Transcription factor AP1 potentiates chromatin accessibility and glucocorticoid receptor binding

Abstract

Ligand-dependent transcription by the nuclear receptor glucocorticoid receptor (GR) is mediated by interactions with coregulators. The role of these interactions in determining selective binding of GR to regulatory elements remains unclear. Recent findings indicate that a large fraction of genomic GR binding coincides with chromatin that is accessible prior to hormone treatment, suggesting that receptor binding is dictated by proteins that maintain chromatin in an open state. Combining DNaseI accessibility and chromatin immunoprecipitation with high-throughput sequencing, we identify the activator protein 1 (AP1) as a major partner for productive GR-chromatin interactions. AP1 is critical for GR-regulated transcription and recruitment to co-occupied regulatory elements, illustrating an extensive AP1-GR interaction network. Importantly, the maintenance of baseline chromatin accessibility facilitates GR recruitment and is dependent on AP1 binding. We propose a model in which the basal occupancy of transcription factors acts to prime chromatin and direct inducible transcription factors to select regions in the genome.

Figure 1. AP1 occupancy at genomic regulatory elements

A. Genomic location of AP1 binding sites identified by ChIP-Seq. Binding profile of AP1 was determined by c-Jun ChIP-Seq after dexamethasone induction (1hr). Promoter regions are defined as +/− 2.5 kb from an annotated transcription start site. B. AP1 binding is associated with regions of baseline accessible chromatin. Most AP1 occupancy (90%, blue) occurs within pre-hormone (constitutive) accessible chromatin. A small fraction of AP1 peaks (~5%) are associated with either inducible chromatin or closed chromatin (orphans). Overlapping sites are defined by an intersection of ≥10bp. Statistical significance was determined by binomial distribution. C. AP1 and GR binding exhibit a high degree of overlap. Venn diagram summarizing global overlap between AP1 and GR binding. Overlapping peaks are defined by an intersection of ≥10bp. A major fraction of GR peaks (51%) overlap sites of AP1 binding. Statistical significance was determined by binomial distribution. D. Global overlap of AP1 and GR binding is statistically significant. Random sampling simulations in silico were employed to compute the chance overlap between GR and AP1 binding and compared with in vivo experimental data (GR and AP1 ChIP-Seq). Genomic regions for simulated binding were DNaseI hotspots windowed into 150 bp or the entire mouse genome windowed into 150 bp. The overlap determined by ChIP-Seq is significantly enriched over random chance. Figure shows the median and error bars show the minimum and maximum values. See Figure S1H for methodology. E. Genomic regions of GR and AP1 show significant overlap. Examples [UCSC browser shots; (Kent et al., 2002)] of AP1 ChIP-Seq in the absence and presence of hormone and DNaseI-Seq and GR ChIP-Seq in the absence and presence of A-fos. The DNaseI and GR ChIP experiments were performed after 1hr hormone treatment. Left panel (45 kb region) shows numerous sites of GR and AP1 overlap, all coincident with open chromatin. Right panel (45 kb region) contains AP1 peaks, independent of GR occupancy but coincident with open chromatin. Black arrows denote GR peaks (left panel) or AP1 peaks (right panel) in constitutively accessible chromatin. Purple arrow denotes GR binding at inducible sites. Red arrow denotes orphan GR binding. Inducible DHS sites are markedly weaker than constitutive DHS sites (John et al., 2011).

Figure 2. Abrogating AP1 binding attenuates GR activity and chromatin accessibility

A. A-fos inhibits expression of GR regulated genes. Schematic of a cell line generated to express A-fos (dominant negative Fos) using a tetracycline (tet-off) conditional expression system (left panel). Expression of A-fos results in the strong inhibition of endogenous AP1 through the formation of DNA-binding incompetent heterodimers. Genome-wide effect of the dominant negative activity of A-fos on gene expression was examined by microarray. A-fos mediated changes in gene expression were calculated at a log₂ difference of ≥1. 48% of all 651 GR-regulated genes (induced and repressed) were compromised in the hormone response. B. A-fos expression attenuates GR binding at sites co-bound by GR and AP1. Black arrows denote GR peaks. C. A-fos expression does not affect GR binding sites that do not co-localize with AP1. Red arrows denote GR peaks. B and C show examples [UCSC browser shots (Kent et al., 2002)] of AP1 ChIP-Seq in the absence and presence of hormone and DNaseI-Seq and GR ChIP-Seq in the absence and presence of A-fos. The DNaseI and GR ChIP experiments were performed after 1hr hormone treatment.

Figure 3. Abrogating AP1 binding attenuates genome-wide GR binding and chromatin accessibility

A-B. A-fos expression reduces both GR and AP1 binding at co-bound sites. Shown are AP1 (A) and GR ChIP (B) q-PCRs performed in the absence or presence of A-fos and treated with hormone for 1hr at select sites. The following regions were analyzed: a control site which is a non-DNaseI hypersensitive site (no DHS) with no GR or AP1 binding (Chr1:155,659,519); a GR only bound site (Chr1:155,663,767) and five sites bound by AP1 and GR. Data represent the average of three biological replicates. Error bars represent standard error of the mean. C-D. A-fos expression reduces global GR binding (C) and chromatin accessibility (D) at GR and AP1 co-bound sites. Shown are box plots of the effect of A-fos on GR binding and chromatin accessibility at sites bound by GR alone or sites bound by both GR and AP1. A-fos selectively affects sites bound by AP1 and compromises GR binding and chromatin accessibility at co-bound regions. E. A-fos expression reduces global chromatin accessibility at AP1 bound sites, independent of GR binding. Shown are box plots of the effect of A-fos on chromatin accessibility at sites not bound by either AP1 or GR or sites bound by AP1 alone. A-fos selectively compromises endogenous AP1 binding and affects chromatin accessibility at sites bound by AP1 (see Figure S3B). For boxplots (C-E), the effect of A-fos on GR binding and chromatin accessibility is expressed as a ratio of tag densities: minus Tet (A-fos expressing) over plus Tet (no A-fos expression). Dotted line is indicative of no A-fos effect. Statistical significance was determined using a KS-test. Boxplots show the median and upper and lower quartiles. Whiskers show the minimum and maximum values. Notches denote the 95% confidence interval of the median.

Figure 4. Motif composition at GR and AP1 co-bound sites

A. Motifs within GR peaks at co-bound sites contain both composite and non-composite elements. The GRE and AP1 motif position-weight matrices derived from de novo motif discovery were used to scan the genome at a p-value of 10⁻³. The presence of a motif within the peak was defined as an overlap between peak and motif by ≥1bp. B. GR peaks at inducible chromatin sites are highly enriched for the GRE motif. Motif composition of GR and AP1 co-bound sites associated with hormone-dependent inducible DHS. The GRE motif is represented in 86% of all GR peaks in inducible sites, in contrast to co-bound regions in constitutively open chromatin (see Figure S4B). C. Composite and non-composite elements show distinct motif distribution profiles at GR and AP1 co-bound sites. Distribution of the GRE and AP1 motif sequence motifs were determined relative to the center of peaks. Motifs were determined from de novo motif discovery at a p-value of 10⁻³ where at least half the motif overlapped the peak. Shown are the distribution profiles for the GRE motif at all GR peaks, AP1 motif at all AP1 peaks, AP1 motif uniquely present at non-composite elements and the AP1 motif uniquely present at composite elements at co-bound sites. D. GR binding is more robust at composite than non-composite elements. Tag density (a measure of DNA binding strength) at GR peaks was determined for GR binding in constitutively accessible chromatin at all co-bound sites containing either the AP1 motif alone (non-composite) or both AP1 and GRE motifs (composite). Statistical significance was determined using a KS-test. E. AP1 binding strength is independent of motif composition. Tag density at AP1 peaks was determined for AP1 binding in constitutive sites at all co-bound sites containing either the AP1 motif alone (non-composite) or both AP1 and GRE motifs (composite). In D-E, boxplots show the median and upper and lower quartiles. Whiskers show the minimum and maximum values. Notches denote the 95% confidence interval of the median. F. DNA motifs at AP1 only versus non-composite sites. De novo motif discovery was performed for AP1 binding sites lacking GR binding and AP1 sites that are co-bound with GR. Shown are results that match known sequence motifs at a significance of P < 10⁻⁴. Non-composite GR and AP1 co-bound sites are enriched by different DNA sequence motifs compared to AP1 only occupied sites. The AP1 motif was highly enriched in both cases (E < 1.6×10⁻³⁵³).