Study Subject
For this study, we used cross-sectional data from the Jackson Heart Study (JHS). The JHS is a single-site community-based cohort study of risk factors for cardiovascular disease among adult African American men and women living in Jackson, Mississippi, Metropolitan area. The study participants consisted of 5,306 individuals recruited, interviewed, and examined by certified technicians in the first exam (2000-2004) (18,19) and followed up in 2 subsequent exams from 2005 to 2008 and 2009 to 2013. The clinic visits encompassed physical examination, blood and urine collection, anthropometry, and collection of data regarding family history, behavioral risk factors, and sociodemographics. There were 3027 participants (Mean age of 55.13±12.84 and 1871 women) who consented to genetic analysis, and their DNA samples were genotyped in the candidate gene association resource (CARe) consortium using the Affymetrix 6.0 platform and later imputed to 1000 genomes phase 1 (20–22). The study was approved by the Institutional Review Board (IRB) of the National Institutes of Health. The IRB approved the protocol of participating institutions (University of Mississippi Medical Center, Jackson State University, and Tougaloo College.
Study Variables
Outcome Variables
Our primary outcome in this study is adiposity (BMI and waist circumference). Adiposity was measured in exam visit 1, and we defined it as a BMI greater than 30 kg/m2 and a waist circumference greater than 102 cm and 88 cm for men and women, respectively. All the above clinical parameters were measured according to standard laboratory and clinical techniques (19).
Independent Variables: SNP Selection Genotyping and Imputation
All 3027 JHS samples were genotyped on the Affymetrix 6.0 based on manufacturer protocol (22). The candidate gene approach was used to select our genetic variants from the entire set of common genetic variants in the MTNR1B gene located on chromosome 11q14.3 and hg19 position base-pair ordinates chr11:92,702,789-92,718,2 (plus-strand orientation). The JHS coordination centers performed the SNPs quality control, and the variants that passed were imputed with 1000G phase 1 using Cosmopolitan reference panel including all races—version 2010-11 data freeze, 2012-03-04 haplotypes (21,22). The imputation was completed using Minimac3 on the Michigan Imputation Server (23); details regarding the reference panel can be found in the 1000 Genomes Project Consortium 2010 (24). Imputed SNPs were filtered for minor allele frequency ≥1%, call rate ≥ 90%, HWE p-value > 10-6, as well as the exclusion of sites with invalid or mismatched alleles for the reference panel (21). For this study, 109 SNPs were genotyped and imputed; we focused on common variants with a minor allele frequency (MAF) ≥ 5%, and with imputation quality ≥ 80%, 30 common variants were selected for downstream analyses. Covariates were age, gender, and 10 principal components to adjust for population stratification due to admixture (17). In additional analyses, we also adjusted for insomnia covariate. The participants were asked if they have insomnia with the answer option of “Yes,” “No,” and “Don’t Know.” Insomnia is clinically defined as the difficulty of falling and staying asleep (25).
Statistical Analysis
Descriptive Statistics:
Study variables were summarized using the mean and standard deviation (SD) for continuous variables and proportions for categorical variables. Continuous variables were first assessed for normality, and then log-transformed if not normally distributed. Analyses for descriptive statistics were performed using statistical software SAS 9.4 (26).
Regression Analysis:
Multivariate logistic regression models were fitted to assess the associations between the dosage of MTNR1B genetic variants and adiposity adjusted for age, gender, and 10 principal components in the adiposity model. Due to the relationship between melatonin signaling and sleep, we examined the modifying effect of insomnia (adiposity-insomnia model) on our adiposity outcome by stratifying by each modifier. Multivariate linear regression models (BMI model and waist circumference model) were fitted to examine the relationships between MTNR1B genetic variants and continuous obesity outcome traits of BMI and waist circumference. The linear and logistic regression models were fitted using ProbAbel v.0.5.0 genetic analysis software (27), assuming population-based design. Although a small subset of JHS participants belongs to a family component, we didn’t adjust for family structure because previous studies have shown minimal impact on power and the inflation of the type I error (28–31). We used false discovery rate (FDR) to correct for multiple testing with an adjusted p-value threshold of 0.05. The NIH dbSNP database was used to annotate the function of the MTNR1B variants that displayed a significant association in the regression models (32).
The variants that were statistically significant after FDR adjustment were used to generate a linkage disequilibrium (LD) plot. The Haploview (Broad Institute, MA, USA) was used to create the LD plot, and we used the Yoruba, Nigeria population as a reference (33). Haploview generated haplotype blocks in the LD plot whenever 95% of informative comparisons were in strong linkage disequilibrium (LD) while ignoring variants with MAF < 0.05 (34).