Introduction

The metabolic syndrome (MetS), is a condition that often clusters together central obesity, elevated blood pressure, lower HDL cholesterol, hypertriglyceridemia and hyperglycemia1. MetS increases type 2 diabetes risk and that of cardiovascular disease by 5-folds and 1.7-folds, respectively2,3. MetS is heritable and polygenic4. Genetics contributes to 16%-85% of Body Mass Index (BMI) variability5 and 37%-81% in that of waist circumference (WC) (e.g6.). MetS is an important public health threat triggering higher disability, health care costs and mortality from all causes7,8,9.

Moreover, obesity may be directly involved in the etiology of vitamin D deficiency, with prior evidence of an inverse relationship between serum 25-hydroxyvitamin D [25(OH)D] concentration and various measures of adiposity10. Conversely, vitamin D3 may influence obesity risk by modulating intracellular calcium homeostasis, due to the fact that higher intracellular calcium triggers lipogenesis and suppresses lipolysis11. Many organs express vitamin D receptor (VDR), a component the super-family termed “nuclear hormone receptor”. The complex made of VDR and 1,25(OH)2D3 modulates transcription of vitamin D responsive genes12 and influences adipocyte differentiation13. The effect of VDR gene polymorphism can potentially be sex-specific as shown in at least one previous study with adiposity phenotypes14.

Epidemiological studies have confirmed associations of VDR polymorphisms with adiposity and other metabolic disturbances6,14,15,16,17,18,19,20,21,22,23. However, studies specifically examining adiposity outcomes either had small sample sizes (<400), (e.g15,16,24.) or were restricted to one sex, (e.g.6,16.) but more importantly were all cross-sectional or case-control by design and none to date have examined these associations among African-American adults.

MEGALIN (aka low-density lipoprotein receptor-related protein-2 [LRP-2]), is the endocytic vitamin D-binding protein receptor which allows vitamin D entry into cells and whose expression is directly regulated by both vitamin D25) and vitamin A26. MEGALIN may influence obesity by mediating the transport of leptin through the blood-brain barrier and modulating its signaling of both leptin and thyroid hormones27 Collectively, leptin and thyroid hormones affect adiposity through energy metabolism regulation28. MEGALIN acting also as the receptor for sex-hormone binding globulin (SHBG), is involved in interactions between estrogen, vitamin D and intracellular calcium within adipocytes, leading to a potentially sex-specific effect of MEGALIN polymorphisms on various phenotypes of obesity, as indicated by findings from previous studies14,29

In this study conducted, we hypothesize that selected VDR and MEGALIN gene polymorphisms have sex-specific associations with several key metabolic disturbances in a longitudinal study of African-American urban adults.

Subjects and Methods

Database

The Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) study is a prospective cohort study, initiated in 2004. It recruited an area probability sample of African Americans and whites residing in 13 neighborhoods of Baltimore, Maryland and aged 30–64 years at baseline. In the baseline visit (visit 1: 2004–2009), screening, followed by recruitment and household interviews were completed during phase 1, while phase 2 consisted of in-depth examinations in a mobile Medical Research Vehicles (MRV)30. Phase 1 of visit 1 included a general household questionnaire and 1 24 hr dietary recall, while phase 2 of that visit collected more in-depth psychosocial data, anthropometric, physiologic and body composition measurements, a fasting blood draw, as well as a second 24 hr dietary recall. The first follow-up visit (visit 2), initiated in 2009, collected similar data as in phase 2 of visit 1 through 2013, with few variations and followed a similar protocol. In both visits, participants provided informed consent form after reviewing a protocol booklet and a video that explained study procedures including future contact efforts. The National Institute on Environmental Health Sciences Institutional Review Board of the National Institutes of Health approved the HANDLS protocol and all methods were performed in accordance with the relevant guidelines and regulations. Participants are remunerated. In this study, we analyzed longitudinal HANDLS data from initial and first follow-up examinations among a sample of African-Americans participating in the HANDLS study, who had complete genetic data. Time elapsed between examination visit 1 (Wave 1:2004–2009) and visit 2 (also known as Wave 3:2009–201331), ranged between <1 y and ~8 y, with a mean of 4.64 ± 0.93 y.

Study subjects

Of the 3,720 baseline participants (mean ± SD age(y) of 48.3 ± 9.4, 45.3% men, and 59.1% African-American), data on genetic polymorphisms were complete for 1,024 participants self-reporting to be African American. However, missing data on covariates reduced our sample to n = 769, and further exclusions resulted in a sample size range of 574 to 598, with 539 participants having complete data on relevant baseline and follow-up outcome measurements (cross-sectional part of the analysis). In the longitudinal part of our analyses, participants who were initially free from metabolic disturbances were selected for each outcome. Their sample sizes ranged from n = 246 (central obesity-free) to n = 466 (hyperglycemia-free) and those who were MetS-free consisted of n = 294 baseline participants (Fig. 1).

Figure 1
figure 1

Participant Flow Chart.

Full size image

Anthropometric measures and metabolic outcome variables

BMI, measured as weight/height2, kg/m2 was computed for each participant based on measured weight and height. Furthermore, WC (in cm.) was measured using a tape that was wrapped around the waist near the navel, starting from the hip bone. Systolic and diastolic blood pressures (SBP and DBP) were measured by averaging 1 right and 1 left sitting non-invasive assessments using brachial artery auscultation using a stethoscope, an aneroid manometer, and an inflatable cuff. After an overnight fast (8–12 hours), a blood draw was taken from an antecubital vein. From this blood draw, fasting glucose (FG), triacylglycerols (TAG), total cholesterol, and HDL-C were assessed using a spectrophotometer (Olympus 5400; Quest Diagnostics).

Classification of health outcomes

General obesity was defined as BMI ≥30 kg/m2, while central obesity (aka abdominal obesity) was based on WC ≥102 cm or 40 inches in men and ≥88 cm or 35 inches in women32. Participants who screened positive on at least 3 of 5 conditions ((1) central obesity (see above); (2) blood pressure ≥130/85 mmHg; (3) dyslipidemia: TAG ≥1.695 mmol/L (150 mg/dl); (4) dyslipidemia: HDL-C < 40 mg/dL in men or <50 mg/dL in women; (5) fasting plasma glucose ≥6.1 mmol/L (110 mg/dl)33.) were classified as MetS-positive1 We examined binary prevalent (V1 and V2) and incident outcomes, namely obesity, central obesity, MetS and its remaining individual components (i.e. hypertension, dyslipidemia-TA, dyslipidemia-HDL and hyperglycemia).

Vitamin D receptor and MEGALIN (LRP2) SNP and SNPHAP

Study participants were genotyped to 907,763 single nucleotide polymorphisms (SNPs) using the Illumina 1 M and 1M-Duo genotyping arrays. Details regarding genotype quality control criteria are provided in Supplemental Methods 1.

For the present study, in the main analysis, we selected VDR and MEGALIN SNPs based on previously published validation studies of adiposity or various health outcomes that were linked to adiposity6,15,16,17,18 and a replication study of similar outcomes among European ancestry participants from the Baltimore Longitudinal Study of Aging (BLSA)14. Three VDR SNPs:; rs1544410 (BsmI: G/A); rs7975232 (ApAI: A/C) and rs731236 (TaqI: G/A), and three MEGALIN SNPs (rs3755166: G/A; rs2075252: C/T; rs2228171: C/T) were chosen for haplotype analysis. The final selected SNPs and their frequencies were published elsewhere34.

VDR and MEGALIN SNPs haplotypes (SNPHAP) were considered main predictors in our analysis. For VDR gene, the BsmI, ApaI and TaqI SNP were combined together to construct SNPHAP, as was done in a previous study34, and their haplotype frequencies in the population were comparable to at least one other study conducted among Whites35. Four SNPHAP were detected in our sample with the SNP combinations being either one of the three: VDR1: GCA [baT], VDR2: AAG [BAt], VDR3: GAA [bAT] and VDR4: AAA [BAT] for one or two alleles. Participants were coded as 0 = having no VDRx haplotype; 1 = having one allele carrying the VDR x haplotype; 2 = having two alleles with the VDRx haplotype. This approach was similarly applied to the three MEGALIN SNP and eight haplotypes were found. However, only two haplotypes were extracted in the present analyses, given that their frequency was greater than 10%. The most common SNPHAP are comparable to our previous study34. Detailed descriptions the SNPHAP are found in Table 1. Furthermore, all available SNPs in and around the VDR and MEGALIN genes were also selected for a supplemental analysis, after passing through eligibility criteria related to reliability of imputation and minor allele frequency. Details on filtering of SNPs is further discussed in Supplemental Method 1–2. Outcomes of interest were MetS (incident, visits 1 and 2).

Table 1 Findings from haplotype analysis: definitions and distributions of SNPHAP for the selected VDR and LRP2 (MEGALIN) SNPs, n = 1,0241.
Full size table

Covariates

Our analyses included the following covariates: baseline age, sex, poverty status, education, smoking, drug use and self-rated health, among fixed or baseline variables. The Healthy Eating Index (HEI-2010) total score, computed using two 24-hr recalls administered at the initial visit, reflected overall dietary quality (see http://appliedresearch.cancer.gov/tools/hei/tools.html) or http://handls.nih.gov/06Coll-dataDoc.htm) and was included in our analyses. Similarly, total energy intake (kcal/d) was included in our models as a potential confounder based on the average of initial 2 24-hr dietary recalls. Finally, 10 principal components were included in order to control for any residual effects of population structure (Supplemental method 1). Covariates were selected based on their known association with the metabolic outcomes of interest. Due to the limited sample size, a sensitivity analysis was conducted for parts of the analysis adjusting only for basic socio-demographic factors, namely age, sex, poverty status and education, as well as the inverse mills ratio.

Statistical analysis

The main part of the analysis was conducted using Stata release 15.036. For each SNP, the Hardy-Weinberg equilibrium assumption was tested using exact test, and pair-wise linkage disequilibrium (LD) was computed and visualized using Haploview version 4.2 package37. To describe study participant characteristics and compare them by sex, t-test and χ2 test were used for continuous and categorical variables, respectively.

Both cross-sectional and longitudinal relationships of VDR and MEGALIN SNPHAP with binary metabolic outcomes, including obesity, central obesity and the MetS were examined. To test cross-sectional associations, multi-variable logistic regression models were conducted for each outcome, controlling for baseline age, sex, poverty status, education, first-visit current smoking and drug use, self-reported health and the HEI-2010 total score, the 10 principal components to adjust for population structure and the inverse mills ratio. For longitudinal analyses, we defined time-to-event from baseline visit (i.e. delayed entry) until outcome or censoring at second visit and constructed multiple Cox proportional hazards models for incident metabolic outcome, overall and after stratifying by sex. Follow-up time was expressed in years. In addition to examining obesity (BMI and WC-based) and MetS as incident outcomes, other components of the MetS were also evaluated as individual incident metabolic outcomes. Linear trend test for associations between haplotype dosage (0, 1, 2 copies) and metabolic outcomes was performed.

Furthermore, selection bias due to the non-random selection of participants with genetic data from target population, was corrected at least in part using a 2-stage Heckman selection model38. At first stage, probit models were constructed to calculate an inverse mills ratio, a function of the predicted selection probability, conditional on key covariates, as previously described39. At a second stage, the inverse mills ratio was entered into the multi-variable logistic or Cox PH models, thus adjusting for selection bias. Stratification was done and effect modification was tested (by adding 2-way interaction terms) by sex for all analyses, including supplemental analyses for single SNPs in and around the VDR and MEGALIN genes. Gender difference in the relationship between MEGALIN gene polymorphisms and metabolic outcomes was an a priori hypothesis40.

Finally, in all our analyses, type I error was set 0.05 prior to correction for multiple testing. A p-value <0.10 was considered as marginally significant. Correction for multiple testing was conducted using a familywise Bonferroni process in which a family was defined by the metabolic outcome and the gene41. Thus, the critical p-value was reduced to 0.05/4 = 0.0125 in the case VDR SNPHAP associations, whereas for MEGALIN SNPHAPs it was reduced to 0.05/2 = 0.025. Correction for multiple testing followed a similar though less stringent approach, whereby a critical p-value per outcome of interest was reduced to 0.01 for overall analysis and 0.02 for sex-specific analysis. For 2-way interaction terms, particularly for testing effect modification by gender, type I error was kept at 0.05 due to reduced statistical power42.

Results

All examined SNPs exhibited Hardy-Weinberg equilibrium (P > 0.002). Variants within each VDR and MEGALIN (LRP2) gene were deemed in low linkage equilibrium (r2 < 0.30). The four selected VDR haplotypes had an overall prevalence ranging from 10.1% for VDR4 to 36.5% for VDR1. A large proportion of African-Americans (65.8%) had 1 copy of VDR3; only 1% had 2 copies of VDR4. Similarly, among the selected MEGALIN haplotypes, MEGALIN1: GCC was the most common (53.3%), with 63.5% having 1 copy and only 4.3% having 2 copies of MEGALIN2: ACC (Table 1).

Table 2 presents baseline characteristics and time-dependent metabolic outcomes (Fig. 1: n2h = 539). Most notably men had higher prevalence of smoking and drug use compared to women as well as higher energy intake, poorer overall dietary quality (HEI-2010 total score), and higher mean fasting blood glucose. In contrast, women had higher mean BMI, WC, HDL-C compared to men, at both waves. Among incident outcomes, central obesity was markedly higher among women compared to men, with no difference noted for incident obesity or incident MetS. Nevertheless, in the cross-sectional data, obesity and central obesity were both significantly more prevalent among women compared to men at both waves, while MetS prevalence was higher among women only at follow-up.

Table 2 Gender differences in baseline characteristics and time-dependent metabolic outcomes among African-Americans with complete genetic, time-dependent metabolic data and baseline covariate data: HANDLS 2004–2009 and 2009–20131.
Full size table

Table 3 shows associations of VDR and MEGALIN haplotypes with incident metabolic disturbances (obesity, MetS, and individual MetS components), stratifying by sex. Among all key results, only survived correction for multiple testing. In fact, among men, incident hypertension was inversely related to the MEGALIN1 haplotype (HR = 0.45, 95% CI:0.23–0.90, p = 0.024). Though not surviving correction for multiple testing, this haplotype was also inversely related to incident MetS among men. Similarly, incident hyperglycemia was linked to VDR1 haplotype in men. Among women, VDR3 was inversely related to incident central obesity, and VDR2 was directly associated with incident hypertension. MEGALIN2 was consistently inversely related to incident hypertension and incident hyperglycemia among women. The latter association differed significantly between sexes.

Table 3 VDR and MEGALIN SNP haplotype (SNPHAP) associations with incident metabolic disturbances: Cox proportional hazards models, (n = 246–466); HANDLS study.
Full size table

Cross-sectional associations between the selected VDR and MEGALIN haplotypes and the main metabolic disturbances are presented in Tables 4 (baseline outcomes) and 5 (follow-up outcomes). There was a linear dose-response direct association between VDR2 and prevalent obesity and MetS at baseline, with no significant sex differences, (OR = 1.60, 95% CI:1.11–2.31, p = 0.012). Moreover, VDR3 was inversely related to prevalent MetS at baseline among men, (OR = 0.40, 95% CI:0.19–0.81, p = 0.011), an association that differed significantly by sex (P < 0.05 for sex × SNPHAP interaction term). Both of these findings survived correction for multiple testing. The associations of VDR and MEGALIN SNPHAPs with follow-up outcomes did not survive correction for multiple testing. Among those, VDR2 was positively associated with prevalent obesity, overall and among women, a finding consistent with the baseline outcome. MEGALIN haplotypes were not associated with prevalent baseline or follow-up outcomes of obesity, central obesity and MetS. A sensitivity analysis that included only basic socio-demographic factors yielded similar results.

Table 4 VDR and MEGALIN SNP haplotype (SNPHAP) associations with prevalent metabolic disturbances (V1): multiple logistic regression models, (n = 539); HANDLS study.
Full size table
Table 5 VDR and MEGALIN SNP haplotype (SNPHAP) associations with prevalent metabolic disturbances (V2): multiple logistic regression models, (n = 539); HANDLS study.
Full size table

Table S2 presents supplemental results for single SNP analyses for VDR and MEGALIN in relation to MetS outcomes, stratifying by gender. For incident MetS, overall, 15 SNP passed correction for multiple testing, of which one had a p = 0.001 (MEGALIN SNP: rs148386284: T allele, HR = 2.63, 95% CI:1.45–4.76). While many other SNPs were also associated with incident MetS, the ones that had p ≤ 0.01 were among women, including MEGALIN SNP rs830966: G allele, HR = 4.28, 95% CI:1.80–10.20) and 6 protective VDR SNPs located near a well-studied SNP names Cdx-I, rs11568250, whose C allele was also inversely related to incident MetS among women (HR = 0.11, 95%CI: 0.02–0.47, p = 0.003). Similarly, both VDR and MEGALIN single SNPs had significant associations with baseline and follow-up MetS outcomes, overall and among men and women, with few SNPs overlapping or being highly correlated with those that affected the incident outcome (e.g. incident MetS vs. follow-up MetS in women: rs830966-rs830969, rs2107301). Moreover, among the selected SNPs for VDR haplotypes, rs731236 (TaqI G > A) was associated with a reduced odds of baseline MetS (OR = 0.69, 95% CI: 0.50, 0.96, p = 0.029). Similarly, rs1544410 (BsmI G > A) was linked to an increased odds of baseline MetS among men (OR = 1.73, 95% CI: 1.01, 3.00, p = 0.045).

Discussion

This study examined associations of selected haplotypes from SNPs in four VDR [rs1544410(BsmI:G/A); rs7975232(ApaI:A/C); rs731236(TaqI:G/A)], and two MEGALIN [rs3755166:G/A; rs2075252:C/T; rs2228171:C/T] gene haplotypes with longitudinal (n = 294–466) and cross-sectional (n = 539) metabolic outcomes among African Americans over a mean period of ~5 y of follow-up. Among key findings, in men, incident hypertension and MetS were inversely related to MEGALIN1 (GCC), and in women, MEGALIN2 (ACC) was consistently inversely related to incident hypertension and incident hyperglycemia. Among men, incident hyperglycemia was positively associated with VDR1 (GCA:baT), and among women, VDR2 (AAG:BAt) was directly associated with incident hypertension and VDR3 (GAA:Bat) was inversely related to incident central obesity. Overall, there was a direct, linear dose-response association between VDR2 (AAG:BAt), obesity and MetS at baseline. Moreover, VDR3 (GAA:bAT) was inversely related to baseline MetS among men, which differed significantly by sex. No associations were detected between MEGALIN haplotypes and outcomes at baseline or follow-up and 3 associations survived correction for multiple testing [VDR2 vs. baseline MetS (overall), VDR3 vs. baseline MetS (men) and MEGALIN1 vs. incident hypertension in men].

Recent studies with cross-sectional or case-control design have examined VDR polymorphisms as risk markers for central adiposity and related metabolic disorders6,14,15,16,17,18,19,20,21,22,23. However, none of these studies included African-American adults in their samples. When examining VDR SNP relationships with adiposity, a recent cross-sectional study (176 randomly selected men aged 25–65 y) found that homozygous BsmI (BB: AA vs. GG) was associated with higher BMI (29.0 vs. 26.8 kg/m2, p = 0.024) and higher WC (101.8 vs. 96.2 cm, p = 0.014)15. A similar finding was observed in another cross-sectional study of 153 women among whom body weight and fat mass were positively associated with the “BB” genotype of VDR SNP BsmI16. Similarly, in a third cross-sectional study, an association was found between a homozygous rare variant of rs3782905 located in the 3‘ VDR region (LD of rs3782905 with BsmI in White Hapmap is ~0.42) and 4.4 cm larger mean WC when compared with the homozygous common variant (Bonferroni-adjusted p = 0.004)6. These consistent findings for BsmI “A” allele dosage increasing the risk for obesity, were replicated with other related phenotypes, including T2D, fasting glucose level, LDL-Cholesterol and coronary heart disease risk in recent studies17,18,19,20. Using data from Baltimore Longitudinal Study of Aging, a previous study found that only the ApaI SNP (“C” allele dosage) significantly increased the odds of higher waist-to-hip ratio over time (P-trend = 0.024)14. These findings are consistent with ours, given that the “B” allele of BsmI corresponds to the “A” risk allele. In fact, findings from our present study indicated a positive association between MetS and the BAt VDR haplotype (i.e. VDR2) in the overall population and an inverse relationship with the bAT (i.e. VDR3) haplotype among men. Many other key results not surviving correction for multiple testing were generally trending in that same direction for various metabolic outcomes. Moreover, MEGALIN polymorphisms influenced central adiposity in Whites residing in Baltimore city: rs2075252 “TT” was associated with elevated waist to hip ratio at one point in time compared with rs2075252 “CC”14. Our finding with MEGALIN1 (GCC for rs3755166:G/A; rs2075252:C/T and rs2228171:C/T) being inversely related to incident hypertension indicates that in fact, a “C” allele for the middle SNP (rs2075252:C/T) may be protective against various obesity-related metabolic disturbance, particularly hypertension incidence rate. However, more studies are needed in diverse samples to replicate this finding.

Few previous studies have examined VDR SNPHAP as predictors of metabolic and cardiovascular outcomes. Despite earlier null findings, (e.g21.) more recent studies indicated that in fact BAt (VDR2) was associated with increased obesity risk, while the haplotype “GAG”, which was rare among our African-American urban adult population, was associated with a reduced risk of obesity22. Moreover, in a population-based study of men and women aged 55–80 y, each baT haplotype copy was associated with a 20% increased likelihood of ECG-confirmed myocardial infarction, adjusting for key confounders23. Similarly, in the BLSA study, an increased risk of longitudinal increase in WC among White women was uncovered with each baT haplotype copy14. In our present study, incident hyperglycemia risk was positively associated with baT, particularly among African-American men. Nevertheless, BAt, a less common haplotype in this population, was associated with a greater risk of incident hypertension in African-American women and with a greater likelihood of prevalent obesity and baseline MetS in the total African-American urban population. The latter finding (i.e. VDR2 vs. baseline MetS) was the only one that survived correction for multiple testing in the total population.

In terms of biological mechanisms, a greater VDR expression in adipocytes decreases energy expenditure markedly leading to increased adiposity. Further, VDR agonists reduce pro-inflammatory cytokines and D3 reduces high-glucose and LPS-induced TNFα and TGFβ release, suggesting a protective mechanism22. Moreover, high BMI is associated with low circulating 25(OH)D due to sequestration in adipose tissue43. Importantly, longer VDR BsmI polyA repeats exhibited less stability and translated less efficiently into VDR protein, resulting in a decreased vitamin D response, muscle cell inhibition and adipocyte differentiation16. In addition to the association of calcium and high Vitamin D intakes, vitamin D may also reduce hepatic synthesis of triglycerides and upregulate adiponectin expression, which in turn could reduce obesity and related metabolic disorders44. In the African-American population, VDR can trigger adiposity by modulating VDR-dependent molecular components of adipogenesis such as PPAR-γ and EBPα and thus inhibiting corresponding adipocyte differentiation45. Further, VDR variants may directly influence the binding of vitamin D and mediate various downstream effects on genes known to be VDR-responsive, thus influencing associated phenotypes6. Particularly for T2D, VDR may act as a transcription factor for \(\beta \) cell insulin secretion regulation, thereby affecting lipid metabolism46.

Our study has several strengths, which include its longitudinal follow-up design, a large sample of a diverse urban population, and the extensive use of advanced statistical methods, through the combination of survival analysis, logistic regression, haplotype analysis and adjustment for selection biases.

Nevertheless, our study has notable limitations. First, our final analytic sample was likely selected in a non-random manner, whereby certain groups were over-sampled when compared with the original African-American sample in HANDLS. A 2-stage Heckman selection model was used to diminish those incurred biases38. Second, first-vist age and between-visit duration varied across participants, which may incur some imbalance in the data structure. Survival analysis methods were used to adjust for this imbalance. Moreover, our study had limited power to examine gene-environment interaction, particularly with 25(OH)D in serum or dietary intakes of vitamin D. Finally, positive results may have been chance findings, while negative findings may have been caused by lack of adequate power.

In conclusion, VDR and MEGALIN gene variations can affect the prevalence of MetS and the incidence of hypertension in a sex-specific manner, respectively, among African-American urban adults. Those study findings provide novel insights into the genetic variants at those gene loci and their association with susceptibility to cardiometabolic risk, including the metabolic syndrome among populations of African descent. Further functional studies of VDR and MEGALIN gene in relation to cardiometabolic risk can provide important validation for our results and can contribute to our understanding of how vitamin D metabolism-related genes can influence metabolic disorders in various populations. In addition, our findings if replicated by others can establish the need for a genetic screening test for VDR and MEGALIN polymorphisms. Thus, further large epidemiologic studies of similar populations are required to replicate our current findings.