SRF Co-factors Control the Balance between Cell Proliferation and Contractility

1Ras-ERK signaling is critical for control of proliferation, invasion, and metastasis. Ras activation stimulates cell-cycle re-entry from quiescence, and Ras genes are mutated in around 20% of human cancers. Ras-ERK signaling plays an important role in this immediate-early (IE) transcriptional response to mitogen stimulation, which activates many genes encoding transcription factors, including Myc and members of the AP1 and Egr families. Many IE genes are controlled by the transcription factor serum response factor (SRF, Srf), which is required for Ras-induced cell-cycle re-entry but not for proliferation per se.

2SRF acts in partnership with two families of signal-regulated cofactors. The three ternary complex factors (TCFs) Elk-1, Net, and SAP-1 (Elk1, Elk3, and Elk4) are Ets domain proteins, regulated by Ras-ERK signaling, whereas the myocardin-related transcription factors (MRTFs) Mkl1 and Mkl2 respond to the Rho-actin pathway. The TCFs and MRTFs interact competitively with the SRF DNA-binding domain. Whether cofactor competition is a general feature of SRF regulation in vivo has been unclear. In fibroblasts, some IE genes appear to be specifically coupled to one pathway or the other, but in smooth muscle cells, platelet-derived growth factor (PDGF) can induce cofactor exchange.

3Genetically, the TCFs are at least partly functionally redundant and may also function independently of SRF. Although they have been implicated in proliferation and cancer, the extent to which the immediate-early transcriptional response is TCF-dependent, and the target genes involved, has not been systematically investigated. In contrast, the MRTFs mediate morphogenetic, adhesive, and motile processes.

4We showed previously that much of the serum-induced immediate transcriptional response is MRTF/SRF-dependent; however, the lack of specific TCF inhibitors and the relatively poor quality of TCF chromatin immunoprecipitation (ChIP) data precluded rigorous analysis of the role of TCF-SRF signaling. Here we used wild-type and triply TCF-deficient mouse embryonic fibroblasts (MEFs) to directly address the role of the TCFs in the transcriptional response to 12-O-tetradecanoyl phorbol-13-acetate (TPA)-induced ERK activation. We show that the majority of the immediate transcriptional response is TCF-dependent, either directly or indirectly. TCF-deficient MEFs proliferate slowly, and TCF-dependent SRF target genes, which include Myc, are predominantly associated with cell signaling, metabolism, and proliferation. Strikingly, the TCFs also act as general negative regulators of cell adhesion, contractility, and motility by inhibiting access of MRTFs to SRF at its target genes.

5To activate TCF-SRF signaling in MEFs, we used the phorbol ester TPA, which activates ERK via protein kinase C (PKC) and RasGRP1. TPA also downregulates Rho-actin signaling and thus allows the identification of genes that are particularly sensitive to TCF activation. In immortalized MEFs, TPA activated classical TCF-SRF targets such as Egr1 but not MRTF-SRF target genes such as Vcl, although all were activated by serum stimulation (Figure S1A). We used RNA sequencing (RNA-seq) to analyze the global response to TPA stimulation, analyzing both total and intronic RNA-seq reads to maximize sensitivity (Figure 1A; Table S1). The TPA-induced gene set is enriched in gene hallmarks (i.e., gene cohorts that change similarly in different contexts; ) associated with the cell cycle, signaling networks, and particular transcription factors (Figure 1B; Table S2A) and gene ontology (GO) terms for genes involved in signaling, metabolism, the cell cycle, and the cytoskeleton (Table S2B). It was also significantly enriched for genes from both the TCF- and MRTF-SRF gene signatures (; Figure 1B), suggesting that many MRTF-SRF target genes are also directly or indirectly controlled by the TCFs (see below). In contrast, TPA-downregulated genes were associated with hallmarks involving specific cellular functions, particularly metabolism; MRTF-SRF target genes were also enriched in this group, consistent with TPA downregulating Rho (Figure 1B; Tables S2A and S2B; see below).

6To investigate the role of TCFs in gene expression and cell behavior, we studied MEFs from embryos lacking all three TCFs (triple knockout [TKO] MEFs; ). Strikingly, TKO MEFs were significantly less proliferative than wild-type MEFs (Figure S1B), with increased numbers of cells in G2/M compared with wild-type cells (Figure S1C). They also exhibited somewhat enhanced ERK activation in response to TPA (Figure S1D), probably reflecting aberrant expression of ERK phosphatases. Nevertheless, TPA induction of classical TCF-linked immediate-early genes was effectively abolished in TKO MEFs (Figure S1E), although serum induction of SRF target genes was less impaired (Figures S1A, S1E, and S1F).

7RNA-seq analysis revealed that the global transcriptional response to TPA was substantially altered in TKO MEFs (Figure 1C). Using an iterative pipeline to compare gene inducibility in the two different backgrounds, we partitioned the genes induced by TPA in wild-type MEFs into TCF-dependent and -independent groups (Supplemental Experimental Procedures). Over 60% of TPA-induced gene expression was TCF-dependent (estimated at 2,142 of 3,470), whereas TPA-downregulated gene expression showed no obvious TCF-dependence (Figure 1D). Thus, consistent with their known biochemical role in gene activation, the TCFs act positively in the transcriptional response to TPA stimulation.

8TCF inactivation also changed the basal expression of approximately 60% of genes detectably transcribed in wild-type MEFs, regardless of their acute response to TPA stimulation (Figure 1E). Consistent with the proliferative defect observed in TKO cells, genes with reduced basal expression were enriched in gene hallmarks and GO terms associated with cell-cycle and proliferative responses (Figure 1E; Tables S2A and S2B). In contrast, genes exhibiting increased basal transcription in TKO cells included a substantial number of genes previously identified as MRTF-SRF targets (383 of 683; ) predominantly associated with cytoskeletal hallmarks and GO terms (Figure 1E; Tables S2A and S2B). Many of these genes were downregulated in response to TPA in wild-type MEFs (Figure 1F). Thus, the TCFs not only mediate the acute response to ERK activation but also negatively regulate MRTF-SRF signaling (see below).

9To identify TCF-dependent SRF target genes, we set out to correlate the RNA-seq data with SRF ChIP sequencing (ChIP-seq). We first identified 2,577 high-confidence SRF binding sites (Figure 2A; Figure S2A; Table S3). In TKO MEFs, ∼40% of these showed reduced SRF binding and increased H3 occupancy, suggesting that cooperative TCF-SRF binding facilitates nucleosome exclusion (; Figure S2B). SRF sites were statistically overrepresented within 100 kb of active genes in MEFs (Figure 2B), coinciding with DNase I-hypersensitive sites (DNase I HS) (Figures S2C–S2E; see GEO: GSM1003831 and GSM1014199). They were associated with the same transcription factor (TF) binding motifs as in NIH 3T3 fibroblasts, although, at TCF-enhanced sites, Ets motifs were more prevalent, suggesting that TCF-SRF interaction facilitates SRF binding (Figure S2F).

10Integration of Hi-C chromosome interaction maps with ChIP-seq data can reliably identify signal-induced genes controlled by remote regulatory sites. We therefore integrated the SRF ChIP-seq data with recently determined Hi-C maps of intrachromosomal interactions in MEFs. We combined the MEF Hi-C datasets using iterative mapping (; Experimental Procedures), identifying 1.3 million Hi-C interactions at 10-kb resolution, of which around 100,000 linked RefSeq-annotated transcription start sites (TSSs) to remote, mostly non-TSS, regions containing DNase I HS (Figure 2C).

11Using principal component analysis (PCA) (; Experimental Procedures), we partitioned the genome into regions of preferential Hi-C interaction (Figure 2D; Figures S2G and S2H). These regions, denoted A and B according to PCA score, encompass the topologically associating domains (TADs) defined previously in MEFs and are analogous to the “A and B compartments” defined previously at 1-Mb resolution. The A regions correlated strongly with the presence of genes and with overall gene expression (Figure S2I). They also showed significant correlation with DNase I HS frequency, an independent indicator of gene activity (Figure S2J). Most importantly, the A regions contained over 80% of the SRF binding sites (2,137 of 2,577; Figure 2E; Figure S2K). Most TSSs linked to remote SRF binding sites by Hi-C were transcriptionally active, and most TPA-inducible TSSs linked to remote sites were also TCF-dependent (Figure 2E; Table S1), suggesting that the integrated SRF ChIP-seq/Hi-C method identifies functionally relevant TSS-SRF interactions.

12Of the 3,470 TPA-inducible genes, 1,231 had TSSs within 10 kb of an SRF site and/or interacted with a remote SRF site in Hi-C (Figure S3A). Activation of 763 was TCF-dependent, so we defined them as high-confidence Direct TCF-SRF targets (Figure 3A; Table S1). The majority of these (472 of 763) had closest interacting SRF sites >10 kb distant, in many cases >100 kb away, and many exhibited multiple long-distance interactions with SRF, including Kbtb2 and Myc, which interacts with five remote SRF sites, the closest at ∼200 kb (Figures 3B and 3C; Table S1; Discussion). TPA-induced chromatin modification at the Myc TSS region is also TCF-dependent (data not shown; C.E., F.G., S.H., A.S., N.M., R.T., G. Kelly, and P. East, unpublished data). The Egr1 TSS region, which contains multiple SRF sites, also interacted with a putative remote SRF-linked enhancer, the TPA-inducible TCF-dependent Etf1 gene ∼70 kb downstream, and other more distant genes (Figure 3C). In all, 174 “remote-controlled” TSSs were linked to SRF sites that were themselves close to TSSs, blurring the distinction between “enhancer” and “promoter” elements (Figures 3B and 3C; Table S1).

13The integrated Hi-C/ChIP-seq data allowed definition of a high-confidence set of 1,062 indirectly TCF-dependent TPA-inducible genes that have TSSs that are neither near SRF sites nor physically interact with them in Hi-C (Figure 3D; see below). Gene set enrichment analysis using the MsigDB database showed that the TCF-dependent Direct and Indirect gene sets were related but distinct (Figure 3E; Tables S4A and S4B). Similar to the TPA-induced population as a whole, Direct genes were enriched in gene hallmarks involved in the cell cycle, signaling networks, and particular transcription factors (Table S4A) and GO categories related to signaling, metabolism, the cell cycle, and the cytoskeleton (Table S4B).

14Interestingly, the TCF-dependent Direct gene set contained 100 genes shown previously to be MRTF-SRF targets in serum-stimulated NIH 3T3 fibroblasts (; Figure S3A; Table S1). This “shared MRTF-TCF” gene set was enriched in gene hallmarks and GO terms involving the cytoskeleton, whereas the remaining “TCF-only” gene set was enriched in proliferative and signaling gene hallmarks and GO terms (Tables S4A and S4B). Finally, approximately 30% of genes whose basal expression was decreased upon TCF inactivation were also physically linked or close to SRF binding sites (1,501 of 5,113; Figure S3B), suggesting that they respond directly to basal levels of TCF activity. The role of the TCFs in the regulation of genes exhibiting enhanced basal activity in TKO MEFs will be considered below.

15To assess to what extent TPA-induced gene expression reflects the action of the ERK-regulated TCF activation domain, we reconstituted TKO MEFs with derivatives of the human Elk-1 TCF (Figure S4A). Re-expression of wild-type Elk-1 but not the transcriptionally inactive mutants Elk-1nonA and Elk-1ΔFW restored significant TPA-inducibility to many Direct and Indirect TCF target genes (Figures 4A and 4B; Table S5). Approximately half of the TPA-inducible TCF-dependent genes (1,016 of 2,142) appeared to be Elk-1 responsive (Figure 4B; Figure S4B). Elk-1 expression preferentially restored the activity of Direct TCF targets, and left TCF-independent genes unaffected (Figure 4C; Figure S4B). Elk-1-responsive genes were enriched for gene hallmarks and GO terms associated with proliferation, cell-cycle control, chromosomal replication, and segregation (Tables S4A and S4B), although Elk-1 expression failed to restore normal proliferation to reconstituted TKO MEFs (Figure 4D). Thus Elk-1, and presumably the other TCFs, is directly involved in control of proliferative gene expression.

16To gain insight into the extent to which Elk-1’s activity is SRF-dependent as opposed to autonomous, we carried out ChIP-seq in reconstituted TKO MEFs (Experimental Procedures). We identified 336 high-confidence Elk-1 binding sites associated with CArG and ETS motifs, bound equally well by the Elk-1nonA and Elk-1ΔFW mutants, and unaffected by TPA stimulation (Figure 4E; Figure S4D, Table S3). Of these Elk-1 sites, 251 coincided with SRF peaks, whereas 85 were apparently SRF-independent “solo” sites (Figure 4F). In general, the SRF-associated Elk-1 peaks coincided with DNase I HS and were located proximal to TSSs, whereas the solo Elk-1 sites were not (Figures S4F and S4G; Table S3).

17Approximately 25% of the detectable SRF/Elk-1 binding sites (89 of 356) were associated with the Direct TCF-SRF target genes, and a further 30 were associated with TPA-inducible SRF target genes that were not scored as TCF-dependent (Figure S3B; Table S3). In contrast, only 2.5% of Elk-1 solo binding sites (9 of 356) were associated with TPA-inducible genes (Table S4A). Elk-1 re-expression in TKO MEFs restored TPA inducibility to genes associated with SRF/Elk-1 sites (94 of 151) or solo Elk-1 sites (3 of 9; Abce1, Trpc7, and Hat1) (Table S5). Thus, Elk-1 acts predominantly in partnership with SRF to control TPA-inducible gene expression.

18Next we sought to understand why Elk-1 expression only partially restored TCF-dependent TPA induction. Although it cannot be ruled out that this reflects gene-specific targeting of Elk-1, we found that the “rescued” genes had TSSs that were in general significantly closer to SRF sites than genes that were refractory to Elk-1 expression (Figure 4G). Previous studies have suggested that strong transcriptional activators work more effectively over distance than weak ones. We therefore compared the strength of transcription activation by Elk-1 with that of its relative SAP-1, with which it is functionally redundant. SAP-1 was significantly more effective than Elk-1 in restoring TPA-inducibility to 19 selected TCF-SRF target genes in TKO MEFs (Figure 4H; Figure S4A) and also substantially enhanced TKO MEF proliferation (Figure 4I). Taken together, these results suggest that the partial rescue of TCF-SRF target genes by Elk-1 expression in TKO MEFs at least partly reflects its relatively weak transcriptional activation capacity.

19The expression of Indirect TCF-dependent target genes must either rely on basal levels of proteins encoded by direct TCF-SRF target genes, or reflect a secondary response to transcriptional regulators produced by immediate-early genes. The Direct TCF-SRF target gene set includes 54 different transcription factors, many of which are themselves sensitive to ERK activation (for references, see Table S6). These include members of the Egr, AP-1, Ets, NF-κB, and STAT families in addition to Myc, p53, and others (Figure 5A), and the Direct gene set is indeed enriched in gene hallmarks associated with these transcription factors (Tables S4A and S4B).

20To investigate the relation between Indirect and Direct TCF target gene regulation, we examined their potential regulatory sequences. Many of these are likely to coincide with DNase I HS, so we performed HOMER motif analysis on all DNase I HS sequences located within 10 kb of or linked to Indirect TSSs by Hi-C, scoring only motifs over-represented with respect to the entire DNase I HS sequence population (Figures 5B and 5C). At TCF targets, TSS-proximal sites were enriched in SRF, Ets, AP-1, NF-κB, Kruppel-related (i.e., Egr-like), and E2F motifs, whereas distal sites were associated with Ets, Homeobox, STAT, p53, and E-BOX motifs; as expected, the SRF consensus was not enriched at Indirect targets (Figure 5C). These data support the view that products of Direct TCF-SRF target genes are required for the response of Indirect TCF targets to TPA stimulation.

21We saw above that many genes exhibit increased basal transcription in TKO MEFs (Figure 1D; Table S1), including hundreds previously identified as MRTF-SRF target genes in NIH 3T3 cells. Indeed, in TKO MEFs, the enhanced transcription of MRTF-SRF targets such as Acta2, Ctgf, and Dstn was inhibited by Latrunculin B (LatB), which inhibits MRTF (; Figure S5A). Moreover, quantitative ChIP showed substantially increased basal levels of MRTF recruitment to SRF targets such as Acta2, Ankrd1, and Slc2a1, which was inhibited by LatB or TPA, which inhibit Rho-actin signaling, but further increased by serum stimulation (Figure 6A; Figure S5A). TKO MEFs also exhibited increased MRTF access to TCF-SRF direct target IE promoters in wild-type cells (Figure S5B). Thus, TCF inactivation potentiates MRTF-SRF signaling.

22Finally we investigated the functional consequences of the increased MRTF-SRF signaling in TKO MEFs. TKO MEFs were larger than wild-type cells, contained increased numbers of parallel F-actin fibers (Figure 6B), and were significantly more contractile than wild-type MEFs, as assessed by the ability to contract collagen gel (Figure 6C). Re-expression of wild-type or transcriptionally inactive Elk-1 derivatives, or wild-type SAP-1 effectively suppressed the hypercontractile phenotype (Figure 6D). Pro-invasive behavior is associated with hypercontractility in cancer-associated fibroblasts. Accordingly, TKO MEFs, but not wild-type MEFs, strongly promoted invasion of 4T1 breast carcinoma cells in organotypic culture, and this was inhibited by small interfering RNA (siRNA)-mediated MRTF depletion (Figure 6E; Figure S5C). In sum, these data show that the TCFs generally inhibit MRTF-SRF signaling and that this reflects direct competition with the MRTFs for access to SRF, rather than the TCFs' ability to induce gene expression (Discussion). As a result, TCF-MRTF antagonism can contribute to control of “activated” fibroblast phenotypes, as seen in carcinoma-associated fibroblasts (Figure 6F; Discussion).

23The ERK-regulated TCF family of SRF partner proteins were first identified as regulators of the Fos gene, but the extent of their role in immediate-early gene expression and its significance for cell proliferation have been unclear. We found that TPA-induced genes in MEFs are enriched for gene hallmarks and GO terms associated with signal transduction, metabolism, transcription, and proliferation. Over 60% of TPA-induced gene expression in MEFs was TCF-dependent, and the proliferation of TCF-deficient MEFs was impaired. Previous studies have implicated the TCFs in control proliferation in response to adhesive and oncogenic stimuli, and our results show that TCF-SRF signaling controls a gene expression program governing proliferation. We also found that TPA significantly downregulated many MRTF-SRF-controlled genes, consistent with the finding that TPA also downregulates Rho.

24We used an integrated SRF ChIP-seq/Hi-C approach to define a set of 763 TPA-induced TCF-SRF Direct target genes, whose TSSs are in close proximity to and/or physically linked to SRF binding sites. The Direct TCF-dependent gene signature provides a picture of the acute transcriptional response to ERK activation and, like the TPA-induced signature as a whole, it is significantly enriched in genes involved in signaling, transcription, and proliferation. Regulated expression of many Direct TCF-SRF target genes in TCF-deficient MEFs in TKO cells could be restored by re-expression of human Elk-1. TCF activity also promoted basal transcription of many genes in MEFs, whether TPA-inducible or not. This might reflect stochastic pulses of ERK activation that occur in many cell types.

25MRTF-SRF signaling regulates target genes involved in cytoskeletal dynamics, and, interestingly, numerous members of the Direct TCF-SRF-dependent gene set were defined previously as MRTF-SRF targets in serum-stimulated NIH 3T3 cells. Thus, at a significant number of genes, both cofactor families can variably access SRF according to cellular context (Figure 6F). Our results imply that the relative levels of TCF and MRTF proteins in different biological contexts can affect the balance between proliferative and cytoskeletal gene expression programs (Figure 6F); this is discussed further below. The Yap-Taz proteins are also implicated in the control of both proliferative and cytoskeletal gene expression, acting at least in part through long-range TEAD/AP-1 elements. Several AP-1 components are encoded by SRF target genes, providing a potential mechanism by which the two pathways might converge.

26We found that TCF-SRF signaling contributes to TPA-induced Myc transcription and identified five distant potential regulatory SRF sites. Myc was first identified as a mitogen-inducible gene over 30 years ago, and its regulation remains poorly understood, but our findings identify a regulatory input for ERK signaling. The E box Myc binding consensus is enriched at putative regulatory sequences of genes that are indirectly TCF-responsive, consistent with Myc playing a role in the secondary response to growth factor stimulation. In spite of this, Myc cooperates with Ras in transformation, perhaps reflecting downregulation of ERK by chronic Ras activation.

27Previous work has shown that the TCFs can regulate transcription independently of SRF, but Elk-1 predominantly acted through SRF in our system. The ability of Elk-1 to restore TPA-inducible transcription in TKO MEFs required its transcriptional activation domain, indicating that Elk-1 acts by providing a primary signal input for activation. Nevertheless, Elk-1 failed to restore the TPA-induced transcriptional response at all TCF-dependent direct SRF targets; it also restored chromatin modifications at only a subset of TCF-dependent SRF target TSSs (C.E., F.G., S.H., A.S., N.M., R.T., G. Kelly, and P. East, unpublished data). Although it is likely that this reflects TCF-specific gene targeting, our results suggest that the strength of transcription activation by the different TCFs may also play a role. In a direct comparison, SAP-1 both activated transcription and restored proliferation more effectively than Elk-1, suggesting that it may restore a greater fraction of TCF-dependent gene expression. However, expression of mouse Elk-1 was also more effective than human Elk-1 in restoring proliferation. Resolution of these issues will require direct comparative analysis of TCF recruitment with genomic targets.

28We identified over 700 TCF-dependent direct SRF target genes by integrating ChIP-seq data from unstimulated MEFs with published MEF Hi-C datasets, an approach used previously to identify NF-κB target TSSs. Although this approach uses the stringent criterion of direct TF-TSS physical interaction to identify candidate TF targets, because of limitations of the Hi-C data it is likely to under- rather than over-estimate the number of true targets. We note, however, that many SRF-interacting TSSs remained unaffected by TPA stimulation and must be somehow refractory to SRF-linked signals, as observed previously. Thus, even with an additional functional constraint, establishing SRF-TSS linkage by Hi-C is not sufficient to predict whether a TSS is signal-regulated.

29The integrated SRF ChiP-Seq/Hi-C approach also allowed us to define more than 1,000 TPA-inducible genes that, although TCF-dependent, are not direct TCF-SRF targets. Induction of these genes must either be dependent on the basal expression of Direct TCF-SRF target genes or arise as a secondary response to their acute activation. The putative regulatory sequences associated with Indirect TCF-SRF targets are significantly enriched for the binding motifs of transcription factors encoded by Direct TCF-SRF targets, supporting the notion that the IE response is a transcription cascade.

30Our data show that TCF-MRTF antagonism is a general feature of SRF regulation, reflecting their competition for binding to SRF, to which their binding is mutually exclusive (,,; Figure 6F). In fibroblasts, TCF-MRTF competition at many SRF sites potentially influences the “activation” state by controlling contractility and pro-invasive behavior. Nevertheless, many SRF target genes appear to be coupled predominantly to one pathway or the other. This preference is likely determined by the quality of the SRF binding site. TCF binds preferentially to sites with well-defined Ets motifs, whereas MRTF-SRF interaction is favored at sites that can be easily bent. Indirect MRTF-TCF antagonism can also occur, however, as in the control of ERK signaling by the MRTF target gene Mig6.

31Our data suggest a model in which the pro-proliferative effects of TCF-SRF signaling and increased contractility brought about by MRTF-SRF signaling are mutually antagonistic. In smooth muscle cells, PDGF stimulation induces exchange of the Elk-1 TCF for myocardin, indicating that this antagonism may itself be influenced by signaling. Although TCF DNA binding activity is not influenced by signal strength, nuclear MRTF levels in resting cells are controlled by cellular G-actin concentration. Thus, TCF-MRTF competition will also be affected by the basal level of Rho signaling, and factors affecting the degree to which a particular stimulus activates Rho- and ERK-dependent pathways will therefore also influence SRF transcriptional outputs (Figure 6F). It is tempting to speculate that it is TCF-MRTF competition that has provided the selective pressure to maintain the single SRF gene during metazoan evolution.

32Wild-type or Elk1−/− Elk3δ/δ Elk4−/− mouse embryo fibroblasts were immortalized by expression of SV40 large T. For stimulation with 50 ng/ml TPA, MEFs were maintained in 0.3% FCS. MEFs were reconstituted with wild-type human Elk-1 and SAP1 and Elk-1nonA and Elk-1ΔFW by retroviral transduction using standard procedures. Pools of MEFs expressing TCFs at similar levels were derived by fluorescence-activated cell sorting (FACS) for GFP marker expression. Fibroblast organotypic cultures and force-mediated matrix remodeling assays were as described previously. High-throughput imaging was used to quantify actin stress fiber length using the Cellomics ArrayScan VTI and compartmental analysis (BioApplications; 3,000 cells/point).

33ChIP and ChIP-seq have been described previously with sequencing on the Hi-seq2500 (150-base read lengths). Antibodies used were as follows: SRF (sc-335, lot no. D3013, Santa Cruz Biotechnology) and MRTF-A (sc-21558, lot no. I1412, Santa Cruz); anti-mouse Elk-1, amino acids (aa) 309–429, was made in-house and affinity-purified against recombinant human Elk-1nonA (aa 309–429). qPCR gene expression analysis was by standard methods. For primers, see Supplemental Information.

34Samples were prepared using the GenElute mammalian total RNA miniprep kit (RTN350-1KT). DNA was removed by DNAase I and rRNA by the Ribo-zero rRNA removal kit (Epicenter). RNA-seq libraries were prepared from 1 μg RNA using the directional mRNA-seq Library Prep v1.0 protocol (Illumina). Libraries were subjected to 72-base single-end sequencing on the Hiseq analyzer, and trimmed to 50 bp. Raw and processed data are in the GEO: GSE75667.

35RNA-seq data were aligned to the mm9 mouse genome using Burrows-Wheeler Aligner (BWA) (default settings). Reads within annotated RefSeq genes (“all reads”) and reads containing intronic sequences (“intronic reads”) were annotated using the bam file obtained by BWA as input. Normalization was against a set of invariant genes across samples, defined as those with differences in read counts within 1σ from the mean difference (μdiff), assuming a quasi-normal distribution of gene read counts. Differential gene expression analysis was performed with Deseq (comparing resting and TPA-induced, wild-type, and TKO MEF samples; adjusted p value (adj-p) ≤ 0.01; minimum change of 10%). The effects of knockout and reconstituted background were estimated by comparing the degree of gene induction under the different conditions and identifying genes that are similarly affected (Supplemental Information).

36ChIP-seq reads were trimmed to 50 bp and aligned using BWA to the mm9 mouse genome (default settings). Raw and processed data are online in the GEO: GSE75667. Candidate SRF and Elk-1 peaks were identified by Model-based Analysis of ChIP-seq (MACS), retaining those called at p < 1E-4 in at least three pseudo-replicate samples. Those whose read distribution was significantly different from Elk-1 ChIP-seq, performed on TKO MEFs transduced with empty vector, were identified using Deseq (minimal increase 1.5-fold, p ≤ 0.05) and retained for further study. This approach has an effectively negligible false discovery rate (FDR). HOMER was used for motif discovery in sequences ± 100 bp from the ChIP-seq peaks or DHS peaks.

37This analysis used the HOMER Hi-C analysis pipeline (http://homer.salk.edu/homer/interactions). Reads from NCBI SRA: SRX554530, and GEO: GSM1648486 and GSM1696042 were aligned to the mouse genome. A “universe” of ∼1.3 million significant interactions at 10-kb-resolution interactions was determined, taking into account linear genomic distance and sequencing depth (p ≤ 0.05 and Z score ≥ 2). Chromosomal regions exhibiting preferential interactions were identified by using the automated PCA analysis on Hi-C data in HOMER (runHiCpca.pl), retaining those with a PCA score >100. Interactions mapping at all RefSeq-annotated TSSs (∼130,000) were retrieved using the annotateInteractions.pl pipeline.

38F.G. and C.E. designed, conducted, and interpreted experiments. F.G., S.H., and A.S. analyzed the genomic data and wrote scripts. N.M. performed the sequence analysis. R.T. conceived the project, suggested and interpreted experiments, and wrote the paper with F.G.