1
Esnault et al
SUPPLEMENTARY MATERIALS
INVENTORY OF SUPPLEMENTARY MATERIALS
1. Supplementary Figures
Supplementary Figure S1 defines the effects of U0126, Latrunculin B and Cytochalasin D on Ras-MAPK, Rho-actin and the Hippo signalling pathway. It shows that U0126 specifically inhibits Ras-MAPK signalling, while LatB inhibits Rho-actin signalling-dependent MRTF nuclear accumulation, and CD specifically activates the MRTFs; and that the Hippo signalling pathway is constitutively active under our assay conditions. Related to Figure 1.
Supplementary Figure S2 shows ChIP-seq control experiments including inhibitor studies, validation by ChIP-qPCR, dimerisation of MRTFs, specificity of MRTF and TCF binding, and motifs associated with SRF sites apparently lacking cofactors. Related to Figures 1 and 2.
Supplementary Figure S3 displays examples of MRTF binding sites where MRTF-B binds apparently in a SRF-independent manner. Related to Figure 2.
Supplementary Figure S4 shows that inducible SRF binding sites exhibit MRTF-dependent nucleosome displacement in response to Rho-actin signalling pathway. Related to Figure 3.
Supplementary Figure S5 shows the validation of the RNA-seq data by Q-PCR and displays examples of potential ncRNA cis-regulatory effect on gene activity. Related to Figure 4.
Supplementary Figure S6 describes the role of SRF/MRTF in controlling expression of genes regulating cytoskeleton remodelling, focal adhesion and the extracellular matrix and shows that SRF deleted cells are impaired in both actin and microtubule assembly. Related to Figure 6.
Supplementary Figure S7. Shows the validation of SRF, MRTF and TCF association with selected genes involved in the circadian rhythm control and confirms that MRTF can synchronise the circadian clock in response to actin dynamics. Related to Figure 7.
2. Supplementary tables
Excel file of the supplementary tables summarising the genomic data. Related to Figure 1, 2, 4, and 6.
Supplementary Table S1 summarises the ChIP-seq data. Related to Figures 1 and 2.
Supplementary Table S2 summarises the RNA-seq data of coding genes. Related to Figure 4.
Supplementary Table S3 summarises the RNA-seq data of ncRNA. Related to Figure 4.
Supplementary Table S4 summarises the relationship between SRF and cofactor binding and the activity of associated genes. Related to Figure 4.
Supplementary Table S5 presents the SRF, MRTF and TCF signatures. Related to figure 6.
Supplementary Table S6 presents the gene ontology analysis. Related to Figure 6
Supplementary Table S7 presents genes regulated by SRF/MRTF involved in cytoskeleton control. Related to Figure 6.
Supplementary Table S8 shows a detailed gene signature enrichment analysis. Related to Figure 6.
3. Extended Methods
4. Supplementary References
SUPPLEMENTARY FIGURE LEGENDS
Supplementary Figure S1. Effects of inhibitors
Cells were maintained in 0.3% FCS and stimulated with 15% FCS with of LatB (0.3µM), CD (2µM) or U0126 as indicated, or treated with drugs alone.
(A) U0126 (10µM), but not LatB (0.3µM), inhibits ERK activation following serum stimulation, as assessed by immunoblot for phospho-ERK. As a loading control the blot was stripped and reprobed with pan-ERK antibody.
(B) ERK inhibition by U0126 (10µM) does not prevent serum-induced nuclear accumulation of MRTF-A, as assessed by immunofluorescence.
(C) Effects of actin-binding drugs on MRTF-A and YAP subcellular localisation. Confocal immunofluorescence microscopy of MRTF-A (green; left panels) or YAP1 (green; right panels) with nuclei and F-actin visualised by DAPI (blue) and Phalloidin (red). MRTF-A is predominantly cytoplasmic in unstimulated cells, and LatB inhibits its serum-induced nuclear accumulation; CD alone is sufficient to induce MRTF-A nuclear localisation. YAP localisation, which is predominantly nuclear, is unaffected by these treatments.
(D) Quantitation of MRTF and YAP1 nuclear localisation and response to signals in (C), with the corresponding signals for similarly-treated confluent cell monolayers, where YAP is more cytoplasmic (Yu et al. 2012) are shown for comparison. ArrayScan VTI and Compartmental analysis (BioApplications). >3000 cells / condition. Mean signal is indicated by Red bars.
(E) Effect of serum stimulation on YAP phosphorylation, assessed by immunoblot with Phospho-YAP (Ser127) antibody. Under our experimental conditions - subconfluent cells grown on plastic - YAP phosphorylation is low and unaffected by serum stimulation; in contrast, in confluent cells YAP phosphorylation is high and decreases upon serum stimulation (see (Yu et al. 2012).
(F) Differential sensitivity of SRF target genes to Rho-actin and Ras-ERK signalling. Endogenous MRTF targets (Acta2 and Srf) and TCF targets (Fos and Egr1) were analysed by qRT-PCR. Expression following specific activation of MRTF by Cytochalasin D (CD; 2µM) is also shown.
Supplementary Figure S2. ChIP-seq control and validation experiments.
(A) SRF ChIP-seq shows that LatB inhibits inducible (red) but not constitutive (black) SRF binding, while U0126 has no effect.
(B) ChIP-qPCR validation of SRF ChIP-seq. 6 constitutive (black) and 49 inducible (red) ChIP-seq peaks. Data are means of 3 independent experiments.
(C) MRTF and SRF signals correlate at inducible SRF sites. SRF peak heights at the 2133 inducible SRF sites are compared to those of MRTF-A (left panels) and MRTF-B (right panels), both expressed as percent of maximum signal compared by scatter plot. Upper panels, high-confidence MRTF sites (MACS p<10-5); lower panels, all MRTF sites.
(D) Serum stimulation and serum + U0126 treatment are pseudoreplicates for MRTF ChIP-seq. Scatter-plot comparing high-confidence MRTF-A and MRTF-B ChIPseq signals (274 MRTF-A, 1176 MRTF-B; MACS p<10-5) from cells stimulated with serum in the presence or absence of U0126.
(E) MRTF-A and MRTF-B dimerise. Left, NIH3T3 cell protein extracts were immunoprecipitated with anti-MRTF-A, anti-MRTF-B, or no antibody, and proteins detected by immunoblotting. Right, antibody specificity was confirmed by immunoblotting of cells treated with SRF, MRTF-A or MRTF-B specific siRNAs (Medjkane et al. 2009).
(F) MRTF-A and MRTF-B predominantly bind the same sites. Left, peak heights for MRTF-A and MRTF-B at sites bound by both proteins, expressed as percent of maximum, were compared by scatter plot (n=1320). Centre, comparison of MRTF-A and MRTF-B raw readcounts at peaks called for MRTF-B only (n=1021). Right, comparison of MRTF-A and MRTF-B raw readcounts at peaks called for MRTF-A only (n=75).
(G) Comparison of SRF ChIP-qPCR data with MRTF-A and MRTF-B qPCR data at 50 SRF ChIP-seq peaks and 3 negative control sites. Sites are ranked in order of increasing SRF ChIP-seq signal. Vertical dashes indicate those peaks called positive by ChIP-seq for SRF, MRTF-A and MRTF-B, and the TCFs (see Figure S2H). Antibodies used for ChIP-qPCR assays are indicated at the top. Note that at 4 peaks called as negative for MRTF-A by ChIP-seq nevertheless validate as MRTF-A positive by ChIP-qPCR, as indicated by the red arrows.
(H) Comparison of TCF ChIP-qPCR data with TCF peaks called by ChIP-seq. 14 SRF-positive ChIP-seq peaks which scored positive for TCF (9 SAP-1, 7 Elk-1, 6 Net; indicated in Figure S2G), and 3 negative control loci were analysed by ChIP-qPCR. Assays are ranked in order of increasing SRF ChIP-seq signal. Antibodies used for ChIP-qPCR assays are indicated at the top. Vertical dashes indicate peaks called as positive for SRF, and the different TCFs by ChIP-seq. Red line marks the maximum negative control value.
(I) Sequence motifs associated "no-cofactor" SRF sites, classified according to whether SRF binding was inducible and LatB-sensitive. Sequences within 100bp of each SRF summit were scanned. The spectrum of motifs associated with each class suggests these sites may represent undetected MRTF- and TCF-specific SRF sites. See Figure 2F.
Supplementary Figure S3. SRF-independent MRTF binding events
Apparently SRF-independent MRTF-B binding events at inducible SRF target genes, at a constitutive gene, and in an intergenic region are shown in orange. MRTF-A binding did not score at MACS p<10-5 at these sites.
Supplementary Figure S4. Inducible SRF binding is associated with signal-regulated nucleosome displacement.
(A) SRF binding correlates with histone H3 displacement on the Fos and Acta2 genes. Tracks show total H3 ChIP-seq signal in resting and serum-stimulated cells, with inhibitors as indicated. Red bars show SRF binding sites, black arrows highlight histone-depleted regions.
(B) SRF binding sites correlate with DNA'ase I sensitivity maxima. Metaprofiles showing DNA'ase I cleavage per base across a 4 kb window centred on the SRF peaks are shown. Inducible sites, red; constitutive sites, black. DNA'ase I data at GEO accession number: GSM1003831.
(C) Scatter plots displaying SRF-ChIP-seq readcounts against H3 ChIP-seq signal (displayed by rank order of decreasing H3 signal; left panels,) or DNA'ase I sensitivity (displayed by rank order of increasing cleavage signal; right panels) in cycling NIH3T3 cells, as determined by others (right panels; wgEncodeEM002916 (2011). Signals at inducible and constitutive SRF sites are shown in red and black respectively. SRF ChIP-seq peak height in resting cells correlates with DNA'ase I sensitivity maxima (left), but the correlation is lost upon serum stimulation (right). Black line, Loess regression fitting curve (20 value moving window). DNA'ase I data at GEO accession number: GSM1003831.
Supplementary Figure S5. RNA-seq: functional validation and ncRNA targets.
(A) RNA changes were assayed by qRT-PCR on 20 endogenous SRF target genes defined by RNA-seq in the indicated culture conditions (Figure 1B). Blue bars qRT-PCR ± SEM, analysis of 3 separate RNA preparations; red bars, RNA-seq readcounts ± halfrange from analysis of 2 independent RNA preparations.
(B) Comparison of RNA-seq and qRT-PCR data. Scatter plots comparing the RNA changes quantified by qRT-PCR and RNA-seq are displayed in function of the cell culture conditions. They show a good correlation between the two methods of quantification (Spearman r, 0.79; p<0.0001).
(C) ncRNA targets in the vicinity of SRF target genes. Top, schematic view of the genomic organisation around Gm15270, Gm13270, Neat1, GM10501, 4930500J02rik and Gm1720 ncRNAs. Scale bars, 20kb. Red bars, ncRNA transcription units; green bars, candidate SRF target genes; vertical lines, SRF ChIP-seq peaks. Bottom, comparison of transcriptional responses of the ncRNAs and adjacent SRF targets. ncRNA candidates for cis-regulation are highlighted with gene expression changes indicated. For functional studies of SRF ncRNA targets, see the following: miR-143 and miR145 (Xin et al. 2009); miR-199a2 and miR214, (Park et al. 2011; Alexander et al. 2013); miR-21, (Kumarswamy et al. 2011); miR-22 (Gurha et al. 2012); Tug1 and Malat1/Neat2, (Gutschner et al. 2013).
Supplementary Figure S6. Cytoskeletal MRTF-SRF targets.
(A) MRTF-SRF signalling in cytoskeletal dynamics. MRTF-SRF target genes (specific proteins or protein functional classes) are indicated by ovals. Left, regulators of the actin dynamics and contractility; right, focal adhesion components.
(B) SRF is required for assembly of F-actin and microtubules, and to maintain nuclear morphology. F-actin and microtubules were visualised using Texas Red-X Phalloidin and a-Tubulin-Alexa488 in wildtype or SRF-deleted MEFs under the indicated conditions. Data were quantified using ArrayScan VTI and Compartmental analysis (BioApplications), >3000 cells / condition.
Supplementary Figure S7. SRF network and circadian clock regulation.
(A) ChIP-qPCR analysis of selected circadian SRF targets detected by ChIP-seq.
(B) CD treatment rapidly activates clock component gene transcription. Analysis by qRT-PCR.
(C) Clock resetting by SRF activation. NIH3T3 cells were treated with 50% FCS and transcripts were quantified by qRT-PCR over 36h.
(D) MRTF-induced clock resetting is SRF-dependent in MEFs. Wildtype and SRF knockout MEFs (pooled cultures from each of 3 embryos) were treated with 2 µM CD and transcripts quantified by qRT-PCR.
SUPPLEMENTARY TABLES
Supplementary Table S1. SRF network ChIP-seq analysis
Summary of the ChIP-seq data from SRF, MRTF-A, MRTF-B, Sap-1, Elk-1 and Net experiments. Chromosomal location of each peak is given, together with the identity of the nearest gene (or genes where the peak is located within a gene feature). Each peak ID can be used on UCSC browser to visualise the genomic area within the GEO database (instructions at http://genome.ucsc.edu). The peak information, the summary of SRF peak detection, the peak signal quantification for all factors in 0.3%FCS, 15%FCS, LatB+15%FCS and U0126+15%FCS, the summary of the detection by MACS of the factors and the relative MRTF and TCF scores are shown. The basis for calling each peak by MACS score and/or coincidence with ChIP-seq signals for other factors is indicated. Shading indicates SRF peaks called at MACS p<10-5 (red) or MACS p<0.05 plus coincidence with a high-confidence MRTF peak (green). Signals are quantified as the number of reads per 15 million. The cofactor score normalises the cofactor signal across each condition to the mean MRTF or TCF signals (MRTFs, mean of 15% FCS and 15% FCS+U0126 signals; TCFs, mean across all conditions) to derive a measure of the relative strength of MRTF and TCF signals at each SRF peak; the ratio of the MRTF and TCF scores is also shown.
Supplementary Table S2. RNA-seq analysis: protein-coding genes
The table presents the RNA-seq analysis of NIH-3T3 cells maintained in 0.3% FCS, stimulated with 15% FCS with or without LatB and/or U0126 treatment, or stimulated with CD, for all Refseq genes (release 47). For each gene, expression was evaluated using all RNAseq reads within the gene feature, or intronic reads only. The effect of serum or CD stimulation is summarised, and serum-induced genes responsive to SRF-linked signals are indicated (serum-induced, sensitive to LatB and/or U0126, or induced by CD; FDR<0.08. Background shading indicates gene activity status: red, inducible; green, repressed; yellow, constitutive; white, inactive. Active genes are defined as all those which show detectable RNA-seq signal in any experimental condition. ChIP-seq data are summarised according to whether the closest SRF site to the gene is within the gene feature, within 70kb, or more distant. The genomic coordinates of the SRF site closest to each gene, and its distance from the TSS, are given (sites within a gene feature are indicated as zero). Cofactor association is defined as those cofactors associated with the SRF peaks for which the gene concerned is the closest; note that for genes associated with multiple peaks, not all will be associated with that cofactor.
Supplementary Table S3. RNA-seq analysis: ncRNAs
The table presents the RNA-seq analysis of NIH-3T3 cells maintained in 0.3% FCS, stimulated with 15% FCS with or without LatB and/or U0126 treatment, or stimulated with CD, for all ncRNA (Ensembl release 69). The ncRNA information section displays ncRNA coordinates, ncRNA biotype, database source, and the identity and proximity of the nearest protein-coding gene. Red shading indicates largest ncRNA in the database, green shading indicates overlapping ncRNAs where multiple ncRNAs are transcribed from the same locus. Expression and ChIP-seq peak data are displayed as in Table S2.
Supplementary Table S4. Relationships between SRF and cofactor binding and the activity of associated genes.