Our previous examination of ΔNp63 binding in NHEK cells demonstrated an active chromatin signature at many of its binding sites consisting of high H3K27ac, H3K9ac, H3K4me1, H3K4me2, and H3K4me3 along with increased chromatin accessibility [18]. These sites while depleted of nucleosomes in NHEK contained nucleosome preferring sequences. However, it is unclear from this analysis if ΔNp63α played an active role in establishing this chromatin architecture or required an apriori permissive chromatin environment to bind. Therefore, to directly test ΔNp63 pioneering capabilities we established an ectopic inducible ΔNp63 in a p63 naive cell line, K562. K562 is a widely used immortalized myelogenous leukemia cell line and is an advantageous choice for these experiments because no p63 isoforms are expressed and there are extensive genomic and epigenomic datasets available from the ENCODE project [19]. Two doxycycline (Dox) inducible cell lines were generated containing the wildtype (WT) HA-tagged ΔNp63α and DNA binding mutant ΔNp63α(R304W). Western blot analysis showed that both WT and mutant ΔNp63α were expressed after Dox induction and the levels comparable to those observed in epithelial cells (Supplementary Figure S1).
ΔNp63α (WT) and ΔNp63α(R304W) were induced in K562 cells with Dox and ChIP-seq experiments performed on two biological replicates with ΔNp63-specific and anti-HA antibodies. Both ChIP-seq experiments for ΔNp63α enriched for DNA sequences containing the p63 binding motif (Supplementary Figure S2). In contrast, ChIP-seq experiments for ΔNp63α(R304W) show limited enrichment and revealed only a few background sites. In total 1980 common sites were identified in experimental replicates using WT ΔNp63α ChIP which were not enriched in the control cells expressing ΔNp63α(R304W) and were chosen for further analysis.
To understand the chromatin characteristics required for ΔNp63α binding we examined the robust dataset of chromatin modification marks and accessibility for K562 that have been generated by the ENCODE project. This provided us with the chromatin characteristics before ΔNp63 was induced in our experiments. Ten histone modifications H3K4me1, H3K4me2, H3K4me3, H3K9ac, H3K9me1, H3K9me3, H3K27ac, H3K27me3, H3K36me3, H3K79me2 with DNase-seq and H2AFZ profiles were examined and clustered into 4 groups (Figure 1A). Groups a, b, c contain a total of 374 sites and appear to represent binding locations occurring in already active chromatin environments with high levels of H3K27ac and H3K4me1 or H3K4me3. The majority of the ΔNp63α bound sites (1606 out of 1980) are in group d and represented genomic segments that were bereft of signal for any histone modification. The average chromatin architecture further demonstrates the absence of histone modifications and is shown in comparison to the active chromatin state at transcriptional start sites (TSS; Figure 1B). Examination of chromatin accessibility by DNase-seq show that the majority of these (1429 of 1980) regions are located at inaccessible sites in K562 as defined by DNase-seq (Figure 1C). Nucleosome occupancy at these binding sites is also enriched before ΔNp63α binding (Figure 1D). Overall these results demonstrate that ΔNp63α can bind at chromatin inaccessible, unmodified, and nucleosome occluded sites.
Comparison of ΔNp63 binding sites between NHEK and ectopic induced binding in K562
To further examine the chromatin characteristics at ΔNp63α binding sites, we compared our K562 targets to the binding sites in a normal ΔNp63α expressing cell type NHEK [18, 20]. Sites were classified as jointly bound, or bound only in each specific cell type. The consequence of ΔNp63 binding can be discerned by comparing the chromatin at sites bound in a ΔNp63 expressing cell line, NHEK, with the chromatin seen at the bound sites in K562. Three binding sites near the TP73 gene exemplify the chromatin differences in the two cell lines (Figure 2A). The sites bound in K562 are often within heterochromatin chromatin states (Gray) or weak-TX (Green). In NHEK, these same sites are shown to be active as Strong Enhancers (Red) or weak Enhancers (yellow). The DNase signal is also higher in 2 of 3 binding sites in NHEK.
Sites bound by ΔNp63 in both cell lines show active histone modifications in NHEK and the absence of similar modifications in K562 (Figure 2B and C). Sites in NHEK are flanked by transcriptional active histone modifications (H3K9ac, H3K27ac, H3K4me1, H3K4me2, H3K4me3), while in K562 these same sites show a very reduced signal. Sites bound only in NHEK (Figure 2B middle, and 2D) have high levels of active histone modifications, while all modifications display low signal in K562. K562 specific sites (Figure 2B bottom and 2E) display relatively low levels of active histone modifications in both NHEK and K562.
Sites bound only in NHEK allow us to address the role of repressive histone modifications (H3K27me3 and H3K9me3). These 8195 sites are all bound in NHEK and thus represent bonafide targets of ΔNp63. These 8195 sites are unbound in K562, but are not enriched for repressive histone modifications, suggesting that active repression is not a strong driving force blocking ΔNp63 binding to these sites in vivo.
ΔNp63α can bind inaccessible, inactive sites in HepG2
To validate that ΔNp63α can target inactive and inaccessible chromatin in any cell type, we developed a ΔNp63α -expressing HepG2 cell line. HepG2 is p63 naïve and is a widely used cell line in the endoderm linage that has been extensively characterized. In total ΔNp63 bound 2939 targets in HepG2 (Supplementary Figure S3). Most sites bound in HepG2 have low levels of histone modifications and accessibility, which is consistent with what is observed in K562 (Figure 3A). In addition, the majority, 65%, are classified as inaccessible (Figure 3B). The chromatin architecture was further examined at the inaccessible bound sites and displayed a low signal compared with the chromatin modifications at TSS (Figure 3C).
ΔNp63 binds at nucleosomes and leads to H3K27ac and nucleosome depletion
Our comparison between K562 and NHEK binding sites suggest that ΔNp63α binding leads to an active histone architecture. To test the consequences of ΔNp63α binding directly we determined if H3K27ac increased at sites bound by ΔNp63α. ChIP-seq for H3K27ac was performed in ΔNp63α and ΔNp63α(R304W) expressing K562 cell lines (Figure 4). Binding of ΔNp63α leads to an increased in H3K27ac at ΔNp63α bound sites while induction of DNA-binding mutant ΔNp63α(R304W) does not change the H3K27ac (Figure 4A). In comparison H3K27ac does not change at TSS in K562 in either cell line (Figure 4B). In addition, these results show the characteristic peak-valley-peak for the H3K27ac surrounding the ΔNp63 binding sites, suggestive of nucleosome depletion at the binding site. Examination of single sites further highlights the starting structure and consequences of ΔNp63α (Figure 4C&D). At these sites the p63BS is located within a well-positioned nucleosome. After induction of ΔNp63α the flanking nucleosomes are acetylated with a dip in signal at the site of binding, suggesting remodeling of the centrally located nucleosome.
ΔNp63 binds nucleosome edges
To further understand the ability of ΔNp63 to bind to inaccessible chromatin we tested ΔNp63 binding to nucleosome DNA with a competitive nucleosome binding assay [21]. 16 templates derived from Widom 601 nucleosome positioning sequence were obtained: 14 nucleosome templates with a high-affinity or intermediate-affinity p63 binding site (p63BS) placed at increasing distance to the nucleosome dyad axis, in an exposed or concealed rotational orientation. Each experiment is internally controlled with 2 nucleosome sequences lacking p63BS. Nucleosome population were obtained after in vitro reconstitution with unmodified histones via salt gradient dialysis on all nucleosome sequences simultaneously, and purified from free DNA with a sucrose gradient.
ΔNp63 was added to 0.25 pmol purified nucleosome at increasing concentration (0 to 2 pmol, 0 to 286 nM). The binding reactions were then separated on a native polyacrylamide gel to detect the ΔNp63-nucleosome complex (Figure 5A). The first lane contained only nucleosomes and was used to measure background and input levels for each experimental replicate. As ΔNp63 concentration increased a supershifted band appears and intensifies at higher amounts of ΔNp63. Meanwhile the nucleosome-only band intensity significantly decreased at higher concentrations of ΔNp63, similar to findings for p53 [21]. DNA was then extracted and purified from the supershifted and nucleosome band and sequenced. The sequencing results were then mapped back to the original 16 nucleosome sequences and compared to the control non-specific sequence in the same lane. This analysis method controls for non-specific binding, gel loading, PCR amplification, and next-generation sequencing. This approach is performed on each non-shifted nucleosome band independently to see what types of nucleosomes are bound by ΔNp63 and shifted, two replicates were conducted (Figure 5B and C). Our analysis demonstrates that nucleosomes containing a high-affinity or intermediate-affinity p63BS located around the nucleosomes boundaries at super-helix location (SHL) 6.5, 7, and in the linker SHL 8 are bound first at the lowest concentrations. While the nucleosomes containing a p63BS located near the dyad are not specifically bound as compared to the control nucleosomes. Examination of the supershifted fragments show similar results (Supplementary Figure S4), and is consistent with ΔNp63 binding only at nucleosome edges.