

and mouse cancers (Ben-Porath et al., 2008; Wong et al.,

2008a; Wong et al., 2008b). The properties of such ESC-like

signatures have thus far not been clearly defined, leaving open

the possibility that they are composed of multiple gene expres-

sion signatures that are the outcomes of functionally indepen-

dent transcriptional regulatory networks. Cancer cells may

share only one or few of these subdivided signatures observed

in ES cells, and thus have relatively less in common with the

‘‘embryonic state’’ than recently suggested.

In the present study, we sought to define how the regulatory

network controlled by Myc relates to the previously defined

core pluripotency network (Boyer et al., 2005; Chen et al.,

2008; Kim et al., 2008; Loh et al., 2006). We first identified

a Myc-centered regulatory network in ES cells and revealed

that this Myc-centered network is largely independent of the

core ES cell pluripotency network. On the basis of these findings,

we propose that the overall ES cell specific gene expression

signature is composed of smaller sets of subsignatures, which

are represented as ‘‘modules’’—modules for the core pluripo-

tency factors (Core module), the Polycomb complex factors

(PRC module), and the Myc-related factors (Myc module). We

provide evidence that these modules are functionally indepen-

dent in ES cells, as well as during somatic cell reprogramming.

With these modules as analytical tools, we observe that ES cells

and cancer cells share Myc module activity, but generally do not

share Core module activity. These findings argue against the

hypothesis that cancer cells often reactivate an embryonic

stem cell gene signature, even as they progress to a more highly

invasive or metastatic state. Instead, the common features of ES

cells and cancer cells reflect in large part the pervasive nature of

the Myc regulatory network.

RESULTS

Construction of a Myc-Centered Protein-Protein

Interaction Network in ES Cells

Previous protein-DNA interaction studies in ES cells indicated

that targets occupied by the core pluripotency factors differ

from genes bound by Myc (Chen et al., 2008; Kim et al., 2008).

A recent RNA interference–based functional screen additionally

suggested the existence of a second network linked functionally

with Myc (Hu et al., 2009). Because coregulators that function

with Myc have not been characterized previously in ES cells,

we first sought to identify protein complexes that contain Myc

with Myc-associated factors in ES cells. Using the in vivo meta-

bolic biotin tagging method (de Boer et al., 2003; Wang et al.,

2006), protein complexes containing tagged Myc in ES cells

were affinity purified and analyzed by mass-spectrometry. We

identified several proteins known to interact with Myc in other

cell types, including Max, Ep400, Dmap1, and Trrap (Figure 1A)

(Cai et al., 2003; Fuchs et al., 2001; McMahon et al., 1998). To

expand and validate the protein-protein interaction network

encompassing Myc, we subsequently generated ES cell lines

expressing tagged Max and tagged Dmap1. ES cells expressing

tagged Tip60 and tagged Gcn5 were also generated because

they are HATs and known interacting partners of Trrap (Ikura

et al., 2000; McMahon et al., 2000). We also generated tagged

E2F4 ES cells, because another E2F family member E2F1 shares

many common targets with Myc (Chen et al., 2008). E2F1 and

E2F4 have many common targets and interchangeable roles in

normal and tumor cells (Xu et al., 2007). Among E2F family

proteins, E2F4 shows strongest expression in ES cells. In

summary, we established ES cell lines expressing tagged Myc,

Max, Dmap1, Tip60, Gcn5, and E2F4 (Figure 1A and Figure S1A

available online) and identified their interacting partner pro-

teins (summarized in Table S1). Figure 1A shows lists of high

confidence interacting partner proteins of each factor tested.

Interactions were independently validated by coimmunoprecipi-

tation (Figure 1C and Figure S1B).

Myc Interacts with the NuA4 HAT Complex in ES Cells

We did not observe overlap of proteins existing between the

core protein interaction network (Wang et al., 2006) and the

Myc-centered protein interaction network (Figure S1C).

Although this may be due to the stringency of our conditions

for recovery of protein complexes, within each network we

observed a high degree of interactions, strongly suggesting

that these two networks, and their protein complexes, are phys-

ically separate. Interestingly, we observed that Myc interacts

with many proteins in a recognized conserved protein complex

known as NuA4 HAT (or the Tip60-Ep400 complex) (Doyon and

Cote, 2004) as shown in Figure 1A (pink cells) and Figure 1B

(proteins in a pink circle). Myc, Max, Dmap1, Tip60, Trrap, and

Ep400 are tightly interconnected within the network; however,

Gcn5 and E2F4 show a lower degree of association, suggesting

their weak or indirect interaction with Myc/NuA4. It has been

suggested that transcription factors, such as Myc, p53, and

E2Fs, require the NuA4 complex to activate downstream targets

in non-ES cell contexts (Ard et al., 2002; McMahon et al., 1998).

Our data (Figure 1 and Table S1) strongly support the view that

Myc interacts with an intact NuA4 HAT complex in ES cells,

also implying that histone 3 and 4 acetylation (AcH3 and AcH4,

respectively) signatures may also be generated in part by the

Myc/NuA4 complex via Tip60 in ES cells. Previous RNAi-based

phenotypic analyses in ES cells revealed that factors in the

NuA4 HAT complex, including Ep400, Dmap1, Tip60, Trrap,

Ruvb1, and Ruvb2, are critical to ES cell identity (Fazzio et al.,

2008) (also our observation, Figures S1D and S1E). These find-

ings imply a crucial role for the Myc/NuA4 complex in ES cells.

Construction of a Myc-Centered Protein-DNA

Interaction Network in ES Cells

To identify genomic targets of Myc and its associated factors

tested in Figure 1, we performed bioChIP-chip (Kim et al.,

2008). Because Tip60 and Gcn5 generate AcH3 and AcH4


histone modification signatures, we also performed ChIP reac-

tions using native antibodies against AcH3 and AcH4. We found

that the six factors we tested (Myc, Max, Dmap1, Tip60, E2F4,

and Gcn5) co-occupy many target promoters in close proximity

(Figure 2A).

To obtain a global view of individual and multiple transcription

factor occupancy, we combined this new data set with previ-

ously published ChIP-chip or ChIP-sequencing data sets (Boyer

et al., 2006; Chen et al., 2008; Ding et al., 2009; Hu et al., 2009;

Kim et al., 2008; Shen et al., 2008) and tested the factor

A

B

C

Gcn5

E2F4

Max

Myc

Tip60

Dmap1

Ldha

Hdgf

Acl6a

Trrap

Brd8

Ep400

Epc1

Epc2

Vps72

Ing3

Actb

Actg1

Trrap Mga Trrap Trrap Trrap Rbl1

Ep400 Trrap Ep400 Tip60 Taf5l E2F4

Myc Mnt Dmap1 Ep400 Gcn5 Tdp1

Max Max Srcap Dmap1 Tad3l Cdc2

Dmap1 Ep400 Brd8 Brd8 Taf6l Tdp2

Brd8 Lmbl2 Yets4 Epc2 Tada1l Ccna2

Epc1 Dmap1 Epc2 Vps72 Krt2 Rb

Epc2 Mycn Epc1 Epc1 Rae1l Tfdp2

Wdr5 Arp6 Ing3 Supt3h Ldha

Ring2 Tip60 Actg2 Taf9

Myc Vps72 Mo4l1 Tcpg

Tip60 Ing3 Actb Pcbp2

Brd8 Actb Actg1 Sf3b3

Mxi1 Actg1 Acl6a Syd

Pcgf6 Acl6a Cpin1 Ldha

Acl6a Ruvb1 Ldha Hdgf

Hsp72 Ruvb2 Hdgf

Myc

complex

Max

complex

Dmap1

complex

Tip60

complex

Gcn5

complex

E2F4

complex

Myc Max Dmap1

Tip60 Gcn5

Trrap E2F4

BirA Dmap1 Gcn5 Max E2F4 Tip60 Myc

In IP In IP In IP In IP In IP In IP In IP

Myc

Dmap1

Brd8

Gcn5

Ep400

Trrap

Max

Ing3

Tip60

E2F4

Figure 1. Myc-Centered Protein-Protein Interaction Network in ES Cells

(A) Schematic representation of the strategy for mapping a Myc-centered protein-protein interaction network in ES cells. High-confidence components of

multiprotein complexes were identified and listed in the table. Pink cells represent NuA4 complex proteins.

(B) Depiction of the features of the Myc-centered protein-protein interaction network. Proteins with green labels are biotin tagged proteins and pink circles

indicate NuA4 complex proteins. Proteins identified by multiple biotin-tagged factors are shown. Entire protein interaction network is shown in Figure S1C.

See also Table S1.

(C) Validation of the interaction network by coimmunoprecipitation.


occupancy or histone modification signatures (see Supple-

mental Information). The numbers of genes that are occupied

by a tested factor or marked by a tested histone modification

signature are summarized in Figure 2B and Table S2 with a hier-

archical clustering image based on target co-occupancy. We

then calculated the degree of target co-occupancy of each

pair of factors. As shown in the target correlation map in

Figure 2D, we observed three major clusters. Factors in Poly-

comb complexes are associated with the H3K27me3 signature

to form a distinct cluster (PRC cluster, blue-colored box in

Figure 2D and blue letters in Figures 2B and 2D). Core pluripo-

tency factors, including Nanog, Sox2, and Oct4 and others,

form an independent cluster (Core cluster, red-colored box

and red letters). Myc forms a cluster with other factors and

AcH3, AcH4, and H3K4me3 signatures (Myc cluster, green-

colored box and green letters).

We calculated the median distances between binding peaks

of each pair of factors using the same cluster information shown

in Figure 2D (except for the PRC cluster because of availability of

the processed data). The target distance map demonstrates

that the factors within the Core or Myc clusters regulate their

common targets in close proximity, whereas the factors

belonging to a different cluster regulate their common targets

in a relatively remote manner (Figure 2E).

Previously, we observed that Myc occupiesmore target genes

than the ES cell core factors (Kim et al., 2008). Similarly, we

observed that the factors in the Myc cluster, such as Max,

nMyc, E2F4, and Dmap1, tend to occupy more targets than

factors in the Core or PRC clusters (Figure 2B), suggesting

more global roles in their target gene regulation. The majority

of binding peaks generated by the factors in the Myc cluster

are more centered at the transcription start site (TSS) compared

to the target binding peaks of the factors in the Core cluster

(Figure 2C). The factors in the Myc cluster may interact with

basal transcription machinery, whereas core factors have both

promoter and upstream enhancer targets, as described

34,740,000 34,750,000 33,290,000 33,300,000 120,290,000 120,300,000 1 00,000 126,710,000 126,720, 0 51,390,000 51,400,000 126,460,000 126,470,000

Mars

Hgs Mrpl12 Slc25a1

Brd8 Cdc23

Cbx3

Hnrnpa2bl

Zfp148 Cdk4

March9 Tspan3l

Myc

Max

Dmap1

Tip60

E2F4

Gcn5

A

D

B

Suz12

Eed

Phc1

Rnf2

E

zh1

H3K

27m

e3

CT

CF

S

mad1

Sta

t3

Klf4

Oct4

N

anog

Sox2

Nac1

Zfp

281

Dax1

Esrr

b

Tcfc

p2l

Ctr

9

Gcn5

Dm

ap1

AcH

4

AcH

3

H3K

4m

e3

E2F

1

E2F

4

cM

yc

Max

nM

yc

Zfx

R

ex1

Tip

60

Cnot3

T

rim

28

12000

10000

8000

6000

4000

2000

0

Nu

mb

er

of

pro

mo

ter

targ

ets

E

0 100 300 600 1000 2000

H3K27me3 CTCF

Smad1 Stat3

Klf4 Oct4 Nanog

Sox2 Nac1

Zfp281 Dax1 Esrrb

Tcfcp2l1 Ctr9

Gcn5 Dmap1

AcH4 AcH3

H3K4me3 E2F1 E2F4 cMyc Max

nMyc Zfx

Rex1 Tip60 Cnot3

Trim28

H3K

27m

e3

CT

CF

S

mad1

Sta

t3

Klf4

Oct4

N

anog

Sox2

Nac1

Zfp

281

Dax1

Esrr

b

Tcfc

p2l1

C

tr9

Gcn5

Dm

ap1

AcH

4

AcH

3

H3K

4m

e3

E2F

1

E2F

4

cM

yc

Ma x

nM

yc

Zfx

R

ex1

Tip

60

Cnot3

T

rim

28

Median

distance

(bp)

C 40

35

30

25

20

15

10

5

0

E2F1

E2F4

cMyc

Max

nMyc

Zfx

Rex1

Tip60

Cnot3

Trim28

-2000

-1600

-1200

-800

-400

TS

S

400

800

1200

1600

2000

Fre

qu

en

cy (

%)

40

35

30

25

20

15

10

5

0

Fre

quency (

%)

H3K27me3

CTCF

Smad1

STAT3

Klf4

Oct4

Nanog

Sox2

Nac1

Zfp281

Dax1

-2000

-1600

-1200

-800

-400

TS

S

400

800

1200

1600

2000

Peak position from TSS (bp)

1.0 0.8 0.6 0.4 0.2 0 -0.2

1

Suz12 Eed

Phc1 Rnf2 Ezh1

Suz12

Eed

Phc1

Rnf2

E

zh1

H3K

27m

e3

CT

CF

S

mad1

Sta

t3

Klf4

Oct4

N

anog

Sox2

Nac1

Zfp

281

Dax1

Esrr

b

Tcfc

p2l

Ctr

9

Gcn5

Dm

ap1

AcH

4

AcH

3

H3K

4m

e3

E2F

1

E2F

4

cM

yc

Ma x

nM

yc

Zfx

R

ex1

Tip

60

Cnot3

T

rim

2 8

H3K27me3 CTCF

Smad1 Stat3

Klf4 Oct4 Nanog

Sox2 Nac1

Zfp281 Dax1 Esrrb

Tcfcp2l1 Ctr9

Gcn5 Dmap1

AcH4 AcH3

H3K4me3 E2F1 E2F4 cMyc Max

nMyc Zfx

Rex1 Tip60 Cnot3

Trim28

Correlation

score

Figure 2. Myc-Centered Protein-DNA Interaction Network in ES Cells

(A) Representative view of Myc, Max, Dmap1, Tip60, E2F4, and Gcn5 occupancy at the target loci.

(B) Number of target promoters bound by each factor or associated with each histone modification signature. Blue represents factors or histone signatures

involved in PRC complexes. Red represents factors involved in ES cell core factors, and green represents Myc and Myc- related factors or histone signatures

(D and E). See also Figure S2 and Table S2.

(C) Relative position of chromosomal target loci of each factor in the Myc cluster (upper panel) and Core cluster (bottom panel) shown in (B) and (C) to the TSS.

(D) Target correlation map: The degree of target co-occupancy of each pair of factors (either transcription factor or histone modification signature) is shown.

Yellow indicates more frequent colocalization of each pair of factors.

(E) Median distance map: Median distances between the loci co-occupied by two tested factors (except PRC complex proteins) shown in (D). Yellow indicates

closer colocalization.


elsewhere (Chen et al., 2008; Loh et al., 2006). In summary, our

data suggest that the factors belonging to each of the distinct

clusters (Core, PRC, and Myc) regulate their own rather similar

downstream targets in close proximity and may be functionally

separated in regulating aspects of ES cell identity.

Target Co-occupancy of Factors within the Myc Cluster

Has a Positive Correlation with Histone H3 and H4

Acetylation Signatures

Our prior work revealed that Myc target promoters correlate

positively with an active H3K4me3 signature and negatively

with a repressive H3K27me3 signature (Kim et al., 2008).

Because Myc is associated with histone acetylation (Frank

et al., 2001), we tested the correlation between target occupancy

of each factor in the Myc cluster and the histone modification

status of their target promoters. As shown in Figure 3A, the

majority of the factors in the Myc cluster harbor significantly

higher levels of H3K4me3, AcH3, and AcH4 signatures on their

target promoters over background (at least >150%). On the

contrary, the H3K27me3 signature is significantly underrepre-

sented on the target promoters of approximately half of the

factors in the Myc cluster. Interestingly, Cnot3 and Trim28 target

promoters show bivalent modifications (both H3K4me3 and

H3K27me3 positive), suggesting that, although these factors

share many common targets with Myc, they may have different

functions compared to the other factors in the cluster.

Additionally, we tested the relationship between the factor co-

occupancy (seven factors in theMyc cluster shown in Figures 2D

and 2E, includingMyc,Max, nMyc, Dmap1, E2F1, E2F4, and Zfx)

and histone modification signatures. As shown in Figure 3B,

target promoters co-occupied by multiple factors in the Myc

cluster show a higher level of histone acetylation than the

common targets of fewer factors. Targets occupied by seven

factors show approximately 400% and 220% of AcH4 and

AcH3 signatures, respectively, over the background level.

Upon the decrease of co-occupancy, the level of these signa-

tures decreased on their common targets. We failed to observe

correlation between co-occupancy and the H3K4me3 signature,

presumably as a result of the abundance of H3K4me3 marks

across many promoters (>60% of all promoters) (Kim et al.,

2008). The repressive signature H3K27me3 displays a reverse

correlation with theMyc cluster factor co-occupancy (Figure 3B).

Modules Defined by Transcriptional Regulatory

Subnetworks in ES Cells

Because we observed a strong positive correlation between

target co-occupancy of the factors in the Myc cluster and

histone acetylation signatures, we examined the correlation

between target co-occupancy and gene expression. As shown

in Figure 4A, targets co-occupied by seven or six factors in the

Myc cluster are more active than the common targets of five or

fewer factors in ES cells (red line) and are repressed upon differ-

entiation (blue line). To test whether the information generated

from the factor co-occupancy in the Myc cluster is functionally

relevant, we compared KEGG pathways (Dennis et al., 2003;

Ogata et al., 1999) enriched in the genes that are common

targets of at least six factors among seven factors in the Myc

cluster (Myc, Max, nMyc, Dmap1, E2F1, E2F4, and Zfx; black

bar in Figure 4A) and the global target genes of Myc. Many

cancer-related pathways (red letters in Figure 4B and Table S3)

are enriched in the genes co-occupied by the factors in the Myc

cluster. In contrast, these cancer-related pathways are not

enriched within the global set of genes occupied by Myc. This

observation strongly suggests that factor co-occupancy in the

Myc cluster does not represent a random subset of Myc targets

and may provide additional information in understanding the

combinatorial function of factors in the Myc cluster in ES cells

and in cancer cells (Figure 4B).

We previously demonstrated that common targets of multiple

factors in the core pluripotency network are significantly active

in ES cells. However, when these factors occupy targets alone

or with few factors, they are not associated with activation of

target genes (Kim et al., 2008). Because the targets co-occupied

by seven factors in the Myc cluster show the strongest gene

activity (Figure 3B and Figure 4A), we classified common target

gene modules in ES cells according to the target co-occupancy

within the clusters shown in Figure 2; the PRC module, the

Core module, and the Myc module (Figure 4C and listed in

Table S3). The definition of each module is as follows; the Core

module is composed of genes co-occupied by at least seven

factors among nine factors shown in the Core cluster (Smad1,

Stat3, Klf4, Oct4, Nanog, Sox2, Nac1, Zfp281, and Dax1), de-

picted in the red box in Figure 2D. The PRC module genes are

the common targets of PRC cluster proteins, Suz12, Eed, Phc1,

and Rnf2 (blue box in Figure 2D). The Myc module is composed

of genes that are common targets of seven factors (Myc, Max,

A

B

350

300

250

200

150

100

50

0

H3K4me3 H3K27me3 AcH4 AcH3

E2F1

Dmap1

Tip60

cMyc

Zfx

nMyc

Gcn5

Rex1

E2F4

Max

Cnot3

Trim28

All

Fre

qu

en

cy (

%)

450

400

350

300

250

200

150

100

50

0

H3K4me3 H3K27me3 AcH4 AcH3

7TFs

6TFs

5TFs

4TFs

3TFs

2TFs

1TF

0TF

All

Fre

qu

en

cy (

%)

Figure 3. Histone Modification Signatures on the Target Promoters

of Myc Cluster Proteins

(A) Histone marks on the target promoters of each factor in the Myc cluster.

‘‘All’’ represents all promoters.

(B) Histone marks and target co-occupancy of seven factors in Myc cluster

(Myc, Max, nMyc, Dmap1, E2F1, E2F4, and Zfx).


nMyc, Dmap1, E2F1, E2F4, and Zfx) in the Myc cluster (green

box in Figure 2D). For construction of the Myc module, we

excluded Tip60, Gcn5, and Rex1, because of their relatively

small number of targets (Figure 2B), and Trim28 and Cnot3,

because of the bivalent signature on their target promoters

(Figure 3A) and the discrepancy of their target similarity within

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

Core module

111

Myc module

503

103

5

PRC module

560

3

1

0

556 497

0 2 4 6 8 10

Ribosome

Cell cycle

Endometrial cancer

DNA replication

Aminoacyl-tRNA biosynthesis

Huntington's disease

Thyroid cancer

Oxidative phosphorylation

Alanine and aspartate metabolism

Parkinson's disease

Homologous recombination

Non-small cell lung cancer

Mismatch repair

Purine metabolism

Glutamate metabolism

Prostate cancer

Non-homologous end-joining

Ubiquitin mediated proteolysis

p53 signaling pathway

Acute myeloid leukemia

Bladder cancer

One carbon pool by folate

Pyrimidine metabolism

Chronic myeloid leukemia

Colorectal cancer

Base excision repair

Folate biosynthesis

Nucleotide excision repair

Pentose phosphate pathway

Proteasome

RNA polymerase

Enrichment Score (-log(p-value))

Myc cluster common

All Myc

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0

1

2

3

4

5

6

7

Ge

ne

exp

ressio

n

Co

-occu

pa

ncy

ES

dES day14

Genes

Core module 0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

ESC-like module_Wong 0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Myc module 0.6

0.5

0.4

0.3

0.2

0.1

0.0

PRC module 0.0

-0.1

-0.2

-0.3

-0.4

-0.5

-0.6

En

rich

me

nt

Sco

re

En

rich

me

nt

Sco

re

WT ES cells vs. day 14 dES cells

ES dES day2 dES day7 dES day14

Core PRC

Myc ESC-like

Avera

ge m

odule

expre

ssio

n

A

B

C

D

E

Figure 4. ES Cell Modules

(A) Gene expression profiles (log2, left y axis) upon J1 ES cell differentiation (wild-type ES cells: ES, differentiated ES cells for 14 days: dES day 14) are shown as

moving window averaged lines (ES; red line, dES day 14; blue line, bin size 100 and step size 1). Randomized genes are sorted (x axis) by the target co-occupancy

of seven factors in theMyc cluster (right y axis). Black bar represents target genes co-occupied by at least six factors among the seven factors in the Myc cluster.

(B) Enrichment of KEGG pathways. All Myc target genes (gray bars, total 3733 genes) and genes co-occupied by at least six factors among the seven factors

tested marked by black bar in (A) (black bars, total 1756 genes). See also Figures S3A and S3B and Table S3.

(C) ES cell modules: Three ES cell modules are defined based on the target co-occupancy within each cluster shown in Figure 2D. See also Table S3.

(D) GSEA analyses show the gene activity of the three ES cell modules (Core, PRC, and Myc modules) as well as the previously defined ESC-like module

(Wong et al., 2008a) upon ES cell differentiation (wild-type ES cells; ES versus 14 days differentiated ES cells; dES).

(E) Average gene expression values (log2) of eachmodule (C) are tested upon ES cell differentiation (ES day0, dES day2, dES day7, and dES day14, respectively).

Data are represented as mean ± SEM. See also Figure S2B and Figure S3C.


the Myc cluster (Figure S2A). Additional gene sets co-occupied

by different combinations of factors in Myc cluster were also

tested but showed no significant difference, because the

majority of target genes among the tested sets are shared (see

below and Figure S2B). Lists of gene sets tested are summarized

in Table S3. Indeed, the Core module includes previously known

factors in core regulatory circuitry, such as Nanog, Oct4, Rest,

Sox2, Tcf3, and Rex1. The PRCmodule includes genes generally

repressed in ES cells, including Hox cluster genes. As shown in

Figure 4C, the overlap between eachmodule is minimal and they

are involved in different pathways (Figures S3A and S3B).

We then tested activity of each module (hereafter referred to

as ‘‘module activity,’’ the averaged expression of all genes in

each module in a given expression data set) in ES cells

compared with the module activity in differentiated cells. Gene

activity of a previously identified Core ESC-like gene module

(hereafter referred to as ‘‘ESC-like module’’) (Wong et al.,

2008a) was also tested. Gene set enrichment analysis (GSEA)

(Subramanian et al., 2005) revealed that the Core, Myc, and

previously identified ESC-like modules are highly active in ES

cells. As anticipated, the PRC module is repressed in ES cells

(Figure 4D). We additionally tested the activity of each module

during a time-course of ES cell differentiation. As shown in Fig-

ure 4E and Figure S3C, in ES cells the Core module is most

active, yet the Myc and ESC-like modules show some activity;

these modules become repressed with time during differentia-

tion, whereas the PRC module shows an opposite pattern.

Functional Separation of Core and Myc Modules

in Partial iPS Cells

Although we observed that both the Core and Myc modules are

active in ES cells, the genes that comprise the Core module are

distinct from those of the Myc module (Figure 4C). To test

whether the modules can be functionally separable, we tested

the module activity of our three ES cell modules, along with the

ESC-like module (Wong et al., 2008a) in other cell types,

including iPS cells, partial iPS (piPS) cells, andmouse embryonic

fibroblasts (MEFs). Global gene expression profiles of ES and iPS

cells are highly similar (Takahashi and Yamanaka, 2006). Relying

on a publicly available data set (Sridharan et al., 2009), we tested

whether the module activity is similar between ES and iPS cells.

Similar to the data shown in Figure 4E, the Core and Myc

modules are highly active in both ES and iPS cells (Figure 5A

and Figure S3D). The PRC module is inactive in both cell types,

as expected. In MEFs, the module activity pattern is similar to

the module activity shown in differentiated ES cells shown in

Figure 4E, suggesting that strongly active Core and Myc

modules, as well as an inactive PRC module, may characterize

the pluripotent state of cells, such as ES and iPS cells.

Previous work has shown that piPS cells exist at an interme-

diate stage in the reprogramming process (Maherali et al.,

2007). The endogenous ES cell core regulators Oct4 and Nanog

are not reactivated in piPS cells, whereas they are reactivated in

fully reprogrammed iPS cells. To test whether the ES cell

modules we have defined are functionally separable in piPS

cells, we analyzed ES module activity using gene expression

data from piPS cells (Figure 5A and Figure S3D) (Sridharan

et al., 2009). We found that the activity of the Myc module in

piPS cells is comparable to that in ES cells and iPS cells, but

the Core module is not reactivated in piPS cells. These data

demonstrate that the regulatory modules defined in ES cells

may be considered functionally separable units, not arbitrary

subdivisions of the overall ES cell signature. Of particular note,

the ESC-like module (Wong et al., 2008a) shows similar module

activity to our Myc module, but not to the Core module in piPS

cells.

ES Cell Module Activity in Cancer

Others have described ESC-like gene modules (Wong et al.,

2008a) or ES-cell like gene expression signatures (Ben-Porath

et al., 2008) that have been widely used in assessment of cancer

gene signatures. With the three ES cell modules we defined as

new analytical tools, we readdress the relatedness of ES cell

and cancer gene signatures as a series of case studies. For

analyses of human data, human orthologs of mouse genes

were used (Table S3).

Myc Induction Does Not Activate the Core Module

in Human Epithelial Cells

We tested ESC-like modules (both Core ESC-like gene and

mouse ESC-like gene modules) (Wong et al., 2008a) and found

that they behave similarly to our Myc module in various settings

(Figure 4E, Figure 5A, and data not shown). Because we

observed that our defined Core and Myc modules can be func-

tionally separated in piPS cells (Figure 5A), we examinedwhether

the induction of Myc may activate the Core module in a different

cellular context. It has been reported previously that the induc-

tion of Myc activates the ESC-like module in adult human

A B

Avera

ge m

odule

expre

ssio

n

ES iPS MEF piPS WT Myc induction

Avera

ge m

odule

expre

ssio

n

Core

PRC

Myc

ESC-like

Core

PRC

Myc

ESC-like

Figure 5. Module Activity in Various Cells

(A) Average gene expression values (log2) (Srid-

haran et al., 2009) of ES cell modules (Core, PRc,

and Myc) and previously defined ESC-like module

are tested in ES cells (ES), iPS cells (iPS), MEFs

(MEF), and partial iPS cells (piPS). See also

Figure S3D.

(B) Average gene expression values (log2) (Bild

et al., 2006) of each module tested in (A) upon

induction of Myc in human epithelial cells (Myc

induction) and in control cells (WT). Human ortho-

logs of genes in three ES cell modules are tested

(listed in Table S3) and data are represented as

mean ± SEM (A and B).


epithelial cells (Wong et al., 2008a). As shown in Figure 5B, upon

reanalysis of this data set (Bild et al., 2006), we find that the Core

module is not activated following Myc induction, whereas the

Myc module is strongly represented. In addition, core factors

in ES cells, such as Nanog and Oct4, are also not activated by

Myc induction (Figure S4A). These observations confirm that

the Myc and Core modules are functionally separable and also

support the view that the overall ES cell expression signature

can be subdivided into functionally distinct units. Our refined

analysis argues against the prior conclusion that Myc induction

leads to activation of an ESC-like gene module in human epithe-

lial cells (Wong et al., 2008a).

ES Cell Modules in Mouse MLL Myeloid Leukemia

Models

We have assessed the relevance of our ES cell modules within

a mouse model of acute myeloid leukemia (AML). Expression

of MLL alleles leading to expression of fusion products, such

as MLL-AF9, MLL-ENL, MLL-AF10, MLL-AF1p, and MLL-

GAS7, initiates leukemia. MLL-associated leukemia models in

mice have served as platforms for purifying and examining the

gene expression profiles of leukemia stem cells (LSCs, also

called leukemia-initiating cells) (Krivtsov et al., 2006). It has

also been suggested that LSCs are present at a higher frequency

in leukemic mice in which AML was initiated by MLL-ENL or

MLL-AF9 as compared with MLL-AF10, MLL-AF1p, and

MLL-GAS7 (Somervaille et al., 2009). We tested the activity of

our defined modules in these leukemias. We first observed that

the Core module is not active in any of the AMLs as compared

to the Core module activity of a control group (Figure 6A). More-

over, we failed to detect an active Core module in AMLs demon-

strated to have high LSC frequency (MLL-ENL and MLL-AF9)

(Figure 6A). In contrast, we observed active Mycmodule expres-

sion in high-frequency LSC AMLs (MLL-ENL and MLL-AF9), but

not in low-frequency LSC AMLs (MLL-AF10, MLL-AF1p, and

MLL-GAS7) or control.

It has been reported that the previously defined ESC-like gene

module (Wong et al., 2008a) is prominent in a MLL-AF10

leukemia cell population enriched for LSCs (c-kit high) as

compared to c-kit low cells (Somervaille et al., 2009). As shown

in Figure 6B, we observed a stronger Myc module activity in the

LSC-enriched population. However, this cell population shows

relatively inactive Core module activity. In both of the tests

shown in Figure 6A and Figure 6B, we observed that the activity

of the previously defined ESC-like gene module (Wong et al.,

2008a) is similar to the activity of the Myc module rather than

the Core module. If the gene expression findings are functionally

relevant to self-renewal of LSCs, our findings undermine the

notion that reactivation of an ESC-like pattern is critical for

LSCs in this setting. In contrast, Myc module activity alone

appears to correlate with LSC frequency in mouse AML models.

Core module activity does not appear to be a major determinant

of LSC frequency in AML.

ES Cell Modules in Human Cancers

To test the activity of ES cell modules more generally, we tested

module activity in gene expression profiles acquired from human

bladder carcinoma samples, including superficial and invasive

carcinomas, as well as a control group of normal urinary tract

cells (Sanchez-Carbayo et al., 2006). Figure 7A represents

each module activity from total 157 patient samples (each

column). Figure 7B represents combined module activity from

different groups of patient samples. In both superficial and inva-

sive carcinomas, the Mycmodule is more active compared to its

level of activity in control samples. However, the Core module

activity is repressed in both grades of cancers. Of note, we

observed a more active Myc module in superficial carcinoma

samples compared to more advanced stage of invasive carci-

noma samples. Heterogeneity of invasive carcinoma samples

may underlie this observation, or the active Myc module may

be critical in initiating invasive behavior, not necessarily active

afterward. Importantly, the previously defined ESC-like gene

module activity is again similar to the activity of the Myc module.

However, the Core module seems to be even more repressed in

carcinoma samples compared to control group (Figure 7A and

Figure 7B).

We next tested module activity within a human primary breast

cancer expression data set (van’t Veer et al., 2002) containing

fifty eight samples from patients who developed distant metas-

tases within 5 years (poor prognosis group), and 39 samples

from patients who continued to be disease free for at least

5 years (good prognosis group). First, we calculated Core

module activity of all samples and further analyzed samples

showing the strongest Core module activity (top 20% of

samples; n = 19), and the weakest Core module activity (bottom

20%; n = 19). As shown in Figure 7C and Figure 7E, no correla-

tion was observed between Core module activity and patient

outcome (interval to metastasis). On the other hand, Mycmodule

A B

-0.08

-0.04

0

0.04

0.08

0.12

c-Kit low c-Kit high

Ave

rag

e m

od

ule

exp

ressio

n

Ave

rag

e m

od

ule

exp

ressio

n

-0.18

0

0.22

MLL-AF1p MLL-AF10 MLL-GAS7 MLL-AF9 MLL-ENL Normal

Core

PRC

Myc

ESC-like

Core

PRC

Myc

ESC-like

Figure 6. ES Cell Modules in Mouse MLL

Myeloid Leukemia Models

(A) Average gene expression values (log2) of ES

cell modules and the previously defined ESC-like

module are tested in various mouse models of

acute myeloid leukemia (AML) initiated by

MLL-AF9, MLL-ENL, MLL-AF10, MLL-AF1p, and

MLL-GAS7 (Somervaille et al., 2009).

(B) Average gene expression values (log2) of each

module are tested in a c-kit high MLL-AF10

leukemia cell population (MLL-AF10 c-kit high)

and a c-kit lowMLL-AF10 leukemia cell population

(MLL-AF10 c-kit low) (Somervaille et al., 2009).

Data are represented as mean ± SEM (A and B).

See also Figure S4.


activity correlates positively with a poor prognosis (Figure 7D).

On average, metastasis occurs within 47 months in breast

cancer patients with strong Myc module activity (top 20%;

n = 19). In contrast, it took on average 89 months for the patients

with weak Myc module activity (bottom 20%) to progress to

metastasis (Figure 7E), suggesting that Myc module activity

predicts patient outcome.We observed that Mycmodule activity

in human breast cancer patient samples is very similar to the

previously defined ESC-like modules (Wong et al., 2008a)

(Figure S5A). Additional analyses using independent breast

cancer data sets also revealed that tumor samples with a more

active Myc module tend to be highly proliferative basal-like

tumors (Figures S5B, S5C, and S5E, middle panel) or ER nega-

tive tumors (Figure S5D and S5E, left panel). These results are

consistent with findings of others demonstrating a correlation

of Myc activity with poor outcome in breast cancer (Wolfer

et al., 2010). Interestingly, we observed that highly proliferative

cells show stronger Myc module activity (Figure 6, Figure 7,

Figure S4, and Figure S5), suggesting a link between the Myc

module activity and cell proliferation.

DISCUSSION

By integrating protein-protein interaction and protein-DNA inter-

action studies, we constructed a Myc-centered transcriptional

regulatory network in an effort to complement the previously

A B

C

Core

PRC

Myc

ESC-like

NU

INV

SUP

Ave

rag

e m

od

ule

exp

ressio

n

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

NU INV SUP

Ave

rag

e m

od

ule

exp

ressio

n

-0.12

-0.08

-0.04

0

0.04

0.08

0.12

Ave

rag

e m

od

ule

exp

ressio

n

-0.12

-0.08

-0.04

0

0.04

0.08

0.12

Inte

rva

l to

me

tasta

sis

(m

on

ths)

0

20

40

60

80

100

120

140

160

Top 20% Bottom 20%

Inte

rva

l to

me

tasta

sis

(m

on

ths)

0

20

40

60

80

100

120

140

160

180

Top 20% Bottom 20%

0

20

40

60

80

100

120

Inte

rva

l to

me

tasta

sis

(m

onth

s)

Core Myc

P = 0.01

P = 0.99

Core Myc

D

E

Top 20%

Bottom 20%

Core

PRC

Myc

ESC-like 0.6 0.4 0.2 0.0-0.2-0.4-0.6

Foldchange

Figure 7. ES Cell Modules in Human Cancers

(A and B) Average gene expression values (log2) of ES cell modules and previously defined ESC-like module are tested in human bladder carcinoma samples

including superficial (SUP), and invasive carcinomas (INV), as well as normal urothelium (NU) as a control group (marked by black bars) (Sanchez-Carbayo

et al., 2006). Each column represents one patient sample (total 157 samples) (A). Averaged module activities within the sample group (NU, INV, and SUP) (B).

Data are represented as mean ± SEM.

(C–E) Average gene expression values (log2) of ES cell Core (C) and Myc (D) module are tested from 97 human breast cancer patient samples (van’t Veer et al.,

2002). (C) Core module activities were calculated, and top and bottom 20% of samples (19 samples each) were further analyzed. Bar graph represents the

corresponding interval to metastases (months, bottom panel). (D) Samples showing top and bottom 20%ofMycmodule activity were further analyzed. Bar graph

represents the corresponding interval tometastases (months) for each patient (bottom panel). (E) For each tested group (C andD), interval to distant metastases is

calculated as mean ± SEM, and p values are from Student’s t tests. See also Figure S5.


identified core regulatory and Polycomb networks in ES cells.

Our approach, analyzed together with data of others, delineates

three major transcriptional regulatory subnetworks in ES cells.

On the basis of the target co-occupancy of factors in each

network, we defined three functionally separable regulatory

modules (Figure 4 and Figure 5) and showed that the overall

ES cell gene transcription program can be subdivided largely

into functionally independent regulatory units.

It is interesting to note that a previous RNAi-based screen

revealed that members of the NuA4 HAT complex (or Tip40-

Ep400 complex) are critical in ES cell identity (Fazzio et al.,

2008). Upon knockdown of some of NuA4 HAT complex pro-

teins, as well as Myc, we also observed that ES cells display

flattened morphology (Figure S1E). Of note, knockdown of

Ep400 or Dmap1 did not change the expression level of Oct4

and Nanog proteins, nor did knockdown of Nanog change the

protein level of Ep400 and Dmap1 (Fazzio et al., 2008; see also

Figure S1D). These data support the conclusion that the Core

and Myc-centered subnetworks in ES cells are separable units

with unique roles in maintaining ES cell self-renewal.

Previous studies have suggested that Myc is critical at an

early stage in somatic cell reprogramming (Sridharan et al.,

2009). Our work suggests that, beyond Myc itself, reactivation

of a larger module composed of more than 500 genes is critical

to achieve partially or fully reprogrammed stem cell–like cells. It

is particularly interesting that the Core module, which is

composed of more than 100 genes, remains inactive in piPS

cells, again implying that the reactivation of an entire functional

module by a limited set of factors is critical to achieving

induced pluripotency. It will be of interest to determine whether

specific small molecules or genes selectively modulate the

activity of the ES cell modules in efforts to identify new chem-

icals or factors not only for replacing Myc or other factors in

somatic cell reprogramming, but also for selection of putative

therapeutic targets in cancer. Because Myc interacts with

NuA4 complex proteins in ES cells, recruitment of the NuA4

HAT complex by Myc may be a critical step in somatic cell

reprogramming.

The relationship between ES cell and cancer signatures has

been a focus of attention given that self-renewal is a hallmark

of both cell types. It has been proposed that the activation of

an ESC-like gene expression program in adult cells may confer

self-renewal to cancer cells or cancer stem cells (Ben-Porath

et al., 2008; Wong et al., 2008a). It is noteworthy that we

observed very similar patterns of module activity between our

Myc module and the previously defined ESC-likes (Core ESC-

like gene module and mouse ESC-like gene module) (Wong

et al., 2008a), but not with our Core module, in situations we

tested. In accordance with this observation, approximately

60% of genes in the previously defined Core ESC-like module

(Wong et al., 2008a) are Myc targets that we identified (Kim

et al., 2008). Notably, 57% of genes in the Core ESC-like module

(Wong et al., 2008a) are common targets of at least five factors

among seven factors in the Myc cluster (Figure 4). In contrast,

less than 2% of genes in the previously defined ESC-like module

are shared with the Core module. These findings argue that the

previously described ESC-like module (Wong et al., 2008a)

conveys information largely contributed by the Myc module,

and, conversely, that the ESC-like module is quite distinct from

the Core module. The simple interpretation that the presence

of ESC-like module activity in cancer reflects dedifferentiation

or regression to an embryonic or ES-like state (Wong et al.,

2008a) is inconsistent with our analysis.

In their recent work, Ben-Porath et al. (2008) compiled 13

partially overlapping gene sets belonging to four groups (ES-ex-

pressed, active NOS [Nanog, Oct4, and Sox2] targets, Polycomb

targets, and Myc targets) that are similar to the modules utilized

in our analysis. They showed that poorly differentiated tumors

show preferential expression of ES cell–specific genes, in

addition to preferential repression of Polycomb target genes.

Interestingly, their analysis revealed that ES-expressed and

Polycomb-target sets show the most significant degree of

enrichment in most tumors, whereas the other gene sets are

not a major determinant of their ES cell-like gene expression

signature. Of special note, we find that 38% and 52% genes in

their ES-expressed gene sets (ES exp1 and ES exp2, respec-

tively) contain the common targets of at least five factors among

seven factors in the Myc cluster, suggesting that a large portion

of genes in their ES-expressed gene sets are, in turn, Myc-

related genes. It is noteworthy that the PRC module defined in

ES cells is also largely repressed in most cancers we tested,

suggesting a role of Polycomb complex proteins and their

targets in cancer initiation and/or progression.

Our analysis is conceptually different from prior approaches in

that we have stringently defined regulatory modules based on

common gene targets of multiple factors. By use of this strategy,

we have defined modules that serve as powerful analytical tools

to interrogate different cellular states and the relatedness of

gene expression signatures of ES cells and cancers. Reanalysis

of prior data sets in this manner raises concern regarding the

hypothesis that cancer cells, or cancer stem cells, recapitulate

regulatory programs characteristic of embryonic stem cells. As

a unifying view, the hypothesis is attractive and has gained

considerable attention in recent literature. Nonetheless, our

findings should temper enthusiasm and stimulate further reas-

sessment of these issues. Moreover, our findings reemphasize

the critical nature of regulatory pathways controlled by Myc in

cancer.


ES Cell Lines and Culture

Mouse J1 ES cell lines were maintained in ES medium as documented in

Supplemental Information.

Protein Complex Pull-Down and Mass Spectrometry

One-step affinity purification and protein complex identification using nuclear

extracts from ES cell lines expressing BirA only (reference) or both BirA- and

biotin-tagged proteins (sample) with streptavidin-agarose were performed

as described elsewhere (Kim et al., 2009; Wang et al., 2006). Further details

are documented in Supplemental Information.

ChIP-chip

At least three biological replicates of ChIP and bioChIP reactions were per-

formed for each factor, as described elsewhere (Kim et al., 2009; Kim et al.,

2008). Detailed procedure and a list of antibodies used for native antibody

ChIP reactions are available in Supplemental Information.


Microarray and Data Processing

Amplification of ChIP samples and microarray hybridizations were performed

as described elsewhere (Kim et al., 2008).

ACCESSION NUMBERS

The raw and processed ChIP-chip data set can be found on the public server

GEOunder the accession number of GSE20551. Further details are available in

Supplemental Information.


Supplemental information includes Extended Experimental Procedures,

five figures, and three tables and can be found with this article online at

doi:10.1016/j.cell.2010.09.010.

ACKNOWLEDGMENTS

We thank Jennifer Trowbridge for critical reading of the manuscript, the Taplin

Biological Mass Spectrometry Facility at Harvard Medical School for mass-

spectrometry and peptide identification, and the Microarray Core Facility at

the Dana Farber Cancer Institute for ChIP sample processing. The project

described is partially supported by Award Number K99GM088384 to J.K.

from the NIH/NIGMS. S.H.O. is an investigator of the Howard Hughes Medical

Institute.

Received: April 22, 2010




REFERENCES

Ard, P.G., Chatterjee, C., Kunjibettu, S., Adside, L.R., Gralinski, L.E., and

McMahon, S.B. (2002). Transcriptional regulation of the mdm2 oncogene by

p53 requires TRRAP acetyltransferase complexes. Mol. Cell. Biol. 22,

5650–5661.

Ben-Porath, I., Thomson, M.W., Carey, V.J., Ge, R., Bell, G.W., Regev, A., and

Weinberg, R.A. (2008). An embryonic stem cell-like gene expression signature

in poorly differentiated aggressive human tumors. Nat. Genet. 40, 499–507.

Bild, A.H., Yao, G., Chang, J.T., Wang, Q., Potti, A., Chasse, D., Joshi, M.B.,

Harpole, D., Lancaster, J.M., Berchuck, A., et al. (2006). Oncogenic pathway

signatures in human cancers as a guide to targeted therapies. Nature 439,

353–357.

Boyer, L.A., Lee, T.I., Cole, M.F., Johnstone, S.E., Levine, S.S., Zucker, J.P.,

Guenther, M.G., Kumar, R.M., Murray, H.L., Jenner, R.G., et al. (2005). Core

transcriptional regulatory circuitry in human embryonic stem cells. Cell 122,

947–956.

Boyer, L.A., Plath, K., Zeitlinger, J., Brambrink, T., Medeiros, L.A., Lee, T.I.,

Levine, S.S., Wernig, M., Tajonar, A., Ray, M.K., et al. (2006). Polycomb

complexes repress developmental regulators in murine embryonic stem cells.

Nature 441, 349–353.

Cai, Y., Jin, J., Tomomori-Sato, C., Sato, S., Sorokina, I., Parmely, T.J.,

Conaway, R.C., and Conaway, J.W. (2003). Identification of new subunits of

the multiprotein mammalian TRRAP/TIP60-containing histone acetyltransfer-

ase complex. J. Biol. Chem. 278, 42733–42736.

Chen, X., Xu, H., Yuan, P., Fang, F., Huss,M., Vega, V.B.,Wong, E., Orlov, Y.L.,

Zhang, W., Jiang, J., et al. (2008). Integration of external signaling pathways

with the core transcriptional network in embryonic stem cells. Cell 133,

1106–1117.

Clarke, M.F., and Fuller, M. (2006). Stem cells and cancer: two faces of eve.

Cell 124, 1111–1115.

Cole, M.D., and Henriksson, M. (2006). 25 years of the c-Myc oncogene.

Semin. Cancer Biol. 16, 241.

de Boer, E., Rodriguez, P., Bonte, E., Krijgsveld, J., Katsantoni, E., Heck, A.,

Grosveld, F., and Strouboulis, J. (2003). Efficient biotinylation and single-

step purification of tagged transcription factors in mammalian cells and

transgenic mice. Proc. Natl. Acad. Sci. USA 100, 7480–7485.

Dennis, G., Jr., Sherman, B.T., Hosack, D.A., Yang, J., Gao, W., Lane, H.C.,

and Lempicki, R.A. (2003). DAVID: database for annotation, visualization,

and integrated discovery. Genome Biol. 4, 3.

Ding, L., Paszkowski-Rogacz, M., Nitzsche, A., Slabicki, M.M., Heninger, A.K.,

de Vries, I., Kittler, R., Junqueira, M., Shevchenko, A., Schulz, H., et al. (2009).

A genome-scale RNAi screen for Oct4 modulators defines a role of the Paf1

complex for embryonic stem cell identity. Cell Stem Cell 4, 403–415.

Doyon, Y., and Cote, J. (2004). The highly conserved and multifunctional NuA4

HAT complex. Curr. Opin. Genet. Dev. 14, 147–154.

Fazzio, T.G., Huff, J.T., and Panning, B. (2008). An RNAi screen of chromatin

proteins identifies Tip60-p400 as a regulator of embryonic stem cell identity.

Cell 134, 162–174.

Frank, S.R., Parisi, T., Taubert, S., Fernandez, P., Fuchs, M., Chan, H.M.,

Livingston, D.M., and Amati, B. (2003). MYC recruits the TIP60 histone

acetyltransferase complex to chromatin. EMBO Rep. 4, 575–580.

Frank, S.R., Schroeder, M., Fernandez, P., Taubert, S., and Amati, B. (2001).

Binding of c-Myc to chromatin mediates mitogen-induced acetylation of

histone H4 and gene activation. Genes Dev. 15, 2069–2082.

Fuchs, M., Gerber, J., Drapkin, R., Sif, S., Ikura, T., Ogryzko, V., Lane, W.S.,

Nakatani, Y., and Livingston, D.M. (2001). The p400 complex is an essential

E1A transformation target. Cell 106, 297–307.

Graf, T., and Enver, T. (2009). Forcing cells to change lineages. Nature 462,

587–594.

Hu, G., Kim, J., Xu, Q., Leng, Y., Orkin, S.H., and Elledge, S.J. (2009). A

genome-wide RNAi screen identifies a new transcriptional module required

for self-renewal. Genes Dev. 23, 837–848.

Huangfu, D., Maehr, R., Guo, W., Eijkelenboom, A., Snitow, M., Chen, A.E.,

and Melton, D.A. (2008). Induction of pluripotent stem cells by defined

factors is greatly improved by small-molecule compounds. Nat. Biotechnol.

26, 795–797.

Ikura, T., Ogryzko, V.V., Grigoriev, M., Groisman, R., Wang, J., Horikoshi, M.,

Scully, R., Qin, J., and Nakatani, Y. (2000). Involvement of the TIP60 histone

acetylase complex in DNA repair and apoptosis. Cell 102, 463–473.

Kang, J., Shakya, A., and Tantin, D. (2009). Stem cells, stress, metabolism and

cancer: a drama in two Octs. Trends Biochem. Sci. 34, 491–499.

Kim, J., Cantor, A.B., Orkin, S.H., and Wang, J. (2009). Use of in vivo

biotinylation to study protein-protein and protein-DNA interactions in mouse

embryonic stem cells. Nat. Protoc. 4, 506–517.

Kim, J., Chu, J., Shen, X., Wang, J., and Orkin, S.H. (2008). An extended

transcriptional network for pluripotency of embryonic stem cells. Cell 132,

1049–1061.

Krivtsov, A.V., Twomey, D., Feng, Z., Stubbs, M.C., Wang, Y., Faber, J.,

Levine, J.E., Wang, J., Hahn,W.C., Gilliland, D.G., et al. (2006). Transformation

from committed progenitor to leukaemia stem cell initiated by MLL-AF9.

Nature 442, 818–822.

Lee, T.I., Jenner, R.G., Boyer, L.A., Guenther, M.G., Levine, S.S., Kumar, R.M.,

Chevalier, B., Johnstone, S.E., Cole, M.F., Isono, K., et al. (2006). Control of

developmental regulators by Polycomb in human embryonic stem cells. Cell

125, 301–313.

Loh, Y.H., Wu, Q., Chew, J.L., Vega, V.B., Zhang, W., Chen, X., Bourque, G.,

George, J., Leong, B., Liu, J., et al. (2006). The Oct4 and Nanog transcription

network regulates pluripotency in mouse embryonic stem cells. Nat. Genet.

38, 431–440.

Maherali, N., Sridharan, R., Xie, W., Utikal, J., Eminli, S., Arnold, K., Stadtfeld,

M., Yachechko, R., Tchieu, J., Jaenisch, R., et al. (2007). Directly reprog-

rammed fibroblasts show global epigenetic remodeling and widespread tissue

contribution. Cell Stem Cell 1, 55–70.

Marson, A., Levine, S.S., Cole, M.F., Frampton, G.M., Brambrink, T.,

Johnstone, S., Guenther, M.G., Johnston, W.K., Wernig, M., Newman, J.,


et al. (2008). ConnectingmicroRNA genes to the core transcriptional regulatory

circuitry of embryonic stem cells. Cell 134, 521–533.

McMahon, S.B., Van Buskirk, H.A., Dugan, K.A., Copeland, T.D., and Cole,

M.D. (1998). The novel ATM-related protein TRRAP is an essential cofactor

for the c-Myc and E2F oncoproteins. Cell 94, 363–374.

McMahon, S.B., Wood, M.A., and Cole, M.D. (2000). The essential cofactor

TRRAP recruits the histone acetyltransferase hGCN5 to c-Myc. Mol. Cell.

Biol. 20, 556–562.

Melton, C., Judson, R.L., and Blelloch, R. (2010). Opposing microRNA families

regulate self-renewal in mouse embryonic stem cells. Nature 463, 621–626.

Meshorer, E., Yellajoshula, D., George, E., Scambler, P.J., Brown, D.T., and

Misteli, T. (2006). Hyperdynamic plasticity of chromatin proteins in pluripotent

embryonic stem cells. Dev. Cell 10, 105–116.

Niwa, H., Burdon, T., Chambers, I., and Smith, A. (1998). Self-renewal of

pluripotent embryonic stem cells is mediated via activation of STAT3. Genes

Dev. 12, 2048–2060.

Ogata, H., Goto, S., Sato, K., Fujibuchi, W., Bono, H., and Kanehisa, M. (1999).

KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 27,

29–34.

Park, I.H., Zhao, R., West, J.A., Yabuuchi, A., Huo, H., Ince, T.A., Lerou, P.H.,

Lensch, M.W., and Daley, G.Q. (2008). Reprogramming of human somatic

cells to pluripotency with defined factors. Nature 451, 141–146.

Reya, T., Morrison, S.J., Clarke, M.F., and Weissman, I.L. (2001). Stem cells,

cancer, and cancer stem cells. Nature 414, 105–111.

Sanchez-Carbayo, M., Socci, N.D., Lozano, J., Saint, F., and Cordon-Cardo,

C. (2006). Defining molecular profiles of poor outcome in patients with invasive

bladder cancer using oligonucleotide microarrays. J. Clin. Oncol. 24, 778–789.

Sato, N., Meijer, L., Skaltsounis, L., Greengard, P., and Brivanlou, A.H. (2004).

Maintenance of pluripotency in human and mouse embryonic stem cells

through activation of Wnt signaling by a pharmacological GSK-3-specific

inhibitor. Nat. Med. 10, 55–63.

Schoenhals, M., Kassambara, A., De Vos, J., Hose, D., Moreaux, J., and Klein,

B. (2009). Embryonic stem cell markers expression in cancers. Biochem.

Biophys. Res. Commun. 383, 157–162.

Shen, X., Liu, Y., Hsu, Y.J., Fujiwara, Y., Kim, J., Mao, X., Yuan, G.C., and

Orkin, S.H. (2008). EZH1 mediates methylation on histone H3 lysine 27 and

complements EZH2 in maintaining stem cell identity and executing pluripo-

tency. Mol. Cell 32, 491–502.

Somervaille, T.C., Matheny, C.J., Spencer, G.J., Iwasaki, M., Rinn, J.L.,Witten,

D.M., Chang, H.Y., Shurtleff, S.A., Downing, J.R., and Cleary, M.L. (2009).

Hierarchical maintenance of MLL myeloid leukemia stem cells employs a

transcriptional program shared with embryonic rather than adult stem cells.

Cell Stem Cell 4, 129–140.

Sridharan, R., Tchieu, J., Mason, M.J., Yachechko, R., Kuoy, E., Horvath, S.,

Zhou, Q., and Plath, K. (2009). Role of the murine reprogramming factors in

the induction of pluripotency. Cell 136, 364–377.

Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,

Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and

Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based

approach for interpreting genome-wide expression profiles. Proc. Natl.

Acad. Sci. USA 102, 15545–15550.

Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells

from mouse embryonic and adult fibroblast cultures by defined factors. Cell

126, 663–676.

van’t Veer, L.J., Dai, H., van de Vijver, M.J., He, Y.D., Hart, A.A., Mao, M.,

Peterse, H.L., van der Kooy, K., Marton, M.J., Witteveen, A.T., et al. (2002).

Gene expression profiling predicts clinical outcome of breast cancer. Nature

415, 530–536.

Wang, J., Rao, S., Chu, J., Shen, X., Levasseur, D.N., Theunissen, T.W., and

Orkin, S.H. (2006). A protein interaction network for pluripotency of embryonic

stem cells. Nature 444, 364–368.

Wolfer, A., Wittner, B.S., Irimia, D., Flavin, R.J., Lupien, M., Gunawardane,

R.N., Meyer, C.A., Lightcap, E.S., Tamayo, P., Mesirov, J.P., et al. (2010).

MYC regulation of a ‘‘poor-prognosis’’ metastatic cancer cell state. Proc.

Natl. Acad. Sci. USA 107, 3698–3703.

Wong, D.J., Liu, H., Ridky, T.W., Cassarino, D., Segal, E., and Chang, H.Y.

(2008a). Module map of stem cell genes guides creation of epithelial cancer

stem cells. Cell Stem Cell 2, 333–344.

Wong, D.J., Segal, E., and Chang, H.Y. (2008b). Stemness, cancer and cancer

stem cells. Cell Cycle 7, 3622–3624.

Xu, X., Bieda, M., Jin, V.X., Rabinovich, A., Oberley, M.J., Green, R., and

Farnham, P.J. (2007). A comprehensive ChIP-chip analysis of E2F1, E2F4,

and E2F6 in normal and tumor cells reveals interchangeable roles of E2F family

members. Genome Res. 17, 1550–1561.

Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J.L.,

Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart, R., et al. (2007). Induced

pluripotent stem cell lines derived from human somatic cells. Science 318,

1917–1920.


CELL_101015

Documents

Transcript of CELL_101015