|dc.description.abstract||Colorectal cancer (CRC) is the fourth most common cancer in the UK. It develops in a step-wise manner from normal tissue to cancerous tumour over many years. The heritability of CRC risk is estimated variously up to 35%. One possible explanation for thus far unexplained heritability is gene-environment interaction (GxE), where the effects of a particular environmental factor are moderated by genotype of an individual. Among numerous environmental factors implicated in CRC, low vitamin D intake and low plasma vitamin D concentration have been associated with increased risk of CRC in observational studies. While supplementation studies have failed to establish a beneficial relationship, these studies can be criticised variously and whether vitamin D can reduce CRC risk therefore remains an open question.
This study utilises in vitro, ex vivo and in vivo approaches, combining wet-lab and in-silico analyses to investigate the differential effect of vitamin D on colorectal cell lines, organoids or tissue in relation to genotype at CRC risk loci.
Vitamin D GxEs were first investigated in an in-house observational study of normal human colorectal mucosa. Genes were first identified which were correlated with expression of the vitamin D receptor gene (VDR). Using an agnostic approach to genomic variants associated both with CRC risk and with expression (cis-eQTLs) of nearby genes themselves correlated with VDR, three GxEs were identified for further study. These were then tested in a human supplementation study, with an interaction between rs7012462 and vitamin D on MYC expression being of particular interest. This GxE was then further explored in CRC cell lines. To demonstrate that this genomic region was definitively responsible for the GxE effect, efforts were then made to delete the region in CRC cell lines using a CRISPR-Cas9 approach.
This study also utilised organoid models derived from adult normal colorectal mucosa (ANMO), adult colorectal cancer (CRCO) and foetal normal colorectal mucosa (FNMO) to investigate vitamin D effects in a model system isolated from many of the vagaries in heterogeneity found when studying a human population. Organoids were established in long-term culture and importantly validated as a model of human colorectal tissue. Vitamin D was shown to induce effects on ANMO phenotype, and to induce transcriptomic changes in ANMO derived from 9 individuals (447 upregulated and 423 downregulated genes, FDR p<0.05). KEGG analysis of differentially expressed genes implicated multiple pathways of relevance to CRC. Stemness-markers LRIG1 and MSI1 along with BMP pathway members BMP4 and SMAD7 were identified as genes of particular interest. Meta-analysis with another recently published ANMO vitamin D study (Fernandez-Barral et al 2019) provided a robust list of vitamin D responsive genes (795 differentially expressed genes). These differentially expressed genes were enriched for intolerance to loss of function mutations, and were enriched for functional mutations in the TCGA CRC dataset. CRCO appeared more resistant or to have a more idiosyncratic response to vitamin D: only 9 differentially expressed genes in CRCO derived from 4 individuals (8 upregulated and 1 downregulated, FDR p<0.05). Meta-analysis with the recently published CRCO vitamin D study did though identify 201 genes differentially expressed, of which 116 genes were common to ANMO.
Gene-environment interactions were also investigated in ANMO, first involving vitamin D pathway gene variants, and then involving CRC risk variants. There was marked variation in CYP24A1 expression (the gene responsible for deactivating the active form of vitamin D) following calcitriol treatment between individuals, with a GxE noted involving rs209955. When CYP24A1 effects were taken into account in testing GxEs involving CRC risk variants, genotype dependent BMP4 expression changes in response to vitamin D were identified in ANMO involving rs35107139 at the 14q22.2 locus. This same GxE was also then identified in CRC cell lines, and in human supplementation and observational studies. Finally, the modifying effect of vitamin D on CRC risk associated with genetic variation at the BMP4 CRC locus suggested a predictive relationship to subsequent CRC onset in the UK Biobank.
Taking a broader approach, this study has used ATACseq and VDR ChIPseq to identify vitamin D dependent open chromatin regions and VDR transcription factor binding regions in CRC cell lines and ANMO. While VDR-RXR motifs were identified in the majority of these regions, this was not among the most enriched motifs. The degree of overlap between different vitamin D dependent epigenetic regions, including from previously published ChIPseq, ChIPexo and FAIREseq studies, with each other and with genomic variants associated with CRC risk was reviewed (along with variants associated with risk of inflammatory bowel disease, rheumatoid arthritis, schizophrenia or intraocular pressure for comparison), showing that these relationships may be tissue, treatment and disease specific.
Finally, this study reviewed effects of 12 weeks vitamin D supplementation on gene expression in adult normal rectal mucosa. While no single gene was differentially expressed at a genome-wide level following supplementation, by taking a gene-correlation network approach, vitamin D induced connectivity changes were identified involving groups (modules) of genes. By reviewing non-preserved modules before and after vitamin D supplementation, hub genes have been identified which may play a key role in modulating vitamin D actions in normal rectal epithelium.
In summary, this project has identified vitamin D effects both at a genome-wide and single gene scale. Vitamin D epigenetic effects have been described, with enrichment of vitamin D dependent epigenetic peaks in proximity to genomic variants associated with CRC. It has also identified in a vitamin D supplementation study that while no single gene was differentially expressed in normal rectal mucosa, groups of genes changed their expression patterns. At the single-gene level, genotype-specific responses to vitamin D have also been identified. In particular GxEs involving vitamin D and rs7012462 on MYC expression and involving rs35107139 on BMP4 expression have been demonstrated. Importantly the modifying effect of vitamin D on CRC risk associated with genetic variation at the BMP4 CRC locus suggests a predictive relationship to subsequent CRC onset in the UK Biobank, and this provides compelling rationale for genotype-stratified RCTs to test protective effects of vitamin D on CRC risk.||en