The Poly(A)-Binding Protein Nuclear 1 Suppresses ... · The Poly(A)-Binding Protein Nuclear 1...

The Poly(A)-Binding Protein Nuclear 1Suppresses Alternative Cleavageand Polyadenylation SitesMathias Jenal,1,5 Ran Elkon,1,5 Fabricio Loayza-Puch,1,5 Gijs van Haaften,1,6 Uwe Kuhn,2 Fiona M. Menzies,3

Joachim A.F. Oude Vrielink,1 Arnold J. Bos,1 Jarno Drost,1 Koos Rooijers,1 David C. Rubinsztein,3 and Reuven Agami1,4,*1Division of Gene Regulation, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands2Institute of Biochemistry and Biotechnology, Martin-Luther-University Halle-Wittenberg, Kurt-Mothes-Strasse 3, 06120 Halle, Germany3Department of Medical Genetics, Cambridge Institute for Medical Research, Wellcome/MRC Building, Addenbrooke’s Hospital,Hills Road, Cambridge CB2 0XY, UK4The Centre for Biomedical Genetics, Utrecht 3584 CG, The Netherlands5These authors contributed equally to this work6Present address: The Department of Medical Genetics, University Medical Center Utrecht, Utrecht 3584 CG, The Netherlands

*Correspondence: [email protected]

DOI 10.1016/j.cell.2012.03.022

SUMMARY

Alternative cleavage and polyadenylation (APA) isemerging as an important layer of gene regulation.Factors controlling APA are largely unknown. Wedeveloped a reporter-based RNAi screen for APAand identified PABPN1 as a regulator of this process.Genome-wide analysis of APA in human cellsshowed that loss of PABPN1 resulted in extensive30 untranslated region shortening. Messenger RNAtranscription, stability analyses, and in vitro cleavageassays indicated enhanced usage of proximalcleavage sites (CSs) as the underlying mechanism.Using Cyclin D1 as a test case, we demonstratedthat enhanced usage of proximal CSs compromisesmicroRNA-mediated repression. Triplet-repeat ex-pansion in PABPN1 (trePABPN1) causes autosomal-dominant oculopharyngeal muscular dystrophy(OPMD). The expression of trePABPN1 in both amouse model of OPMD and human cells elicitedbroad induction of proximal CS usage, linked tobinding to endogenous PABPN1 and its sequestra-tion in nuclear aggregates. Our results elucidatea novel function for PABPN1 as a suppressor of APA.

INTRODUCTION

The addition of an adenine tail (poly(A)) to messenger RNA

(mRNA) 30 ends is a cotranscriptional nuclear process that is

important for nuclear export, translation, andmRNA stability (Le-

may et al., 2010). mRNA polyadenylation is a two-step process

consisting of an endonucleolytic cleavage and addition of an un-

templated poly(A) tail. Adjacent cis-acting motifs direct cleavage

position and efficiency. Mainly, an upstream hexamer polyade-

nylation signal (PAS) and a downstream UG-rich sequence

538 Cell 149, 538–553, April 27, 2012 ª2012 Elsevier Inc.

recruit the cleavage and polyadenylation specificity factor

(CPSF) and the cleavage-stimulating factor (CSTF), respectively,

both required for the polyadenylation process (Bentley, 2005;

Zhang et al., 2010). The canonical PAS is AAUAAA, which

appears in about half of the cleavage sites (CSs). More than

ten variants of the canonical PAS have been reported, and the

most prominent among them is AUUAAA (Beaudoing et al.,

2000; Tian et al., 2005).

mRNA poly(A) tails are added by a highly evolutionary con-

served poly(A) polymerase (PAP) and recruit poly(A)-binding

proteins (PABPs), which come in two flavors, the cytosolic

PABPC and the nuclear PABPN1. The nuclear PABPN1 is

structurally and functionally different from PABPC and is

thought to function during poly(A) tail addition to mRNAs.

In vitro, efficient polyadenylation requires CPSF, PAP, and

PABPN1 (Kuhn et al., 2009; Wahle, 1995). Whereas CPSF

binds to PAS to cleave mRNAs and modestly stimulates polya-

denylation, PABPN1 addition to CPSF/PAP strongly stimulates

polyadenylation.

Most mammalian mRNAs contain a 30 untranslated region

(30 UTR) that plays a key role in regulating mRNA stability, local-

ization, and translation efficiency (Andreassi and Riccio, 2009;

Fabian et al., 2010). These functions are mediated mostly

through binding of microRNAs (miRNAs) and RNA-binding

proteins (RBPs) to regulatory elements embedded in the target

30 UTRs. miRNAs are small, noncoding RNAs that guide miRNP

complexes to 30 UTRs by sequence-specific interaction with 30

UTRs based on a 6–8 seed sequence at their 50 ends (Fabian

et al., 2010). miRNAs are predicted to target the majority of

human protein-coding genes, enabling them to play numerous

regulatory roles in many physiological and developmental pro-

cesses (Kedde and Agami, 2008). RBPs, like miRNAs, are key

components in gene expression regulation. They can either

bind specific sequence elements on mRNAs (e.g., the human

Pumilio RBP family) or looser consensus sequences, such as

U- and A/U-rich motifs, to control expression in a negative or

positive manner (Filipowicz et al., 2008). Moreover, RBPs can

mailto:[email protected]

http://dx.doi.org/10.1016/j.cell.2012.03.022

bind 30 UTRs and interfere or stimulate miRNA function (Kedde

et al., 2007, 2010). Thus, it appears that 30 UTRs can be viewed

as regulatory platforms that determine mRNA stability, subcel-

lular localization, and rate of protein synthesis (van Kouwenhove

et al., 2011).

The number of human and mouse genes reported to use alter-

native cleavage and polyadenylation (APA) to generate multiple

mRNA isoforms with different 30 UTR lengths is continuously

increasing and already passes the 50% estimate (Fu et al.,

2011; Jan et al., 2011; Lutz, 2008; Ozsolak et al., 2010; Shepard

et al., 2011; Tian et al., 2005). Therefore, APA is emerging as an

important layer in gene regulation as it modulates the inclusion of

target sites for miRNAs and RBPs in 30 UTRs. Already more than

a decade ago, it was shown that changes in the concentration of

CstF could affect APA of IgM heavy-chain transcripts during B

cell development (Takagaki and Manley, 1998; Takagaki et al.,

1996). Recently, it was observed that whereas primary human

cells mostly use distal PASs, usage of proximal sites, which

results in 30 UTR shortening, is enhanced during cell activation

(Sandberg et al., 2008), early in development (Ji et al., 2009),

and in cancer (Mayr and Bartel, 2009; for a recent review, see

Di Giammartino et al., 2011). However, factors that regulate

APA during cancer progression and differentiation are still largely

unknown.

Oculopharyngeal muscular dystrophy (OPMD) is an auto-

somal-dominant form of late-onset (patients aged >40 years),

slowly progressive myopathy, characterized by bilateral eyelid

ptosis (drooping), dysphagia (swallowing difficulties), and prox-

imal limb weakness (Ruegg et al., 2005). Beyond the age of 70,

all patients are symptomatic. OPMD was reported in >30 coun-

tries worldwide, with a low estimated occurrence that can be as

high as 1:600 in specific communities (Blumen et al., 1997).

OPMD is caused by short expansions of a (GCG)6 triplet repeat

to (GCG)8–13 in the PABPN1 gene (trePABPN1), which results

in the expansion of a polyalanine stretch from 10 to 12–17

alanines in the amino terminus of the protein (Brais et al.,

1998). Disease mutant (mut) trePABPN1 induces nuclear protein

aggregation and forms filamentous nuclear inclusions, a patho-

logical hallmark of OPMD (Calado et al., 2000; Thome et al.,

1997). How the triplet-expansion mutations affect PABPN1

activity and cellular behavior is currently unknown. To study

OPMD, a transgenic mouse model was developed that

expresses mut trePABPN1(A17) under the control of the

muscle-specific human skeletal actin promoter (Davies et al.,

2005). trePABPN1(A17) mice appear normal at birth but develop

a progressive muscle weakness that is accompanied by the

formation of trePABPN1(A17)-containing aggregates in the nuclei

of skeletal myocytes and by accelerated apoptosis (Davies and

Rubinsztein, 2011; Davies et al., 2005).

Here, using a wide-scale RNAi screen focused on human

RBPs, we identified PABPN1 as a suppressor of APA. Loss of

PABPN1 resulted in a global enhancement of proximal CSs

usage in independent human cellular systems. In amousemodel

for OPMD, as well as in human cells, the expression of a disease

mut trePABPN1(A17) also resulted in an extensive enhancement

of proximal CSs. Our results elucidate a role for PABPN1 in sup-

pressing APA sites and predict a role for unbalanced APA in

OPMD.

RESULTS

AnRNAi Screen Identifies PABPN1as aRegulator of APAWe designed a reporter system (named pSTAR; a plasmid

system to identifyAPA regulators) to identify factors that regulate

the usage of APA sites. We cloned downstream to the Renilla

luciferase in psiCHECK2 (psiC) an �1 kb region of the 30 UTRof the mixed leukemia lymphoma gene (MLL), recently discov-

ered to strongly repress luciferase expression, mainly at the

level of RNA synthesis (Figures 1A; Figure S1A available online;

Gomez-Benito et al., 2011). Physical engagement of RNA poly-

merase (Pol) II with transcription of this MLL region results in

repression and low luciferase counts. Next, we cloned the prox-

imal alternative PAS region of Cyclin D2 (CCND2; Mayr and

Bartel, 2009) upstream of the MLL insert and used the mutated

PAS as control (pSTAR-D2WT, pSTAR-D2mut; Figures 1A, S1A,

and S1B). We expected that the Renilla luciferase repression

by MLL-30 UTR in pSTAR would be weakened due to early

cleavage mediated by the CCND2 PAS in pSTAR-D2WT. Figures

1A and S1C show that the MLL-30 UTR repressor reduced the

Renilla/Firefly (R/F) ratio by more than 10-fold by lowering the

Renilla luciferase expression. As expected, the insertion of

CCND2 wild-type (WT) proximal PAS significantly relieved the

repression, and this relief was less effective when that PAS

was mutated (Figure 1A). Sequence analysis of transcript ends

downstream of the Renilla gene in pSTAR-D2WT revealed the

exact expected CS (Figure S1B). Similar effects were observed

with the p27 (CDKN1B) 30 UTR distal PAS region (Figures 1B

and S1D), albeit this PAS showed a stronger effect, which is

consistent with it being the strongest functional PAS of the p27

gene (Figures 1B and S4C). Therefore, in both cases our pSTAR

reporter system successfully monitored PAS activity.

Next, we performed an RNA interference (RNAi) library screen

to identify RBPgenes required for APA.We reasoned that knock-

ing down genes involved in APA should attenuate the difference

in R/F ratios measured for the WT and mutant proximal CCND2

PAS constructs. Thus, we transfected pSTAR-D2WT and -D2mut

into U2OS cells and reverse transfected siRNA smart pools de-

signed to target 489 RBP genes (Figure S1E). The R/F ratio of

pSTAR-D2WT was compared to that of -D2mut (Figure 1C). Genes

were ranked according to the relative difference between R/F

ratios for pSTAR-D2WT and -D2mut (Table S1). The top-scoring

candidate in this screen (i.e., the candidate that minimized the

gap between pSTAR-D2WT and -D2mut) was PABPN1 (Figure 1C,

red point). To validate this result, we tested each individual small

interfering RNA (siRNA) from the PABPN1 smart pool. All four

siRNAs effectively silenced PABPN1 expression and diminished

the difference between pSTAR-D2WT and -D2mut (Figures 1E

and 1D). siRNA #1 and #2 were the most efficient ones. This

effect of PABPN1 knockdown (kd) was confirmed at the mRNA

level (Figure S1F). Detailed examination of the effect of PABPN1

kd showed that in addition to its effect on the relative pSTAR-

D2-WT/mut counts, it also lowered the basal R/F ratio (Figure 1E).

However, the latter effect was observed also in the control

pSTAR vector (Figure 1F, upper panel), indicating it is not related

to PABPN1’s role in regulating the usage of the CCND2-PAS

insert. Intriguingly, the attenuation of the gap between pSTAR-

D2WT and -D2mut caused by PABPN1 kd did not result from

Cell 149, 538–553, April 27, 2012 ª2012 Elsevier Inc. 539

Figure 1. Functional Genetic Screen Identifies PABPN1 as a Regulator of PAS Usage

(A) Our pSTAR system consists of psiCHECK2 (psiC) dual luciferase vectors (Figure S1A), containing the WT or mut proximal CCND2 PAS (D2WT or D2mut;

Figure S1B) upstream of the MLL-30 UTR fragment (240–1143 bp; named repressor [R]). These systems were transfected into U2OS cells for 72 hr. The relative

R/F ratio (average ± standard deviation [SD]) was determined.


a compromised relief of the Renilla luciferase repression in the

pSTAR-D2WT construct: in PABPN1 kd cells, the R/F ratio

between pSTAR-D2WT and pSTAR (7.0) was similar to the control

(5.5; Figures 1F, upper panel and 1G). Instead, in the absence of

PABPN1, the effect of pSTAR-D2mut on the repression relief was

markedly reduced (Figures 1F, upper panel and 1G). As an indi-

cation for the specificity of PABPN1-mediated regulation of APA,

no major effect of PABPN1 kd was observed on the stronger

distal p27-PAS (Figure 1F, middle panel; PABPN1 kd confirmed

by quantitative reverse transcription-polymerase chain reaction

[qRT-PCR]; lower panel) nor on the proximal FGF2 and Dicer1

PASs (Figure S1G). To further ascertain PABPN1’s involvement

in the regulation of CCND2-PAS usage, we constructed

siRNA-resistant PABPN1 expression vectors for siRNAs #1

and #2. Expression of the siRNA-resistant PABPN1 proteins

restored in the presence of PABPN1 siRNAs #1 and #2 the differ-

ence in R/F ratio of WT versus mut CCND2-PAS to levels

observed by WT PABPN1 overexpression (Figure 1H). We

confirmed PABPN1 kd and reconstitution by siRNA-resistant

constructs with immunoblotting (Figure 1I). Altogether, our

results indicate that in the absence of PABPN1, early cleavage

occurred efficiently in both the pSTAR-D2WT and -D2mut

constructs, suggesting that loss of PABPN1 activated a dormant

CS in the CCND2-proximal PAS region to mediate an early

cleavage in the mut PAS construct.

Loss of PABPN1 Induces Global 30 UTR ShorteningTo examine the role of PABPN1 in PAS selection on a global

scale, we applied a deep-sequencing-based technique adapted

from Beck et al. (2010) and named 30Seq (Figure S2A). 30Seqallows precise mapping of 30-end cleavage and poly(A) sites at

a nucleotide resolution based on the alignment of sequenced

reads that span the CS and therefore cover the beginning of

the untemplated poly(A) tail (Figures S2B and S2C). We applied

30Seq to four independent U2OS-transfected control and two

PABPN1 kd samples. Analyzing the 30Seq dataset, we identified

23,635 CSs, of which 58% mapped to 30 UTRs of 9,345 tran-

scripts (Figure S2D). 2,990 (32%) transcripts contained more

than one CS in their 30 UTR (Figure S2E). We benchmarked these

CSs against annotated CSs recorded in the poly(A) database

(Lee et al., 2007) or in Ozsolak et al. (2010) and found 11,407

(83%) of the 30 UTR CSs to overlap with at least one of these

resources but only 2,227 (22%) of other CSs (Figure S2F). There-

fore, in subsequent analyses, we included only CSs that mapped

(B) Same experiments as in (A), but a region encompassing the p27-PAS (p27WT

(C) U2OS cells were transfected with pSTAR-D2WT and -D2mut constructs, reverse

Figure S1E). Standardized differences between log2 of R/F ratios measured in

PABPN1 kd (in red) strongly minimized the gap between pSTAR-D2WT and -D2m

(D and E) The four individual siRNAs of the PABPN1 smart pool used in (C) were

relative to b-actin). (E) shows relative R/F ratios (average ± SD) for pSTAR-D2WT

(F) pSTAR-D2WT and -D2mut were transfected into U2OS cells. After 16 hr, cells we

ratios were determined (mean ± standard error of the mean [SEM]; upper and mid

by qRT-PCR.

(G) Quantification of the R/F ratios shown in (F) for WT and mut CCND2-PAS (n =

(H) In addition to the reporter constructs used in (F) (upper panel), expression

cotransfected. Data are represented as average ± SD.

(I) Immunoblot analysis with PABPN1 and b-actin antibodies of whole-cell extrac

See also Figure S1 and Table S1.

to 30 UTRs. We next searched for enriched sequence motifs in

the region of ±50 nt with respect to the CSs. The most enriched

signal in this region was the canonical AAUAAA PAS (corre-

sponding to AATAAA on the DNA level). Also, all major variants

of the canonical signal (Tian et al., 2005) were highly enriched

too, with AUUAAA being the second most prominent motif,

appearing together with the canonical signal in �60% of the

mapped CSs (Figure S2G). As an additional indication for the

precision of the CS mapping by 30Seq, the location distribution

of the major PAS signals showed a sharp peak at the docu-

mented position �20 nt upstream of the CSs (Figure S2H) (Tian

et al., 2005). De novo motif discovery analysis also detected

enriched T/TG motifs, which, in accordance with previous

reports (Hu et al., 2005; Lutz, 2008; Nunes et al., 2010), peaked

downstream of the CSs (Figure S2I).

Next, we examined the effect of PABPN1 kd on PAS selection.

We identified 572 transcripts that showed a significant shift in

CS usage in PABPN1 kd versus control samples (p < 0.001;

chi-square test; Table S2A). In more than 90% of these tran-

scripts, the shift in CS usage was toward the proximal PASs

(p < 10�99; binomial tail), indicating an extensive 30 UTR short-

ening in the absence of PABPN1 (Figures 2A and 2B). We vali-

dated by quantitative polymerase chain reaction (qPCR) and

northern blot the increase in the level of the short relative to

the long isoform for selected genes (Figures 2C and S2M). The

northern blot showed for the majority of the examined genes

that PABPN1 kd results in elevated (absolute) levels of the short

isoform, indicating that the increase in the short/long ratio in the

absence of PABPN1 was associated with enhanced cleavage

at the proximal PASs (Figure S2M, right panel). To quantify the

relative usage of each PAS in a transcript, we defined the PAS

usage index (PUI; Extended Experimental Procedures). Compar-

ison of the proximal PUI distribution in PABPN1 kd versus control

samples further demonstrated the significantly enhanced usage

of proximal PASs in the absence of PABPN1 (Figures 2D and 2E).

To confirm the effect of PABPN1 kd on PAS selection, we

applied 30Seq to a second cell line (RPE-1). Here too, PABPN1

kd resulted in a broad 30 UTR shortening (Figures S2N–S2P).

Of the 357 transcripts with a significant shift in CS usage, 291

(82%) showed 30 UTR shortening in the PABPN1 kd samples

(p = 3.97 * 10�35) (Figure S2Q).

These results strongly suggest that PABPN1 kd causes a

significant shift toward proximal PAS usage. Therefore, we

sought to identify sequence features that potentially differ

and p27mut; Figure S1D) was used.

transfected with siRNA smart pools targeting 489 RBP genes (384-well format;

the pSTAR-D2WT and -D2mut constructs are shown for each targeted gene.ut.

reverse transfected. (D) shows the relative PABPN1 mRNAs levels (qRT-PCR;

and -D2mut in the individual siRNA samples.

re reverse transfected with the indicated siRNAs. Seventy-two hours later, R/F

dle panels). Lower panel: PABPN1 mRNA levels in transfected cells measured

15; mean ± SEM; one-tailed t test).

vectors for empty control, PABPN1 WT, or siRNA-resistant PABPN1 were

ts (WCE) from the cell populations used in (H).


Figure 2. PABPN1 kd Induces Extensive 30 UTR Shortening

(A) Examples of transcripts showing significant 30 UTR shortening, manifested by a prominent increase in the proximal CS level relative to the distal one in

PABPN1 kd compared to control samples. Transcripts with red-colored reads are encoded on the reverse DNA strand, thus the distal CS is on the left and the

proximal CS is on the right side.


between proximal and distal sites. Analyzing the set of tran-

scripts containing exactly two 30 UTR CSs, we found a signifi-

cantly higher prevalence of the canonical PAS, but not of its

variants, in distal versus proximal sites (49% versus 27%,

respectively; Table S3). This difference was even more pro-

nounced in the subset of transcripts, which showed significant

30 UTR shortening in the PABPN1 kd samples (Table S3), sug-

gesting that PABPN1 kd enhances the usage of noncanonical

proximal PASs.

Loss of PABPN1 Enhances mRNA Cleavage Mediatedby Proximal PASsPABPN1 kd significantly increased the usage of proximal CSs

relative to distal ones in hundreds of transcripts, and for several

transcripts, a clear upregulation of the short form was observed

by northern blots (Figure S2M). However, it is possible that

PABPN1 kd globally destabilized transcript isoforms with longer

30 UTRs (e.g., due to broad activation of miRNA activity or RBPs).

To examine this possibility, we measured the stability of short

and long transcript isoforms for several genes that showed

a significant shift toward the proximal PAS in PABPN1 kd

samples. We used 30-end qRT-PCR to distinguish between the

stability of isoforms with short and long 30 UTRs (generated by

cleavage at proximal and distal CSs, respectively). To check

the validity of the 30-end qRT-PCR method, we first examined

the relatively unstable p27 transcript. Standard qRT-PCR

measurement of p27 mRNA stability showed that p27 mRNA

fromU2OS cells has a half-life of�2 hr (Figure 3A). No significant

change in p27 stability was observed in PABPN1 kd cells (Fig-

ure 3A). When the same total RNA samples were examined

with 30-end qRT-PCR covering the p27-PAS sequence, similar

p27 stability was found, and again no significant effect of

PABPN1 kd was seen (Figure 3B, left panel). 30-end qRT-PCR

applied to b-actin mRNA expectedly revealed a highly stable

transcript (Figure 3B, right panel). Thus, stability estimates ob-

tained by 30-end qRT-PCR agree with those obtained with stan-

dard qRT-PCR. Moreover, PABPN1 kd did not significantly

change the stability of these two single PAS-containing genes.

Next, we applied 30-end qRT-PCR to compare the stability of

the long and short 30 UTR transcript isoforms of PSMD8,

SUPT6H, PDRG1, E2F1, and CCND1 that showed a strong shift

toward the proximal PAS following PABPN1 kd. Regardless of

the stability of the 30 end observed in control cells (e.g.,

PSMD8 is highly stable, whereas PDRG1 is relatively unstable),

PABPN1 kd had no significant effect on the stability of all five

transcripts (Figures 3C, 4D, and S3A). qRT-PCR demonstrated

(B) Five hundred and seventy-two transcripts showed a significant shift (p < 0.0

tween PABPN1 kd and control samples. In 523 of these transcripts, absence o

resulting in overall 30 UTR shortening. (p value calculated using binomial tail.)

(C) To confirm the results of the 30Seq, the same samples as in (B) were sub

average ± SD.

(D) Comparison between the distribution of the PUI of the proximal sites in the PAB

PABPN1 kd resulted in a global increase of proximal PAS usage (p value was c

showed significant shift in CS usage.

(E) Themedian of the proximal sites PUI distribution, calculated over all transcripts

proximal CSs in each sample. PABPN1 kd resulted in a marked increase in this

See also Figure S2 and Tables S2A and S3.

an efficient PABPN1 kd (Figure 3D). We conclude that a global

destabilization of longer transcripts is not a major cause of the

broad shift toward proximal PASs observed in the absence of

PABPN1.

Next, we examined the transcription along MAPKAPK3,

PDRG1, and E2F1, which showed significantly elevated usage

of the proximal relative to the distal CS in PABPN1 kd cells. If

proximal PAS usage is indeed activated, reduced levels of Pol

II occupancy should be detected in the 30 UTR downstream of

the proximal PAS following loss of PABPN1 expression.

Conversely, if destabilization of the longer transcript causes

the relative increase in the usage of the proximal versus the distal

PAS, no change in Pol II levels along the 30 UTR should be

observed in PABPN1 kd samples. Chromatin immunoprecipita-

tion (ChIP) from extracts of control and PABPN1 siRNA-trans-

fected cells showed that for these genes, but not for p27 (nega-

tive control), PABPN1 kd significantly reduced Pol II interaction

with distal 30 UTRs relative to coding sequence (CDS) (Figure 3E).

This indicates an early termination of Pol II transcription in the

absence of PABPN1 and supports the conclusion that a reduced

level of PABPN1 results in activation of proximal CSs.

PABPN1 Regulation of Proximal CSsThe extensive activation of proximal CSs in the absence of

PABPN1 can result from enhanced activity of the pre-mRNA

30-end processing machinery. A recent study identified >90

proteins that constitute this processing complex (Shi et al.,

2009). Global upregulation of the set of genes encoding these

factors accompanies the broad 30 UTR shortening observed

during enhanced cellular proliferation (Ji et al., 2009 and unpub-

lished data). To examine whether a similar induction of 30-endprocessing genes underlies the 30 UTR shortening in the absence

of PABPN1, we used the 30Seq data to derive gene expression

estimates and compared the expression of these genes between

control and PABPN1 siRNA-transfected U2OS cells. However,

no global change in the expression of genes involved in the

30-end processing was detected in PABPN1 kd samples (Fig-

ure S3B). This indicates that the absence of PABPN1 does not

activate proximal CSs through global induction of the 30-endprocessing machinery.

Another possible explanation for activated proximal CS usage

in the absence of PABPN1 is a direct interaction of PABPN1 with

proximal PAS regions, which are typically weaker than distal

signals. Such binding of PABPN1may compete with PAS recog-

nition and/or mRNA cleavage complexes and thus can reduce

cleavage efficiency at these sites. To examine this model, we

01; chi-square test; see Extended Experimental Procedures) in CS usage be-

f PABPN1 enhanced the usage of the proximal site relative to the distal one,

jected to qRT-PCR. Results were normalized to siControl #1 and shown as

PN1 kd and control samples. The significant shift to the right demonstrates that

alculated using Wilcoxon test). The analysis included the 572 transcripts that

with more than one 30 UTR CS, was taken as a global measure for the usage of

measure.


tested PABPN1 interaction with proximal PAS regions in vivo.

We immunoprecipitated PABPN1 from U2OS cells that were

subjected to 4-thiouridine and ultraviolet (UV) crosslinking

treatments. Then we performed qRT-PCR with reverse oligos

immediately downstream of the proximal CSs of SUPT6H, MAP-

KAPK3, and hnRNPUL2 in order to distinguish unprocessed

transcripts from cleaved and polyadenylated isoforms (Fig-

ure S3C). Such primer setting is essential because PABPN1

strongly associates with poly(A) tails inside nuclei, and this

binding may overshadow PABPN1 binding to regions adjacent

to proximal CSs. As control for the specificity of the PABPN1

binding, we used qRT-PCR primers mapping to coding regions

(Figure S3C). The proximal PAS regions of all three genes exam-

ined were significantly enriched for PABPN1 binding (Figure 3F).

We then examined whether PABPN1 binds to proximal PAS

regions in vitro. We synthesized the proximal hnRNPUL2 PAS

including 30 nt down- and upstream and used RNA-electromo-

bility shift assays (R-EMSA) to examine the binding of recombi-

nant PABPN1 (rPABPN1) to this radiolabeled proximal PAS

region (Kuhn et al., 2009). We observed rPABPN1 binding that

was largely dependent on the integrity of the PAS sequence,

as mutating the triple A of the PAS to triple C clearly reduced

rPABPN1 binding to this region (Figure 3G). rPABPN1 binding

to the WT PAS region was efficiently competed with cold WT,

but not mutated, PAS sequence (Figure 3H). This indicates

specific binding to the PAS region. Similar results were obtained

with the proximal MAPKAPK3 PAS region and increasing

rPABPN1 amounts (Figures 3I and S3D). These results suggest

a direct effect of PABPN1 on proximal PAS usage.

Next, we performed in vitro cleavage assays to test whether

PABPN1 directly regulates PAS-mediated cleavage. We used

in vitro synthesized RNA probes comprising the proximal PAS

regions of MAPKAPK3, hnRNPUL2, and SUPT6H and the distal

PAS region of MAPKAPK3. In these probes, the CS of the

MAPKAPK3 distal PAS region is located �10 nt from the 30 end,and the CSs of MAPKAPK3, hnRNPUL2, and SUPT6H proximal

PAS regions at �25–30 nt from the 30 end (Figure S3E). As

controls, we used probes with mutated PAS sequence (AAA to

CCC, designated WT to mut). When incubating these probes

Figure 3. PABPN1 Regulation of Proximal CSs

(A and B) U2OS cells were transfected with PABPN1 siRNAs (kd #1 and #2) fo

measured by standard qRT-PCR for p27-CDS (A) or by 30-end qRT-PCR for p27

(C) The stability of mRNA isoforms terminated at the distal CS was measured by

a destabilization of the longer transcript or in a stabilization of the shorter isofom

(D) qRT-PCR was performed to check PABPN1 kd in the same samples as in (A

(E) Pol II ChIPs from extracts of U2OS cells transfected with control or PABPN1 siR

and in CDS (mean ± SEM; one-tailed t test). Lower panel: schematic location of

(F) WCE of 4-thiouridine-treated and UV-crosslinked U2OS cells were immunopre

(primers: Figure S3C). Relative enrichment of PABPN1 over IgG was determined

(G) R-EMSA performed with RNA oligonucleotides encompassing 30 nt up- and

bacterial rPABPN1 was used.

(H) WT hnRNPUL2 PAS oligonucleotides were competed with increasing amount

hnRNPUL2 PAS was efficiently competed with cold WT oligonucleotide but not

(I) R-EMSA performed with RNA oligonucleotides surrounding the proximal WT o

(J) In vitro cleavage assays with probes encompassing the proximal MAPKAPK3,

nuclear extracts from HeLa cells transfected with control (C) or PABPN1 (#2) siR

mented with 50 nM rPABPN1, respectively). The arrowmarks the cleaved product

levels (average ± SD) and a western blot of PABPN1 (same HeLa nuclear extrac

See also Figure S3.

with nuclear extract of HeLa cells transfected with control

siRNAs (Figure 3J, lane C), we observed efficient and specific

cleavage of the canonical distal MAPKAPK3 region but very

limited cleavage of the proximal PAS regions. PABPN1 kd (#2)

enhancedcleavageof all proximal probesbut not of thecanonical

distal one. As no cleavage products of proximal or distal probes

were seen when the control PAS mut probes were used, and

as the lengths of the cleavage products fit the expected sizes

(Figure S3E), we conclude that this cleavage is mediated by the

30-end processingmachinery and that loss of PABPN1 enhances

it. Supporting this notion, adding rPABPN1 to the cleavage reac-

tion markedly reduced PABPN1-induced cleavage efficiency

(Figure 3J, lane #2+P). PABPN1 kd was confirmed by qRT-PCR

and western blotting (Figure 3J, upper right corner). This result

strongly supports a direct role for PABPN1 in suppressing PAS-

mediated cleavage at proximal sites.

PABPN1 Contributes to Efficient 30 UTR-MediatedRepressionLoss of PABPN1 in U2OS cells increased the usage of CCND1

proximal PAS by �2-fold (Figures 4A–4C). This shift was not

accompanied by any stability changes of the long or the short

transcript isoforms (Figure 4D). CCND1 is a target of the miR-

17-19 cluster (Figure 4E), which is expressed in U2OS cells (Fig-

ure 4F). Thus, we examined whether loss of PABPN1 attenuates

CCND1 sensitivity to miR-17-19. We transfected U2OS cells

with a construct containing the miR-17-19 cluster (miRVec-17-

19, expressing miR-17, -18a, -20a, -19a, and -19b-1; Fig-

ure S4A), or a control vector, and reverse transfected these cells

with control and PABPN1 #1 and #2 siRNAs. Transfection of

miRVec-17-19 increased miR-20a levels and suppressed

CCND1 protein expression in control but not in PABPN1

siRNA-transfected cells (Figures 4F and 4G). To further examine

the PABPN1 effect on miR-17-19-mediated repression of the

CCND1 30 UTR, we constructed dual luciferase vectors contain-

ing either the full-length CCND1 30 UTR (psiC-D1-FL) or a short

CCND1 30 UTR that terminates at the proximal CS (psiC-D1-S).

The transfection of miRVec-17-19 resulted in �3.5-fold relative

inhibition of the full-length CCND1 30 UTR compared to

r 72 hr and incubated with actinomycin D (ActD) for 6 hr. mRNA levels were

-PAS and b-Actin PAS (B). Data are shown as average ± SD.

30-end qRT-PCR (average ± SD). In none of the cases did PABPN1 kd result in

s (Figure S3A).

), (B), and (C). Data are shown as average ± SD.

NAs. Pol II binding wasmeasured in 30 UTRs downstream of the proximal PAS

the primers.

cipitated with PABPN1 or control IgG antibody, and qRT-PCR was performed

(mean ± SEM; one-tailed t test).

downstream of the proximal WT or mut hnRNPUL2 PAS. Fifty nanomolar of

s of cold WT or mut hnRNPUL2 PAS oligonucleotides. rPABPN1 binding to WT

with the mut oligonucleotide.

r mut MAPKAPK3 PAS and increasing rPABPN1 amounts.

hnRNPUL2, and SUPT6H PASs and the distal MAPKAPK3 PAS, incubated with

NAs (� and +P: probes incubated without nuclear extract or extracts supple-

s. Upper right corner: qRT-PCR showing the relative PABPN1 to b-actin mRNA

ts as used for the cleavage assay). BG: background band.


Figure 4. PABPN1 kd Induces CCND1 Proximal PAS Usage and Reduces Its Targeting by miRNAs(A) A UCSC web browser screenshot showing the shift in PAS usage of CCND1 in PABPN1 kd cells.

(B) The ratios of proximal to distal peak area shown in (A), normalized to the average of the four controls.

(C) qPCR with cDNA from (A) (average ± SD).

(D) 30-end qRT-PCR to measure the stability of the mRNA isoforms generated using the proximal and the distal CCND1 PAS. Data are shown as average ± SD.

(E) A scheme of the CCND1 30 UTR, proximal and distal PASs, and conserved target sites for miR-17/20a, 19, and 15/16 and let-7 (targetscan; Grimson

et al., 2007).


miRvec-Ctrl transfection (Figure S4B, left panel). Moreover,

whereas the full-length CCND1 30 UTR was highly repressed

by miR-17-19, the short CCND1 30 UTR form was completely

resistant (Figure S4B, left panel). PABPN1 kd significantly

reduced miR-17-19 effect on the full-length construct (p <

0.00001; one-tailed t test; Figure 4H, left panel). To control for

general effects on miRNA activity, we used p27 30 UTR that

showed no APA in our dataset (Figure S4C) and contained

two miR-222 target sites (le Sage et al., 2007). We cotransfected

the psiC-p27 30 UTR with miRVec-222 and observed an �2-fold

relative inhibition of p27 30 UTR compared to miRvec-Ctrl trans-

fection (Figure S4B, right panel). A construct containing muta-

tions in the two miR-222 target sites (p27-30 UTR-DM) was resis-

tant to miR-222-mediated suppression (Figure S4B, right panel).

In contrast to CCND1, PABPN1 kd did not affect miR-222-

mediated repression of p27 30 UTR (Figure 4H, right panel), indi-

cating that PABPN1 does not generally interfere with miRNA

function in our system. These results provide an example for

the role of PABPN1 in suppressing proximal PAS usage, thereby

allowing the regulation of the target gene by its full 30 UTR

sequence.

Extensive 30 UTR Shortening Caused by an OPMD-Associated PABPN1 MutantShort triplet-repeat expansion mutations in the PABPN1 gene

(trePABPN1) cause the autosomal-dominant OPMD disease.

As the effect of trePABPN1 on the function of the endogenous

WT PABPN1 protein is unknown, we examined whether tre-

PABPN1(A17) interacts with endogenous WT PABPN1. We mildly

expressed a GFP-tagged trePABPN1(A17) in U2OS cells and de-

tected nuclear dots with immunostaining (Figure 5A, marked

with #). Immunohistochemistry with PABPN1 antibody revealed

a colocalization of endogenous PABPN1 with GFP-tre-

PABPN1(A17) dots in transfected cells (#), whereas a rather

homogenous staining in untransfected cells was seen (Figure 5A,

marked with arrows). This suggests that trePABPN1(A17) seques-

ters endogenous PABPN1. If this effect is direct, one

should expect binding of trePABPN1(A17) to PABPN1. Indeed,

coimmunoprecipitation reveals interaction between GFP-tre-

PABPN1(A17) and endogenous WT PABPN1 (Figure 5B, lane 6).

Next, we examined whether trePABPN1(A17) also affects APA,

as we found that loss of WT PABPN1 significantly induced the

usage of proximal PASs. We transfected U2OS cells with an

expression vector for trePABPN1(A17) or a control vector and

performed 30Seq. trePABPN1(A17) expression resulted in 30

UTR shortening: of the 134 transcripts that showed a significant

shift in CS usage, 103 transcripts showed an induction of the

proximal CS usage (p = 1.6 * 10�10; binomial tail; Figure 5C).

(F) U2OS cells were transfected with miR-17-19 or control vectors for 96 hr. miR-2

in BJ primary foreskin cells where the miR-17-19 cluster is hardly expressed. Da

(G) Immunoblot showing CCND1 in control and PABPN1kd (#1, #2) cells, with CC

Asterisk: nonspecific band.

(H) U2OS cells were transfectedwithmiRVec-control, miRVec-17-19 (left panel), o

and #2 siRNAs. The relative R/F ratios of CCND1 short (psiC-D1-S) versus full-le

PABPN1 kd significantly reduced miR-17-19-mediated repression of the CCND1

CCND1 30 UTR form from �3-fold to �2-fold (n = 9; mean ± SEM; one-tailed t te

See also Figure S4.

Proximal PUI analysis also demonstrated that trePABPN1(A17)

expression resulted in enhanced usage of proximal CSs (Figures

5D and 5E). Importantly, we observed a significant overlap

between the genes that showed 30 UTR shortening in response

to trePABPN1(A17) overexpression and in response to endoge-

nous PABPN1 kd (60 transcripts; p = 1.7 * 10�7; hypergeometric

tail). These results indicate that trePABPN1 functions in a domi-

nant-negative manner by interacting with WT PABPN1 and

sequestering it to nuclear dots, and its effect on APA is similar

to the effect of inhibiting WT PABPN1 by RNAi.

To study the effect of trePABPN1 function on APA in vivo, we

applied 30Seq to skeletal muscle tissues from an OPMD mouse

model that expresses trePABPN1(A17) (Davies et al., 2005). This

30Seq dataset included samples from OPMD mice aged

20 weeks or 12 months and their matching controls. We de-

tected 10,849 CSs, 55% of which mapped to 30 UTRs. Motif

analysis identified the canonical PAS as themost enriched signal

in the CS regions, and its location distribution sharply peaked at

the expected position �20–25 nt upstream of the mapped CSs

(Figures S6A and S6B). In line with our findings in human cells,

murine muscle tissue expressing the trePABPN1(A17) mut

showed an extensive 30 UTR shortening at both the age of

12 months (Figures 6A–6D) and the age of 20 weeks (Figures

S6C–S6E). In the analysis of the 12-month-old mice, a total of

305 transcripts showed a significant shift in CS usage in the

comparison between trePABPN1(A17) and control muscle tissues

(Table S2B). Strikingly, 273 of them (90%, p = 3.9 * 10�49; bino-

mial tail) showed elevated usage of proximal sites in the tre-

PABPN1(A17) mice (Figure 6B). These results detect for the first

time wide-spreading abnormal APA in a model of a human

genetic disorder and further strongly pinpoint PABPN1 as an

important regulator of this gene regulation process.

Assuming that in many cases 30 UTR shortening would atten-

uate miRNA-mediated transcript degradation, we compared

transcript expression levels between the trePABPN1(A17) and

control samples. We found that at the age of 12 months, the

expression of transcripts with enhanced proximal PAS usage

in trePABPN1(A17) mice was significantly elevated in these mice

compared to their expression in the WT controls (Figure 6E). A

similar trend, but with reduced statistical significance, was

observed for the 20-week-old mice (Figure S6F) and in U2OS

cells expressing trePABPN1(A17) (Figure S5).

DISCUSSION

APA is a layer of gene regulation that to date remains largely

unexplored. Yet, it is becoming evident that a large portion of

human genes contain multiple CSs and PASs in their 30 UTRs.

0a was detected by qRT-PCR and normalized to GAPDH and tomiR-20a levels

ta are shown as average ± SD.

ND1 ratios (miR-17-19 to control transfected cells) indicated below the blots.

r miRVec-222 (right panel) and reverse transfectedwith control and PABPN1 #1

ngth (psiC-D1-FL) or p27 double mutant (DM) versus WT 30 UTRs are shown.

full-length construct, resulting in a reduction of the relative short/long ratio of

st).


Figure 5. OPMD-Associated trePABPN1(A17) Induces 30 UTR Shortening

(A) U2OS cells mildly transfected with GFP-trePABPN1(A17) for 2 days (0.8-fold compared with endogenous WT PABPN1, as determined by quantitative analysis

of WCE bands in B, lane 4, upper panel). Cells were fixed and immunostained with PABPN1 antibody.

(B) Immunoprecipitations (a-GFP or control antibody) of WCE from GFP and GFP-trePABPN1(A17) transfected cells were immunoblotted to detect the indicated

proteins with a-PABPN1 (upper panel). a-CCND2 (D2) and a-GFP: controls (middle and lower panels). PABPN1 is present in anti-GFP immunoprecipitations from

cells transfected with trePABPN1(A17)-expressing construct.

(C) Numbers of transcripts with 30 UTR shortening or lengthening in U2OS cells expressing trePABPN1(A17) compared to control cells. (p value was calculated

using binomial tail.)

(D) The median of the proximal PUI distribution as a global measure for the level of proximal CS usage in each sample.

(E) Comparison between the proximal PUI distribution calculated in the trePABPN1(A17) and control U2OS cells. (p value calculated using Wilcoxon test.)


Therefore, it is conceivable that the majority of human genes are

subjected to APA regulation. The maturation of deep-

sequencing methods specifically designed to probe transcript

30 ends significantly enhances our ability to globally characterize

alternative CSs and PASs and hence appreciate the full scope of

APA regulation under various physiological stresses and patho-

logical conditions. Here, we developed an RNAi screen to iden-

tify APA regulators and pinpointed PABPN1 as an important

player in this process. We have adapted a deep-sequencing

technique, 30Seq, and applied it to obtain genome-wide views

of PABPN1’s role in determination of PAS usage.

PABPN1 Controls Proximal PAS UsagePrevious in vitro data indicated that PABPN1 binds to poly(A)

tails and controls their extension (Kuhn et al., 2009). It stimulates

polyadenylation by the PAP enzyme but also restricts polyade-

nylation once the poly(A) tail reaches a certain size (�250 nt in

humans). Here, we show that PABPN1 regulates APA by sup-

pressing the usage of weak proximal PASs (Figure 7, upper

and middle panel). This suggests that PABPN1 acts prior to

cleavage and polyadenylation of mRNAs to determine proper

rate of proximal PAS usage.

Our in vitro and in vivo data suggest that PABPN1 directly

interacts with PAS regions and by that competes with the

cleavage and polyadenylation complexes on the recognition of

weak proximal and noncanonical PASs. Whereas PABPN1 is

likely to bind also to canonical PAS regions, the competition

with polyadenylation and cleavage complexes on such regions

is predicted to be much lower due to the significantly stronger

recruitment of the cleavage and polyadenylation machinery to

canonical PAS regions. Our in vitro cleavage assays support

this notion. Canonical distal PAS regions were more active and

less affected by reduced PABPN1 levels than proximal PAS

regions. Nevertheless, our results cannot exclude indirect

effects of PABPN1 on PAS selection, e.g., through PABPN1

association with Pol II (Bear et al., 2003).

PABPN1 and OPMDAGCG repeat extension in the PABPN1 gene from 6 to 8–13 trip-

lets (trePABPN1) causes OPMD, which is characterized by

progressive degeneration of particular muscles in adults (Tava-

nez et al., 2005). trePABPN1 proteins induce nuclear protein

aggregation and form filamentous nuclear inclusions (Calado

et al., 2000; Thome et al., 1997). Our data support the induction

of nuclear inclusion by trePABPN1(A17) and further suggest that

by binding to WT PABPN1 and sequestering it to inclusion

bodies, trePABPN1(A17) dominant negatively affects its WT

protein function (Figure 7, lower panel). Indeed, the global effect

on CS selection that we elicited by ectopic expression of tre-

PABPN1(A17) resembled the effect elicited by PABPN1 kd.

A transgene OPMDmouse model was developed by express-

ing mut trePABPN1(A17) under the control of the muscle-specific

human skeletal actin promoter (Davies et al., 2005). In this mouse

model, the expression of trePABPN1(A17) leads to a dystrophic

phenotype in the presence of two endogenous WT PABPN1

copies and mimics the progressive muscle weakness seen in

OPMD patients. This observation argues as well for a toxic

gain of function for trePABPN1(A17) (Davies et al., 2005).

We found in this mouse model a global induction of proximal

PAS usage mediated by trePABPN1(A17). Also, it seems that

aggregates of the mut protein coaggregate with WT PABPN1,

which would result in depletion and reduced activity of WT

PABPN1 in OPMD muscle cells.

Our results imply that OPMD is associated with misregulated

APA, which results in unbalanced formation of alternative

mRNA 30ends. They also predict that OPMD symptoms in hu-

mansmaybe to a certain degree a result of aberrant gene expres-

sion due to a change in 30-end formation. Though our results

predict a change in APA during OPMD development, it could

be expected that endogenously expressed trePABPN1 in

OPMD patients would present a milder APA phenotype com-

pared to the mouse model (due to higher transgene expression)

or compared to experimental conditions in which WT PABPN1

expression is dramatically reduced. To what extent trePABPN1

mutants induce aberrant APA in patient cells, and which genes

and phenotypes are affected, remain to be investigated.

Possible Role of PABPN1 in Proliferation and Cancer?Regulation of PAS usage and progressive 30 UTR shortening and

lengthening were observed during immune cell reactivation

(Sandberg et al., 2008), cancer progression (Mayr and Bartel,

2009), and development (Ji et al., 2009). To date it is unknown

whether PABPN1 is involved in these processes. Given that

OPMD patients were so far not reported to be at higher risk of

developing cancer and OPMD mouse models are not suscep-

tible to tumor growth, it would be interesting to examine levels

of PABPN1, its posttranscriptional modification status, and its

function in cancer or during normal cellular differentiation.

EXPERIMENTAL PROCEDURES

Cell Culture

U2OS osteosarcoma and HeLa cells and RPE-1 hTERT-immortalized retinal

pigment epithelial cells were maintained in DMEM or RPMI 1640, respectively,

supplemented with 10% FCS in a 5% air-humidified atmosphere at 37�C.

Transfections and Luciferase Assay

Transfection constructs were cloned as described in the Extended Experi-

mental Procedures. Cells were transfected with psiCHECK2 constructs

(Fugene6; Roche). After 72 hr, dual luciferase activity assays (Promega)

were performed. Cells were transfected with PABPN1 siRNAs (final concentra-

tion: 25 nM; Dharmafect3; Dharmacon).

Coimmunoprecipitation

GFP-trePABPN1(A17) or GFP was overexpressed in U2OS cells for 2 days and

immunoprecipitated with GammaBind G Sepharose beads (Ge Healthcare)

using a mouse GFP antibody (Roche) or an unrelated mouse Cox-2 antibody

(sc-19999, Santa Cruz Biotechnology). Cell extracts were cleared in lysis

buffer (50 mM HEPES, 150 mM NaCl, 0.5% Nonidet P-40 [pH 7.5], protease

inhibitor [PI; Roche]) and incubated with GFP or Cox-2 antibody overnight

at 4�C. Next, the extract/antibody mix was combined with the GammaBind

G Sepharose beads and incubated for 3 hr at 4�C. After washing with lysis

buffer (described above), the beads were resuspended in sample buffer and

subjected to immunoblotting.

R-EMSA

Radiolabeled RNA probes (Extended Experimental Procedures) were gener-

ated by in vitro transcription (miRNA probe construction kit; Ambion). RNA

probe (15,000 cpm) was incubated with 50 nM rPABPN1 in 10 ml binding buffer

(50 mM Tris-HCl [pH 8.0], 10% glycerol, 0.2 mg/ml BSA, 0.01% Nonidet P-40,


Figure 6. Extensive APA Misregulation in an OPMD Mouse Model

(A) Examples of three genes showing a significant increase in proximal relative to distal CS usage in muscle tissues from 12-month-old trePABPN1(A17) compared

to control WT mice.

(B) Three hundred and five transcripts showed a significant shift in CS usage between trePABPN1(A17) and WT mice, with 90% shifting toward proximal CSs.

(p value was calculated using binomial tail.)

(C) Comparison between the PUI distribution of the proximal CSs in trePABPN1(A17) and WT mice. (p value calculated using Wilcoxon test.)

(D) The median of the proximal PUI distribution as a global measure for proximal CS usage in each sample, turning from negative in WT to positive in tre-

PABPN1(A17) mice.


Figure 7. A Model for PABPN1’s Mode of Action in

Regulation of APA

Schematic model for PABPN1’s role in suppressing the

usage of proximal weaker PASs (upper and middle

panels). Ectopic expression of trePABPN1(A17) reduces

the endogenousWT PABPN1 function by excluding it from

the soluble nucleoplasm (lower panel).

0.5 mM DTT, 100 mM KCl, 2 mM MgCl2) for 15 min at room temperature (RT)

andwas runwith 2 ml loading buffer (50%glycerol, 0.1% bromophenol blue) on

a native gel (6% polyacrylamide [37.5:1], 40% glycerol, and 13 TBE) for 3 hr at

220 V. The gel was exposed to a Phosphor Imaging Plate and developed on

the BAS-2500 system (both Fujifilm).

In Vivo Binding Assay

Cells were grown overnight in medium containing 1 mM 4-thiouridine

(4SU; SigmaAldrich). Next, the cells were irradiated (365 nmUV light crosslink-

ing; 0.15 J/cm2; Stratalinker 2400) and harvested in lysis buffer (50 mMHEPES

[pH 7.5], 140 mM NaCl, 1% Triton X-100, 0.1% NaDeoxycholate, PI

[Roche], RNaseOUT [Invitrogen]). Cleared cell lysates were treated with RNA

fragmentation buffer (Ambion) for 3 min at 70�C and incubated with a PABPN1

antibody (2428-1; Epitomics) or an unrelated IgG antibody (sc-2027, Santa

Cruz Biotechnology) overnight at 4�C. GammaBind G Sepharose beads (GE

Healthcare) were preblocked with yeast tRNA (Invitrogen) and RNase-free

BSA (Ambion). Next, preblocked G Sepharose beads were combined with

antibody/protein lysate mix and incubated for 4 hr at 4�C. After intensive

washing, the beads were resuspended in lysis buffer (with proteinase K)

and incubated for 1 hr at 50�C. RNAwas extracted (Trizol; Invitrogen) and sub-

(E) Comparison of changes in expression levels in trePABPN1(A17) versus WTmuscles samples (12-mo

containing genes that showed a significant shift toward proximal CSs in the trePABPN1(A17) tissue an

measured in our dataset (p value calculated using Wilcoxon test).

See also Figures S5 and S6 and Table S2B.

Cell 149

jected to gene-specific cDNA synthesis and qPCR

(primer: Extended Experimental Procedures).

In Vitro Cleavage Assay

Assay was performed as described by Wahle and Keller

(1994; Extended Experimental Procedures). 1 3 107

HeLa cells were resuspended in 500 ml buffer (10 mM

Tris-HCl [pH 7.8], 1.5 mM MgCl2, 10 mM KCl, 0.5 mM

DTT, PI; Roche), homogenized (25G needle), and centri-

fuged at 2000 3 g for 15 min. Nuclei were resuspended

in 200 ml solution (25% glycerol, 20 mM Tris-HCl [pH

7.8], 1.5 mM MgCl2, 0.2 mM EDTA, 0.6 M NaCl, 0.5 mM

DTT, PI; Roche). Capped (Vaccinia capping system; New

England Biolabs) and 32P-labeled RNAs were prepared

with Ambion’s probe construction kit and gel purified.

Cleavage reactions were prepared in 12,5 ml containing

50,000 cpm of labeled RNA, 2 mM DTT, 1.5 mM MgCl2,

0.8 mM 30dATP, 0.1 mg/ml tRNA, 3% PEG 6000, 50 mM

KCl, 20 mM creatine phosphate, 40 U RNaseOUT (Invitro-

gen), and 12.5 ml nuclear extracts, incubated at 30�C for

2 hr, and stopped with proteinase K and SDS. RNA was

extracted (Trizol; Invitrogen), precipitated, separated on

a 10% acrylamide-7 M urea gel, and visualized by

autoradiography.

30-End qRT-PCR

Total RNA was extracted (RNeasy Mini kit; QIAGEN), and

3 mg was heat fragmented (fragmentation buffer; Ambion)

for 5 min at 70�C. The fragmented total RNA was precipi-

tated, the first cDNA strand synthesized (SuperScript III

reverse transcriptase; P7-t25-vn oligo-dT primer; Invitrogen), and qRT-PCR

performed as described above (P7-primer; gene-specific forward primers;

Extended Experimental Procedures).

Animal Studies

Animal work was performed under the jurisdiction of appropriate UK Home

Office Licenses.

ACCESSION NUMBERS

Raw sequence data and wig files that summarize read coverage along the

chromosomes (normalized to 10 M mapped reads) are available in the Gene

Expression Omnibus (GEO) database under accession number GSE27452.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures,

six figures, and six tables and can be found with this article online at

doi:10.1016/j.cell.2012.03.022.

nth-old mice) between two sets of transcripts: a target set

d a background (BG) set containing the rest of the genes

, 538–553, April 27, 2012 ª2012 Elsevier Inc. 551

http://dx.doi.org/doi:10.1016/j.cell.2012.03.022

ACKNOWLEDGMENTS

We thank all members of the Agami group including Bob Rosier for technical

help and discussions. We thank Roderick Beijersbergen, Pasi Halonen, and

Ben Morris from the NKI Robotics and screening center. Further we are grate-

ful to Arno Velds and Ron Kerkhoven at the NKI Central genomics facility for

deep sequencing our samples. This work was supported by funds from the

ERC (European Research Council), KWF (Dutch cancer foundation), Horizon

and VICI-NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek),

and CBG (Centre for Biomedical Genetics) to R.A., the SNF (The Swiss

National Science Foundation) to M.J., the VENI-NWO to G.v.H., and the Well-

come Trust Principal Research Fellowship to D.C.R.

Received: July 1, 2011

Revised: September 15, 2011

Accepted: March 2, 2012

Published online: April 12, 2012

REFERENCES

Andreassi, C., and Riccio, A. (2009). To localize or not to localize: mRNA fate is

in 30UTR ends. Trends Cell Biol. 19, 465–474.

Bear, D.G., Fomproix, N., Soop, T., Bjorkroth, B., Masich, S., and Daneholt, B.

(2003). Nuclear poly(A)-binding protein PABPN1 is associated with RNA poly-

merase II during transcription and accompanies the released transcript to the

nuclear pore. Exp. Cell Res. 286, 332–344.

Beaudoing, E., Freier, S., Wyatt, J.R., Claverie, J.M., and Gautheret, D. (2000).

Patterns of variant polyadenylation signal usage in human genes. Genome

Res. 10, 1001–1010.

Beck, A.H., Weng, Z., Witten, D.M., Zhu, S., Foley, J.W., Lacroute, P., Smith,

C.L., Tibshirani, R., van de Rijn, M., Sidow, A., et al. (2010). 30-end sequencing

for expression quantification (3SEQ) from archival tumor samples. PLoS ONE

5, e8768.

Bentley, D.L. (2005). Rules of engagement: co-transcriptional recruitment of

pre-mRNA processing factors. Curr. Opin. Cell Biol. 17, 251–256.

Blumen, S.C., Nisipeanu, P., Sadeh, M., Asherov, A., Blumen, N., Wirguin, Y.,

Khilkevich, O., Carasso, R.L., and Korczyn, A.D. (1997). Epidemiology and

inheritance of oculopharyngeal muscular dystrophy in Israel. Neuromuscul.

Disord. 7 (Suppl 1), S38–S40.

Brais, B., Bouchard, J.P., Xie, Y.G., Rochefort, D.L., Chretien, N., Tome, F.M.,

Lafreniere, R.G., Rommens, J.M., Uyama, E., Nohira, O., et al. (1998). Short

GCG expansions in the PABP2 gene cause oculopharyngeal muscular

dystrophy. Nat. Genet. 18, 164–167.

Calado, A., Kutay, U., Kuhn, U., Wahle, E., and Carmo-Fonseca, M. (2000).

Deciphering the cellular pathway for transport of poly(A)-binding protein II.

RNA 6, 245–256.

Davies, J.E., and Rubinsztein, D.C. (2011). Over-expression of BCL2 rescues

muscle weakness in a mouse model of oculopharyngeal muscular dystrophy.

Hum. Mol. Genet. 20, 1154–1163.

Davies, J.E., Wang, L., Garcia-Oroz, L., Cook, L.J., Vacher, C., O’Donovan,

D.G., and Rubinsztein, D.C. (2005). Doxycycline attenuates and delays toxicity

of the oculopharyngeal muscular dystrophy mutation in transgenic mice. Nat.

Med. 11, 672–677.

Di Giammartino, D.C., Nishida, K., and Manley, J.L. (2011). Mechanisms and

consequences of alternative polyadenylation. Mol. Cell 43, 853–866.

Fabian, M.R., Sonenberg, N., and Filipowicz, W. (2010). Regulation of mRNA

translation and stability by microRNAs. Annu. Rev. Biochem. 79, 351–379.

Filipowicz,W., Bhattacharyya, S.N., and Sonenberg, N. (2008). Mechanisms of

post-transcriptional regulation by microRNAs: are the answers in sight? Nat.

Rev. Genet. 9, 102–114.

Fu, Y., Sun, Y., Li, Y., Li, J., Rao, X., Chen, C., and Xu, A. (2011). Differential

genome-wide profiling of tandem 30 UTRs among human breast cancer and

normal cells by high-throughput sequencing. Genome Res. 21, 741–747.


Gomez-Benito, M., Loayza-Puch, F., Oude Vrielink, J., Odero, M.D., and

Agami, R. (2011). 30UTR-mediated gene silencing of the Mixed Lineage

Leukemia (MLL) gene. PLoS ONE 6, e25449.

Grimson, A., Farh, K.K., Johnston, W.K., Garrett-Engele, P., Lim, L.P., and

Bartel, D.P. (2007). MicroRNA targeting specificity in mammals: determinants

beyond seed pairing. Mol. Cell 27, 91–105.

Hu, J., Lutz, C.S., Wilusz, J., and Tian, B. (2005). Bioinformatic identification of

candidate cis-regulatory elements involved in human mRNA polyadenylation.

RNA 11, 1485–1493.

Jan, C.H., Friedman, R.C., Ruby, J.G., and Bartel, D.P. (2011). Formation,

regulation and evolution of Caenorhabditis elegans 30UTRs. Nature 469,

97–101.

Ji, Z., Lee, J.Y., Pan, Z., Jiang, B., and Tian, B. (2009). Progressive lengthening

of 30 untranslated regions of mRNAs by alternative polyadenylation during

mouse embryonic development. Proc. Natl. Acad. Sci. USA 106, 7028–7033.

Kedde, M., and Agami, R. (2008). Interplay between microRNAs and RNA-

binding proteins determines developmental processes. Cell Cycle 7, 899–903.

Kedde, M., Strasser, M.J., Boldajipour, B., Oude Vrielink, J.A., Slanchev, K., le

Sage, C., Nagel, R., Voorhoeve, P.M., van Duijse, J., Orom, U.A., et al. (2007).

RNA-binding protein Dnd1 inhibits microRNA access to target mRNA. Cell

131, 1273–1286.

Kedde, M., van Kouwenhove, M., Zwart, W., Oude Vrielink, J.A., Elkon, R., and

Agami, R. (2010). A Pumilio-induced RNA structure switch in p27-30 UTR

controls miR-221 and miR-222 accessibility. Nat. Cell Biol. 12, 1014–1020.

Kuhn, U., Gundel, M., Knoth, A., Kerwitz, Y., Rudel, S., and Wahle, E. (2009).

Poly(A) tail length is controlled by the nuclear poly(A)-binding protein regulating

the interaction between poly(A) polymerase and the cleavage and polyadeny-

lation specificity factor. J. Biol. Chem. 284, 22803–22814.

le Sage, C., Nagel, R., Egan, D.A., Schrier, M., Mesman, E., Mangiola, A., Anile,

C., Maira, G., Mercatelli, N., Ciafre, S.A., et al. (2007). Regulation of the

p27(Kip1) tumor suppressor by miR-221 and miR-222 promotes cancer cell

proliferation. EMBO J. 26, 3699–3708.

Lee, J.Y., Yeh, I., Park, J.Y., and Tian, B. (2007). PolyA_DB 2: mRNA polyade-

nylation sites in vertebrate genes. Nucleic Acids Res. 35, D165–D168.

Lemay, J.F., Lemieux, C., St-Andre, O., and Bachand, F. (2010). Crossing the

borders: poly(A)-binding proteins working on both sides of the fence. RNABiol.

7, 291–295.

Linhart, C., Halperin, Y., and Shamir, R. (2008). Transcription factor andmicro-

RNA motif discovery: the Amadeus platform and a compendium of metazoan

target sets. Genome Res. 18, 1180–1189.

Lutz, C.S. (2008). Alternative polyadenylation: a twist on mRNA 30 end forma-

tion. ACS Chem. Biol. 3, 609–617.

Mayr, C., and Bartel, D.P. (2009). Widespread shortening of 30UTRs by alterna-tive cleavage and polyadenylation activates oncogenes in cancer cells. Cell

138, 673–684.

Nunes, N.M., Li, W., Tian, B., and Furger, A. (2010). A functional human Poly(A)

site requires only a potent DSE and an A-rich upstream sequence. EMBOJ. 29,

1523–1536.

Ozsolak, F., Kapranov, P., Foissac, S., Kim, S.W., Fishilevich, E., Monaghan,

A.P., John, B., and Milos, P.M. (2010). Comprehensive polyadenylation site

maps in yeast and human reveal pervasive alternative polyadenylation. Cell

143, 1018–1029.

Ruegg, S., LehkyHagen,M., Hohl, U., Kappos, L., Fuhr, P., Plasilov, M.,Muller,

H., and Heinimann, K. (2005). Oculopharyngeal muscular dystrophy - an

under-diagnosed disorder? Swiss Med. Wkly. 135, 574–586.

Sandberg, R., Neilson, J.R., Sarma, A., Sharp, P.A., and Burge, C.B. (2008).

Proliferating cells express mRNAs with shortened 30 untranslated regions

and fewer microRNA target sites. Science 320, 1643–1647.

Shepard, P.J., Choi, E.A., Lu, J., Flanagan, L.A., Hertel, K.J., and Shi, Y. (2011).

Complex and dynamic landscape of RNA polyadenylation revealed by

PAS-Seq. RNA 17, 761–772.

Shi, Y., Di Giammartino, D.C., Taylor, D., Sarkeshik, A., Rice, W.J., Yates, J.R.,

3rd, Frank, J., and Manley, J.L. (2009). Molecular architecture of the human

pre-mRNA 30 processing complex. Mol. Cell 33, 365–376.

Takagaki, Y., andManley, J.L. (1998). Levels of polyadenylation factor CstF-64

control IgM heavy chainmRNA accumulation and other events associatedwith

B cell differentiation. Mol. Cell 2, 761–771.

Takagaki, Y., Seipelt, R.L., Peterson, M.L., and Manley, J.L. (1996). The poly-

adenylation factor CstF-64 regulates alternative processing of IgM heavy chain

pre-mRNA during B cell differentiation. Cell 87, 941–952.

Tavanez, J.P., Calado, P., Braga, J., Lafarga, M., and Carmo-Fonseca, M.

(2005). In vivo aggregation properties of the nuclear poly(A)-binding protein

PABPN1. RNA 11, 752–762.

Thome, F.S., Rodrigues, A.T., Bruno, R., Barros, E.J., and Goldani, J.C. (1997).

CAPD in southern Brazil: an epidemiological study. Adv. Perit. Dial. 13,

141–145.

Tian, B., Hu, J., Zhang, H., and Lutz, C.S. (2005). A large-scale analysis of

mRNA polyadenylation of human and mouse genes. Nucleic Acids Res. 33,

201–212.

van Kouwenhove, M., Kedde, M., and Agami, R. (2011). MicroRNA regulation

by RNA-binding proteins and its implications for cancer. Nat. Rev. Cancer 11,

644–656.

Wahle, E. (1995). Poly(A) tail length control is caused by termination of proces-

sive synthesis. J. Biol. Chem. 270, 2800–2808.

Wahle, E., and Keller, W. (1994). RNA Processing: A Practical Approach,

Volume 2, S.J. Higgins and B.D. Hames, eds. (Oxford, UK: Oxford University

Press), pp. 1–34.

Zhang, X., Virtanen, A., and Kleiman, F.E. (2010). To polyadenylate or to

deadenylate: that is the question. Cell Cycle 9, 4437–4449.


Supplemental Information

EXTENDED EXPERIMENTAL PROCEDURES

Constructs903 bp of the MLL-30 UTR (240–1143 bp fragment) were cloned into the psiCHECK-2 luciferase reporter (Promega) downstream to

the Renilla luciferase (= pSTAR), using XhoI/SalI and NotI restriction sites.

For the pSTAR-D2WT luciferase reporter construct, 266 bp of the proximal CCND2 PAS region were fused to 903 bp of the MLL 30

UTR (240–1143 bp fragment) and cloned downstream to the Renilla luciferase in the psiCHECK2 vector (Promega) using XhoI/SalI

and NotI restriction sites.

For the pSTAR-p27WT construct the p27 distal PAS region was introduced into pSTAR, using the QuikChange II Site-Directed

Mutagenesis kit (Agilent Technologies).

For the pSTAR-DICER1WT luciferase reporter construct, 278 bp of the proximal DICER1 PAS region were fused to 903 bp of the

MLL 30 UTR (240–1143 bp fragment) and cloned downstream to theRenilla luciferase in the psiCHECK2 vector (Promega) using XhoI/

SalI and NotI restriction sites.

For the pSTAR-FGF2WT luciferase reporter construct, 282 bp of the proximal FGF2 PAS region were fused to 903 bp of the MLL 30

UTR (240–1143 bp fragment) and cloned downstream to the Renilla luciferase in the psiCHECK2 vector (Promega) using XhoI/SalI

and NotI restriction sites.

The luciferase constructs containing the mutated CCND2, p27, DICER1, and FGF2 PAS signals were generated with site-directed

mutagenesis (Agilent Technologies).

For the PABPN1 WT expression plasmid the PAPBN1 coding region was cloned into pcDNA5-FRT-V5-HIS TOPO TA expression

vector (Invitrogen).

The PABPN1R #1 and PAPBN1R #2 resistant mutant expression vectors were generated with site-directed mutagenesis (Agilent

Technologies).

For the construction of miRVec-17-19, we used the described miRVec-17&18, miRVec-18,19a&20 and miRVec-19b&92

(Voorhoeve et al., 2006). miRVec-17-19 was cloned by PvuII digestion of miRVec-17&18, PvuII/NheI digestion of miRVec-

18,19a&20 and NheI digestion of miRVec-19b&92. Fragments were ligated and used as template for PCR (primer: ‘‘17’’ fw; ‘‘92’’ rev).

This PCR fragment was subsequently used as template for PCR with ‘‘17’’ fw and ‘‘19’’ rev primers followed by BamHI/EcoRI

cloning into miRVec. The expression of miR-17, -18, and -19 were confirmed by RNase Protection Assay (data not shown).

The full-length 30 UTR of CCND1 and its short form were cloned into the psiCHECK2 luciferase reporter (Invitrogen) downstream of

the Renilla luciferase using XhoI/SalI and NotI restriction sites.

The cloning of psi-C-p27-30UTR and psi-C-p27-30UTR-DMhas been described previously (Kedde et al., 2010; le Sage et al., 2007).

The cloning of pHM6-PABPN1(A17) and pEGFP-C1-PABPN1(A17) has been described previously (Bao et al., 2002).

50-30 primer sequences are listed in Table S6.

ImmunoblottingWhole-cell extracts were separated on 10% SDS-PAGE gels and transferred to Immobilon-P membranes (Millipore). The following

antibodies were used: Anti-PABPN1 (2428-1; Epitomics), anti-CCND1 (sc-718; Santa Cruz Biotechnology), anti-CCND2 (sc-181

[C-17]; Santa Cruz Biotechnology), anti-b-Actin (ab8227; Abcam), and anti-GFP (Roche Applied Science) were used as primary

and either swine anti-rabbit or goat anti-mouse horseradish peroxidase–conjugated IgG (Dako) as secondary antibodies.

Western blots were developed with enhanced chemiluminescence (ECL; Amersham Biosciences) and exposed to film (Kodak)

or developed with the Molecular Imager ChemiDoc XRS System (Bio-Rad Laboratories). Densitometric analysis: Image lab 3.0.1

(Bio-Rad Laboratories).

qRT-PCRTotal RNAwas extracted using the RNeasyMini kit (QIAGEN) or Trizol (Invitrogen). cDNAwas synthesized with SuperScript III reverse

transcriptase (Invitrogen) and oligo(dT) primers. For miRNA qRT-PCR, total RNA was extracted using Trizol (Invitrogen). cDNA was

synthesized with random primers (Taqman High Capacity cDNA kit; Applied Biosystems). qRT-PCR was performed using SYBR

Green PCRmaster mix (Applied Biosystems), Light Cycler 480 Sybr Green I Master mix (Roche) or TaqMan UNGmaster mix (Applied

Biosystems) and the Chromo 4 system (Bio-Rad Laboratories). 50-30 oligonucleotide sequences are listed in Table S6. qRT-PCR

primers used for the detection of miR-20a and GAPDH were purchased from Applied Biosystems.

R-EMSA50-30 oligonucleotide sequences are listed in Table S6.

In Vivo Binding50-30 oligonucleotide sequences are listed in Table S6. The SUPT6H, MAPKAPK3, and hnRNPUL2 prox reverse primers were de-

signed to bind downstream of the proximal CS to exclude detection of PABPN1 bound to proximal poly(A) tails.

Cell 149, 538–553, April 27, 2012 ª2012 Elsevier Inc. S1

In Vitro Cleavage AssayAssays were performed as described by Wahle and Keller (1994), using 50-30 oligonucleotide sequences listed in Table S6.

ChIP AssayCells were transfected with control or PABPN1 #2 siRNA and after 72 hr crosslinked with 1% formaledehyde for 10 min. Crosslinking

was stopped with 0.125 M glycine for 5 min at RT. Cells were washed with ice-cold 13 PBS, resuspended and incubated with cell

lysis buffer (50 mM Tris-HCl [pH 8.0], 85 mM KCl, 0.5% Nonidet P-40, PI; Roche) for 10 min. Nuclear pellets were spun down at

5,000 rpm for 5 min, resuspended in nuclear lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris HCl [pH 8.0]), and incubated for

10 min at 4�C. Chromatin was sonicated for 30min (pulses every 30 s) and centrifuged at 14,000 rpm for 10 min to remove the debris.

Lysates were incubated with 5 mg RNA Pol-II (Millipore) or IgG antibody (Santa Cruz) overnight at 4�C and then incubated with 40 mg

agarose protein G beads (GE Healthcare) for 1 hr at 4�C. Beads were washed five times. DNA complexes were eluted and

reverse crosslinked in 0.3MNaCl at 65�C for 4 hr. Protein was digested with proteinase K at 45�C for 1 hr, genomic DNA was purified

(QIAGEN PCR columns), and enrichment of RNA Pol II was shown by qRT-PCR. 50-30 oligonucleotide sequences are listed in Table

S6.

30-End qRT-PCRcDNA was synthesized using the P7-t25-vn oligo-dT primer with the following sequence: 50-CAAGCAGAAGACGGCATACGA

GATTTTTTTTTTTTTTTTTTTTTTTTTVN-30.Subsequently, qRT-PCR was performed using P7-primer and 50-30 oligonucleotide sequences listed in Table S6.

Northern BlotTwo micrograms of mRNA was separated on a 1.5% formaldehyde-denaturing agarose gel and transferred onto nylon membrane

(Hybond-N+; AmershamBiosciences). After UV crosslinking, themembranewas hybridized using RNAprobes in NorthernMax buffer

(Ambion). Riboprobes were synthesized (miRNA probe construction kit; Ambion). 50-30 oligonucleotide sequences are listed in

Table S6.

Construction of 30Seq LibrariesTotal RNA was extracted (RNeasy Mini kit; QIAGEN) and mRNA selected (Oligotex mRNA Kit; QIAGEN). Five hundred nanograms of

mRNA was heat fragmented (fragmentation buffer; Ambion) for 5 min at 70�C and precipitated, and first strand cDNA was synthe-

sized (SuperScript III reverse transcriptase; Invitrogen; P7-t25-vn oligo-dT primer). Single-strand cDNA was converted to double-

strand cDNA (Invitrogen) and end repaired for 30 min at 20�C (NEBNext End Repair Enzym Module; New England Biolabs; supple-

mented with 1 ml Klenow DNA polymerase; Invitrogen). After purification (QIAquick PCR Purification Kit; QIAGEN), 1 ml 45 mM an-

nealed P5-splinkerette was ligated overnight at 16�C to the cDNA. The P5-splinkerette-ligated cDNA was purified and size selected

for �220 bp long fragments (E-Gel iBase Power System; E-Gel Size select 2% gels; Invitrogen). The size-selected fraction was PCR

amplified to generate the final 30Seq library with P5 primer, P7 primer and Phusion Hot Start DNA polymerase (Finnzymes). After a last

purification (QIAquick PCR Purification Kit; QIAGEN) the 30Seq library was sequenced on the Illumina Genome Analyzer (U2OS si-

PABPN1 and control samples) or the Illumina HiSeq system (RPE-1 and U2OS tPABPN1(A17) samples and all mouse samples). 50-30 oligonucleotide sequences are listed in Table S6. The size-selected fraction was PCR amplified to generate the final 30-Seq library

using P5 primer and P7 primer and Phusion Hot Start DNA polymerase. PCR steps were: 98�C for 30 s; 18 cycles of 98�C for 10 s,

65�C for 30 s, 72�C for 30 s; 72�C for 5 min.

Confirmation of 30Seq Results with qPCRTo confirm the results gained from the 30-Seq, qPCR with the 30Seq libraries was performed using SYBR Green PCR master mix

(Applied Biosystems), with the P7 primer and the 50-30 oligonucleotide sequences listed in Table S6.

Analysis of 30Seq DataSequenced reads were aligned against the human (hg18) and mouse (mm9) genomes using Bowtie (Langmead et al., 2009). Up to

two mismatches were allowed in the reads’ seed region (the reads’ first 28 nt). To allow the alignment of reads which span CSs and

thus contain the start of the untemplated poly(A) tail, Bowtie’s -e parameter was increased to allow mismatches in all bases after the

seed region. Only uniquely mapped reads were used in subsequent analysis. Numbers of total and uniquely mapped reads are

provided in Table S4. Raw sequence data and wig files that summarize read coverage along the chromosomes (normalized to

10 M mapped reads) are available at GEO DB (accession number GSE27452). To map reads to genes and genomic regions (e.g.,

30 UTRs, introns, CDSs, etc.), gene coordinates and annotations were extracted from the human and mouse Ensembl-Gene tables

of the UCSC browser (Rhead et al., 2010). To cover novel CSs located downstream to current transcript annotations, we extended

the 30 UTR of each transcript by 1,000 bp.

S2 Cell 149, 538–553, April 27, 2012 ª2012 Elsevier Inc.

Mapping of Poly(A) CSsTo identify poly(A) CSs we searched for uniquely mapped reads which contained untemplated stretch of As (Figure S2B). To identify

CSs while reducing false calls that stem from priming of the oligo-dT primer to internal A-rich regions within transcripts, we required

that to support a CS, reads should contain a stretch of at least 8 As, and that at least 5 of the first 8 As in the stretch shouldmis-match

the corresponding bases on the transcript reference sequence. The location of the cleavage was taken as the location where the A

stretch started. Since the location of the CS often fluctuated around a major site, for each gene and sample, we calculated a ‘‘CS

profile’’ that recorded the number of reads supporting a cleavage at each position along the transcript (Figure S2C). We considered

the local maxima of these CS profiles as the CS locations, and required spacing of at least 50 nt between consecutive CSs (in case of

lower spacing between CSs, the stronger, that is, the one supported by a higher number of reads, was chosen). Only CSs supported

by at least 10 readswere considered in subsequent analyses. A total of 23,635, 28,263, and 10,849 CSswere identified in the analysis

of the U2OS, RPE-1 andmouseOPMDdatasets, respectively. The genomic coordinates of these CSs are provided in Table S5. 58%,

43%, and 55% of the CSs in these three datasets were mapped to 30 UTRs.

Differential Usage of CSsIn order to identify significant shifts in CS usage between samples, we first identified regions of ‘‘read peaks’’ along transcripts.

For this goal we first divided each chromosome into 5 nt bins, and counted how many reads started at each bin. To define peak

boundaries, which are common across all analyzed samples in a dataset, we took the union of the bin counts and calculated the

maximum count at each bin over the samples. Peaks were defined as intervals of consecutive bin runs with maximum read count

of least 3 reads (Figure S2J). Adjacent peaks with distance below 60 nt were merged. Next, we counted the number of reads in

each peak in each sample (Figure S2K). To statistically detect transcripts that showed a significant change in PAS usage between

the compared samples, for each peak in a transcript, we summed up the number of counts measured in each sample. Only peaks

mapping to 30 UTRswere considered in the analysis. Next we tested each transcript, which containedmore than one 30 UTR cleavage

site for relationship between peak levels (that is, CS usage) and condition (that is, control or PABPN1-manipulated) using chi-square

tests (Figure S2L). We used p = 0.001 as a cut-off for significance. All tests were carried out using R (http://www.r-project.org/).

To determine whether shift was toward proximal or distal CS, for each transcript in each condition, we calculated a weighted

mean of CS index (weighting each CS [j] according to its reads coverage [wj]):

hCS Ji=Xn

j =1

wj 3 j:

A decrease in < CS_J > indicates 30 UTR shortening and vice versa.

PUIWe defined the PUI in order to measure, per condition, the relative usage of each 30 UTR CS within a transcript. For a given transcript

with N 30 UTR CSs, the PUI of the jth CS is defined by:

PUIj = log2

�Ej

hEi�; j = 1::N;

where Ej is the level of the peak associated with the jth cleavage site, and < E > is the geometric mean of the levels of all the peaks

associated with the N cleavage sites. PUI distributions calculated over proximal CSs were compared between samples using the

nonparametric Wilcoxon test. All transcripts which showed significant APA shift between the compared samples were included in

this analysis. As a global measure for the usage of proximal CSs in a sample, we took the median of the distribution of proximal

site PUIs in that sample, taking into account all transcript with more than one 30 UTR CS.

Motif AnalysisSequences surrounding the CSs (±50 nt) were extracted from the human and mouse genomes using the Galaxy utility (Woollard,

2010). Second-order Markov random sequences were generated based on frequencies observed in the real sequences. Enriched

sequence motifs in CS regions were sought using the Amadeus package (Linhart et al., 2008). In addition, all possible 6-mers

were examined for enrichment in the regions [�45, �5] and [0, 40] upstream and downstream to the location of the CS, using

geometric distribution tail.

Gene Expression Analysis30Seq allows not only to identify CSs and quantify their usage but also to obtain estimates for gene expression levels. As an estimate

for expression level we took for each gene the total number of readsmapping to its 30UTR. Quantile normalization was applied before

comparing expression levels between samples.


http://www.r-project.org/

SUPPLEMENTAL REFERENCES

Bao, Y.P., Cook, L.J., O’Donovan, D., Uyama, E., and Rubinsztein, D.C. (2002). Mammalian, yeast, bacterial, and chemical chaperones reduce aggregate forma-

tion and death in a cell model of oculopharyngeal muscular dystrophy. J. Biol. Chem. 277, 12263–12269.

Kedde, M., van Kouwenhove, M., Zwart, W., Oude Vrielink, J.A., Elkon, R., and Agami, R. (2010). A Pumilio-induced RNA structure switch in p27-30 UTR controls

miR-221 and miR-222 accessibility. Nat. Cell Biol. 12, 1014–1020.

Langmead, B., Schatz, M.C., Lin, J., Pop, M., and Salzberg, S.L. (2009). Searching for SNPs with cloud computing. Genome Biol. 10, R134.

le Sage, C., Nagel, R., Egan, D.A., Schrier, M., Mesman, E., Mangiola, A., Anile, C., Maira, G., Mercatelli, N., Ciafre, S.A., et al. (2007). Regulation of the p27(Kip1)

tumor suppressor by miR-221 and miR-222 promotes cancer cell proliferation. EMBO J. 26, 3699–3708.

Linhart, C., Halperin, Y., and Shamir, R. (2008). Transcription factor andmicroRNAmotif discovery: the Amadeus platform and a compendium of metazoan target

sets. Genome Res. 18, 1180–1189.

Rhead, B., Karolchik, D., Kuhn, R.M., Hinrichs, A.S., Zweig, A.S., Fujita, P.A., Diekhans, M., Smith, K.E., Rosenbloom, K.R., Raney, B.J., et al. (2010). The UCSC

Genome Browser database: update 2010. Nucleic Acids Res. 38, D613–D619.

Voorhoeve, P.M., le Sage, C., Schrier, M., Gillis, A.J., Stoop, H., Nagel, R., Liu, Y.P., van Duijse, J., Drost, J., Griekspoor, A., et al. (2006). A genetic screen impli-

cates miRNA-372 and miRNA-373 as oncogenes in testicular germ cell tumors. Cell 124, 1169–1181.

Wahle, E., and Keller, W. (1994). RNA Processing: A Practical Approach, Volume 2, S.J. Higgins and B.D. Hames, eds. (Oxford, UK: Oxford University Press),

pp. 1–34.

Woollard, P.M. (2010). Asking complex questions of the genome without programming. Methods Mol. Biol. 628, 39–52.


Figure S1. A Reporter System Identifies PABPN1 as a Regulator of APA, Related to Figure 1

(A)We cloned DNA inserts encompassing PAS sequences in between a repressor 30 UTR (MLL-30UTR, designated R) and theRenilla luciferase gene (hRluc) of the

psiC luciferase vector (construct named pSTAR).

(B) WT andmut PAS regions of CCND2 that were used in this study (upper panel). The sequence of transcript ends downstream of the Renilla gene in the pSTAR-

D2WT luciferase construct was analyzed by sequencing and indicates the expected proximal CS (lower panel).

(C) psiC and pSTAR vectors were transfected into U2OS cells. Seventy-two hours later, the cells were harvested and Renilla and Firefly activities were measured.

The relative Renilla and Firefly counts from psiC and pSTAR are depicted as average ± SD.

(D) WT and mut PAS regions of p27 that were used in this study.

(E) pSTAR-D2WT and pSTAR-D2mut were transfected into U2OS cells. The next day the cells were reverse transfected with siRNA smart pools from Dharmacon

designed to target 489 RBP genes in a 384-well format. Seventy-two hours later the cells were harvested and theRenilla and Firefly luciferase activities measured.

(F) TheRenilla and FireflymRNA levels of pSTAR-D2WT (D2-PASWT) and pSTAR-D2mut (D2-PASmut) in control and PABPN1 kd samples weremeasured by qRT-

PCR and depicted as R/F ratios (mean ± SEM; one-tailed t test).

(G) The same experiment was performed as in (C). This time either pSTAR-FGF2WT (FGF2-PAS) and pSTAR-FGF2mut (FGF2-PASmut, left panel) or DICER1-PAS

(pSTAR-DICER1WT) and DICER1-PASmut ((pSTAR-DICER1mut, right panel) were used and R/F ratios are depicted as average ± SD.


Figure S2. Quantification with 30Seq of the Effect of PABPN1 kd on CS Usage, Related to Figure 2

(A) Overview of the 30Seq procedure used (adapted from Beck et al., 2010, for details see Extended Experimental Procedures).

(B) 30Seq allows precise mapping of poly(A) CSs by identifying sequenced reads that align to the 30 ends of transcripts, span the CSs and therefore cover the

beginning of untemplated poly(A) tails. The panel shows an example of such reads that map to the 30 end of E2F1. The canonical PAS upstream the CS is also

indicated.

(C) CS profiles were calculated for each transcript, representing the number of reads supporting cleavage at each position along the transcript. Local maxima of

these profiles were considered as the CSs. Only CSs supported by at least 10 reads and separated by at least 50 nt from each other were considered in the

analysis (see Extended Experimental Procedures). The figure shows the two CSs identified in the 30 UTR of PSMD8.

(D) Distribution of the 23,635 CSs identified in the U2OS 30Seq dataset into genomic regions.

(E) Distribution of the number of 30 UTR CSs per transcript. Thirty-two percents of the transcripts contain more than one CS in their 30 UTRs.(F) Overlap between the CSs identified in our U2OS 30Seq dataset and two comprehensive resources on CSs data in human: Ozsolack et al. (Ozsolak et al., 2010)

and poly(A) database (DB) (Lee et al., 2007). CSs were regarded as overlapping if their distance was up to 10 nt. The degree of overlap was significantly higher for

the CSs which mapped to 30 UTRs.(G) The canonical PAS and its major variants were significantly enriched in the [�40 pb,�5 bp] region relative to the CSs. Enrichment was calculated against 2nd-

order markov random sequences. (Enrichment Factor is the ratio between the signal prevalence in the true and random sequences; p values were calculated

using hypergeometric tail).

(H) Location distribution of the top six PASs with respect to CS position. The distribution of the canonical signal (AATAAA) and most of its variants showed a very

sharp peak at �20 nt upstream of the CS. The prevalence of each signal is indicated. The numbers reported here are different than in panel G as signals can

appear several times in each CS and all appearances were counted here.

(I) De novo motif discovery analysis was done using the AMADEUS tool (Linhart et al., 2008). In addition to the canonical PAS and its variants, AMADEUS also

identified T- and TG-rich motifs whose location distribution peaked downstream to the CSs.

(J) ‘‘Read peaks’’ were defined along transcripts. The number of reads in each peak indicates the strength of the usage of its corresponding PAS and CS. As an

example the figure shows the two peaks (proximal and distal) identified in the 30 UTR of E2F1.

(K) Number of reads measured for the proximal and distal peaks of E2F1 in each of the control and PABPN1 kd samples in U2OS cells.

(L) Chi-square tests were applied to identify significant shifts in CS usage between samples. Here shown as an example, the test for relationship between PAS

usage (proximal or distal) and condition (control or PABPN1 kd) in the case of E2F1 30 UTR.(M) Northern blot analysis of control (C) and PABPN1 #1 and #2 kd samples. The ratios of short and long 30 UTR isoforms of CHMP1A, MAPKAPK3, hnRNPUL2,

PSMD8, PDRG1, and CCND1 have been measured. The table on the right side depicts the short isoform ratios of siRNA control and siRNA PABPN1 #2.

(N) Immunoblot analysis (PABPN1 and control b-actin protein) to confirm PABPN1 kd in RPE-1 cells transfected with control and PABPN1 #2 siRNAs.

(O) Examples of three genes that showed in RPE-1 cells significant increase in proximal CS levels relative to the distal sites in the PABPN1 kd samples compared

to control ones.

(P) Comparison of proximal PUI distribution between PABPN1 kd and control samples measured in RPE-1 cells (for a more detailed explanation, see legend of

Figures 2D and 2E).

(Q) In 82% of the cases of significant CS shift between PABPN1 kd and control samples, the shift was toward proximal CSs in the PABPN1 kd sample (reflecting

extensive 30 UTR shortening).


Figure S3. Related to Figure 3

(A) 30-end qRT-PCR was applied to measure the stability of transcript isoforms determined at the proximal PAS (shorter 30 UTR) of genes which showed a

significant shift toward proximal CS in the PABPN1 kd samples. In none of the cases PABPN1 kd caused any change in stability. Data are shown as average ±SD.

(B) No global change in expression of 30-end processing genes was observed in PABPN1 kd samples compared to control.

(C) Primer setting used for measurement of in vivo binding of PABPN1 near hnRNPUL2, SUPT6H, and MAPKAPK3 proximal CS in U2OS cells by RNA immu-

noprecipitation.

(D) Same exposure as shown in Figure 3I, only that the full probe band is presented.

(E) PAS regions used as templates for the cleavage assay presented in Figure 3J. In red is the T7 promoter, in green the PAS sequence, and CS is marked with

a triangle.



(A) A schematic drawing of the miRVec-17-19 construct.

(B) A representative of raw data used for calculating the bars that correspond to siCtrl in Figure 4H. R/F ratios of miRVec-Ctrl and miRVec-17-19 (in case of

CCND1) and R/F ratios of miRVec-Ctrl. andmiRVec-222 (in case of p27) are presented (as average ±SD), respectively. Note that psiC-D1-FL and p27-30 UTR-WT

show lower R/F levels and higher sensitivity to miR-mediated repression compared to psiC-D1-S and p27-30 UTR-DM, respectively.

(C) A screenshot of the UCSC browser showing p27-30 UTR with only one single major CS.



Comparison of the change in expression level in trePABPN1(A17) versus control cells between two sets of transcripts: a target set which contained the geneswhich

showed significant shift toward proximal CSs in the trePABPN1(A17) sample and a background (BG) set which contained the rest of the genes measured in our

dataset (p value was calculated using Wilcoxon test).


Figure S6. Characterization of Signals in the Region of CSs Identified in the Murine 30Seq Dataset, Related to Figure 6

(A) De novomotif discovery identified two highly enrichedmotifs in the [+50,�50] region encompassing the CSs relative to 2nd-order Markov sequences. The top

motif corresponds to the canonical PAS (AATAAA) and its major variant (ATTAAA); the 2nd motif corresponds to U/UG-rich downstream elements (DSEs).

(B) Location distribution of the canonical PAS and its major variants with respect to CS position.

(C) 30 UTR shortening tendency was observed also for the 20-week-old trePABPN1(A17) mice (binomial tail).

(D and E) Enhanced usage of proximal CSs in 20-week-old trePABPN1(A17) mice reflected by increase in proximal PUI median (panel D; three independent OPMD

and WT mice were examined) and proximal PUI distribution (panel E) (Wilcoxon test).

(F) Same as Figure 6E, but here for 20-week-old mice.


The Poly(A)-Binding Protein Nuclear 1 Suppresses ... · The Poly(A)-Binding Protein Nuclear 1...

Documents

Transcript of The Poly(A)-Binding Protein Nuclear 1 Suppresses ... · The Poly(A)-Binding Protein Nuclear 1...