Characterising the determinants of hypoxia inducible transcription factors binding to chromatin

Hypoxia regulates many hundreds of genes that play important roles in numerous physiological and pathophysiological processes. The hypoxia inducible transcription factors (HIFs) are central to the transcriptional activation of these hypoxia-regulated genes. However, to date, little is known about the determinants of HIF-1 and HIF-2 binding site selection. Both HIF-1 and HIF-2 appear to bind as HIF-α/HIF-1β heterodimers, and recognise the same core consensus DNA motif, the hypoxia response element (HRE). However, each has its own distinct but partially overlapping set of binding sites that accounts for only a subset of the total accessible HRE motifs. Here, I have utilised ChIP-seq to systematically compare pan-genomic HIF-1α, HIF-2α and HIF-1β DNA binding in multiple cell lines and have related this to RNA-seq analyses and other publically-accessible next-generation sequencing datasets. I show that endogenous HIF-α subunits exhibit a high-degree of binding site concordance with HIF-1β, consistent with largely canonical binding of intact heterodimers. Despite cell-type specific differences in HIF-1 and HIF-2 binding site occupancy, each isoform exhibits a remarkable rigidity in its preference to bind either promoter-proximal or promoter-distal sites, respectively. These specific distribution patterns are unaffected by the absence of the other HIF-α subunit, suggesting that they do not result from exclusion of one isoform by competition with the other, but rather are discrete properties of each. Furthermore, hypoxia regulated genes neighbouring sites that are shared by both HIF-1 and HIF-2 are more likely to be regulated by HIF-1 when the site is closer to the gene and by HIF-2 when further away, indicating that post-binding mechanisms of transcriptional regulation also follow a similar pattern. Comparison of sites preferentially bound by HIF-1 and HIF-2, respectively, revealed associations with distinct histone environments, distinct accessory transcription factor binding motifs and distinct patterns of transcription factor binding site occupancy, suggesting that each may be influencing specific HIF-1 and HIF-2 binding site selection. In particular, both the AP-1 motif and AP-1 binding site occupancy were enriched within HIF-2 binding sites compared to HIF-1 sites. Intervention on AP-1 DNA-binding using the dominant-negative protein, AFos, attenuated HIF binding, specifically at sites co-occupied by AP-1 and HIF. This indicates that a cooperative relationship exists between the two transcription factors. However, binding of both HIF-1α and HIF-2α were affected suggesting that while AP-1 binding may account for the ability of HIF to bind some HRE motifs but not others, it is not a determinant of differential binding between the two isoforms. Overall, this work reveals remarkably distinct and functionally relevant patterns of HIF-1 and HIF-2 binding across the genome, and provides insight into underlying mechanisms of binding.