Combinatorial Recruitment of CREB, C/EBPβ and c-Jun Determines Activation of Promoters upon Keratinocyte Differentiation

Background Transcription factors CREB, C/EBPβ and Jun regulate genes involved in keratinocyte proliferation and differentiation. We questioned if specific combinations of CREB, C/EBPβ and c-Jun bound to promoters correlate with RNA polymerase II binding, mRNA transcript levels and methylation of promoters in proliferating and differentiating keratinocytes. Results Induction of mRNA and RNA polymerase II by differentiation is highest when promoters are bound by C/EBP β alone, C/EBPβ together with c-Jun, or by CREB, C/EBPβ and c-Jun, although in this case CREB binds with low affinity. In contrast, RNA polymerase II binding and mRNA levels change the least upon differentiation when promoters are bound by CREB either alone or in combination with C/EBPβ or c-Jun. Notably, promoters bound by CREB have relatively high levels of RNA polymerase II binding irrespective of differentiation. Inhibition of C/EBPβ or c-Jun preferentially represses mRNA when gene promoters are bound by corresponding transcription factors and not CREB. Methylated promoters have relatively low CREB binding and, accordingly, those which are bound by C/EBPβ are induced by differentiation irrespective of CREB. Composite “Half and Half” consensus motifs and co localizing consensus DNA binding motifs are overrepresented in promoters bound by the combination of corresponding transcription factors. Conclusion Correlational and functional data describes combinatorial mechanisms regulating the activation of promoters. Colocalization of C/EBPβ and c-Jun on promoters without strong CREB binding determines high probability of activation upon keratinocyte differentiation.

The mechanism that allows for the selective regulation of genes involved in diverse cell functions such as proliferation and differentiation using the same transcription factors is still unknown. Differential expression of AP-1 family members c-Jun/JunB, Fra2/cFos and C/EBPs family members C/EBPa and C/EBPb contribute to the modulation of gene activities upon differentiation [11,22], [13,25]. Since the DNA binding domains of these transcription factors within their families are identical [26], exchanges of the transcription factor bound to DNA will occur at the regulatory elements of genes that are (or are not) activated by differentiation. One possible explanation for selective activation of promoters by transcription factors is the binding of heterodimers of AP-1 and/or C/EBPb to composite elements [27][28][29][30]. In this case, the c-Jun-C/EBPb heterodimer represses transcription [31], while promoters bound by C/EBPb alone can be activated [32].
CREB is another transcription factor important for keratinocyte proliferation [2], [33], [34], [23]. In different cell types CREB is bound to promoters of nearly the same set of genes responsible for cell survival and cell cycle progression [35]. Inhibition of CREB leads to the repression of skin tumor initiation in mice and repression of cell cycle progression [33]. The role of CREB in keratinocyte differentiation is not well studied, although it has been shown that inhibition of CREB by A-CREB dominant negative represses both CREB and AP-1 reporter activities [34]. While some studies reported that CREB protein level is induced by keratinocytes differentiation [34], other show that it is repressed [2].
Our previous work has demonstrated that DNA methylation is important for activation of some tissue specific genes involved in keratinocyte differentiation [45]. Notably, although methylation inhibits CREB binding to its consensus binding site, C/EBPb and c-Jun can bind the methylated CREB binding site. Also, methylation of the consensus C/EBP DNA binding site increases its affinity to C/EBP [45]. Thus, the DNA methylation status of promoters may play a role in differential recruitment of CREB, c-Jun and C/EBPb.
We hypothesized that CREB, c-Jun and C/EBPb are functioning differently when they bind to promoters in different combinations. Because CREB is important for cell survival, we hypothesized that C/EBPb and c-Jun regulate genes involved in cell survival and proliferation if they co-localize with CREB and regulate genes involved in differentiation when they do not. We test this hypothesis by analyzing how genome-wide promoter binding of transcription factors CREB, C/EBPb and c-Jun in different combinations correlate with either the activation or repression of promoters and their target genes upon keratinocyte differentiation.

Upon Keratinocyte Differentiation RNAP is Preferentially Recruited to Promoters bound by a Combination of C/ EBPb and cJun
In order to determine factors correlating with activation of promoters as measured by RNA polymerase II (RNAP) binding, we determined localization of three transcription factors -CREB, C/EBPb and c-Jun, as well as RNAP, in proliferating and differentiated keratinocytes using chromatin immunoprecipitation followed by microarray hybridization. Despite close to 100% immunoprecipitation efficiency, c-Jun had lower DNA binding levels than the other proteins ( Figure S1).
Upon keratinocyte differentiation, RNAP binding increased at 797 promoters and decreased at 841 promoters ( Figure 1A). Out of 17928 genes for which we have both promoter binding data and mRNA expression levels, we observed that 1326 (7.4%) have increased mRNA levels and 1458 (8.1%) genes are repressed upon keratinocyte differentiation. This increase correlates with induction of RNAP binding ( Figure 1A). In contrast, after keratinocyte differentiation the binding of CREB, C/EBPb and c-Jun does not change dramatically ( Figure S1A). The highest increase of binding was observed for, C/EBPb (Figure S1 A, B). Scatterplots of changes in RNAP binding versus changes in transcription factor binding upon differentiation show that induction of RNAP upon differentiation is overrepresented among promoters with induction of C/EBPb binding ( Figure S1B). We did not observe this for c-Jun or CREB. Induction of RNAP is also preferentially observed in promoters bound by C/EBPb and c-Jun, while promoters bound by CREB are predominantly repressed upon differentiation ( Figure S1C). Thus, changes of RNAP binding upon differentiation are not determined only by binding of CREB, C/EBPb or c-Jun to promoters.
We next analyzed if promoters bound by different combinations of transcription factors are preferentially repressed or induced upon differentiation. CREB, C/EBPb and cJun bind distinct set of promoters in differentiated keratinocytes ( Figure 1B). Among those, promoters bound by C/EBPb and c-Jun are overrepresented in those which also have induction of RNAP binding upon differentiation ( Figure 1B). Figure 1C shows the fraction of promoters bound by different combination of transcription factors in differentiated keratinocytes where RNAP binding or mRNA levels are repressed (white bars) or induced by differentiation (black bars). Induction of RNAP binding and mRNA by differentiation is the lowest for promoters bound by CREB alone or in combination with c-Jun or C/EBPb, and the highest for C/ EBPb alone, C/EBPb, cJun and CREB, C/EBPb, c-Jun bound promoters. In particular, 20% of promoters bound by C/EBPb and cJun are induced upon differentiation, five times more than expected by chance ( Figure 1C).
We found the same absolute values of RNAP induction between groups ( Figure S2A); however, induction of mRNA upon differentiation is highest when promoters are bound by c-Jun or C/EBPb and lowest for genes with CREB-bound promoters (Figure S2 B).
Transcription factors are known to recruit histone acetyltransferases that change chromatin accessibility and transcriptional activation. Therefore, we next investigated whether H3K9 acetylation preferentially altered upon differentiation for promoters bound by specific combinations of transcription factors. As expected, H3K9 acetylation correlates with RNAP binding in differentiated ( Figure S3A) and undifferentiated keratinocytes. Although, relatively wide range of RNAP binding for given H3K9 acetylation level suggests that RNAP binding is also regulated by other factors. Likewise, changes in RNAP binding upon differentiation positively correlate with changes in H3K9 acetylation ( Figure S3B). We found 800 promoters with induced H3K9 acetylation and 186 that also have induced RNAP binding upon differentiation. Similar to what we observed for RNAP binding, H3K9 acetylation is preferentially induced on promoters bound by C/EBPb and cJun and promoters bound by CREB, cJun and C/ EBPb ( Figure S3C). Promoters bound by C/EBPb and CREB and promoters bound by cJun and CREB have lower levels of H3K9 acetylation induction compared to promoters bound by C/EBPb and c-Jun ( Figure S3D).

Colocalization of C/EBPb or c-Jun with CREB Determine Genes which Expression and Induction upon Differentiation is Dependent on C/EBPb and c-Jun Functions
C/EBPb binding to promoters is induced upon keratinocyte differentiation and RNAP binding is preferentially induced to promoters bound by C/EBPb and cJun. In order to determine genes that are dependent upon C/EBPb or cJun function we expressed dominant negatives A-C/EBP and A-Fos using a tetracycline driven expression system [17], [10]. A-C/EBP and A-Fos have been extensively used to study these proteins in a variety of systems [10,17,34,36,38]. We find that A-C/EBP most strongly represses genes whose promoters are bound by C/EBPb and not CREB ( Figure 2A). Likewise, A-Fos dominant negative preferentially represses genes whose promoters are bound by c-Jun alone or in combination with C/EBPb (Figure 2A). Because differentiation preferentially induces the expression of genes whose promoters are bound by C/EBPb and c-Jun, we tested if inhibition of C/EBPb or cJun would preferentially repress them. We found that A-C/EBP and A-Fos both repress 25% of promoters induced in differentiation, over three times what would be expected by chance, suggesting that differentiation is dependent on these transcription factors. Similar to what is observed for all promoters, A-Fos and A-C/EBP preferentially inhibit genes induced by differentiation if their promoters are bound only by cJun or C/ EBPb ( Figure 2B); however the number of induced genes whose promoters are also bound by combinations of transcription factors is relatively low, so many of the differences do not reach statistical significance.

CREB Binding is Relatively Low for Promoters bound by CREB, C/EBPb and cJun and Induced by Differentiation
Because CREB is rarely bound to promoters induced by differentiation, we hypothesized that CREB binding affinity might be lower in promoters bound by CREB C/EBPb and cJun and induced by differentiation. Comparison of transcription factor binding distributions in all promoters and promoters bound by CREB, C/EBPb and cJun in differentiated keratinocytes revealed lower levels of CREB binding in promoters bound by all three of these transcription factors and induced by differentiation ( Figure 3A). We arbitrarily selected four genes based on different combinatorial recruitment of transcription factors and induction of RNAP binding upon differentiation and show the ChIP-chip binding patterns for RNAP, CREB, C/EBPb and c-Jun across their promoters ( Figure S4). Claudin4, SPRP1A and DNAse1L3 with no or low CREB binding were induced by differentiation. In contrast, Ribosomal protein 19 with high CREB binding was not. We confirmed binding of transcription factors by Chip-PCR for SPRP1A and DNAse1L3. CREB binding was low in comparison with C/EBPb and cJun. ( Figure 3B).

Colocalization of C/EBPb and cJun with CREB is Associated with a High Probability of RNAP Binding in Proliferating or Differentiated Keratinocytes
Colocalization of C/EBPb and cJun with CREB makes promoters refractory to induction with calcium and to repression by dominant negatives. CREB is known to bind nearly the same set of promoters in different cells [35] suggesting that CREB is involved in housekeeping functions of the cell. These promoters frequently contain CpG islands and typically are highly active. We hypothesized that when C/EBPb or c-Jun co-localize with CREB, these promoters are strongly bound by RNAP and are not induced by differentiation because other CG binding factors like CREB are already activating them.
Indeed, scatter plots of transcription factors and RNAP DNA binding ( Figure 4A) showed that 93% of promoters bound by CREB, 84% of promoters bound by C/EBPb and 78% of promoters bound by c-Jun are also bound by RNAP in differentiated keratinocytes ( Figure 4A, Figure S5). Analysis of RNAP binding in groups of promoters bound by different combination of transcription factors showed that colocalization of C/EBPb or c-Jun with CREB corresponds to a high probability of RNAP recruitment (Figure 4 B). 44% of c-Jun only and 72% of C/EBPb only bound promoters bind RNAP (Table S1). In contrast, when C/EBPb or c-Jun colocalizes with CREB, about 90% of these promoters bind RNAP (Table S1). Similarly, in undifferentiated keratinocytes colocalization with CREB determines high probability of RNAP recruitment ( Figure S5, Table  S1). Likewise, when c-Jun colocalizes with C/EBPb, the distribution of RNAP binding is the same as is observed for promoters bound by C/EBPb only ( Figure 4B). For promoters induced by differentiation we found higher RNAP binding for promoters bound by CREB than bound by C/EBPb and/or c-Jun ( Figure 4C). Despite low CREB binding ( Figure 3A), promoters induced by differentiation and bound by CREB, C/EBPb and c-Jun ( Figure 3A) have higher RNAP binding than, promoters induced by differentiation and bound by C/EBPb and c-Jun only ( Figure 4C).

Gene Promoters that are Methylated and bound by C/ EBPb are Preferentially Induced by Differentiation
It was shown that DNA methylation is important for activation of some tissue specific genes involved in keratinocytes differentiation [45]. C/EBPb and c-Jun can bind methylated CREB binding site, although CREB can not [45]. Thus, the DNA methylation status of the promoters may affect their activation through the combinatorial recruitment of CREB, c-Jun and C/ EBPb. Indeed, C/EBPb and c-Jun preferentially bind methylated promoters ( Figure 5A). Fraction of methylated promoters was 54% for c-Jun and C/EBPb and less than 14% for promoters where CREB is bound (Table S1).
Similar to what is observed for all promoters, unmethylated promoters containing C/EBPb and c-Jun are mostly induced by differentiation while those that are also bound by CREB are not ( Figure 5B). The fraction of genes for which mRNA is induced by differentiation is higher for methylated in comparison to unmethylated promoters bound by C/EBPb. 30% of methylated promoters compared to 10% for unmethylated that are bound by both CREB and C/EBPb are induced by differentiation. (Figures 5B, 5C right panels). Similar to what is observed for promoters induced by differentiation and bound by CREB, c-Jun and C/EBPb (Figure 3), methylated promoters have relatively low CREB binding ( Figure S6). To address functionality of DNA methylation in gene activation upon differentiation, we treated cells with DNA demethylation agent 5-azacytidine (5-AZA) and measured mRNA levels in treated v.s. untreated cells in differentiating keratinocytes. 5-AZA preferentially represses genes whose promoters are bound by C/EBPb and/or c-Jun but not CREB ( Figure S7). Notably, among genes repressed by 5-AZA, bound by C/EBPb and inhibited by A-C/EBP are markers of keratinocyte differentiation desmocolline and small proline rich protein like 9.

Consensus and Composite Motifs Containing Two Parts of Consensus Motifs are Enriched in the Groups of Promoters bound by Different Combinations of Transcription Factors
The presence of specific DNA sequences which bind transcription factors is the major mechanism of transcription factor recruitment to the promoters. We hypothesized that specific combinations of DNA binding motifs recruit corresponding combinations of CREB, c-Jun and C/EBPb. To identify these combinations we calculated motifs overrepresented in sets of promoters bound by transcription factors alone and in different combinations. First, we used all promoters as a background and report the top two enriched motifs in Figure 6A (top seven rows). We found that the most enriched motifs correspond to the known consensus DNA binding sequences. The first DNA binding motif identified for CREB-only was ETS-CREB composite motif, also described in [46]. The second DNA binding motif identified for CREB was ACTACAnnTCCCA and represents a composite ZFP143-RBPJ binding site [47]. For promoters bound by c-Jun and C/EBPb we found AP1 (TGACTCA) and C/EBPb (TTGCGCAA ) consensus sequences. For promoters bound by combination of CREB and C/EBPb we found ACTA-  CREB, c-Jun and C/EBPb can bind promoters via separate binding sites; compete for the same DNA binding motifs or bind DNA as heterodimers. In improve identification of motifs that can recruit two transcription factors, we used as a background promoters bound by one transcription factor (CREB or c-Jun or C/EBPb, lines 8-13) and calculated motifs enriched in promoters bound by combination of two transcription factors. This analysis identified composite motif containing half C/EBP and half CRE site and C/EBP consensus site for promoters bound by C/EBPb and CREB relative to CREB. Inversely, when promoters bound by C/EBPb were used as a background for the same test site we identified CREB binding site and CTACANNTCCC sequence. Likewise, the same calculation performed for the promoters bound by c-Jun and C/EBPb identified TRE site relative to the C/EBPb set. Interestingly second identified motif (GACAAGTTAGGTCA) was the same when promoters bound by C/EBP or c-Jun were used as a background. The same motif was identified for c-Jun and C/EBPb bound promoters relatively to c-Jun. Motifs enriched in CREB -c-Jun bound promoters relative to c-Jun or CREB represent known binding motifs for corresponding transcription factors. The same calculations performed for combination of three transcription factors identified consensus motifs for c-Jun and CREB and TCCCANNNCCTC for C/EBPb.

Consensus DNA Binding Motifs are Preferentially Colocalize in Promoters bound by Combinations of Corresponding Transcription Factors
Data in Figure 6A suggest that consensus and combination of consensus motifs for transcription factors are enriched in the groups of promoters bound by combination of corresponding transcription factors. We asked if consensus motifs co-occur in the same promoters. Consensus motifs for CREB (CRE) TGACGTCA, C/EBPb TTGCGCAA and c-Jun (TRE) TGA(C/G)TCA were used. We found that promoters containing combinations of consensus DNA binding motifs are overrepresented in the promoters bound by corresponding combinations of transcription factors ( Figure 6B). This analysis suggests that combinatorial sited and combination of consensus transcription factor binding sites lead to specific

Discussion
Our results suggest that during keratinocyte differentiation, the activation of gene expression is regulated by promoter-specific combinations of CREB, c-Jun and C/EBPb. Specifically, C/EBPb binding is induced upon differentiation and promoters bound by C/EBPb, either alone or in combination with c-Jun with no or low CREB binding are preferentially induced by differentiation. Colocalization of c-Jun and C/EBPb with CREB on the promoter corresponds to a high probability of RNAP recruitment in differentiated or proliferating keratinocytes. Moreover, relatively large fraction of genes where CREB is bound to promoters alone or in combination with C/EBPb has higher expression in proliferating compared to differentiated keratinocytes.
Dominant negatives A-C/EBP and A-Fos inhibited expression of genes whose promoters are not bound by CREB. The role of C/EBPs and AP-1 in keratinocytes differentiation and proliferation is well established [2], [5][6][7][8], [22], [25]. Our data confirmed our initial hypothesis that C/EBPb and c-Jun regulate keratinocyte differentiation when they do not colocalize with CREB. The fact that promoters bound by combination of C/EBPb and c-Jun are preferentially induced by differentiation is new and was not expected. This data also suggest that CREB works as a dominant transcription factor both in inducing gene expression and preventing it from being affected by other transcription factors. The existence of small group of promoters that are bound by CREB but not RNAP suggests that some regulatory mechanisms prevent them from being activated. CREB is known to bind the same set of promoters in different cell types suggesting that it is involved in housekeeepig cell functions [35]. It is also known to mediate IL-1 induced Fos expression in keratinocytes [48], stimulate transcription of gluconeogenic genes in liver [49], involved in memory formation [50] as well as other cell type specific functions. Concordantly, our data shows that promoters bound by CREB, c-Jun and C/EBPb are preferentially induced upon differentiation. Thus, it is the combinatorial recruitment of transcription factors that determines whether a gene is going to be repressed, activated or experience no change in response to specific stimulus.
Our data identified only part of combinatorial logic which regulates genes expression during keratinocytes differentiation. Existence of promoters bound by the same combination of transcription factors and either induced or repressed by differentiation suggest that other factors (besides CREB identified in this study) bound to promoters together with C/EBPb and c-Jun will determine if specific gene is ultimately induced, repressed or do not change upon differentiation. Many other transcription factor are involved in keratinocytes differentiation [1], [2], [3], [4] and, it would be interesting to identify whether they are functioning on the same promoters cooperatively activated by C/EBPb and c-Jun.
Notably, analysis of combinatorial recruitment of CREB, c Jun and C/EBPa generated similar results ( Figure S8 and Table S1).
Presence of specific DNA sequences in the promoter determines recruitment of corresponding transcription factors for regulation of gene expression. CREB, c-Jun and C/EBPb can compete for the same binding sites or bind simultaneously at different sequences. c-Jun, C/EBPb and CREB bind to TGACTCA, TTGCGCAA and TGACGTCA consensus sequences respectively. CREB and c-Jun can bind the same TGACGTCA [13], [35], [51] sequence and C/ EBPa -c-Jun heterodimer binds TTGCGTCAT sequence [30], whose core element, CGTCA, also can be bound by CREB [28,29,35]. DNA methylation can regulate differential binding of transcription factors. For example, methylation of CRE inhibits binding of CREB but promotes binding of C/EBPb and C/EBPa and does not influence c-Jun binding [45]. Data presented in this paper suggest that combinatorial recruitment of transcription factors induces activation of genes during differentiation in a different manner for methylated compared to unmethylated promoters. CREB binding to methylated promoters bound by C/EBPb was low and these promoters were more often induced by differentiation than unmethylated promoters bound by C/ EBPb and CREB.
To understand how sequences of promoters determine preferential recruitment of CREB, c-Jun and C/EBPb in different combinations, we identified DNA motifs overrepresented in promoters bound by one, two or all three of these transcription factors relative to all promoters or relative to promoters bound by one or two transcription factors. As expected, we found that consensus binding sites are overrepresented in promoters bound by single transcription factor or in groups where corresponding transcription factors co-localize with other transcription factors. As expected, when promoters bound by a particular transcription factor were used as a background, subsets of promoters bound both by this and another transcription factor where enriched in the other transcription factor's consensus sequence. The composite motif between CREB and C/EBPb binding motif identified in our analysis was similar to the one reported in [28,29]. Several studies suggest that CREB and C/EBPb do not interact but compete for the same sequence [28], [26]. Interestingly, novel long motif CCCACCATGCTTTTGGTCA is identified in promoters bound by C/EBPb and c-Jun. Similar composite C/EBPb -c-Jun motif GACAAGTT(T/A)GGTCA is enriched in promoters bound by C/EBPb and c-Jun relatively either to c-Jun or C/EBPb bound promoters. This suggests that it is the C/EBPb/c-Jun protein complex that binds to this motif, similar to what is described in papers [30,31].
This paper uncovered combinatorial rules of transcription factor recruitment that determine activation of gene expression upon differentiation.

Ethics Statement
All mouse experiments were approved by the Animal Ethics Committee at the National Institutes of Health, under the approved protocol number LM-076. Newborn mice were handled and humanely sacrificed by carboxyl dioxide followed by secondary physical method in accordance with the National Institutes of Health Institutional Guidelines (NCI, NIH, Bethesda, MD, USA).

Primary Keratinocytes Cultures
Keratinocytes were cultured as describe in [52] from new-born wild type FVB mice or from mice expressing A-C/EBP or A-Fos dominant negatives of C/EBP's or AP-1 in keratinocytes under the control of tetracycline inducible repressor [10], [17]. Primary keratinocytes were seeded at a density of one mouse epidermis per 10 cm dish or equivalent in calcium and magnesium free S-MEM (GIBCO Laboratories, Grand Island, N.Y), supplemented with 8% Chelex (Bio-Rad, Richmond, CA) treated FBS (Atlanta Biologicals, Inc) and 0.2 mM calcium (CaCl 2 ). After 24 hours, cultures were washed twice with PBS and switched to medium with 0.05 mM calcium (low calcium). At the same time doxycyclin added to the media of A-C/EBP or A-Fos cultures was removed to induce expression of A-C/EBP or A-Fos. 24 hours later, cultures were switched to the same medium with 0.4 mM calcium (Hi calcium) for two days to induce differentiation or maintained in parallel as undifferentiated keratinocytes in the low calcium medium. For demethylation experiments, 5-azacytidine (5-aza) was added 6 hr after seeding to the culture medium to a final concentration of 1 mM. Cell culture medium was replaced every day.

Chromatin Immunoprecipitation (ChIP)
ChIP experiments were performed following the protocol from Farnham's group (http://farnham.genomecenter.ucdavis.edu/ protocol.html). Primary cultured cells were chemically cross-linked by adding 0.6% formaldehyde (Sigma) directly to the medium. Cells were allowed to cross-link for 10 minutes with gentle swirling at room temperature. The cross-linking reaction was stopped by adding 125 mM glycine and dishes were swirled for 5 minutes at room temperature. Cells were washed twice with ice-cold PBS and harvested in ice-cold PBS containing protease (Complete mini, Roche) and phosphatase inhibitors (1 mM NaF, 1 mM Na 3 VO 4 ). Cells were pelleted by centrifugation at 4uC for 5 minutes at 300 g. Cells where incubates in cell lysis buffer (5 mM PIPES pH 8.0 85 mM KCL 0.5% NP40 1 mM NF 1 mM NaVa Roche protease inhibitors cocktail) and resuspended in nuclear lysis buffer (50 mM Tris-Cl pH 8.1 10 mM EDTA, 1% SDS proteases and phosphates inhibitors as above) and sonicated to average size of 500 bp. After centrifugation supernatant was diluted 3 times by dilution buffer (0.01% SDS, 1.1% Trition6100, 1.2 mM EDTA, 16.7 mM Tris-Cl pH 8.1, 167 mM NaCl). 300 l of sonicated chromatin preparation with protein concentration 1-2 mg/ml (determined using BCA, PIERCE) was incubated overnight with antibodies. ChIP was performed using antibodies against RNA pol II (RNAP) (20 mg/ml final, Covance, 8WG16) that recognizes the unphosphorylated form of RNAP, CREB (2 mg/ml of the each antibody from Santa Cruz (sc-186) and Upstate (06-863)), H3K9 acetyl from Upstate 06-942, c-Jun from Santa Cruz (10 ug/ml, sc-1694) and C/EBPb from Santa Cruz (10 mg/ml, sc-150). A fraction of lysate was left untreated to serve as an input control. Immunocomplexes were captured using protein G agarose beads (Invitrogen) blocked with 1 g/l yeast tRNA and BSA (Sigma) and washed twice with the buffer containing 2 mM EDTA, 100 mM Tris-Cl pH 8.0 and 0.18% Sarkosyl, and four times with the Chip wash buffer (100 mM Tris-Cl pH 8.5, 500 mM LiCl, 1% NP40, 1% deoxycholic acid) and two times with TE. After incubating with RNAse A and Proteinase K, DNA was eluted using QIAquick PCR Purification Kit. From 1-2610 6 primary keratinocytes for RNAP, CREB, and C/EBPb, 0.5-2 ng of DNA were typically immunoprecipitated. DNA quantification was performed using Picogreen DNA quantification kit (Invitrogen, USA). PCR was performed using RedMix Tag polymerase (Sigma).

ChIP DNA Amplification and Hybridization
Primers conjugated with Cy3 or Cy5 (Sigma Genosys, USA) were used for the amplification of the input or immunoprecipitated DNA using round A/B/C random amplification protocol: http://research.stowers-institute.org/gertonlab/protocols/ RandomPCRamplification.pdf. Only one round of amplification was used. After amplification 5-6 mg of ChIP or control DNA was purified using PCR Purification Kit (Qiagen), isopropanol precipitated, and vacuum dried for 5 minutes. DNA was dissolved in 3 ml water, mixed with Component A and Hybridization buffer (Nimblegen) according to manufacturer's instructions. Amplified ChIP DNA was hybridized overnight to Nimblegen MM5 min Mouse promoter microarrays containing 400,000 oligos interrogating 26,264 promoters in MAUI hybridization station (BioMicro Systems) and washed according to manufacturer instruction. Arrays were dried by centrifugation and scanned using Axon 4000 B scanner. Images were processed with NimbleScan (Nimblegen) using default settings.

Calculation of Transcription Factors DNA Binding
Nimblegen MM5 min promoter arrays contain probes for 26,264 promoter regions spanning approximately 21000 bp…+500 bp relative to the transcription start site. We limited our data analysis to 20,328 promoters. We excluded promoters that are located on the X or Y chromosomes, 3,940 promoters were further eliminated that either had unsequenced DNA segments larger than 150 bp or which contained large regions of DNA sequence ($150 bp) which were identical to regions in 10 or more of other promoters. Binding of given transcription factor was defined as an average enrichment of ChIP DNA over input DNA for the whole 21000…+500 promoter region. Binding from individual replicates were averaged. Promoters were defined as bound by several transcription factors if binding of these factors were higher than specific thresholds.

Analysis of DNA Methylation
MeDIP experiments were described in Rishi [45].
Affymetrix Gene Expression Profiling mRNA expression profiling with Affymetrix microarrays (Mouse genome 430 2.0 array) was performed by NCI microarray core facility (Frederick). We compared the mRNA expression levels of genes determined using mRNA expression arrays to the binding of proteins to the promoters regions of those genes, determined by ChIP-chip data collected using Nimblegen promoter chips. Using GenBank Accessions, Gene Symbols, UniGene Clusters and UniGene IDs the mRNA expression data was mapped to the same promoter set as the ChIP-chip data. Genes in the expression data which shared a common identifier with promoters in the ChIP-chip data were assigned to their matching promoters in the ChIP-chip set. When multiple data points from the mRNA expression data were all mapped to the same promoter in the ChIP-chip data, the average of these points was assigned to that promoter. 17,930 of the 20,328 promoters used in our ChIP-chip analysis were successfully assigned mRNA expression values. Threshold for mRNA changes upon differentiation -more than 1.4 or 0.5 in Ln2 scale.

DNA Motif Analysis
DNA Motif analysis was performed using CisFinder http:// lgsun.grc.nia.nih.gov/CisFinder/ [53]. For calculation of motifs colocalization, number of motifs colocalizing in groups of promoters bound by different combinations of CREB, C/EBPb and c-Jun was normalized to number of colocalizing motifs in all promoters.

Statistical Analysis
Excel ''Chitest'' function was used to calculate significance of observed values from expected ones based on the total number of events. Excel two tailed ''Ttest'' function was used to determine whether two samples are likely to have come from the same two underlying populations that have the same mean.

GEO Accession Number
GSE48383. Table S1 Numbers and fractions of promoters bound by different combinations of transcription factors and induced or repressed by differentiation. Reference groups used to calculate fractions in different groups are also presented. Above row 64 is the data for C/EBPb, CREB and c-Jun and below row 67 is the data for C/EBPa, CREB and c-Jun. (XLS) Author Contributions