Serum response factor binding sites differ in three human cell types

Abstract

The serum response factor (SRF) is essential for embryonic development and maintenance of muscle cells and neurons. The mechanism by which this factor controls these divergent pathways is unclear. Here we present a genome-wide view of occupancy of SRF at its binding sites with a focus on those that vary with cell type. We used chromatin immunoprecipitation (ChIP) in combination with human promoter microarrays to identify 216 putative SRF binding sites in the human genome. We performed independent quantitative PCR validation at over half of these sites that resulted in 146 sites we assert to be true binding sites at over 90% confidence. Nearly half of the sites are bound by SRF in only one of the three cell types we tested, providing strong evidence for the diverse roles for SRF in different cell types. We also explore possible mechanisms controlling differential binding of SRF in these cell types by assaying cofactor binding, DNA methylation, histone methylation, and histone acetylation at a subset of sites bound preferentially in smooth muscle cells. Although we did not see a strong correlation between SRF binding and epigenetics modifications, at these sites, we propose that SRF cofactors may play an important role in determining cell-dependent SRF binding sites. ELK4 (previously known as SAP-1 [SRF-associated protein-1]) is ubiquitously expressed. Therefore, we expected it to occupy sites where SRF binding is common in all cell types. Indeed, 90% of SRF sites also bound by ELK4 were common to all three cell types. Together, our data provide a more complete understanding of the regulatory network controlled by SRF.

Figure 1.

SRF binding in three cell types. For each cell type and each putative target, the intensity of red indicates size of the Z-score obtained from the arrays. The right panel displays a subset of genes bound specifically by SRF in neurons (top) and smooth muscle cells (bottom). Gene names followed by an asterisk have been validated by quantitative PCR or were previously known.

Figure 2.

CArG position is conserved between human and mouse. We compared positions of the SRF consensus binding site in mouse and human promoters and observed a strong correlation (R² = 0.497). This correlation is driven primarily by consensus sites within 1–2 kb of the TSS. We display sites validated by quantitative PCR as squares and untested sites as triangles.

Figure 3.

DNA methylation SRF binding sites found only in smooth muscle cells (SMC). In SMC (black) and Jurkat cells (gray) we observed a depletion of unmethylated DNA in the promoter of six genes. Except for the caldesmon gene (CALD1), depletion is similar between the two cell types. Depletion is a measure of fold difference between uncut and cut sample (see Methods).

Figure 4.

Cell-type dependent binding and the ELK4 cofactor. For SRF binding sites independent of cell type (black squares), binding of SRF and ELK4 at the site is correlated. For cell-type dependent SRF binding sites (grey squares) no correlation is observed. Binding of each factor was determined by quantitative PCR as described (Methods) and is displayed as a log of the enrichment observed at each site compared to a panel of negative controls.