In accordance with the Declaration of Helsinki, the Human Research Ethics Committee of the South African National Blood Service (SANBS HREC clearance certificate number 2010/01) and the Ethics Committee, Faculty of Science, University of Johannesburg (2010/06/03) approved the study. After written informed consent, the SANBS collected blood from randomly selected healthy Black (n = 50; age 17–62 years; 25 males and 25 females) and White (n = 50; age 17–69 years; 25 males and 25 females) South Africans living in Gauteng, SA. Samples were collected across all seasons, though no White individuals were collected in winter for functional analysis. UVI was obtained from the South African Weather Service weather station in Irene, Pretoria, Gauteng. As 25(OH)D₃ has a half-life of 2–4 weeks in circulation (27), the approximate UVI that each individual could have been exposed to was calculated as the 4-week average before blood collection, using the average hourly UVI between 11.00 a.m. and 14.00 p.m. across the years of sample collection (2011–2014).

Liquid chromatography tandem mass spectrometry (LC-MS/MS) facilitated quantification of 25(OH)D₃ concentration in the Department of Clinical Biochemistry, University Hospital of South Manchester (UK), including four human serum pools from the Vitamin D External Quality Assessment Scheme (DEQAS, UK). A concentration of ≥50 nmol/L was considered normal/sufficient (28). 25(OH)D₃ concentration was below the detection limit (3 nmol/L) in four Black samples, while sample was insufficient for six Blacks and eight Whites.

To identify putative functional loci that could influence VDR expression and function through genetic and/or epigenetic mechanisms, a bioinformatics workflow was developed (Methods S1.1, Figures S1 and S2, and Table S1 in Supplementary Material).

Monocytes were isolated and gDNA extracted, at time zero, as previously described (17). VDR methylation analysis by bisulfite pyrosequencing was outsourced to EpigenDx, Inc. (MA, USA). Selected sites typed included 10 CpGs in CpG island (CGI) 1066 spanning enhancer U3 (chr12:48340628-48340806, hg19), 56 CpGs in CGI 1062 spanning the primary promoter (chr12:48299359-48298799), 12 CpGs in CGI 1061 spanning exon 3 (chr12:48258845-48259024), and 18 CpGs in CGI 1060 spanning exon 9 (chr12:48238512-48238799). CpGs were numbered in the 5′–3′ direction.

Genotyping by pyrosequencing was outsourced to EpigenDx, Inc. (MA, USA). Typed SNPs included GC rs7041, rs4588, and rs146681395; TLR1 rs5743551 (A7202G), rs146940675, rs4833095 (N248S), rs111807776, rs143576765, rs5743618 (I602S), rs151036585, rs5743613 (P315L), rs185747096, rs146782074, and rs200631178; TLR2 rs3804099 (T597C); TiRAP rs8177374 (S180L) and rs141792148; VDR rs11168312, rs11568820 (Cdx-2), rs182743714, rs184448883, rs4516035 (GATA), rs2228570 (FokI), rs187018098, rs71951818, rs1544410 (BsmI), rs7975232 (ApaI), rs731236 (TaqI), and rs4987032; CYP24A1 rs6068812; and DMNT3A rs1550117 and rs112621472.

To estimate TLR2/1 pathway efficacy, VDR mRNA, VDR protein, CAMP mRNA, hCAP-18 peptide, and CYP24A1 mRNA, hereafter referred to as functional variables, were quantified following different treatments of monocyte/macrophage cultures that were established as previously described (17). Some monocytes were retained for functional analysis at time zero (baseline). The rest were settled in culture for 16 h prior to 24 h treatment with the vehicle control, 1,25(OH)₂D₃ (10 nM, Sigma Aldrich, St Louis, MO, USA), the TLR2/1 elicitor Pam₃CSK₄, (6.5 µg/ml culture media, EMC microcollections, Tuebingen, Germany), or both the elicitor and 1,25(OH)₂D₃.

The relative level of VDR, CAMP, and CYP24A1 mRNA was quantified by RT-qPCR and VDR protein and intracellular hCAP-18 peptide by flow cytometry as previously described (17). Gene normalization was performed against two stably expressed reference genes: ubiquitin C (UBC) and tyrosine-3-monooxygenase/tryptophan-5-monooxygenase activation protein, zeta polypeptide (YWHAZ). Gene expression was quantified using the comparative CT method according to the MIQE guidelines, using inter-run calibrators and qBASE^PLUS software. To compensate for variability in fluorescence readings between experiments on the flow cytometer, the median fluorescence intensity (MFI) of broad-spectrum calibration beads was used to normalize data and thereby provide a calibrator for instrument-related variation in the flow cytometry readings over time. For hCAP-18, mouse IgG₁ anti-human hCAP-18 primary antibodies (10 µg/ml, Abcam, Cambridge, UK) and APC-conjugated goat anti-mouse IgG1 secondary antibodies (2 µg/mL, Abcam, Cambridge, UK) were used. Western blotting facilitated tracing of hCAP-18 processing and secretion (Methods S1.2 in Supplementary Material).

IBM^® SPSS^® Statistics (v 23; SPSS Inc., Chicago, IL, USA) and SIMCA (v 14; Umetrics, Umea, Sweden) facilitated statistical analysis. Normal distributions were obtained by natural logarithm (ln) transformation of all functional data, except VDR protein. Correlation coefficients were computed using Pearson or Spearman’s rho. A general linear model was used for multivariate analysis of variance to assess main effects and factor interaction. Mann–Whitney U tested methylation differences. Pearson’s Chi-square test for independence assessed SNP distribution. Orthogonal projections to latent structures discriminant analysis (OPLS-DA, Methods S1.3 and Figures S3 and S4 in Supplementary Material) facilitated the study of the multivariate effect of VDR methylation, SNPs, and environment on TLR2/1-VDR signaling.

Plasma 25(OH)D₃ concentration was quantified by LC-MS/MS (Figure 1). Race had a significant main effect on 25(OH)D₃ concentration (P < 0.001). Overall, Blacks were deficient (<50 nmol/L) with a significantly lower 25(OH)D₃ concentration (P < 0.010) than Whites, who had normal levels (≥50 nmol/L). Race interacted significantly with season (P < 0.010), showing a lower 25(OH)D₃ concentration in Blacks in winter (P < 0.050) and spring (P < 0.001), but not in summer and autumn.

To assess the impact of VDR methylation on TLR2/1-VDR signaling, VDR methylation was quantified by bisulfite pyrosequencing. Regional methylation (Figure 2A) was compared between Blacks and Whites across the VDR in key CGIs identified through a bioinformatics workflow (Methods S1.1, Figures S1 and S2, and Table S1 in Supplementary Material). Comparing regional methylation, Whites had significantly lower levels at CGI 1062 (P < 0.001) and CGI 1060a (CpG 1-5, 5′ of TaqI, Figure 2A, P < 0.010) than Blacks, but higher levels at CGI 1060b (CpG 7-18, 3′ of TaqI, Figure 2A, P < 0.001). Significant racial differences in site-specific methylation (Results S2.2 and Figure S5 in Supplementary Material) were common in CGI 1062 (27/56) and 1060 (10/18), but less so in CGI 1066 (1/10) and1061 (3/12).

To assess genetic variation between individuals in the TLR2/1-VDR signaling pathway, SNPs in several genes of the pathway and in the de novo methyltransferase enzyme, DNMT3A, were genotyped by pyrosequencing. Several SNPs, including DNMT3A SNPs, were monomorphic in the study population. Except for VDR BsmI and TaqI, the frequency distribution of polymorphic SNPs differed significantly between Blacks and Whites (Table 1). The 1000 Genomes Deep Catalog of Human Genetic Variation confirmed race-specific genotype frequency distribution, also for BsmI and TaqI though not for ApaI, comparing Africans (Yoruba) and Caucasians (Table S2 in Supplementary Material).

To assess the efficacy of the TLR2/1-VDR signaling regarding functional variables in the pathway (VDR mRNA, VDR protein, CAMP mRNA, hCAP-18, and CYP24A1 mRNA), multivariate analysis of the effect of race, 25(OH)D₃ status, season, and treatment (with/without in vitro 1,25(OH)₂D₃ supplementation and/or TLR2/1 elicitation) was performed (Table 2). Treatment had a significant main effect on functional variables (Figure 3), while season, race, and 25(OH)D₃ status showed several complex interactions regarding VDR mRNA and VDR protein (Figure 4), CAMP mRNA and CYP24A1 mRNA (Figure 5) and hCAP-18 (Figure 6). TLR2/1 elicitation induced VDR protein (Figure 3B, P < 0.001), while 1,25(OH)₂D₃ induced CAMP mRNA and CYP24A1 mRNA (Figures 3C,E, P < 0.001). Considering interactions, 25(OH)D₃-deficient Blacks had significantly lower VDR mRNA in summer than deficient Whites or Whites and Blacks with a normal 25(OH)D₃ status (P < 0.050, Figure 4A). In contrast, 25(OH)D₃-deficient Whites had significantly lower VDR mRNA in autumn than deficient Blacks or Whites and Blacks with a normal 25(OH)D₃ status (P < 0.050). VDR protein dropped significantly in summer and autumn for Whites and Blacks, respectively, showing a significant race difference in summer (P < 0.050, Figure 4B). CAMP mRNA increased significantly from summer to autumn in Whites, being significantly higher than Blacks (P < 0.050, Figure 5A). Whites with a normal 25(OH)D₃ status had significantly higher CYP24A1 mRNA than 25(OH)D₃-deficient Whites or Blacks and Blacks with a normal 25(OH)D₃ status (P < 0.050, Figure 5B). A notable decrease in hCAP-18 was observed in Whites from spring through summer to autumn being significantly higher in spring and significantly lower in autumn, compared to Blacks (P < 0.050, Figure 6A). All individuals with a normal 25(OH)D₃ status showed a similar, significant decrease in hCAP-18 from spring to autumn (P < 0.050, Figure 6B). Blacks with a normal 25(OH)D₃ status had significantly more hCAP-18, than normal Whites and deficient Blacks or Whites (P < 0.050, Figure 6C). To confirm that the intracellular decrease in hCAP-18 in response to seasons with higher 1,25(OH)₂D₃ reflects hCAP-18 processing and LL-37 secretion, we performed Western blotting on 20 additional randomly selected healthy Black (n = 10) and White (n = 10) South Africans. These individuals, collected in winter and for whom no other variables were quantified, were also included in 25(OH)D₃ quantification (shown in Figure 1). Western blotting showed individual-specific hCAP-18 processing and LL-37 secretion, which depended on 25(OH)D₃ status, extent of 1,25(OH)₂D₃ supplementation, and incubation time (Figure 7). For example, an individual with a sufficient 25(OH)D₃ status (>50 nmol/L, Figure 7A) had the highest level of intracellular LL-37 under control condition and already secreted LL-37 upon TLR2/1 elicitation at 16 h, and secreted even more with moderate 1,25(OH)₂D₃ supplementation (10 nM) or at 24 h. However, with excessive (50 nM) supplementation, the individual secreted less (16 h) or none (24 h). In contrast, 25(OH)D₃-deficient individuals secreted LL-37 slower and required at least 50 nM of 1,25(OH)₂D₃ to secrete at 16 h (Figure 7C). An individual with severe 25(OH)D₃ deficiency (23.2 nmol/L) secreted LL-37 only after 24 h in the presence of 50 nM 1,25(OH)₂D₃ (Figure 7D). It should be noted that secreted LL-37 was undetectably low in 13 of the 20 randomly selected individuals subjected to Western blotting.

To further explore the relation between independent variables (UVI, 25(OH)D₃ concentration, age and regional methylation) and dependent functional variables, following different treatments, correlation analysis was performed (Table 3). UVI correlated moderately with circulating 25(OH)D₃, which in turn showed a moderate, negative correlation with regional methylation of the VDR promoter CGI 1062. Regarding inter-CGI correlations, CGI 1066 and 1060b correlated negatively with CGI 1062, 1061, and 1060a. The significant correlation between 25(OH)D₃ and methylation was maintained irrespective of season, with partial correlations controlling for seasonal changes in UVI yielding similar correlation coefficients (data not shown). Considering functional variables, regional methylation showed moderate correlation with VDR protein under certain treatments: CGI 1066 and 1060b (positive for all treatments but baseline); 1060a (negative for all treatments but baseline); 1062 and 1061 (negative for 1,25(OH)₂D₃ and/or TLR2/1 elicitor). As expected, each functional variable showed correlation between all treatments (not shown). In addition, VDR mRNA showed a number of moderate positive correlations with CAMP mRNA, across a number of treatments, similarly so, CAMP with CYP24A1 mRNA. Age showed a moderate negative correlation with hCAP-18 peptide with TLR2/1 elicitor + 1,25(OH)₂D₃ (r = −0.330, P < 0.01, n = 99).

To identify the main variables underlying differential levels of VDR and downstream targets (CAMP, hCAP-18, and CYP24A1), the multivariate OPLS-DA statistical method was performed and validated as described (Methods S1.3 in Supplementary Material). Evaluation of the loadings S-plots (Methods S1.3 and Figure S4 in Supplementary Material) and descriptive assessment of the scores space for each model, identified combined effects of genetics, VDR methylation, vitamin D status, and UVI on the efficacy of TLR2/1-VDR signaling, as assessed by the level of functional variables produced in response to various treatments. Functional variables were categorized as above/below average and X (independent) variables that significantly (P ≤ 0.050) and/or measurably (≥1.5-fold or ≤0.667) discriminate mean values for above/below average response were recorded with their correlation (Table 4). Methylation at CGI 1062, CpG 3 significantly and most notably discriminate mean values for above and below average VDR protein level, particularly with TLR2/1 elicitation. Other sizeable methylation–function interactions observed that were significant and occurred in at least two treatments of a functional variable included 1060 CpG 6 [positive impact on VDR mRNA with 1,25(OH)₂D₃ supplementation or TLR2/1 elicitation], 1062 CpG 23 (negative impact on hCAP-18 at control and elicitation, with or without supplementation), and CGI 1060a across several neighboring CpGs [1060 CpG 1-5, negative impact for 1,25(OH)₂D₃ supplementation, TLR2/1 elicitation or both, clustering by 25(OH)D₃ status in the absence of supplement and by the TaqI VDR SNP with/without supplement]. Methylation at CGI 1066, CpG 1-6 also influenced VDR protein level positively following 1,25(OH)₂D₃ treatment, with or without TLR2/1 elicitor, showing clustering based on race and SNPs in VDR (BsmI, ApaI, and TaqI), TLR1 (I602S), and TIRAP (S180L), particularly in the presence of supplement. Compared to VDR protein, CAMP mRNA level was inversely impacted by CGI 1066, CpG 1-6 methylation, following 1,25(OH)₂D₃ treatment, while it clustered only based on 25(OH)D₃ status and TaqI. Methylation of the CpG located at TaqI was the only methylation site with significant impact on VDR mRNA level following 1,25(OH)₂D₃ or TLR2/1 elicitor, but not both. Considering environmental factors, UVI and 25(OH)D₃ significantly influenced VDR mRNA level following treatment with 1,25(OH)₂D₃, TLR2/1 elicitor, or both. VDR protein level at baseline and CAMP and CYP24A1 mRNA upon elicitation were significantly influenced by 25(OH)D₃. 25(OH)D₃ status and TaqI most consistently showed clustering during descriptive assessment of the scores space of computed models. Model construction for hCAP-18 was less favorable and we were unable to identify clustering. Clustering by season, although tested, was not observed.