Latent Enhancers Activated by Stimulation in ...Latent Enhancers Activated by Stimulation in...

Latent Enhancers Activatedby Stimulation in Differentiated CellsRenato Ostuni,1,2,* Viviana Piccolo,1,2 Iros Barozzi,1,2 Sara Polletti,1,2 Alberto Termanini,1 Silvia Bonifacio,1

Alessia Curina,1 Elena Prosperini,1 Serena Ghisletti,1 and Gioacchino Natoli1,*1Department of Experimental Oncology, European Institute of Oncology (IEO), Via Adamello 16, 20139 Milan, Italy2These authors contributed equally to this work

*Correspondence: [email protected] (R.O.), [email protected] (G.N.)http://dx.doi.org/10.1016/j.cell.2012.12.018

SUMMARY

According to current models, once the cell hasreached terminal differentiation, the enhancer reper-toire is completely established and maintained bycooperatively acting lineage-specific transcriptionfactors (TFs). TFs activated by extracellular stimulioperate within this predetermined repertoire, landingclose to where master regulators are constitutivelybound. Here, we describe latent enhancers, definedas regions of the genome that in terminally differenti-ated cells are unbound by TFs and lack the histonemarks characteristic of enhancers but acquire thesefeatures in response to stimulation. Macrophagestimulation caused sequential binding of stimulus-activated and lineage-determining TFs to theseregions, enabling deposition of enhancer marks.Once unveiled, many of these enhancers did notreturn to a latent state when stimulation ceased;instead, they persisted and mediated a faster andstronger response upon restimulation. We suggestthat stimulus-specific expansion of the cis-regula-tory repertoire provides an epigenomic memory ofthe exposure to environmental agents.

INTRODUCTION

Mammals contain about 200 specialized cell types whose

distinct transcriptional outputs reflect the selection during devel-

opment of unique combinations of coding and regulatory ele-

ments. Indeed, the repertoire of cis-regulatory elements active

in each distinct lineage, as identified by chromatin marks or

occupancy by transcriptional regulators (Barski et al., 2007;

Heintzman et al., 2007; Wang et al., 2008), is only a small fraction

of all possible genomic regulatory elements, with minimal over-

lap between different cell types (Ernst et al., 2011; Heintzman

et al., 2009; Visel et al., 2009).

Cell-type-specific usage of the cis-regulatory information

reflects the combinatorial activity of sequence-specific tran-

scription factors (TFs) involved in lineage determination and

maintenance of cell identity (Natoli, 2010). Enhancers active in

a given tissue are typically co-occupied by multiple TFs enforc-

ing cell-type-specific gene expression programs (Chen et al.,

2008; He et al., 2011; Junion et al., 2012; Siersbæk et al.,

2011; Song et al., 2011; Tijssen et al., 2011). Shaping this

genomic cis-regulatory landscape entails multiple TF activities,

including recruitment of chromatin modifiers, nucleosome re-

modelers, and histone chaperons (Hu et al., 2011; Ram et al.,

2011), eventually resulting in the demarcation of short DNA

stretches that contain accessible TF binding sites (TFBS) brack-

eted by positioned nucleosomes (He et al., 2010; Schones et al.,

2008). Enhancers are marked by monomethylated K4 on the

histone H3 N-terminal tail (H3K4me1) (Heintzman et al., 2007,

2009) and have been classified into poised and active enhancers

based on the absence or presence of histone acetylation, re-

spectively (Creyghton et al., 2010; Rada-Iglesias et al., 2011).

Each cell-type-specific enhancer repertoire provides a unique

cis-regulatory context in which TFs activated by environmental

stimuli operate to modify gene expression. For instance, STAT1

was recruited to preexisting H3K4me1-marked enhancers in

HeLa cells within minutes after interferon gamma (IFNg) stimula-

tion (Heintzmanet al., 2009). Inmammary epithelial cells estrogen

receptor alpha (ERa) was recruited to accessible chromatin sites

established by the forkhead TF FoxA1 (Hurtado et al., 2011).

FoxA1 was also found to organize the enhancer repertoire of

prostate cancer cells, thus directing androgen receptor (AR)

recruitment to predefined genomic sites (Wang et al., 2011).

Transforming growth factor beta (TGFb) stimulation of different

cell types resulted in the recruitment of Smad2/3 to distinct

sets of regulatory elements determined by cell-type-specific

TFs (Mullen et al., 2011). Finally, recruitment of stimulus-induced

TFs such as NF-kB and Lxr in activated macrophages occurs at

cell-type-specific enhancers controlled by the master regulator

Pu.1 (Ghisletti et al., 2010; Heinz et al., 2010; Natoli et al., 2011).

So far, the prevailing concepts are that the enhancer repertoire

represents a predetermined landscape generated and enforced

by the same TFs that control cell identity, and that all transcrip-

tional regulatory events in the differentiated cell occur within this

landscape. This model implies that external stimuli that activate

a transient response cannot change the repertoire of genomic

regulatory elements but solely act on the one that is already

available. Clearly, this represents a homeostatic mechanism

that guarantees robustness in maintenance of cell identity in

spite of a changing environment.

Nevertheless, the behavior of some cell types can be persis-

tently affected by transient stimuli to the point that the resulting

Cell 152, 157–171, January 17, 2013 ª2013 Elsevier Inc. 157

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.cell.2012.12.018

http://crossmark.dyndns.org/dialog/?doi=10.1016/j.cell.2012.12.018&domain=pdf

Figure 1. Latent Enhancers Unveiled by LPS Stimulation of Macrophages

(A) Classification of macrophage enhancers based on their response to LPS stimulation. Numbers refer to the 24 hr time point.

(B) A representative genomic region showing latent enhancers induced by LPS stimulation.

(C) Heatmap showing the dynamic behavior of histone marks and Pu.1 at latent enhancers induced by a 4 hr or 24 hr LPS treatment.

(legend continued on next page)

158 Cell 152, 157–171, January 17, 2013 ª2013 Elsevier Inc.

cell, while retaining its identity, acquires new functional proper-

ties and a corresponding transcriptional output, which together

can define a new cellular subtype. This phenomenon, often

indicated as plasticity, is well known in cells of the myeloid

lineage such as macrophages, neutrophils, and dendritic cells,

whose exposure to different stimuli results in subtypes with

specific properties (Biswas and Mantovani, 2010; Fridlender

et al., 2009; Lawrence and Natoli, 2011).

In this study, we addressed the hypothesis that stimulation-

induced plastic changes in cell behavior may be associated

with a partial reprogramming of the available cis-regulatory infor-

mation. To this aim we carried out a systematic analysis of the

impact of stimuli affecting macrophage properties and transcrip-

tional outputs (Mosser and Edwards, 2008) on their enhancer

repertoire. We classifiedmacrophage enhancers in several cate-

gories based on their response to distinct stimuli, thus deter-

mining general principles regulating the dynamic usage of the

genomic regulatory information. As expected, a large number

of pre-existing, poised enhancers were specifically activated in

response to distinct stimuli, and in a reciprocal fashion many

active elements were repressed upon stimulation. Each stimulus

also allowed the surfacing of a distinct subset of latent

enhancers, namely genomic regulatory elements that were

unbound by TFs and unmarked in unstimulated cells. These

enhancers recruited the macrophage master regulator Pu.1

only in response to stimulation and conditionally to the recruit-

ment of stimulus-activated TFs such as STAT1 and STAT6,

thus indicating a mechanisms of enhancer induction and ac-

tivity that is radically different from that of classical enhancers

established during differentiation by cooperative binding of

lineage-determining TFs. Washout of the stimulus resulted in

the rapid loss of acetylation and Pu.1 occupancy, whereas

residual H3K4me1 was sustained and associated with a faster

and stronger induction upon restimulation, thus providing an

epigenomic memory of the initial stimulation.

RESULTS

Latent Enhancers Are a Distinct Class of cis-RegulatoryElementsTFs activated by macrophage stimulation with a canonical

inflammatory agent (lipopolysaccharide [LPS]), such as NF-kB

and IRFs, land at regulatory elements predefined by Pu.1 and

constitutively marked by H3K4me1 (Barish et al., 2010; Escou-

bet-Lozach et al., 2011; Ghisletti et al., 2010; Heinz et al.,

2010). Although this is the most common occurrence, these

data do not rule out the possibility that stimulation may also

modify the pre-existing regulatory landscape. We therefore

generated a panel of ChIP-seq data sets exploring the regulatory

repertoire of mouse bone marrow-derived macrophages stimu-

lated with LPS for 4 and 24 hr.

Data obtained with each antibody at different time points

showed higher correlation (e.g., R2 = 0.936 considering

(D) Proximity of latent enhancers to LPS-inducible genes.

(E) GO categories of genes assigned to LPS-activated latent enhancers (as infer

(F) In vivo induction of two latent enhancers in LPS-induced peritoneal macroph

See also Figure S1 and Tables S1, S2, and S3.

H3K4me1 in untreated and 4 hr LPS-treated cells) compared

to data obtained with a different antibody (e.g., R2 � 0.1 when

comparing H3K4me1 with H3K4me3 in untreated cells), indi-

cating high reproducibility. Overall, response to stimulation

was highly dynamic, allowing the classification of �65,000

regulatory elements in multiple classes (Figure 1A; Table S1

available online). Constitutive enhancers were associated with

both H3K4me1 and H3K27Ac in unstimulated cells. Of the

14,902 constitutive enhancers we detected, 2,261 showed

increased acetylation after a 24 hr LPS treatment (constitutive-

activated enhancers; Table S1). Poised enhancers showed basal

H3K4me1 without H3K27Ac. Most of them (37,629) were unaf-

fected by LPS stimulation at either time point, whereas a com-

paratively small subset (poised-activated enhancers) acquired

H3K27Ac (3,904 at 4 hr and 3,045 at 24 hr). Poised enhancers

unaffected by stimulation likely represent a heterogeneous set

that may also include constitutively repressed elements (Bonn

et al., 2012). Repressed enhancers (11,150 at 4 hr and 9,004

at 24 hr) displayed constitutive H3K27Ac that was either lost

or greatly reduced upon stimulation (in most of these cases

H3K4me1 was unaffected).

The most unexpected behavior was observed at genomic

regions that were unmarked in untreated macrophages and

acquired enhancer features upon LPS stimulation (Figures 1A–

1C; Table S1). We named these putative regulatory elements,

whose dynamic behavior differed from all previously identified

enhancer classes, latent enhancers to highlight the notion that

they are inactive, unbound, and unmarked in the basal state,

being selectively unveiled by stimulation.

Latent enhancers were identified on the basis of two restric-

tive criteria: (1) the lack of H3K4me1, H3K27Ac, and Pu.1 in un-

stimulated cells even when using relaxed statistical thresholds

(see Extended Experimental Procedures); and (2) the presence

of an LPS-induced H3K4me1 peak using a restrictive threshold

(p < 10�10 using MACS). Although the induction of H3K27Ac

was not included among the criteria (because acetylation is

not invariably detected at enhancers), in the overwhelming

majority of cases acquisition of H3K4me1 was accompanied

by the appearance of H3K27Ac (Figures 1B and 1C). Fre-

quently, histone acetylation was only transiently increased at

4 hr and then returned to low-to-undetectable levels at 24 hr

(see Figure 1B and the upper area of Figure 1C). At the same

regions where this behavior was detected, in most cases

H3K4me1 remained steadily elevated at 24 hr, indicating an

uncoupling of these two marks at later time points. Transient

H3K4me1 was detected at a comparatively small number of

regions.

Using these criteria, 514 latent enhancers were identified at

4 hr and 1,002 at 24 hr post-LPS stimulation (Table S1). Although

LPS-induced latent enhancers represented a relatively small

fraction of the total enhancer repertoire, they made up 15.8%

of the total pool of enhancers activated by LPS in this system

(1,002/6,308).

red from a GREAT analysis) (McLean et al., 2010).

ages. Data are expressed as mean ± SEM.


We also considered additional features known to be asso-

ciated with enhancers, including increased accessibility of

genomic DNA (Gross and Garrard, 1988), binding of histone

acetyltransferases (HATs) (Heintzman et al., 2007), RNA Pol II,

and noncoding transcription (De Santa et al., 2010; Kim et al.,

2010). None of these marks could be used to identify latent

enhancers in untreated macrophages, as virtually no overlap

could be observed between latent enhancers and regions of

basal genomic DNA accessibility (measured by formaldehyde-

assisted isolation of regulatory elements [FAIRE]; 0.79%) (Giresi

et al., 2007), basal p300 (0.37%) (Ghisletti et al., 2010), RNA Pol II

(0.6%), and noncoding chromatin-associated transcripts (Bhatt

et al., 2012) (Figures S1A and S1B). Although we cannot rule

out the existence of additional, as yet unknown features that

may be associated with latent enhancers prior to stimulation,

these data support the definition of latent enhancers as a class

of regulatory elements that cannot be identified in unstimulated

cells using currently available markers and appear in terminally

differentiated cells conditionally to stimulation.

To evaluate the contribution of latent enhancers to the

LPS-induced gene expression program, we generated ChIP-

seq data sets for RNA Pol II in naive and LPS-stimulated

macrophages and analyzed the genomic distribution of latent

enhancers relative to the transcription start site (TSS) of LPS-

regulated genes. We assigned enhancers from different classes

to the nearest gene induced in response to stimulation (Table

S2). Proximity to LPS-responsive genes was different among

enhancer classes, latent enhancers being generally associated

with LPS-inducible genes (p < 13 10�15 in a chi-square test, Fig-

ure 1D). Latent enhancers were located closer to the assigned

genes than both poised-activated and constitutive enhancers

(p = 1.5 3 10�8 and p = 2.8 3 10�25) (Figure S1C). We then

scored latent enhancers for the associated functional categories

of the nearby genes using an independent approach (McLean

et al., 2010). Strong functional enrichments were observed (Fig-

ure 1E), suggesting that latent enhancers contribute to the

activation of specific components of the LPS response.

Genes are activated by LPS with very different kinetics.

Therefore, we analyzed the association of latent enhancers

with genes activated at different times after LPS stimulation.

Considering the seven recently reported kinetic classes of

LPS-induced genes (Bhatt et al., 2012), we found that the

classes of genes with faster activation kinetics were underrep-

resented among the inducible genes associated with latent

enhancers (p = 8.9 3 10�6 in a chi-squared test). Therefore,

latent enhancers may be mainly dedicated to the activation of

slowly induced genes. Conversely, no obvious bias was

observed for genes subjected or not to LPS tolerance (i.e., genes

resistant to a second LPS stimulation) (Foster et al., 2007).

Additional features that support the functionality of latent

enhancers include: (1) their sequence conservation (Fig-

ure S1D), (2) the enrichment for specific TFBS (Table S3), and

(3) their ability to promote expression of a reporter gene

driven by a minimal promoter (14/24, 58.8% of tested en-

hancers) (Figure S1E). Finally, latent enhancer unveiling could

also be observed in vivo in LPS-elicited peritoneal macrophages

(Figure 1F), thus further supporting their potential biological

relevance.


A Complex Repertoire of Latent Enhancers Unveiled byDistinct StimuliThese data suggest that the emergence of latent enhancers may

represent a general feature of the response of differentiated cells

to external stimuli. Macrophages are equipped with a plethora of

receptors allowing them to sense, and react to, a broad panel of

agonists (Medzhitov and Horng, 2009). Thus, we treated macro-

phages with different stimuli with varying degrees of functional

overlap, and analyzed alterations of the enhancer repertoire

and transcriptional outputs. Stimuli included (1) two Toll-like

receptor (TLR) agonists specific for TLR2 (macrophage-acti-

vating lipopeptide-2 [MALP-2]) and TLR9 (unmethylated CpG-

containing oligonucleotide) (Kawai and Akira, 2011); these

stimuli differ from LPS (acting on TLR4) mainly because of their

inability to induce IRF3-dependent transcription of the interferon

beta (IFNb) gene, which generates a secondary wave of induc-

ible transcription in this system (Doyle et al., 2002); (2) interleukin

4 (IL-4) and IFNg, two cytokines that induce macrophage polar-

ization toward an anti-inflammatory (M2) or pro-inflammatory

(M1) phenotype, respectively (Mosser and Edwards, 2008); (3)

tumor necrosis factor-alpha (TNFa) and interleukin 1-beta (IL-

1b), two inflammatory cytokines; and (4) the immunoregulatory

cytokine TGFb.

Each stimulus induced histone acetylation at a discrete sub-

set of poised enhancers ranging from 2,142 (TGFb) to 3,371

(MALP-2) elements (Figure 2A; Table S4). An even higher number

of constitutively acetylated enhancers that ranged between

4,174 (IL-1b) and 8,046 (IFNg), displayed loss or reduction of

acetylation upon stimulation. All stimuli were also able to unveil

distinct and comparatively small sets of latent enhancers

(Figures 2A and 2B; Table S4), indicating that the unveiling of

selected regulatory elements is a hallmark of the cellular reaction

to changing environments. As seen with LPS, induction of

H3K4me1 usually occurred concurrently with histone acetylation

and often with recruitment of Pu.1 (Table S4).

The data sets described above allowed us to identify 2,140

latent enhancers that emerged in response to at least one

stimulus. Becausewe used a limited panel of agonists, represen-

tative of just a fraction of the environmental complexity macro-

phages can be exposed to, and given that for all stimuli except

LPS just a single time point was analyzed, it appears evident

that the hidden repertoire of regulatory elements that may

emerge in response to stimulation is much larger.

We next interrogated our ChIP-seq data sets to determine to

what extent stimuli induced shared rather than unique sets of

latent enhancers. The overlap between different stimuli closely

reflected their functional similarity. TLR agonists activated

largely overlapping groups of latent enhancers, with each stim-

ulus also retaining the capacity to unveil distinct and specific

regulatory regions (Figure 2B). A significant fraction of the latent

enhancers common to TLR agonists was also activated by proin-

flammatory cytokines (TNFa and IL-1b). The immunoregulatory

cytokine TGFb, apart from inducing a unique repertoire of latent

enhancers, also activated enhancers responsive to inflammatory

agents, suggesting a possible integration of multiple signals at

these regions. IFNg and IL-4 were able to induce two large

sets of unique and nonoverlapping latent enhancers (Figures

2B and S2), consistent with their opposing functions as

Figure 2. The Complex Repertoire of Latent Enhancers Induced by Distinct Macrophage Activators

(A) Genomic snapshot showing poised-activated and latent enhancers induced in response to a panel of macrophage activators.

(B) Circos visualization of latent enhancers induced by multiple stimuli.

(C) Correlation matrix of the number of reads (log2) for latent enhancers activated by each stimulus. Pearson correlation values are shown by color and intensity of

shading. Variables were ordered by PCA (principal component analysis).

(D) Heat-map showing the distance between each latent enhancer (represented by an individual line) and the closest gene regulated by the same stimulus.

See also Figure S2 and Tables S4, S5, and S6.

M1 and M2 macrophage polarizers, respectively. Overall, CpG

DNA, MALP-2, and TNFa were highly correlated, whereas LPS

and IFNg were anticorrelated to IL-4 (Figure 2C). Latent

enhancers induced by a given stimulus were frequently at a short

distance from a gene regulated by the same stimulus (Figure 2D)

and, as shown above for LPS, they were significantly closer to

candidate target genes than poised-activated (p = 2.2 3 10�16)

or constitutive (p = 1.04 3 10�13) enhancers. Finally, different

functional categories were enriched in sets induced by different

stimuli (Table S5).

To gain insight into the process of unveiling of latent

enhancers, we identified TFBS overrepresented in the sets


Figure 3. Sequence-Specific TFs Recruitment to Latent Enhancers

(A) TFBS overrepresented in latent enhancers unveiled by different stimuli. For all agonists, PWMs identified by de novo motif discovery and relative E values

(obtained using MEME) are shown at the top line. Below each motif the two top-ranking Jaspar matrices are shown.

(B and C) ChIP-seq analysis of JunB, Stat6, and Stat1 recruitment to latent enhancers induced in response to LPS, IL-4, and IFNg, respectively. Heatmaps show

the normalized number of reads in the genomic area of each latent enhancer before and after stimulation. Regions are sorted based on tag density after

stimulation. The green flags on the right highlight the presence of a statistically significant peak at 4 hr (MACS, p % 1 3 10�5). The PWMs obtained by de novo

motif discovery on the peaks of each TF at the respective latent enhancer set are shown in (C).

(D) A representative snapshot showing JunB recruitment to an LPS-induced latent enhancer.

See also Figure S3.

induced by different stimuli (Figure 3A; Table S6). Latent

enhancers induced by TLR agonists were mainly enriched in

AP-1 sites, whichmay relate to the reported ability of AP-1 family

TFs to promote chromatin opening (Biddie et al., 2011). Consis-

tently, the AP-1 protein JunB was inducibly recruited to the

majority (79.6%) of LPS-dependent latent enhancers (Figure 3B).

De novo motif discovery on sequences bound by JunB in latent

enhancers retrieved a canonical AP-1 matrix (Figure 3C). A

representative snapshot is shown in Figure 3D. Enhancers

induced by TGFb were associated with a SMAD-type binding

site and recruited Smad3 (Figure S3), whereas different types


of STAT sites were overrepresented in the IL-4 and IFNg sets

(Figure 3A). IL-4-induced latent enhancers were associated

with an inverted repeat 50-TTCNNNNGAA-30 that matches the

STAT6 binding motif identified in ChIP-seq experiments (Elo

et al., 2010). Consistently, analysis of STAT6 genomic distribu-

tion in IL-4-treated (4 hr) macrophages revealed an extensive

association with IL-4-induced latent enhancers (Figure 3B), and

de novo motif discovery on these sequences retrieved the

same inverted repeat (Figure 3C). IFNg-induced latent enhancers

were associated with an IRF-type STAT site. Homodimeric

STAT1 binds canonical STAT binding sites (50-TTCCNGGAA-30)

(Robertson et al., 2007); however, within a trimeric complex with

STAT2 and IRF9, which can be activated by IFNg (Matsumoto

et al., 1999) binding occurs at IRF-type binding sites formed

by direct repeats of 50-GAAA-30 units (Taniguchi et al., 2001).

A Stat1 ChIP-seq confirmed its broad association with latent

enhancers activated by IFNg (Figure 3B), and the underlying

binding site identified by de novo motif discovery was indeed

an IRF-type site (Figure 3C), thus confirming that the STAT1 iso-

form involved in the formation of latent enhancers in response to

IFNg is probably a complex with IRFs.

In addition, binding sites similar but not perfectly identical to

those for themacrophagemaster regulator Pu.1 were overrepre-

sented in all latent enhancer sets.

Mechanistic Aspects of Latent EnhancersWe focused on IL-4 and IFNg because they both operate through

an experimentally tractable two-step pathway that includes

a receptor-associated JAK family tyrosine kinase and a STAT

family TF whose phosphorylation triggers nuclear translocation

and transcriptional activation (Darnell et al., 1994).

We first analyzed the kinetics of increase of H3K4me1,

H3K27Ac, and Pu.1 at representative enhancers. With both

IL-4 and IFNg, signals increased slowly and in a progressive

manner and their kinetics were similar (Figures 4A and 4B).

Consistently, FAIRE indicated that basal accessibility of latent

enhancers was low and identical in macrophages and in

Hepa1-6 hepatocytes, an unrelated cell line used for comparison

(Figures 4A and 4B). In response to stimulation, FAIRE signals

increased in a progressive fashion that was indicative of an

ongoing nucleosomal reorganization, eventually leading to an

increased central region of high accessibility bracketed on

both sides by modified nucleosomes (Figure 4C).

STAT6 and STAT1 recruitment to latent enhancers showed

peculiar kinetic properties (Figures 4A and S4A). The slow

increase in STAT1 and STAT6 occupancy, which resembled

the slow changes in accessibility and the deposition of histone

modifications, was dissociated from bulk changes in TF phos-

phorylation, as this was already maximal at 10 min poststimula-

tion (Figure 4D). However, consistent with its rapid activation and

nuclear translocation, STAT6 was recruited with fast kinetics to

other regulatory regions, such as the promoter of the IL-4-induc-

ible Ccl12 gene (Figure S4B). Therefore, although STAT6 is fully

competent for DNA binding at early time points, it could not be

quickly recruited to latent enhancers.

To determine the kinetic properties of latent enhancers at

genome scale, we generated ChIP-seq data sets at multiple

time points after IL-4 (Figure 4E) or IFNg treatment (Figure S4D).

Genomic data confirmed that the activation of latent enhancers

is a relatively slow process, particularly if compared to the

kinetics of TF activation (Figure 4D) and to the kinetics of TF

recruitment to poised enhancers, which was often maximal at

15 min poststimulation (enhancer marked by asterisks in Fig-

ure 4E). Moreover, kinetics of STAT TF recruitment often pre-

ceded the deposition of H3K4me1 and in some cases also

Pu.1 recruitment (Figures 4E and S4D).

Mechanistically, the kinetic discrepancy between STAT1 or

STAT6 activation and their recruitment to latent enhancers,

together with the slow increase in accessibility and histone

mark deposition described above, indicate that the STAT TF

binding sites in latent enhancers are not immediately available

for binding. Clearly, the time lag between stimulation and

maximal occupancy must be interpreted in the context of a pop-

ulation analysis, and be taken as evidence of a progressively

increasing fraction of cells in which STAT binding sites in latent

enhancers are exposed. TFBS at latent enhancers were not

covered by positioned nucleosomes in unstimulated macro-

phages, as indicated by homogeneous H3 signals (Figure S4C).

Upon stimulation, nucleosomal occupancy was increased on

both sides of a centrally located, H3-depleted area of the

enhancer (Figures S4C and 4C), likely because of the barrier

effect exerted by newly recruited TFs (Jiang and Pugh, 2009; Se-

gal and Widom, 2009; Valouev et al., 2011). Lack of a well-posi-

tioned nucleosome in the basal state implies that nucleosomes

at latent enhancers are allowed to undergo lateral movements

leading to stochastic fluctuations in their positions. In turn,

such movements may create windows of opportunity for TF

binding. Once the STAT TF are bound they may recruit Pu.1

and together stabilize the open and accessible state by recruit-

ing chromatin-remodelers (Cui et al., 2004; Huang et al., 2002;

Liu et al., 2002), as demonstrated at the promoters of inflamma-

tory genes (Ramirez-Carrozzi et al., 2009).

To determine the role of STAT TFs in latent enhancer unveiling,

we first used a Jak kinase inhibitor (Jak-i) that blocked STAT TF

phosphorylation without affecting other pathways activated by

these cytokine receptors (such as the ERK/MAPK pathway) (Fig-

ure 5A). Jak-i blocked the induction of H3K4me1 and H3K27Ac

in response to IL-4 and IFNg (Figures 5B and 5C). Pu.1 recruit-

ment was also entirely dependent on Jak-Stat signaling.

We next analyzed at genome scale the impact of Stat6 and

Stat1 gene loss on IL-4 and IFNg induction of latent enhancers.

De novo histone mark deposition and Pu.1 recruitment to latent

enhancers were entirely dependent on STAT TFs activation, as

virtually no latent enhancer could be detected in their absence

(Figures 5D and 5E). A representative snapshot is shown in

Figure 5F.

We noticed that new Pu.1 peaks were in some cases located

within inducible genes (Figure S5A) and therefore we considered

the possibility that in those specific cases RNA-Pol II-dependent

nucleosome displacement and chromatin opening may increase

accessibility of latent sites. Blockade of RNA Pol II elongation

by the Cdk9 inhibitor DRB blocked Pu.1 recruitment to one

such enhancer locatedwithin theNos2 gene (Figure S5B). There-

fore, in some cases RNA Pol II may play an active role in

promoting the surfacing of latent enhancers contained within

partially inaccessible chromatin. Conversely, the protein syn-

thesis inhibitor puromycin only marginally impaired H3K4me1

induction (Figure S5C).

A most intriguing observation was the lack of constitutive Pu.1

binding at latent enhancers in spite of the presence of Pu.1 sites.

Indeed, Pu.1 is considered a pioneer factor able to invade nucle-

osomal chromatin and to promote both the formation of nucleo-

some-free regions and the deposition of H3K4me1 at adjacent

nucleosomes (Ghisletti et al., 2010; Heinz et al., 2010). Lack of

basal Pu.1 binding at latent enhancers may be explained by

different mechanisms, including: (1) an actively enforced repres-

sive chromatin state that renders the underlying binding sites


Figure 4. Kinetic Analysis of Latent Enhancers

(A) Kinetics of H3K4me1 and H3K27Ac deposition, Pu.1 and Stat1 (or Stat6) recruitment, DNA accessibility (FAIRE) at representative IL-4- (left) and IFNg-induced

(right) latent enhancers. Hepa1-6, mouse hepatocyte cell line. Data are expressed as mean ± SEM.



unavailable for binding, and (2) the presence of low affinity

binding sites that cannot support recruitment of Pu.1 without

cooperation provided by a partner TF activated by stimulation.

An additional possibility is that Pu.1’s pioneering function may

be dependent on cooperative binding or synergy with constitu-

tively expressed TFs that bind adjacent sites, which may be

absent at the latent enhancers. Regarding the first mechanism,

we investigated by ChIP the presence of negatively acting his-

tone marks. Levels of H3K27me3, H3K27me2, and H3K9me3

were undetectable or very low at all latent enhancers tested

(Figures 6A and S6A), indicating that these marks were not

involved in maintenance of basal repression.

Consistent with the second mechanism, Pu.1 binding sites

within latent enhancers displayed lower affinities as compared

to those in constitutively bound enhancers. First, using in vitro

determined binding affinities (Wei et al., 2010), we computed

the Pu.1 binding affinities for the consensus sites found in the

different classes of enhancers. The sites in latent enhancers

showed much lower affinities for Pu.1 relative to constitutive

enhancers (Wilcoxon test p value 7.063 10�54) and poised-acti-

vated enhancers (p value 1.01 3 10�65). Instead poised-

activated and constitutive enhancers showed similar binding

affinities (p value 0.02) (Figure 6B). Second, we computed the

distributions of best matches for the Pu.1 position weight matrix

(PWM) derived from ChIP-seq data (Heinz et al., 2010) for each

class of enhancers. The cumulative distribution of PWM scores

for constitutive, steady and poised enhancers showed nearly

overlapping distribution frequencies (Wilcoxon test p value

0.017) (Figure 6C), whereas latent enhancers showed a sig-

nificant divergence in sequence composition (p values 6.931 3

10�117 and 1.113 3 10�95 relative to poised and constitutive

enhancers, respectively) (Figure 6C). Overall, these data indicate

that the composition of Pu.1 binding sites in latent enhancers

diverges from canonical high affinity sites enough to prevent

Pu.1 binding in basal conditions; however, these low affinity sites

allow Pu.1 recruitment in the presence of cooperating stimulus-

activated TFs.

Latent Enhancers and Short-Term TranscriptionalMemoryWe therefore investigated whether newly activated latent en-

hancers could be maintained in the absence of a sustained

stimulation. Cells were treated with IL-4 or IFNg for either 4 hr

or 24 hr, washed and incubated in fresh medium for additional

48 hr before harvesting (Figure 6D). Complete loss of STAT1 or

STAT6 phosphorylation after washout confirmed efficient signal

termination (Figure 6D). Histone acetylation, accessibility and

Pu.1 occupancy all returned to prestimulation levels upon cyto-

kine removal (Figure 6E). STAT1 and STAT6 binding to latent

(B) Data from multiple latent enhancers activated by either IL-4 or IFNg are show

(C) A representative IL-4-induced latent enhancer was analyzed using multiple

summit of the inducible Pu.1 peak. Data are expressed as mean ± SEM.

(D) Western blots with phospho-specific STAT6 and STAT1 antibodies showing

(E) ChIP-seq analysis of IL-4-dependent latent enhancer induction kinetics. A re

indicate a poised enhancer where Stat6 recruitment was already maximal at 150

progressive. The box plots indicate relative tag counts at all mapped IL-4 latent

See also Figure S4.

enhancers were also lost after washout (Figure 6E). These obser-

vations indicate that following the initial recruitment, sustained

Pu.1 occupancy required the constant presence of cytokine-

activated STAT1 or STAT6, thus suggesting a strong functional

cooperativity. As a consequence of the removal of STAT1/Pu.1

or STAT6/Pu.1 from DNA, histone acetylation and accessibility

of latent enhancers were completely lost. Instead, H3K4me1

persisted for days after the washout (Figures 6E and S6B), indi-

cating an autonomous sustainment of this mark. We generated

H3K4me1 ChIP-seq data sets to determine the fraction of latent

enhancers that were retained after extensive washout. We found

that of the 1,002 newly formed H3K4me1 peaks that were

detected at 24 hr of IFNg treatment and persisted after 72 hr of

continuous stimulation, 307 (30.6%) were also detected after

a 48 hr washout (Figure 6F). These data show that a large fraction

of latent enhancers remains stably marked in the absence

of continuous stimulation. Therefore, by uncovering a specific

and previously hidden repertoire of enhancers, each stimulus

generates an epigenomic signature of its exposure.

Residual marking by H3K4me1 was indicative of a poised

state of these enhancers after cytokine washout. When primed

cells were restimulated with the same agonist used in primary

stimulation, faster acetylation kineticswere observed (Figure 7A).

Moreover, final levels of acetylation were consistently higher

than those achieved in response to primary stimulation. These

data are consistent with the notion that the H3K4me1 mark

primes enhancers for recruitment of HATs (Jeong et al., 2011).

Finally, genes adjacent to latent enhancers were induced faster,

and often to higher levels in response to secondary stimulation

(Figure 7B). These enhancing effects were not due to a stronger

induction of the Jak/Stat signal transduction pathway in restimu-

lated cells, because levels of phospho-STATs were comparable

to those found in unprimed macrophages (Figure 7C).

As shown above (Figure 2), stimuli differed in their ability to

cause the emergence of latent enhancers. However, the pres-

ence of complex combinations of TFBS may in principle enable

cross-reactivation once the latent enhancer has surfaced. To

test this hypothesis, we primed macrophages with IL-4 and after

cytokine washout, we restimulated them with a panel of different

agonists. In some cases, latent enhancers were promiscuously

induced by multiple stimuli and retained increased responsive-

ness to the same agonists upon restimulation (Figure 7D, left).

At other enhancers induced selectively by IL-4, responsiveness

to secondary stimulation was exquisitely restricted to IL-4 (Fig-

ure 7D, middle). Finally, in other cases, reactivation of a surfaced

IL-4-induced latent enhancerwashighly promiscuous (Figure 7D,

right) indicating that although the primary exposure of these

elements was highly selective, the very same enhancers could

be subsequently acted upon by TFs activated bymultiple stimuli.

n. Each line represents a different enhancer.

primers spanning a �3 kb region. The central position (0) corresponds to the

their activation kinetics.

presentative genomic region is shown on the left. The asterisks (bottom right)

, as opposed to the latent enhancer shown on its left, where recruitment was

enhancers.


Figure 5. STAT TFs Requirement for the Unveiling of Latent Enhancers in Response to IL-4 and IFNg Stimulation

(A) Inhibition of STAT tyrosine phosphorylation but not MAPK ERK phosphorylation by a selective Jak kinase inhibitor (Jak-i).

(B and C) Inhibition of histone marks deposition and Pu.1 recruitment by Jak-i at individual enhancers (B) or at a panel of IL-4- or IFNg-induced latent enhancers

(C). Data in (B) are expressed as mean ± SEM.



DISCUSSION

In this study, we report a distinct class of transcriptional en-

hancers whose main distinguishing features are: (1) lack of basal

marking and TF binding in fully differentiated cells, (2) acquisition

of enhancer marks and accessibility upon stimulation, (3) depen-

dence on both stimulus-activated TFs and lineage determining

TFs, (4) selective retention of H3K4me1 upon signal termination,

and (5) faster and/or augmented response upon secondary stim-

ulation. Although the repertoire of available cis-regulatory ele-

ments in a given cell type appears to be a specific property of

its differentiated state and to be directly controlled by the

enhancer-organizing activity of lineage-determining TFs (Natoli,

2010), the repertoire of surfaced latent enhancers is determined

by the functional cooperation between stimulus-activated TFs

and lineage-restricted TFs, thus integrating cell-type-specific

and stimulus-specific inputs.

We suggest that the unveiling, and the subsequent retention, of

an agonist-specific repertoire of latent enhancers can be consid-

ered an epigenomic footprint (or scar) of the stimuli the cell has

previously been exposed to. The surfacing of latent enhancers

also implies an expansion of the repertoire of accessible regula-

tory elements,which effectively changes thewaycells respond to

a subsequent stimulation. It has been previously reported that in

macrophages that underwent a prolonged exposure to LPS,

although some genes cannot be reactivated upon washout and

secondary LPS stimulation, many others were hyperinduced

(Foster et al., 2007), a phenomenon that may relate to the expan-

sionof the enhancer repertoirewedescribed here. In this context,

it is interesting to note that once unveiled, some latent enhancers

become responsive to stimuli that were unable to induce their

exposure in naive cells. Therefore, not only does the surfacing

of latent enhancers change the way cells respond to repeated

challenges with the same stimulus, but also the way a precondi-

tioned cell will respond to stimuli it never encountered before.

Mechanistically, latent enhancers can be distinguished from

those poised enhancers whose basal inactivity is associated

with both H3K4me1 and H3K27me3 (Rada-Iglesias et al.,

2011). Our data suggest that lack of marking of latent enhancers

in the basal state reflects the fact that these enhancers are

ignored because of a combination of two main factors: nucleo-

some occlusion and a comparatively low affinity for Pu.1, which

would otherwise be able to invade nucleosomal DNA and induce

enhancer formation (Ghisletti et al., 2010; Heinz et al., 2010).

Upon stimulation, Pu.1 can gain access to such sites (and

promote the organization of an active enhancer) because of

the functional cooperation provided by stimulus-activated TFs.

However, the slow kinetics of surfacing of latent enhancers is

not compatible with a simple model of cooperative binding,

because in this case Pu.1 recruitment and enhancer organization

would occur immediately after the nuclear translocation of

inducible TFs such as STAT1 or STAT6. Conversely, the time

lag between activation of the STAT TFs and their recruitment

(D–F) STAT6 (D) and STAT1 (E) requirement for IL-4- or IFNg-mediated induction

de novo deposition of histone marks and Pu.1 recruitment at latent enhancers,

signed-rank test). A representative snapshot from Stat6�/� macrophages is show

See also Figure S5.

(together with Pu.1) to latent enhancers suggests the existence

of an additional regulatory layer that may be due to nucleosomal

interference with DNA binding. Reorganization of the nucleo-

somal spacing at latent enhancers may reflect either an active

process mediated by chromatin remodelers or random fluctua-

tions in the position of nonfixed nucleosomes. In the first case,

TFs such as the STAT TF and AP-1 endowed with the ability to

recruit chromatin-remodeling complexes (Cui et al., 2004; Huang

et al., 2002; Liu et al., 2002; Biddie et al., 2011) would initially

make short contacts with partially occluded binding sites within

latent enhancers, thus promoting (in combination with Pu.1) their

opening via an iterative process involving the repetition of tran-

sient binding events followed by progressively increased site

exposure. The second possibility is that either nucleosome turn-

over or spontaneous lateral movements of nucleosomes that are

not constrained because of the lack of either adjacent barriers or

positioning signals (Valouev et al., 2011) would create windows

of opportunity for TF binding, thus eventually leading to the stabi-

lization of the open state. Upon stimulus termination and release

of the bound TFs, loss of the barrier activity they exerted would

allow spontaneous nucleosome repositioning and loss of acces-

sibility. Nevertheless, H3K4me1 marking will persist and pro-

mote a faster reinduction at restimulation, which may reflect

a direct reading of this mark by HATs (Jeong et al., 2011).

Our experimental system does not allow determining

the potential for transmission of the unveiled repertoire of

latent enhancers across DNA synthesis. Therefore, retention of

the H3K4me1 mark—and the associated functional conse-

quences—after stimulus termination can only be discussed

here in the context of short-term transcriptional memory mech-

anisms. In any case, such short-term epigenomic memory may

have great functional implications in postmitotic cells such as

neurons. Indeed, it will be interesting to determine if brain

memory circuits, in which epigenetic phenomena have been

suggested to play an important role (Levenson and Sweatt,

2005), also involve latent enhancers.

EXPERIMENTAL PROCEDURES

Antibodies, Reagents, Cell Culture Procedures, Cloning,

Transfections, and RNA Analyses

For information on antibodies, reagents, cell culture procedures, cloning,

transfections, and RNA analyses, please refer to the Extended Experimental

Procedures.

Chromatin Immunoprecipitation and Sequencing

ChIP was carried out using either a previously described protocol (Ghisletti

et al., 2010) or a high-throughput protocol (Garber et al., 2012). Briefly,

5–15 3 106 (ChIP-seq for histone marks and Pu.1) or 100 3 106 (ChIP-seq

for other TFs and RNA Pol II) fixed macrophages were lysed with RIPA buffer

and, after chromatin shearing by sonication, incubated overnight at 4�Cwith protein G Dynabeads (Invitrogen) that were previously coupled with

3–10 mg of antibody. For ChIP-qPCR experiments, 10%–20% of cells indi-

cated above were used, and IP was performed using 1–3 mg of antibody. After

immunoprecipitation, beads were recovered using a 96-well magnet, washed,

of latent enhancers. Box plots show the consequence of STAT TF absence on

as measured by ChIP-seq (p values calculated with paired-samples Wilcoxon

n in (F).


Figure 6. Mechanistic Dissection of Latent Enhancer Formation

(A) Repressive histone marks at representative latent enhancers. The first column on the left is a positive control for each modification. For each modification, the

enhancers were ranked based on decreasing signal intensity. Data are expressed as mean ± SEM.

(B) Analysis of Pu.1 binding sites at latent enhancers as compared to constitutively bound Pu.1 sites (in constitutive and poised-activated enhancers). The Z score

was calculated using Pu.1 affinity data obtained from protein binding microarrays (Wei et al., 2010). The p values shown were obtained using a Wilcoxon test.

(C) Cumulative distribution for PWM scores in latent enhancers and two control groups. The reference PWM used was generated from in vivo binding data (Heinz

et al., 2010). The p values were generated by a Wilcoxon test. Ecdf, empirical cumulative distribution function.



Figure 7. A Short-Term Memory Associated with Latent Enhancers

(A) Acetylation kinetics at latent enhancers upon restimulation.

(B) Induction of genes adjacent to latent enhancers upon restimulation.

(C) STAT TF activation upon secondary stimulation.

(D) Selective versus promiscuous activation of latent enhancers by alternative stimuli. Three different representative behaviors are shown. The enhancer on the

left was activated in a promiscuous manner, the one in the middle was selectively activated by IL4 both at primary and secondary stimulation, and the one on the

right was selectively activated by IL4 at primary stimulation but promiscuously at restimulation. ChIP data in (D) refer to a priming of 24 hr and a (re)stimulation of

4 hr. Data in (A), (B), and (D) are expressed as mean ± SEM.

and DNA was eluted and decrosslinked overnight at 65�C. DNA was then

purified with Solid-phase reversible immobilization (SPRI) beads (Agencourt

AMPure XP, Beckman Coulter) and quantified with PicoGreen (Invitrogen).

For ChIP-QPCR experiments, 0.4 ml of purified DNA was used for amplification

on an ABI 7500 machine. Primers used for ChIP-QPCR are in Table S7. ChIP

DNA was prepared for Solexa GAII or HiSeq2000 sequencing following stan-

dard protocols. Basic data processing was carried out using the Fish-The-

ChIP pipeline (Barozzi et al., 2011). A detailed description of the computational

analyses is provided in the Extended Experimental Procedures.

ACCESSION NUMBERS

Raw data sets have been submitted to the Gene Expression Omnibus data-

base (http://www.ncbi.nlm.nih.gov/geo/) and are available under the acces-

sion GSE38379. R was used to compute statistics and generate plots.

(D) Cytokine washout results in signal termination and loss of STAT TF phospho

(E) Pu.1 binding to latent enhancers, acetylation and increased accessibility, bu

mean ± SEM.

(F) Two representative ChIP-seq snapshots showing H3K4me1 behavior at two

See also Figure S6.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures, six

figures, and seven tables and can be found with this article online at http://

dx.doi.org/10.1016/j.cell.2012.12.018.

ACKNOWLEDGMENTS

We thank D. Voehringer and A. Planas for the generous gift of Stat6- and Stat1-

deficient bone marrows, respectively; B. Amati for critical comments to the

manuscript; I. Amit (Weizmann Institute, Rehovot) for training in the HT-ChIP

protocol; L. Rotta, M. Riboni, S. Minardi (Cogentech, Milan), and S. Bianchi

(IIT@SEMM) for thepreparationandprocessingofsequencing libraries;K.Ferrari

and D. Pasini (IEO) for sharing reagents. Sequence data in fastq format were

prepared by H. Muller and the Computational Research Unit of the Center for

rylation.

t not H3K4me1, require continuous cytokine signaling. Data are expressed as

IFNg-induced latent enhancers upon cytokine washout.


http://www.ncbi.nlm.nih.gov/geo/



Genomic Sciences of IIT@SEMM and hosted on the IIT@SEMM computational

infrastructure. Training inHT-ChIPwas supported by a short-termEMBO fellow-

ship (ASTF 210–2012) to R.O. This study was supported by the European

Research Council (advanced grant 268757-NORM to G.N.). R.O. and S.P.

carried out the ChIP-seq and RNA-Seq experiments, their validation and func-

tional assays with the help of S.B., A.C., E.P., and S.G. V.P. and I.B. analyzed

data with the help of A.T. G.N. designed the study with the help of R.O. and

S.G. and wrote the paper.

Received: May 29, 2012

Revised: October 12, 2012

Accepted: December 13, 2012

Published: January 17, 2013

REFERENCES

Barish, G.D., Yu, R.T., Karunasiri, M., Ocampo, C.B., Dixon, J., Benner, C.,

Dent, A.L., Tangirala, R.K., and Evans, R.M. (2010). Bcl-6 and NF-kappaB cis-

tromes mediate opposing regulation of the innate immune response. Genes

Dev. 24, 2760–2765.

Barozzi, I., Termanini, A., Minucci, S., and Natoli, G. (2011). Fish the ChIPs:

a pipeline for automated genomic annotation of ChIP-Seq data. Biol. Direct

6, 51.

Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G.,

Chepelev, I., and Zhao, K. (2007). High-resolution profiling of histone methyl-

ations in the human genome. Cell 129, 823–837.

Bhatt, D.M., Pandya-Jones, A., Tong, A.J., Barozzi, I., Lissner, M.M., Natoli,

G., Black, D.L., and Smale, S.T. (2012). Transcript dynamics of proinflamma-

tory genes revealed by sequence analysis of subcellular RNA fractions. Cell

150, 279–290.

Biddie, S.C., John, S., Sabo, P.J., Thurman, R.E., Johnson, T.A., Schiltz, R.L.,

Miranda, T.B., Sung, M.H., Trump, S., Lightman, S.L., et al. (2011). Transcrip-

tion factor AP1 potentiates chromatin accessibility and glucocorticoid

receptor binding. Mol. Cell 43, 145–155.

Biswas, S.K., and Mantovani, A. (2010). Macrophage plasticity and interaction

with lymphocyte subsets: cancer as a paradigm. Nat. Immunol. 11, 889–896.

Bonn, S., Zinzen, R.P., Girardot, C., Gustafson, E.H., Perez-Gonzalez, A., Del-

homme, N., Ghavi-Helm, Y., Wilczy�nski, B., Riddell, A., and Furlong, E.E.

(2012). Tissue-specific analysis of chromatin state identifies temporal signa-

tures of enhancer activity during embryonic development. Nat. Genet. 44,

148–156.

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.

Creyghton, M.P., Cheng, A.W., Welstead, G.G., Kooistra, T., Carey, B.W.,

Steine, E.J., Hanna, J., Lodato, M.A., Frampton, G.M., Sharp, P.A., et al.

(2010). Histone H3K27ac separates active from poised enhancers and

predicts developmental state. Proc. Natl. Acad. Sci. USA 107, 21931–21936.

Cui, K., Tailor, P., Liu, H., Chen, X., Ozato, K., and Zhao, K. (2004). The

chromatin-remodeling BAF complex mediates cellular antiviral activities by

promoter priming. Mol. Cell. Biol. 24, 4476–4486.

Darnell, J.E., Jr., Kerr, I.M., and Stark, G.R. (1994). Jak-STAT pathways and

transcriptional activation in response to IFNs and other extracellular signaling

proteins. Science 264, 1415–1421.

De Santa, F., Narang, V., Yap, Z.H., Tusi, B.K., Burgold, T., Austenaa, L., Bucci,

G., Caganova, M., Notarbartolo, S., Casola, S., et al. (2009). Jmjd3 contributes

to the control of gene expression in LPS-activated macrophages. EMBO J. 28,

3341–3352.

De Santa, F., Barozzi, I., Mietton, F., Ghisletti, S., Polletti, S., Tusi, B.K., Muller,

H., Ragoussis, J., Wei, C.L., and Natoli, G. (2010). A large fraction of extragenic

RNA pol II transcription sites overlap enhancers. PLoS Biol. 8, e1000384.


Doyle, S., Vaidya, S., O’Connell, R., Dadgostar, H., Dempsey, P., Wu, T., Rao,

G., Sun, R., Haberland, M., Modlin, R., and Cheng, G. (2002). IRF3 mediates

a TLR3/TLR4-specific antiviral gene program. Immunity 17, 251–263.

Elo, L.L., Jarvenpaa, H., Tuomela, S., Raghav, S., Ahlfors, H., Laurila, K.,

Gupta, B., Lund, R.J., Tahvanainen, J., Hawkins, R.D., et al. (2010).

Genome-wide profiling of interleukin-4 and STAT6 transcription factor regula-

tion of human Th2 cell programming. Immunity 32, 852–862.

Ernst, J., Kheradpour, P., Mikkelsen, T.S., Shoresh, N., Ward, L.D., Epstein,

C.B., Zhang, X., Wang, L., Issner, R., Coyne, M., et al. (2011). Mapping and

analysis of chromatin state dynamics in nine human cell types. Nature 473,

43–49.

Escoubet-Lozach, L., Benner, C., Kaikkonen, M.U., Lozach, J., Heinz, S.,

Spann, N.J., Crotti, A., Stender, J., Ghisletti, S., Reichart, D., et al. (2011).

Mechanisms establishing TLR4-responsive activation states of inflammatory

response genes. PLoS Genet. 7, e1002401.

Foster, S.L., Hargreaves, D.C., and Medzhitov, R. (2007). Gene-specific

control of inflammation by TLR-induced chromatin modifications. Nature

447, 972–978.

Fridlender, Z.G., Sun, J., Kim, S., Kapoor, V., Cheng, G., Ling, L., Worthen,

G.S., and Albelda, S.M. (2009). Polarization of tumor-associated neutrophil

phenotype by TGF-beta: ‘‘N1’’ versus ‘‘N2’’ TAN. Cancer Cell 16, 183–194.

Garber, M., Yosef, N., Goren, A., Raychowdhury, R., Thielke, A., Guttman, M.,

Robinson, J., Minie, B., Chevrier, N., Itzhaki, Z., et al. (2012). A high-throughput

chromatin immunoprecipitation approach reveals principles of dynamic gene

regulation in mammals. Mol. Cell 47, 810–822.

Ghisletti, S., Barozzi, I., Mietton, F., Polletti, S., De Santa, F., Venturini, E.,

Gregory, L., Lonie, L., Chew, A., Wei, C.L., et al. (2010). Identification and char-

acterization of enhancers controlling the inflammatory gene expression

program in macrophages. Immunity 32, 317–328.

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.

Gross, D.S., and Garrard, W.T. (1988). Nuclease hypersensitive sites in chro-

matin. Annu. Rev. Biochem. 57, 159–197.

He, A., Kong, S.W., Ma, Q., and Pu, W.T. (2011). Co-occupancy by multiple

cardiac transcription factors identifies transcriptional enhancers active in

heart. Proc. Natl. Acad. Sci. USA 108, 5632–5637.

He, H.H., Meyer, C.A., Shin, H., Bailey, S.T., Wei, G., Wang, Q., Zhang, Y., Xu,

K., Ni, M., Lupien, M., et al. (2010). Nucleosome dynamics define transcrip-

tional enhancers. Nat. Genet. 42, 343–347.

Heintzman, N.D., Stuart, R.K., Hon, G., Fu, Y., Ching, C.W., Hawkins, R.D.,

Barrera, L.O., Van Calcar, S., Qu, C., Ching, K.A., et al. (2007). Distinct and

predictive chromatin signatures of transcriptional promoters and enhancers

in the human genome. Nat. Genet. 39, 311–318.

Heintzman, N.D., Hon, G.C., Hawkins, R.D., Kheradpour, P., Stark, A., Harp,

L.F., Ye, Z., Lee, L.K., Stuart, R.K., Ching, C.W., et al. (2009). Histone modifi-

cations at human enhancers reflect global cell-type-specific gene expression.

Nature 459, 108–112.

Heinz, S., Benner, C., Spann, N., Bertolino, E., Lin, Y.C., Laslo, P., Cheng, J.X.,

Murre, C., Singh, H., and Glass, C.K. (2010). Simple combinations of lineage-

determining transcription factors prime cis-regulatory elements required for

macrophage and B cell identities. Mol. Cell 38, 576–589.

Hu, G., Schones, D.E., Cui, K., Ybarra, R., Northrup, D., Tang, Q., Gattinoni, L.,

Restifo, N.P., Huang, S., and Zhao, K. (2011). Regulation of nucleosome land-

scape and transcription factor targeting at tissue-specific enhancers by BRG1.

Genome Res. 21, 1650–1658.

Huang, M., Qian, F., Hu, Y., Ang, C., Li, Z., and Wen, Z. (2002). Chromatin-re-

modelling factor BRG1 selectively activates a subset of interferon-alpha-

inducible genes. Nat. Cell Biol. 4, 774–781.

Hurtado, A., Holmes, K.A., Ross-Innes, C.S., Schmidt, D., and Carroll, J.S.

(2011). FOXA1 is a key determinant of estrogen receptor function and endo-

crine response. Nat. Genet. 43, 27–33.

Jeong, K.W., Kim, K., Situ, A.J., Ulmer, T.S., An, W., and Stallcup, M.R. (2011).

Recognition of enhancer element-specific histone methylation by TIP60 in

transcriptional activation. Nat. Struct. Mol. Biol. 18, 1358–1365.

Jiang, C., and Pugh, B.F. (2009). Nucleosome positioning and gene regulation:

advances through genomics. Nat. Rev. Genet. 10, 161–172.

Junion, G., Spivakov, M., Girardot, C., Braun, M., Gustafson, E.H., Birney, E.,

and Furlong, E.E. (2012). A transcription factor collective defines cardiac cell

fate and reflects lineage history. Cell 148, 473–486.

Kawai, T., and Akira, S. (2011). Toll-like receptors and their crosstalk with other

innate receptors in infection and immunity. Immunity 34, 637–650.

Kim, T.K., Hemberg, M., Gray, J.M., Costa, A.M., Bear, D.M., Wu, J., Harmin,

D.A., Laptewicz, M., Barbara-Haley, K., Kuersten, S., et al. (2010). Widespread

transcription at neuronal activity-regulated enhancers. Nature 465, 182–187.

Lawrence, T., and Natoli, G. (2011). Transcriptional regulation of macrophage

polarization: enabling diversity with identity. Nat. Rev. Immunol. 11, 750–761.

Levenson, J.M., and Sweatt, J.D. (2005). Epigenetic mechanisms in memory

formation. Nat. Rev. Neurosci. 6, 108–118.

Liu, H., Kang, H., Liu, R., Chen, X., and Zhao, K. (2002). Maximal induction of

a subset of interferon target genes requires the chromatin-remodeling activity

of the BAF complex. Mol. Cell. Biol. 22, 6471–6479.

Matsumoto, M., Tanaka, N., Harada, H., Kimura, T., Yokochi, T., Kitagawa, M.,

Schindler, C., and Taniguchi, T. (1999). Activation of the transcription factor

ISGF3 by interferon-gamma. Biol. Chem. 380, 699–703.

McLean, C.Y., Bristor, D., Hiller, M., Clarke, S.L., Schaar, B.T., Lowe, C.B.,

Wenger, A.M., and Bejerano, G. (2010). GREAT improves functional interpre-

tation of cis-regulatory regions. Nat. Biotechnol. 28, 495–501.

Medzhitov, R., and Horng, T. (2009). Transcriptional control of the inflamma-

tory response. Nat. Rev. Immunol. 9, 692–703.

Mosser, D.M., and Edwards, J.P. (2008). Exploring the full spectrum of macro-

phage activation. Nat. Rev. Immunol. 8, 958–969.

Mullen, A.C., Orlando, D.A., Newman, J.J., Loven, J., Kumar, R.M., Bilodeau,

S., Reddy, J., Guenther, M.G., DeKoter, R.P., and Young, R.A. (2011). Master

transcription factors determine cell-type-specific responses to TGF-b sig-

naling. Cell 147, 565–576.

Natoli, G. (2010). Maintaining cell identity through global control of genomic

organization. Immunity 33, 12–24.

Natoli, G., Ghisletti, S., and Barozzi, I. (2011). The genomic landscapes of

inflammation. Genes Dev. 25, 101–106.

Rada-Iglesias, A., Bajpai, R., Swigut, T., Brugmann, S.A., Flynn, R.A., and Wy-

socka, J. (2011). A unique chromatin signature uncovers early developmental

enhancers in humans. Nature 470, 279–283.

Ram, O., Goren, A., Amit, I., Shoresh, N., Yosef, N., Ernst, J., Kellis, M., Gym-

rek, M., Issner, R., Coyne, M., et al. (2011). Combinatorial patterning of chro-

matin regulators uncovered by genome-wide location analysis in human cells.

Cell 147, 1628–1639.

Ramirez-Carrozzi, V.R., Braas, D., Bhatt, D.M., Cheng, C.S., Hong, C., Doty,

K.R., Black, J.C., Hoffmann, A., Carey, M., and Smale, S.T. (2009). A unifying

model for the selective regulation of inducible transcription by CpG islands and

nucleosome remodeling. Cell 138, 114–128.

Robertson, G., Hirst, M., Bainbridge, M., Bilenky, M., Zhao, Y., Zeng, T., Eu-

skirchen, G., Bernier, B., Varhol, R., Delaney, A., et al. (2007). Genome-wide

profiles of STAT1 DNA association using chromatin immunoprecipitation

and massively parallel sequencing. Nat. Methods 4, 651–657.

Schones, D.E., Cui, K., Cuddapah, S., Roh, T.Y., Barski, A., Wang, Z., Wei, G.,

and Zhao, K. (2008). Dynamic regulation of nucleosome positioning in the

human genome. Cell 132, 887–898.

Segal, E., and Widom, J. (2009). What controls nucleosome positions? Trends

Genet. 25, 335–343.

Siersbæk, R., Nielsen, R., John, S., Sung, M.H., Baek, S., Loft, A., Hager, G.L.,

and Mandrup, S. (2011). Extensive chromatin remodelling and establishment

of transcription factor ‘hotspots’ during early adipogenesis. EMBO J. 30,

1459–1472.

Song, L., Zhang, Z., Grasfeder, L.L., Boyle, A.P., Giresi, P.G., Lee, B.K., Shef-

field, N.C., Graf, S., Huss, M., Keefe, D., et al. (2011). Open chromatin defined

by DNaseI and FAIRE identifies regulatory elements that shape cell-type iden-

tity. Genome Res. 21, 1757–1767.

Taniguchi, T., Ogasawara, K., Takaoka, A., and Tanaka, N. (2001). IRF family of

transcription factors as regulators of host defense. Annu. Rev. Immunol. 19,

623–655.

Tijssen, M.R., Cvejic, A., Joshi, A., Hannah, R.L., Ferreira, R., Forrai, A., Bellis-

simo, D.C., Oram, S.H., Smethurst, P.A., Wilson, N.K., et al. (2011). Genome-

wide analysis of simultaneous GATA1/2, RUNX1, FLI1, and SCL binding in

megakaryocytes identifies hematopoietic regulators. Dev. Cell 20, 597–609.

Valouev, A., Johnson, S.M., Boyd, S.D., Smith, C.L., Fire, A.Z., and Sidow, A.

(2011). Determinants of nucleosome organization in primary human cells.

Nature 474, 516–520.

Visel, A., Blow, M.J., Li, Z., Zhang, T., Akiyama, J.A., Holt, A., Plajzer-Frick, I.,

Shoukry, M., Wright, C., Chen, F., et al. (2009). ChIP-seq accurately predicts

tissue-specific activity of enhancers. Nature 457, 854–858.

Wang, D., Garcia-Bassets, I., Benner, C., Li, W., Su, X., Zhou, Y., Qiu, J., Liu,

W., Kaikkonen, M.U., Ohgi, K.A., et al. (2011). Reprogramming transcription

by distinct classes of enhancers functionally defined by eRNA. Nature 474,

390–394.

Wang, Z., Zang, C., Rosenfeld, J.A., Schones, D.E., Barski, A., Cuddapah, S.,

Cui, K., Roh, T.Y., Peng, W., Zhang, M.Q., and Zhao, K. (2008). Combinatorial

patterns of histone acetylations and methylations in the human genome. Nat.

Genet. 40, 897–903.

Wei, G.H., Badis, G., Berger, M.F., Kivioja, T., Palin, K., Enge, M., Bonke, M.,

Jolma, A., Varjosalo, M., Gehrke, A.R., et al. (2010). Genome-wide analysis of

ETS-family DNA-binding in vitro and in vivo. EMBO J. 29, 2147–2160.


Supplemental Information

EXTENDED EXPERIMENTAL PROCEDURES

AntibodiesAntibodies against H3K4me1 (ab8895), H3K27Ac (ab4729), H3 (ab1791), H3K9me3 (ab8898) and Smad3 (ab28379) were from

Abcam. Anti-H3K4me3 (#39159), and H3K27me3 (#39155) were from Active Motif. Antibodies for H3K27me2 (#9755), STAT1

(used for Western blot, #9172), phospho-STAT1 (Tyr701, #9171), STAT6 (used for Western blot, #9362), phospho-STAT6 (Tyr641,

#9361), Erk1/2 (#9102), phospho-Erk1/2 (Thr202/Tyr204, #9101) were from Cell Signaling Technology. Anti-RNA Pol II (sc-899),

anti-STAT6 (used for ChIP, sc-981), anti-STAT1 (used for ChIP, sc-592), anti-JunB (sc-46) were from Santa Cruz. Anti-actin

(A4700) was from Sigma-Aldrich. The anti-Pu.1 rabbit polyclonal antibody was generated in-house against the N-terminus of Pu.1

(aa. 1-100; NP_035485.1).

Cells and ReagentsAnimal experiments were performed in accordance with the Italian Laws (D.L.vo 116/92), which enforce the EU 86/609 Directive.

Macrophage cultures from bone marrows of C57/BL6 mice (Harlan) were generated as described (De Santa et al., 2009). Stat1�/�,Stat6�/� andwild-type control bonemarrowswere obtained fromA.M. Planas (IDIBAPS, Barcelona) andD. Voehringer (Ludwig-Max-

imilian Universitat Munchen), respectively. LPS from E.Coli serotype EH100 (Alexis) was used at 10ng/ml unless otherwise indicated.

Recombinant IL-4 (10ng/ml), IFN-g (100ng/ml), TNFa (10ng/ml), TGF-b1 (1 ng/ml), IL-1b (10ng/ml) were from R&D Systems. MALP-2

(1 ng/ml) was from Alexis. CpG-ODN 1668 (50-TCCATGACGTTCCTGATGCT-30) was synthesized by Primm (Milan, Italy) and used at

100nM. Poly(I:C) (10mg/ml) and Gardiquimod (GRD, 1mg/ml) were from InvivoGen. Jak inhibitor I (Jak-i 10mM, fromCalbiochem), DRB

(76mM, Sigma) and puromycin (10mg/ml, Sigma) were administered to cells 1h before stimulation.

In Vivo Stimulation7 week-old C57/BL6 mice were injected intraperitoneally with 100 mg of LPS (or PBS as a control). After 24 hr, peritoneal exudates

cells were collected and CD11b+ F4/80+ (as measured by flow cytometry) monocytes/macrophages isolated by adherence.

Cloning, Transfections, and Luciferase AssayTransient transfections were performed in a 24-well format using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s

protocol. The genomic regions of interest (Table S7) were cloned (KpnI and SacI restriction sites) in the pGL3-promoter luciferase

reporter vector (Promega).

RT-qPCR and RNA-SeqTotal RNA was extracted from 1-5x106 cells using the RNeasy Mini Kit (Quiagen), quantified with Nanodrop (ThermoScientific) and

reverse-transcribed using M-MuLV Reverse Transcriptase (Finnzymes). 10 ng of cDNA were used for amplification on an ABI 7500

machine. Primers used for RT-qPCR are listed in Table S7. mRNA-Seq library preparation from 1 mg of total RNAwas performed with

the TruSeq RNA Sample Prep kit (Illumina) according to the manufacturer’s instruction and sequenced on a Solexa GAII following

standard protocols.

Chip-Seq AnalysesShort reads obtained either from Illumina Genome Analyzer II or Illumina HiSeq 2000 were quality-filtered according to the Illumina

pipeline. Analysis of the data sets was automated using the Fish the ChIPs pipeline (Barozzi et al., 2011). This includes alignment to

the mm9 genome using Bowtie v0.12.7 (Langmead et al., 2009). All the reads with a unique match to the genome with two or fewer

mismatches (-m 1 –v 2) were retained. Peak calling was performed using MACS v1.4 (Zhang et al., 2008) with default parameters and

bw= 100 (except for the pol_II for which bw= 300was used). Each IP was compared to input DNAderived from bonemarrow-derived

macrophages (GEO accession: GSM499415). In order to identify regions completely devoid of H3K4me1 (see section Classification

of Enhancers) we used a threshold of 13 10�3 for the peak calling of the IP versus the input DNA. When calling differentially enriched

regions among treated and untreated samples, we filtered the resulting regions keeping only those found enriched against the input

as well. All the lists were annotated over Ensembl genes (Flicek et al., 2012, table downloaded from the UCSC (Fujita et al., 2011)

genome browser on 2011, June 6th) taking advantage of GIN (Cesaroni et al., 2008, priority set to ‘‘gene’’ and promoter definition

to ‘‘-20000’’).

Samples were collected in two separate ‘‘batches,’’ one for LPS and one for all the remaining stimuli. For this reason, two different

untreated samples were generated and used separately for the analyses.

Genes Showing Dynamic RNA Pol II Behavior after LPS StimulationFor the 4h LPS-treated time point, lists of induced and repressed Ensembl genes were generated based on the behavior of RNA Pol II

over the locus (defined from�2.5kbp from the annotated TSS to +2.5kbp from the annotated transcriptional end site, TES). For each

locus the number of reads within/without the locus in the treated sample was compared to the number of reads within/without the

locus in the untreated sample using a chi-square test. These p-values were corrected for multiple hypotheses testing using the Bon-

ferroni correction. Each locus was also assigned a sign, representing the log2-ratio of the number of reads (normalized according to

Cell 152, 157–171, January 17, 2013 ª2013 Elsevier Inc. S1

the sequencing depth). Loci with a corrected p-value lower than 1 3 10�3, positive sign and overlapping one or more induced RNA

Pol II regions identified by MACS were classified as induced. On the contrary, loci with a p-value lower than 1 3 10�3, negative sign

and overlapping one or more repressed RNA Pol II regions identified by MACS were classified as repressed.

RNA-Seq AnalysisAfter quality filtering according to the Illumina pipeline, paired-end reads were aligned to the mm9 reference genome and to theMus

musculus transcriptome (Ensembl build 63, Flicek et al., 2012) using TopHat (Trapnell et al., 2009). We allowed up to twomismatches

and specified a mean distance between pairs (-r) of 250 bp.

Transcript abundance was quantified using Cufflinks 1.2.1 (Trapnell et al., 2010). Differentially expressed genes were called using

Cuffdiff 1.2.1 (Trapnell et al., 2010). For subsequent analyses we considered the information at the level of genes. Genes showing an

FPKM (fragments per kilobase of exon per million fragments mapped) equal or higher than one in at least one sample (untreated or

any of the 7 stimuli) were retained for further analyses. Differentially expressed genes were defined using an FDR % 0.05. No

threshold was applied on the fold change. During transcript quantification we used options –N (which specifies for upper-quartile

normalization) and -u (which allows a better weighting of the multi-mapping reads).

Tracks for the UCSC genome browser (Fujita et al., 2011) were generated using the uniquely alignable reads. Tracks were linearly

rescaled to the same sequencing depth.

Classification of EnhancersFirst of all, we considered two different thresholds to score for the presence of a peak in each ChIP, namely 1 3 10�5 when consid-

ering a ChIP versus the input DNA; and 13 10�10 when scoring a difference (either induction or repression) between twoChIPs. In the

following paragraphs, ChIPs for H3K4me1, H3K27ac and PU.1 will be referred to as me1, ac and PU.1.

In order to define different categories of enhancers, we centered our analysis onme1 signals. Separately for both theme1 LPS time

points and for the different stimuli the me1-enriched regions (1 3 10�5 against the input) were merged together and used as view-

points for the following evaluation steps: (1) If any repressive event involving ac was found in the region, the region was defined as

repressed; (2) If neither me1 (at a very relaxed threshold, namely 1 3 10�3) nor ac nor PU.1 are found enriched against input in the

basal (untreated) state while me1 and ac are induced by stimulation, the region was defined as latent; (3) If me1 is present in the basal

state while ac is not and at the same time ac is induced by stimulation, such region was defined as poised activated; (4) If me1 is

present in the basal state while ac is not, and any stimulation considered did not affect either me1 or ac, the region was defined

as poised not activated; (5) A region that shows basal level of me1 and ac and in which ac is significantly greater after stimulation

is defined as constitutive activated; regions with the same basal state but not showing any difference in ac after stimulation are

defined as constitutive not activated.

Heatmap of LPS-Inducible Latent EnhancersThe heatmap in Figure 1C shows the counts of the reads in latent enhancers for each IP in untreated, 4h and 24h LPS. The number of

reads for each region was normalized on the sequencing depth of the least-sequenced sample. Besides, since these regions are not

homogeneous in length, counts were normalized on the size of the region in kbp. After that, considering each antibody independently,

counts exceeding the 95th percentile were set to this value. This was done in order to avoid any bias due to the outliers. Values were

then set to the range 0-1, still considering each antibody independently. Finally, regions were sorted according to their acetylation

levels.

Enhancer-Gene AssignmentEach enhancer was assigned to the nearest TSS or TES considering inducible genes only, as determined by changes in Pol_II occu-

pancy (for LPS) or by RNA-seq (for all the other stimuli).

Functional Annotations of the Enhancers Using GREATFor each class of enhancers, GREAT 1.8 (McLean et al., 2010) was run against the whole mm9 genome as background.

In the case of the LPS response, GREAT results were summarized as a heatmap (Figure 1E). We considered all the terms signif-

icantly enriched in at least one class of enhancers (Hypergeometric FDR% 0.05) for the following ontologies: GO Biological Process,

GO Molecular Function, PANTHER Pathways. FDR values were log10-transformed and the results were clustered (along different

classes of enhancers as well as functional terms) using complete linkage and 1 - Pearson Correlation as measure of distance.

Conservation Analysis of LPS-Inducible EnhancersFor each enhancer we looked for the DNA fragment with the maximum number of aligned species in a 30-way alignment (Blanchette

et al., 2004) and used this number as a proxy for the region. Starting from these values, we were able to build class-specific distri-

butions. For each class, we then randomly sampled the murine genome (excluding annotated gaps) for 10 times the number of

regions in the class, with respect to the specific genomic distribution (introns, exons, intergenic) and size of the enhancers. These

random samples were used as class-specific controls. A Mann-Whitney test was used to evaluate the statistical significance of

the differences observed in each class compared to its random matched sample.

S2 Cell 152, 157–171, January 17, 2013 ª2013 Elsevier Inc.

CircosIn order to show the relationships between stimuli in latent regions we used Circos v0.56 (Krzywinski et al., 2009). Compared to other

visualization methods, the circular ideogram layout has the advantage to display stimulus-region associations combining both color-

coded links and a heatmap. Circos is generally used for displaying variation in the structure of the genome. We generated

a completely custom configuration file in order to accommodate this non-canonical type of data. We adapted the ideogram to

our purpose, dividing it into 8 parts, thereby generating an ‘‘artificial karyotype.’’ The biggest part corresponds to the entire pool

of latent regions for the 7 stimuli and the seven smaller ones specifically match each stimulus. Before generating the custom Circos,

latent regions were sorted according to the different combinations of stimuli. We achieved a cleaner image using this sorting proce-

dure along with the ribbon visualization.

Motifs Enrichment AnalysisIn order to identify over-represented motifs corresponding to known TF binding sites, Clover (Frith et al., 2004) was run on latent and

poised enhancers activated by LPS.We chose Clover because it provides a statistics irrespectively of the distribution of size (in bp) of

the input regions. This way we were able to perform an analysis on the entire repertoires of latent and poised activated regions.

Regions were scanned with 597 models (position weight matrices, PWM) collected from the literature (Portales-Casamar et al.,

2010; Badis et al., 2009; Berger et al., 2008; Bucher, 1990; Jolma et al., 2010) using four different backgrounds: chromosome 19,

the entire set of RefSeq TSSs, the constitutive and the poised-not-activated enhancers. Significantly over-representedmotifs (empir-

ical p-value % 0.01) were very consistent irrespectively of the background used (Table S3).

For the other stimuli, we focused on the 100 bp surrounding the summit of the PU.1 binding sites (as determined by ChIP-seq). For

each stimulus-specific set of latent enhancers, Pscan (Zambelli et al., 2009) was performed using basal PU.1 peaks in constitutive

regions as background. The PWMs used in the Clover analysis were also used as input models for Pscan.

De Novo Motif DiscoveryWe ran MEME v4.6.1 (Bailey and Elkan, 1994) on the same PU.1-centered latent regions used in the Pscan analysis (results are

summarized in Table S6) as well as on the 100 bp surrounding the summit of JunB, Stat6 and Stat1 latent-overlapping peaks. We

then used TOMTOM (Gupta et al., 2007) in order to assess similarity of the motifs identified to known TF binding preferences.

Heatmaps Showing Inducible TFs at Latent EnhancersThe heatmaps in Figure 3C show the counts of the reads in latent enhancers for each IP in untreated and for each different stimulus

(4h). The number of reads for each region was normalized on the sequencing depth of the least-sequenced sample in every heatmap.

Besides, since these regions are not homogeneous in length, counts were normalized on the size of the region in kbp. Considering

each antibody independently, counts exceeding the 95th percentile were then set to this value. Finally, regions were sorted according

to the tag counts at 4h.

Time Course NormalizationAfter normalization for sequencing depth, the read counts from the ChIP-seq data sets were normalized along the different time

points using the following formula: (# reads – min) / (max�min). Min and max stand for the minimum and maximum number of reads

along the time course separately for each latent considered.

Analysis of the PU.1 Binding Sites in Different Classes of EnhancersFirst of all latent, poised activated and constitutive enhancers for the 7 stimuli were pooled with LPS-inducible regions within the

same classification. We derived PU.1-centered regions (+/� 50bp) and run MEME v4.6.1 (Bailey and Elkan, 1994) on them. We

then selected the motif showing the highest similarity to the canonical PU.1 site (Wei et al., 2010). Starting from precise binding loca-

tions as reported by MEME, we extracted the genomic DNA sequence underlying the PU.1 binding motif for the 1,050 latent, 2,400

constitutive and 2,213 poised activated enhancers. 12-nt long regions were centered on the highly conserved GGAA core

(50-NNNNNGGAANNN-30). We then assessed binding affinities for these regions using two different approaches. In the first

approach, we represented PU.1- affinity to each binding site as the affinity of the overlapping 8-mer with the highest value (as defined

experimentally by protein-binding microarrays) (Wei et al., 2010). In the second approach, we computed the score of each region to

the PU.1 PWM indicated by Heinz et al. (2010) and used the relative percentage to the best score achievable by the consensus as

representative for each region. Results from the first approach were shown as empirical cumulative distribution functions (ecdfs)

while results from the second approach were shown as box plots. In both cases, Mann-Whitney tests were carried out to assess

statistical significance of the differences.

SUPPLEMENTAL REFERENCES

Badis, G., Berger, M.F., Philippakis, A.A., Talukder, S., Gehrke, A.R., Jaeger, S.A., Chan, E.T., Metzler, G., Vedenko, A., Chen, X., et al. (2009). Diversity and

complexity in DNA recognition by transcription factors. Science 324, 1720–1723.

Bailey, T.L., Elkan, C. (1994). Fitting a mixture model by expectation maximization to discover motifs in biopolymers. Proc. Int. Conf. Intell. Syst. Mol. Biol. 2,

28–36.


Berger, M.F., Badis, G., Gehrke, A.R., Talukder, S., Philippakis, A.A., Pena-Castillo, L., Alleyne, T.M., Mnaimneh, S., Botvinnik, O.B., Chan, E.T., et al. (2008).

Variation in homeodomain DNA binding revealed by high-resolution analysis of sequence preferences. Cell 133, 1266–1276.

Blanchette,M., Kent,W.J., Riemer, C., Elnitski, L., Smit, A.F., Roskin, K.M., Baertsch, R., Rosenbloom, K., Clawson, H., Green, E.D., et al. (2004). Aligningmultiple

genomic sequences with the threaded blockset aligner. Genome Res. 14, 708–715.

Bucher, P. (1990). Weight matrix descriptions of four eukaryotic RNA polymerase II promoter elements derived from 502 unrelated promoter sequences. J. Mol.

Biol. 212, 563–578.

Cesaroni, M., Cittaro, D., Brozzi, A., Pelicci, P.G., and Luzi, L. (2008). CARPET: a web-based package for the analysis of ChIP-chip and expression tiling data.

Bioinformatics 24, 2918–2920.

Flicek, P., Amode,M.R., Barrell, D., Beal, K., Brent, S., Carvalho-Silva, D., Clapham, P., Coates, G., Fairley, S., Fitzgerald, S., et al. (2012) ENSEMBL 2012. Nucleic

Acids Res. 40, D84–D90.

Frith, M.C., Fu, Y., Yu, L., Chen, J.-F., Hansen, U., andWeng, Z. (2004). Detection of functional DNAmotifs via statistical over-representation. Nucleic Acids Res.

32, 1372–1381.

Fujita, P.A., Rhead, B., Zweig, A.S., Hinrichs, A.S., Karolchik, D., Cline, M.S., Goldman, M., Barber, G.P., Clawson, H., Coelho, A., et al. (2011). The UCSC

Genome Browser database: update 2011. Nucleic Acids Res. 39, D876–D882.

Gupta, S., Stamatoyannopoulos, J.A., Bailey, T.L., and Noble, W.S. (2007). Quantifying similarity between motifs. Genome Biol. 8, R24.

Jolma, A., Kivioja, T., Toivonen, J., Cheng, L., Wei, G., Enge, M., Taipale, M., Vaquerizas, J.M., Yan, J., Sillanpaa, M.J., et al. (2010). Multiplexedmassively parallel

SELEX for characterization of human transcription factor binding specificities. Genome Res. 20, 861–873.

Krzywinski, M., Schein, J., Birol, I., Connors, J., Gascoyne, R., Horsman, D., Jones, S.J., andMarra, M.A. (2009). Circos: an information aesthetic for comparative

genomics. Genome Res. 19, 1639–1645.

Langmead, B., Trapnell, C., Pop,M., and Salzberg, S.L. (2009). Ultrafast andmemory-efficient alignment of short DNA sequences to the human genome. Genome

Biol. 10, R25.

Portales-Casamar, E., Thongjuea, S., Kwon, A.T., Arenillas, D., Zhao, X., Valen, E., Yusuf, D., Lenhard, B., Wasserman, W.W., Sandelin, A. (2010). JASPAR 2010:

the greatly expanded open-access database of transcription factor binding profiles. Nucleic Acids Res. 38, D105–D110.

Trapnell, C., Pachter, L., and Salzberg, S.L. (2009). TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 1105–1111.

Trapnell, C., Williams, B.A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M.J., Salzberg, S.L., Wold, B.J., and Pachter, L. (2010). Transcript assembly and

quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat. Biotechnol. 28, 511–515.

Zambelli, F., Pesole, G., Pavesi, G. (2009). Pscan: finding over-represented transcription factor binding site motifs in sequences from co-regulated or co-ex-

pressed genes. Nucleic Acids Res. 37(Web Server issue), W247–W252.

Zhang, Y., Liu, T., Meyer, C.A., Eeckhoute, J., Johnson, D.S., Bernstein, B.E., Nusbaum, C., Myers, R.M., Brown, M., Li, W., and Liu, X.S. (2008). Model-based

analysis of ChIP-Seq (MACS). Genome Biol. 9, R137.


Figure S1. Chromatin and Functional Features of Latent Enhancers, Related to Figure 1

(A) Lack of enhancer features in unstimulated cells at a latent enhancer in the Ptgs2 locus.

(B) Chromatin associated RNAs (data from Bhatt et al., 2012) and RNA Pol II at intergenic enhancers.

(C) Latent enhancers are closer to assigned genes than other categories of enhancers.

(D) Sequence conservation analysis of LPS-induced latent enhancers (upper box plot, in red). In themiddle (orange) and lower (yellow) panels similar analyses are

reported for comparison relative to poised enhancers activated by LPS and constitutive enhancers.

(E) Functional validation of latent enhancers using a luciferase assay in Raw264.7 cells. A.L.U.: absolute luciferase units.


Figure S2. Two Genomic Regions Showing Examples of Latent Enhancers with Different Specificities, Related to Figure 2

The black arrows indicate a representative latent enhancer that is either selectively activated by IFNg (left panel, near the Cd86 gene) or by multiple proin-

flammatory stimuli (right).


Figure S3. Role of Smad3 in the Formation of Latent Enhancers, Related to Figure 3

ChIP analysis of recruitment of the TF Smad3 to representative latent enhancers induced by TGFb. Data are expressed as mean ± SEM.


Figure S4. Kinetics of Latent Enhancer Formation, Related to Figure 4

(A) Recruitment kinetics of cytokine-activated STATs to a representative set of latent enhancers.

(B) Rapid Stat6 recruitment to the promoter of a representative early gene induced by IL4 stimulation (Ccl12). Data are expressed as mean ± s.e.m.

(C) Analysis of H3 occupancy at a representative latent enhancer. H3 ChIP data (shown as percentage of input or as fold change) obtained with primer pairs

spanning the same region shown in Figure 4C. The central position (0) corresponds to the summit of the inducible Pu.1 peak. Data are expressed asmean ± s.e.m.

(D) Time-resolved ChIP-seq analysis (H3K4me1, Pu.1 and Stat1) of latent enhancer formation in response to IFNg stimulation. A representative snapshot is shown

on the left, with an emerging latent enhancer indicated by an arrow. It should be noticed that in this early window of time, H3K4me1 deposition only starts to be

evident, while Stat1 recruitment begins at early time points and progressively increases. Box plots on the right indicate the relative tag counts at all latent

enhancers activated by IFNg.


Figure S5. Role of RNA Pol II and New Protein Synthesis in Latent Enhancer Unveiling, Related to Figure 5

(A) A genomic snapshot showing an LPS-dependent latent enhancer (see black arrow) surfaced within the transcribed region of an inducible gene (Upp1).

(B) Inducible Pu.1 recruitment at a latent enhancer located within the Nos2 gene in the presence or absence of the RNA Pol II elongation inhibitor DRB. Data are

expressed as mean ± s.e.m.

(C) Effect of the protein synthesis inhibitor puromycin on de novo induction of H3K4me1 at IL-4- or IFNg-activated latent enhancers.


Figure S6. Repressive Marks and Persistence of Latent Enhancers, Related to Figure 6

(A) H3K27me3 andH3K9me3ChIP on a representative panel of latent enhancers unveiled by LPS. The first column on the left in each graph is a positive control for

the relative modification. Data are expressed as mean ± s.e.m.

(B) Persistent H3K4me1 at latent enhancers after stimulus washout. Macrophages were primed with IL-4 for 8 hr, followed by cytokine washout and incubation in

fresh medium for the indicated period. 5 days after the washout, morphological changes and cell death were evident, impeding longer-term experiments.

H3K4me1 was detected by ChIP-qPCR on three representative latent enhancers. Data are expressed as mean ± s.e.m.


Latent Enhancers Activated by Stimulation in ...Latent Enhancers Activated by Stimulation in...

Documents

Transcript of Latent Enhancers Activated by Stimulation in ...Latent Enhancers Activated by Stimulation in...