The most arresting finding in this study is that the number of VDR binding sites in primary CD4+ cells is strongly correlated with 25-hydroxyvitamin D levels. The previous VDR ChIP-seq experiments using MCLs and LCLs found an increase in VDR binding site occupancy following treatment with supraphysiological levels of calcitriol [9
]. Our finding of a far greater number of VDR binding sites in sufficient vitamin D samples than insufficient samples suggests that this effect also occurs with different in vivo
levels of vitamin D. In vivo
levels of 25-hydroxyvitamin D are directly associated with the number of VDR binding sites.
VDR binding sites are enriched for markers of active transcription and open chromatin; 25(OH)D ≥75 samples seemed to be less enriched for these markers than 25(OH)D <75, perhaps reflecting binding to open chromatin state in 25(OH)D <75 samples.
We have confirmed that the observation of significant overlap between VDR binding and genomic regions implicated in autoimmune diseases in LCLs is also seen in primary CD4+ cells [9
]. Gene ontology analysis suggests that VDR binding in conditions of 25-hydroxyvitamin D sufficiency may be more directly related to immune cell function. This is supported by the observed higher levels of VDR binding near genes expressed specifically in T-regulatory and T-helper cells in 25(OH)D ≥75 but not 25(OH)D <75 samples.
We found a lack of classical VDR binding motifs within the VDR ChIP-seq peaks. In the ChIP-seq studies in MCLs and LCLs the authors identified classical DR3 motifs at differing proportions of sites (32% in MCLs, 67% in LCLs) with SP1-like and ETS-like non-classical peaks identified in the MCL ChIP-seq study (23% and 12% respectively) [9
]. We found enrichment of CTCF motifs in several of our samples but were unable to identify any previously described VDR motifs. One possibility is that in vivo
VDR binding is modulated by protein-protein interactions with co-factors: SP1 and ETS1 are known to modulate VDR binding, and there is some evidence that interactions between SP1 and VDR may enable modulation of genes that lack a classical VDR recognition motif [43
]. Several other proteins are known to bind in association with VDR, including NR4A1 and c-MYC [45
]. CTCF is known to modulate DNA binding via protein-protein interactions with other nuclear receptors [47
]. However, it is unlikely that protein-protein interactions with transcription factors with specific recognition sequences can explain most of these motifless binding sites since one would have expected to find that motif through MEME-ChIP analysis. It may be that in response to physiological levels of 25-hydroxyvitamin D most VDR binding occurs at motifless binding sites similar to those identified by ENCODE [35
], supported by the increased overlap with DNase I peaks. Another possibility is that the lack of motifs may reflect the fact that these CD4+ cells were not stimulated with 1,25D3
, as the previous LCL ChIP-seq did not find classical motifs prior to stimulation [9
]. Alternatively, current motif-finding methods may be insufficient to locate true VDR binding motifs. Further research will be needed in more lymphocyte subsets to delineate further the role of non-classical binding sites in VDR binding. It would also be useful to obtain 1,25D3
, parathyroid hormone and calcium measurements for future study.
The overlap between genomic regions associated with many autoimmune diseases and VDR binding in primary CD4+ cells strongly suggests a role for vitamin D in many of these diseases, as already seen for MCLs and LCLs [9
]. This is strengthened by the observation that this effect tends to be stronger in individuals sufficient for 25-hydroxyvitamin D. Interestingly, the magnitude of enrichment for autoimmunity increased as the flanks of the region surrounding implicated SNPs was reduced. This further suggests that this is not a chance finding and that VDR binding may have a functional role in modulating adaptive immunity in autoimmune diseases. We also controlled for genomic architectural features that could bias our results and observed that the results were not substantively altered. Future functional work should focus on the effects of VDR binding on nearby gene expression and targeted sequencing in patients with autoimmune conditions to identify possible rare variants affecting VDR binding.