Vitamin D Receptor (VDR) mRNA and VDR Protein Levels in Relation to Vitamin D Status, Insulin Secretory Capacity, and VDR Genotype in Bangladeshi Asians

This study cohort consisted of 143 apparently healthy Bangladeshi adults (subjects with previously diagnosed diabetes were excluded), screened as having a fasting or a postprandial capillary glucose within the upper range of normoglycemia by previously defined criteria (19). These subjects have been extensively phenotyped for risk factors for type 2 diabetes including insulin secretion index (calculated from basal and 30-min glucose and insulin levels during an oral glucose tolerance test [OGTT] [22]). Glucose tolerance was normal in 75.4% of subjects and impaired in 16.4%; 8.2% of subjects were found to have diabetes by 1985 World Health Organization criteria current at the time of initial recruitment (23). DNA from the study subjects was digested with FokI, and fragments were separated by standard agarose-gel electrophoresis; FokI genotypes were designated by the lowercase letter f for presence of the cutting site and by the uppercase F for its absence; ApaI, TaqI, and BsmI genotyping was available from the original study (19).

Forty-one subjects, aged 31–65 years (mean age ± SD, 48.32 ± 9.9), from the original cohort gave informed consent to provision of a further blood sample for the present study, as approved by the District Ethical Committee. Of these, 39% were men and 61% were women. A total of 12.2% had developed diabetes and 4.9% proven IHD since the original study. The data on VDR genotype from the original and current studies was combined for analysis (see below). Serum 25(OH)D concentration was measured by radioimmunoassay (IDS Immunodiagnostics) (intra- and interassay coefficients of variation 8.8% and 10.8%, respectively); levels of 25(OH)D <20 ng/ml defined vitamin D insufficiency (24). 1,25-Dihdroxyvitamin D was measured by specific radioimmunoassay after immunoextraction [Gamma-B 1,25(OH)D RIA kit; IDS] and measured both bioactive forms of calcitriol (1,25-dihydroxyvitamin-D2 and -D3) equally well; intra- and interassay coefficients of variation were <8% and <10%, respectively. The reference range for normal adults is 48–110 pmol/l.

PBMCs were isolated from freshly heparinized venous blood using standard RPMI solutions and Ficoll-Hypaque gradient (Pharmacia, Uppsala, Sweden) separation. The cells were resuspended at 1 × 10⁶ cells/ml in cold PBS and counted in Neubauer chambers. Aliquots containing 10⁷ cells were used in both the VDR mRNA and VDR protein studies.

Total cellular RNA was extracted from 10⁷ PBMCs using the RNeasy mini RNA isolation kit (Qiagen, Crawley, UK), according to the manufacturer’s recommendations, and quantified using a GeneQuant RNA/DNA Calculator (Pharmacia Biotech) with storage at −70°C.

Forward (ATCTGCATCGTCTCCCCAGAT) and reverse (AGCGGATGTACGTCTGCAGTG) oligonucleotide primers and a TaqMan probe (TGATTGAGGCCATCCAGGACCGC) were designed for the VDR gene (accession no. AC004466) using Primer Express software (PE-Applied Biosystems, Warrington, U.K.). The RT-PCR assays were carried out with an ABI Prism 7700-sequence detector as described (25) and were analyzed using sequence detector software (version 1.6.3). Absolute mRNA copy numbers were determined from a VDR-specific standard curve obtained by serially diluting a single-stranded sense oligonucleotide specifying the 98-bp VDR amplicon. Absolute mRNA levels were expressed as VDR mRNA copy numbers per microgram of total RNA (25).

Total 1,25-dihydroxyvitamin-D3 receptor protein (VDR protein) concentrations in PBMC homogenates were measured by specific micro enzyme-linked immunosorbent assay against standard curves (correlation coefficient, 0.999) for purified human VDR protein; intra-assay coefficient of variation was <6.6% (26). Results from two samples with results below the limit of sensitivity were expressed as being at that limit (1 μg/mg of total extract protein).

SPSS for Windows (version 10) was used to examine frequencies, calculate correlations, examine variation of means with genotype by one-way ANOVA, calculate odds ratios, perform χ² analyses, and carry out multiple regression analyses (stepwise to P < 0.05) for the determination of best-fit models. Where VDR genotype was used in multiple regression analysis, all four polymorphisms were included. Pairwise linkage disequilibrium between FokI and the three previously studied polymorphisms of the VDR gene was assessed using the estimate haplotype program (27). The distribution of VDR genotypes was examined for compliance with Hardy-Weinberg equilibrium for the group as a whole, for the vitamin D-insufficient subgroup, and for the extension study subgroup, and the proportions in these groups were examined for the Wahlund effect.

The distribution of the insulin secretion index in the initial study (n = 143) was normalized by log transformation, as in our previous report (showing it to vary with ApaI, BsmI, and TaqI polymorphisms of the VDR gene) (19). Multiple regression analysis showed the significant independent determinants of the insulin secretion index at initial OGTT to be diabetic status (normal, impaired, or diabetic glucose tolerance) (P < 0.001), BMI (P < 0.001), and TaqI genotype (P < 0.001) in the best-fit model obtained (P < 0.001) (Table 1). If diabetic status was excluded from analysis, on the basis that insulin secretion index is a determinant of diabetic status, then the best-fit model (P = 0.002) gave age (P = 0.002), TaqI genotype (P = 0.001), and BMI (P = 0.011) as independent determinants of the insulin secretion index, the constant remaining significant for each of these models. Both models showed TaqI t homozygotes to be associated with the highest levels of insulin secretion (by ∼12%) (Table 2). In vitamin D insufficiency, the FokI polymorphism appeared as an additional independent determinant (Table 1).

Examination for linkage disequilibrium between VDR restriction fragment length polymorphisms (RFLPs) demonstrated that the ApaI, BsmI, and TaqI polymorphisms were in linkage equilibrium with FokI (P = 0.07, 1.0, and 0.15, respectively), whereas there was tight linkage disequilibrium between the ApaI, BsmI, and TaqI genotypes (P < 0.001), as previously shown (19). There was no significant deviation, using χ² tests, from Hardy-Weinberg equilibrium in this group or in the subgroup whose PBMCs were studied; furthermore, there was no evidence of the Wahlund effect for any of the four VDR genotypes.

The mean ± SD serum 25(OH)D, measured on 39 of 41 subjects at the time of reexamination for these VDR studies (insufficient serum, n = 2), was 9.3 ± 4.4 ng/ml, range 1.6–25.8; the serum 1,25(OH)₂D, measured on 34 of 41 re-bled subjects (insufficient serum, n = 7) was 80.1 ± 27.2 pmol/l, range 38.8–150.1. The serum 25(OH)D concentration did not vary with VDR genotype, whether examined in the group as a whole (n = 143) or in just those with vitamin D insufficiency (n = 93). Neither vitamin D status nor serum 1,25(OH)₂D (activated vitamin D) levels, measured at the time of blood sampling for further studies on the VDR (n = 41), varied with VDR genotype, nor did serum 25(OH)D and 1,25(OH)₂D levels correlate with each other (P = 0.98).

The best-fit model (P = 0.024) on multiple regression analysis (n = 39) showed VDR mRNA to be the only independent determinant of insulin secretion index (Table 1).

The values used in the analyses were the means derived from two separate experiments, in which samples from all 41 subjects studied were measured together. However, the results did not differ significantly from the means derived from the three other measurements made for each sample. Determinants of VDR mRNA copy number were examined using stepwise multiple regression analysis (Table 1). The best-fit model obtained, P = 0.004, showed FokI (P = 0.044) and TaqI (P = 0.04) genotypes and insulin secretion index (P = 0.042) to be independent determinants of VDR mRNA in the PBMCs; neither vitamin D status nor activated vitamin D were significant factors. The TaqI TT and FokI FF genotypes were associated with the highest VDR mRNA copy numbers by ∼25.8% and ∼19.7% compared with tt and ff genotypes, respectively (Table 2), although this did not differ significantly with VDR genotype on simple one-way ANOVA.

Cellular VDR protein concentrations were analyzed after we had excluded the results on three samples, where the protein content of the extracts was too low for accuracy to be ensured. Independent determinants of VDR protein in the PBMC extracts in the best-fit model using multiple regression analysis (P = 0.006), and including 1,25-(OH)₂D and vitamin D status in the analysis, were TaqI genotype (P = 0.005) and 1,25(OH)₂D (P = 0.03) (Table 1); these findings were not altered by reanalysis including the three discarded VDR protein values (see above). The TaqI genotype TT was associated with the highest VDR protein levels (by ∼72.9% compared with that for tt) (Table 2). Messenger RNA copy number approached significance as a determinant of VDR protein concentration in the PBMCs (P = 0.06). Insulin secretion index (P = 0.34), factors relating to either body build or diabetes, and the other VDR genotypes (ApaI, P = 0.28; FokI, P = 0.11) did not affect these findings, although BsmI did approach significance (P = 0.07).

Ischemic heart disease, diabetes, or hypertension (or combinations of more than one of these conditions) had developed in eight extension study subjects during the years between the initial and the current studies. The distribution of the four VDR genotypes was compared (for ApaI, BsmI, TaqI, and FokI) between those with and without these problems. The prevalences of the VDR genotypes ff and tt were increased from 11.1% to 16.6% and from 6.6% to 25% in subjects with these disorders. The relative risk of these disorders (95% CI) for absence of the tt genotype was reduced to 0.15 (−0.4 to 1.1) and for the ff genotype to 0.29 (0.15 to 0.49).

We have demonstrated an association between VDR polymorphism and insulin secretory capacity and have shown that this gene is also a significant determinant of the amount of VDR mRNA and VDR protein expressed in PBMCs.

A feature of those subjects re-bled for the VDR mRNA and protein studies is that all but one of the subjects was 25(OH)D insufficient, probably reflecting the fact that these subjects were restudied in the autumn and confirming the inadequacy of the British summer and diet for maintenance of vitamin D repletion in this population (28). Although 25(OH)D levels were low, normal levels of calcitriol [hormonally activated 1,25(OH)₂D] were maintained and there was, therefore, no simple correlation between 25(OH)D and 1,25(OH)₂D levels. We have reported insulin secretion to vary with vitamin D status and, independent of that status, to vary with VDR genotype (19,28). Several studies, including our work on MMP system activation (21), suggest that VDR genotype disease associations may be modulated by vitamin D status. Associations between VDR genotype and rickets, colorectal cancer, tuberculosis, hepatitis B, and leprosy are all more marked in subjects with vitamin D deficiency or in populations where vitamin D deficiency is common (5,16,19,21,29–31). Furthermore, phytohemagglutinin-stimulated growth of PBMCs varies with FokI genotype when studied at half the maximal effective concentration of 1,25(OH)₂D, whereas no such variation was found at the maximal concentration of activated vitamin D (32).

Further to our previous study (19), we have now used multiple regression analysis to allow for the effects of factors known to affect insulin secretion other than VDR genotype, age, body build, and vitamin D status. We have, in the same way, included factors that are known to influence VDR mRNA formation such as insulin secretion (33) and that influence VDR expression, such as activated vitamin D. The best determinants of insulin secretion were TaqI genotype, BMI, and age, although simple analysis of variation of the mean with genotype had suggested in our initial study (19) that the strongest association was with the ApaI genotype. Because ApaI, BsmI, and TaqI polymorphisms are in tight linkage disequilibrium, the fact that our reanalysis suggests TaqI genotype to be a significant determinant of VDR mRNA formation, rather than ApaI, is not at variance with our original observations. The data are complicated by the fact that disease associations have been described with four different RFLPs, as defined by FokI, BsmI, ApaI, and TaqI, and with a mononucleotide repeat (A_n) in the 3′ untranslated region (5,34). The BsmI and ApaI RFLPs are intronic (between exons 8 and 9), and the TaqI polymorphism is located in exon 9, but none of these changes lead to amino acid substitutions. Although it has been postulated that these RFLPs might affect VDR mRNA stability, this has not been confirmed, nor do they appear to affect ligand binding (35–37). These inconsistent results on the relationship between VDR genotype and VDR expression may have several explanations. In analyzing our data on VDR expression for possible effects of VDR polymorphisms, we have allowed for factors that influence VDR expression; indeed, had we relied simply on variation of our VDR data with VDR genotype, we might have concluded that associations did not exist, adding further confusion to this field. Furthermore, several, but not all, other studies have used less precise methods of mRNA analysis on a variety of tissues (PBMCs, parathyroid, fibroblasts, and cell lines) (37–44). Nonetheless, some positive results have been found in the latter studies, including variation in TaqI allele or associated haplotype-specific mRNA levels (37,39,41). In contrast to the other VDR polymorphisms, the FokI polymorphism results in the incorporation of three extra amino acids in the NH₂ terminal of VDR protein, which influences transcriptional activity by modulating interaction with the transcription factor IIB (TFIIB) (43,44).

In those subjects with 25(OH)D insufficiency in the original study (serum levels <20 ng/ml), FokI was a determinant of insulin secretion in addition to TaqI genotype and age. This is consistent with linkage equilibrium between FokI and TaqI polymorphism (unlike the strong linkage disequilibrium that exists between the ApaI, BsmI, and TaqI polymorphisms). It is therefore of interest that we observed significant variation of VDR mRNA with the FokI polymorphism. Because PBMC production of activated vitamin D increases in vitamin D deficiency (45), FokI genotype may contribute to feedback control of expression of the 25-hydroxyvitamin D-1-α-hydroxylase (CYP27B1) gene, located in the same chromosomal region as the VDR gene and active in PBMCs (46,47). Alternatively, the FokI polymorphism may interfere with formation of VDR heterodimers with the retinoid X receptor to form effector complexes (48). This study does suggest that TaqI genotype contributes to the determination of both VDR mRNA and VDR protein levels independently, whereas FokI genotype and insulin secretory capacity also contribute to the determination of VDR mRNA levels. Homozygosity for the VDR TaqI t genotype is associated with lower levels of PBMC VDR mRNA and protein but with higher levels of insulin secretion. This might reflect differences in the effects of ligand-activated receptors between different tissues, but the correlation of PBMC VDR mRNA with insulin secretory capacity indicates that this is unlikely.

We have found circulating levels of activated vitamin D [1,25(OH)₂D] to contribute to VDR protein levels, confirming earlier in vitro studies in PBMCs (45,49), whereas total VDR mRNA (the amplicon used in this study amplifies all mRNA species) was a borderline determinant of VDR protein levels in the PBMCs. In contrast, the insulin secretory capacity was a predictor of PBMC mRNA but not protein. These observations are not unexpected, as the relation between mRNA and its protein will depend on several factors including the half-life of the mRNA compared to the protein and on posttranscriptional processing. Indeed, tissue expression of VDR is directed by a number of promoters and transcription initiation from exons 1a or 1d, and alternative splicing can generate at least 10 different VDR transcripts, each of which was found in all of the tissues examined by Crofts et al. (50). Furthermore, we have previously studied mouse fibroblast and intestinal cell lines and showed that the addition of calcitriol [1,25(OH)₂D] increased VDR protein levels without changes in mRNA (51); similar changes have been reported in humans for healthy macrophages (49).

Many VDR disease associations reported are specifically with the TaqI polymorphism or associated haplotypes. It is with the TaqI polymorphism that we have found variation in insulin secretion and in TIMP1 responses to increases in circulating MMP9 (21). The apparent variation in risk for IHD and type 2 diabetes with TaqI and FokI polymorphisms therefore warrants further investigation in larger populations. Further work is needed to define the mechanisms by which such variations in VDR synthesis lead to variation in cellular function and to determine whether such variation is abolished by adequate vitamin D repletion. The prescient suggestion, made 25 years ago, that requirements for vitamin D may vary between individuals (52) is now supported by clinical findings in tuberculosis, uremia, and colonic adenoma (16,30,53,54). It is clear that susceptibility to a wide range of disorders associated with VDR polymorphism, including type 1 and type 2 diabetes, could be reduced if adequacy of vitamin D repletion could be redefined to include those with the highest requirements (6–8,19,20,55). Furthermore, appropriate supplementation and maintenance in populations where vitamin D status is commonly inadequate is recommended (28,56–58).

National Health Service R&D, NE Thames Region (now North Thames Region), and Diabetes U.K. (previously the British Diabetic Association) provided grant support.