Abstract
Gene regulatory mechanisms underlying all known prostate cancer risk loci were characterized.
Major finding: Gene regulatory mechanisms underlying all known prostate cancer risk loci were characterized.
Approach: Proxy SNPs in LD with lead SNPs were annotated with prostate cancer functional genomics data.
Impact: Analysis of functional genomic data reveals and defines gene regulatory mechanisms at risk loci.
One hundred risk loci have been identified in prostate cancer, but expression quantitative trait locus (eQTL) effects at these loci and the molecular mechanisms underlying these effects have been investigated in isolated cases. Whitington, Gao, and colleagues developed a high-throughput approach to identify gene regulatory mechanisms underlying the 100 known prostate cancer risk loci. This approach combined data from 295 chromatin immunoprecipitation and sequencing (ChIP-seq) experiments with genotype and transcriptome data from 602 prostate tumor samples to annotate linkage disequilibrium (LD) proxy SNPs, which are proxy SNPs in LD with lead SNPs, investigate the effects of LD proxy SNPs on ternary transcription factor (TF)–TF–DNA complexes by including position weight matrices (PWM) to model complex binding specificities, and identify potential target genes of regulatory risk variants at risk loci. This integrated approach showed that 73 LD proxy SNPs disrupted TF binding according to ChIP-seq and PWM data, 56 of which occurred in open chromatin regions according to DNase-seq data, while 57 genes showed significant eQTL associations with LD proxy SNP genotypes. Additional interrogation of ChIP-seq data and PWM analysis revealed that LD proxy SNPs frequently ablate ternary complexes between androgen receptor (AR), the TF forkhead box A1 (FOXA1), and the TF homeobox B13 (HOXB13) as well as the binding of other TFs and the transcriptional regulator CCC–binding factor (CTCF). Disruption of AR-FOXA1, AR-HOXB13, and CTCF binding by LD proxy SNPs was validated by allelic-specific ChIP and qPCR, and knockdown of FOXA1 or androgen treatment resulted in decreased expression of the AR-FOXA1 target gene keratin 8 type II (KRT8). Similarly, knockdown of CTCF resulted in increased expression of the putative CTCF target gene C-terminal binding protein 2 (CTBP2). Consistent with these findings, the ChIP-seq and PWM predictions of enhancer activities were validated by allele-specific enhancer reporter assays and eQTL (aseQTL) analyses. These results describe and validate an integrated approach to identify gene regulatory mechanisms at cancer susceptibility loci.