Integration of VDR genome wide binding and GWAS genetic variation data reveals co-occurrence of VDR and NF-κB binding that is linked to immune phenotypes

Abstract

Background: The nuclear hormone receptor superfamily acts as a genomic sensor of diverse signals. Their actions are often intertwined with other transcription factors. Nuclear hormone receptors are targets for many therapeutic drugs, and include the vitamin D receptor (VDR). VDR signaling is pleotropic, being implicated in calcaemic function, antibacterial actions, growth control, immunomodulation and anti-cancer actions. Specifically, we hypothesized that the biologically significant relationships between the VDR transcriptome and phenotype-associated biology could be discovered by integrating the known VDR transcription factor binding sites and all published trait- and disease-associated SNPs. By integrating VDR genome-wide binding data (ChIP-seq) with the National Human Genome Research Institute (NHGRI) GWAS catalog of SNPs we would see where and which target gene interactions and pathways are impacted by inherited genetic variation in VDR binding sites, indicating which of VDR's multiple functions are most biologically significant.

Results: To examine how genetic variation impacts VDR function we overlapped 23,409 VDR genomic binding peaks from six VDR ChIP-seq datasets with 191,482 SNPs, derived from GWAS-significant SNPs (Lead SNPs) and their correlated variants (r ² > 0.8) from HapMap3 and the 1000 genomes project. In total, 574 SNPs (71 Lead and 503 SNPs in linkage disequilibrium with Lead SNPs) were present at VDR binding loci and associated with 211 phenotypes. For each phenotype a hypergeometric test was used to determine if SNPs were enriched at VDR binding sites. Bonferroni correction for multiple testing across the 211 phenotypes yielded 42 SNPs that were either disease- or phenotype-associated with seven predominately immune related including self-reported allergy; esophageal cancer was the only cancer phenotype. Motif analyses revealed that only two of these 42 SNPs reside within a canonical VDR binding site (DR3 motif), and that 1/3 of the 42 SNPs significantly impacted binding and gene regulation by other transcription factors, including NF-κB. This suggests a plausible link for the potential cross-talk between VDR and NF-κB.

Conclusions: These analyses showed that VDR peaks are enriched for SNPs associated with immune phenotypes suggesting that VDR immunomodulatory functions are amongst its most important actions. The enrichment of genetic variation in non-DR3 motifs suggests a significant role for the VDR to bind in multimeric complexes containing other transcription factors that are the primary DNA binding component. Our work provides a framework for the combination of ChIP-seq and GWAS findings to provide insight into the underlying phenotype-associated biology of a given transcription factor.

Keywords: ChIP-seq; Cistrome; DR3 motif; GWAS; Immune function; Linkage disequilibrium; NF-κB; Nuclear receptor superfamily; SNP; VDR.

Fig. 1

Schematic workflow and key findings of the study. (Top panel) shows the SNPs and ChIP-seq datasets. Examples of possible overlap of Lead SNPs and SNPs in LD with Lead SNPs at genomic VDR loci are shown with examples of LD blocks. (Bottom panel) shows the number of SNPs identified from individual data sets (GWAS catalog, HapMap3 and 1000 genome project). It also shows the number of SNPs under VDR peaks and SNPs in a DR3-type motif

Fig. 2

Function of a SNP contained within a VDR binding peak associated with Toll-like receptor 1 (TLR1). The top panel shows the presence of rs5743565 (blocked out), which is in LD with rs2101521 (orange arrow) associated with self-reported allergy in GWAS study. VDR peaks are shown as Red block custom tracts. The rs5743565 is associated with expression of the genes TLR10, TLR6 and TLR1. The bottom panel shows the ENCODE ChIP-seq data for binding of different transcription factors in the region harboring rs5743565 and VDR binding sites (in red). The presence of rs5743565 in the binding region is predicted to alter the association of multiple transcription factors and the expression of associated genes, and therefore been assigned a score in RegulomeDB of < = 2

Fig. 3

The co-occurrence of VDR and NF-κB1 binding at sites of significant genetic variation. a The wordcloud characterizes the transcription factors binding at the location of significantly associated Lead SNPs (and not linked to a DR3 motif) were identified by RegulomeDB; 95 different transcription factors, representing eight unique subgroups, overlapped with, and their function was altered by, these SNPs. The font size of transcription factor related to the number of times it was identified to associate with a significant SNP site. b The intersection of VDR and NF-κB ChIP-Seq from CEPH cell lines. The VDR ChIP-seq in the unstimulated and ligand-stimulated states in the CEPH cell lines GM10855 and GM10861 were intersected to generate a consensus cistrome for VDR binding sites in the unstimulated and stimulated states. Similarly, a consensus NF-κB cistrome was generated by intersecting the ChIP-Seq from the cell lines GM12878, GM12891, GM12892, GM15510, GM18505, GM18526, GM18951, GM19099, GM19193. These consensus cistromes were then intersected to reveal the unique and shared binding sites, and the overlap with the identified SNPs. c The effect of rs10174949 genotype on NF-κB binding in ChIP-seq from HapMap cell lines. The altered allele of rs10174949 was predicted to decrease the strength of predicted affinity of NF-κB1 by HaploReg. The ENCODE ChIP-seq data sets of NF-κB for HapMap cell lines for which genotype data was available were downloaded into Integrative Genomics Viewer (IGV). Population ID and the genotype for each cell lines is shown. Cell lines with homozygous reference allele are shown in blue, heterozygous samples are shown in green and sample with homozygous altered allele is shown in red. Samples with the homozygous altered allele (AA) completely lost NF-κB binding compared to the homozygous (GG) or heterozygous (AG) reference alleles