Background
Vitamin D deficiency is associated with the risk of several common cancers, the strongest evidence supporting a link between vitamin D and colorectal cancer [
1,
2]. However, a causal association has yet to be convincingly demonstrated, because the available observational evidence may be participant to several potential confounders. Environmental risk factors associated with CRC are also associated with vitamin D status (i.e. co-causality; e.g. physical activity), while CRC or its treatment may itself lower plasma vitamin D levels (i.e. reverse causation). However, a recent randomised-control trial (RCT) reported an association between supplementation, vitamin D receptor genotype, and risk of colorectal adenoma, supporting the premise that the beneficial effect may be causal [
3]. Meanwhile, vitamin D-related genetic variation has been shown to influence the association between 25-OHD level and CRC survival [
4‐
6], with a recent meta-analysis of RCT data strongly supporting a causal effect for vitamin D supplementation on CRC mortality [
7,
8].
Differences in gene expression have been reported in CRC and adenoma tissue relative to normal colorectal tissue [
9‐
12], with genes involved in metabolism, transcription, and translation and cellular processes commonly altered [
13]. Recent transcriptome wide association studies confirm the importance of gene expression in carcinogenesis [
14,
15]. Vitamin D broadly influences gene expression through activation of the ligand-activated transcription factor
VDR, which has been shown to influence cancer cell growth in vitro [
16]. Therefore, investigation of gene expression in the colorectum in the context of vitamin D status or supplementation may provide fresh insight into mechanisms underlying the relationship between CRC and vitamin D. Recent evidence suggests one mechanism may be that 1,25-dihydroxyvitamin D3 modulates immune and inflammatory pathway genes in large bowel epithelium [
17]. However, differential expression in response to high dose 1,25-dihydroxyvitamin D3 may not accurately reflect the relationship between vitamin D status and gene expression at normal or low vitamin D levels, or in response to regular vitamin D3, the most commonly used vitamin D supplement.
We investigated whether circulating vitamin D concentration is associated with differential gene expression in rectal normal mucosa using a 2-Phase approach with validation of putative biomarkers in an independent study dataset. We directly assayed total 25-OHD, which reflects both dietary intake and skin synthesis of vitamin D [
18,
19] and investigated its relationship with gene expression in normal mucosa, assessed by microarray. In the Phase 1 correlative study we sought to identify a prioritised list of differentially expressed genes associated with 25-OHD level. In Phase 2, we conducted a study in human volunteers who were supplemented with oral vitamin D to determine whether the corresponding transcriptomic response was induced in vivo. Using blood peripheral blood mononuclear cells (PBMC) transcriptomic analysis, we also identified potential blood biomarkers that indirectly indicate a beneficial response in the host rectal mucosa.
Discussion
This study reveals demonstrable differences in gene expression patterns in normal rectal mucosa correlated with plasma 25-OHD level. These differences are consistent with beneficial effects on processes relevant to colorectal carcinogenesis. Furthermore, we show that oral supplementation with vitamin D induces changes in the prioritised gene list. This indicates that the beneficial expression “signature” is not static, but rather can be modified by oral vitamin D supplementation, at least within the timescale tested here. Although we were not able to directly test cancer endpoints, there is considerable published evidence supporting the premise that enrichment of this favourable gene-set imparts anti-tumour effects.
Homeodomain-interacting protein kinase 2 (
HIPK2) is a known tumour suppressor gene [
60] and the Protein Phosphatase 1 Catalytic Subunit Gamma gene (
PPP1CC), a published molecular marker of CRC [
61]. We found expression changes in these genes in blood to have predictive value in reflecting rectal mucosa response to supplementation. Hence, these may have utility as blood biomarkers of a beneficial epithelial response to supplementation. In addition, the effect on gene expression in blood PBMCs appears robust, since we replicated the effect in a large, independent, expression dataset, namely the BEST-D trial in which subjects were administered oral vitamin D supplementation (2000/4000IU 12 months).
We devised a 2-Phase in vivo approach, firstly to identify differentially expressed genes in the rectal epithelium associated with plasma 25-OHD and determine the GO terms and processes linked to that prioritised gene list. However, our ultimate aim was to establish whether these transcriptomic responses could be recapitulated by oral vitamin D supplementation, thereby demonstrating a modifiable transcriptomic landscape. Many of the top-ranked genes associated with higher 25-OHD level have links with CRC, for instance
CNN1 [
36],
COX7A1 [
37], PEG3 [
62],
PIP5K1C [
38],
TAGLN [
63] and
DAAM2 [
64]. Furthermore, we highlight a number of genes within processes relevant to tumourigenesis which are associated with 25-OHD level and influenced by supplementation. The directions of effect of these genes were consistent with tumour suppressor activity. Enrichment of pathways involved in cell migration and cell death validate published in vitro data which demonstrate vitamin D-induced growth arrest and apoptosis of CRC cell lines, modulation of the
Wnt signalling pathway, DNA repair and immunomodulation [
16]. Published clinical data also corroborate our current findings, for example Protiva et al., reported upregulation of genes involved in cell adhesion in response to 1,25(OH)2D3 [
17], while Bostick, reported increased cell differentiation and apoptosis in the normal human colorectal epithelium [
65]. Taken together, these data suggest possible mechanisms underlying the widely reported link between vitamin D deficiency and increased CRC risk [
1,
2]. It also might explain the recently reported beneficial impact of supplementation on CRC survival outcomes [
8].
Despite compelling published observational and pre-clinical data, the link between vitamin D and risk of cancer and several other traits remains controversial. Indeed, several large intervention trials have shown no benefit on cancer endpoints (VITAL Trial [
66], Vitamin D Assessment (ViDA) study [
67,
68] and Baron et al. [
69]). However, participants in these trials were predominantly sufficient for vitamin D at the trial outset, thereby potentially blunting beneficial effects [
70]. We have previously rehearsed potential reasons why previous study designs might have failed to detect real effects [
70]. To counter potential confounding effects, we conducted a Mendelian randomisation study but this also did not demonstrate a beneficial effect of circulating vitamin D on CRC risk. However, available genetic instrumental variables are weak and explain only a small portion of variance of 25OHD levels [
71,
72].
In this study, 18% of participants receiving vitamin D supplementation exhibited a response in the colorectal epithelium (the putative target tissue). Unmeasured variables may account for the marked inter-individual variation in response including ethnic or genetic background, dietary, lifestyle, pharmacological (e.g. concurrent medication) or pathological effects (e.g. unknown viral infection during course of supplementation). Variation in the activity of the vitamin D enzymes or carriers (e.g. CYP24A1, GC and CYP27B1 undetected in the current dataset) may impact responses in the mucosa relative to 25-OHD change, and we noted differential CYP2R1 change in the rectum after supplementation between those with/without a response to our candidate gene-set (FC 0.80 vs. 1.04, P = 0.007). Irrespective of the cause of inter-individual variation in response, if our hypothesis holds that expression changes translate to cancer endpoints, this low response rate would adversely impact on statistical power of trials conducted to date which have tested the effect of vitamin D supplementation on clinical endpoints. Such trials routinely perform subgroup analyses based on change in circulating 25-OHD level, yet the current study reveals poor correlation between plasma level and mucosal gene expression changes, suggesting non-linear responses to vitamin D may introduce further heterogeneity to clinical endpoints. Future work, including GWAS and machine learning approaches will aim to define whether ‘response’ can be determined at baseline is required.
Crucially, we have identified blood biomarkers that reliably identify participants who respond to vitamin D supplementation by inducing gene expression changes in the target tissue. The value of these biomarkers is replicated in a larger independent expression dataset. Further work is required to assess the utility of respective blood protein assays (e.g. ELISA) and reproducibility of these blood biomarkers in identifying mucosal response across a larger cohort. Nevertheless, these exciting and novel findings provide rationale for a trial of vitamin D and CRC prevention using easily assayed blood gene expression signatures as intermediate biomarkers of response.
Whilst the 2-Phase design of an intervention study informed by our correlative dataset has many positive attributes, the study has a number of limitations. First, the low median level of 25-OHD and narrow positively skewed distribution of 25-OHD in the Phase 1 cohort may have masked some true associations between vitamin D level and gene expression. Failure to identify individual gene significance after adjustment for genome-wide multiple testing may also indicate inadequate sample size, physiological autoregulation maintaining constant gene expression despite differences in circulating 25-OHD, differences between plasma and rectal mucosa concentrations of 25-OHD or 1,25-OHD [
73] or heterogeneity in cell type in sampled mucosal tissue.
This intervention study is larger than many published studies of gene expression and vitamin D supplementation [
17,
74‐
77], yet may still have limited power to achieve individual gene significance. Phase 2 participants were recruited as a subset of the Phase 1 cohort, which may influence gene-set enrichment test results. However, we did not select those who received supplementation based on 25-OHD level or baseline gene expression, which could have led to overfitting of the data, but instead took an unselected group. If anything, this approach could blunt the observed effect of supplementation, as mucosal response to supplementation may be capped in those with specific 25-OHD or favorable patterns of gene expression at baseline. Despite this, we observed significant enrichment of the candidate gene-set derived in Phase 1 in those receiving supplementation. Unmeasured variation in environmental exposures (e.g. diet or UVB exposure) may have influenced responses, which should be accurately detailed in future studies of vitamin D supplementation. Sampling of rectal mucosa but not colonic mucosa avoided the use of cleansing bowel laxatives which may influence gene expression [
78], yet limits the generalisability of our findings to more proximal colonic mucosa. Sampling after 12 weeks of supplementation may not adequately capture early or later gene expression changes yet more frequent or delayed sampling would provide additional practical and ethical challenges. Finally, we recognise responses to vitamin D supplementation and the biomarker utility of
HIPK2 and
PPP1CC require further mechanistic study (e.g. protein expression) to validate and expand these findings towards a fuller understanding of both biological mechanisms and biomarker potential.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.