|Home | About | Journals | Submit | Contact Us | Français|
Adipose (adp) is an obesity gene in Drosophila and mice with crucial functions in fat metabolism. We investigated the correlation between genetic variation of the WDTC1 locus, the ortholog of adp, and human obesity. Five WDTC1 single nucleotide polymorphisms (SNPs) were genotyped in 935 and 1115 adults of two ethnically diverse US populations. In the Boston Puerto Rican population, we demonstrated that two WDTC1 SNPs strongly associated with obesity. Homozygote and heterozygote carriers of the major allele i22835A, representing about 96% of the population, had significantly higher mean BMI (31.5 and 31.0 kg/m2, respectively) than non-carriers (28.6 kg/m2). Conversely, homozygotes of the minor allele i22835G were leaner and were 74% less likely to be overweight or obese (OR=0.26, P=0.003) compared to homozygote carriers of the major allele. Haplotype analyses based on two SNPs further supported these findings. In addition, we found a strong interaction of monounsaturated fatty acid (MUFA) intake by genotype in this population. As dietary MUFA intake increased, minor allele carriers of SNPs i22835A>G had higher BMIs, whereas major allele carriers had lower BMIs. A White population also exhibited a pattern of association between WDTC1 genotypes and obesity although of a different nature. Those WDTC1 variants associated with obesity likely have experienced strong positive selection in human history, when food supply was unpredictable. Given the high frequency of the major alleles in both populations, we suggest that WDTC1 variation may be an important risk factor contributing to obesity in these populations.
Obesity is a global health problem associated with increased risk of type 2 diabetes and coronary heart disease. This affliction is controlled by multiple genetic factors and complex interactions between genetic and environmental factors. The genetic component has been extensively investigated through quantitative trait loci mapping, candidate gene approach (1) and genome-wide association studies (2). In addition, model systems, such as Drosophila and C. elegans, are well founded as powerful tools to identify genes associated with human diseases and to gain understanding of their biological functions and molecular mechanisms (2-5). Drosophila often serves as a model for human diseases and it was in this capacity that adp was identified as an obesity gene in Drosophila (6). Its product, Adp, containing six WD40 protein-protein interaction domains and three tetratricopeptide repeats, is predicted to be a key player in fat metabolism (6, 7). adp mutants are obese, starvation-resistant, and less active (6, 7). In the Drosophila model, adp is primarily expressed in the body fat (6, 7). Null flies have increased triglyceride storage in the body fat, whereas transgenic over-expressors of adp exhibit reduced fat storage. Similarly, heterozygous adp knockout mice displayed obesity and insulin resistant phenotypes resembling those of the null flies, while transgenic mice over-expressing adp in fat pads are lean and display wild type metabolic phenotypes (6, 7). The human ortholog of adp protein, WD and tetratricopeptide repeats 1 (encoded by WDTC1), was identified as a conserved and single copy gene in humans (6). Because the role of WDTC1 in human obesity has yet to be demonstrated, we investigated in this study the association between WDTC1 genetic variation and obesity in two US ethnic diverse populations: a Puerto Rican Hispanic immigrant population living in the Boston area (8,9) and a North American White population living in the Minneapolis and Salt Lake City areas (10). Although Puerto Rican Hispanics have been identified as a vulnerable group at increased risk for age-related chronic diseases (8, 9), both populations have high prevalence of obesity, underlying the importance of investigating the genetic basis for obesity in both populations.
This study sample was comprised of 264 men and 671 women who were self-identified Puerto Ricans living in the greater Boston metropolitan area and for whom full data records for demographics, biochemical characteristics and genotypes were collected. These subjects were recruited by investigators from the Boston Puerto Rican Center for Population Health and Health Disparities to participate in a longitudinal cohort study on stress, nutrition, health and aging--the Boston Puerto Rican Health Study (8), (http://hnrcwww.hnrc.tufts.edu/departments/labs/prchd/). The detailed description of the population was reported previously (11). Written informed consent was obtained from each participant and the protocol was approved by the Institutional Review Board at Tufts University.
This study sample comprises 536 men and 579 women who participated in the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) and for whom full data records for anthropometric measurements and genotype data exist. Detailed design and methodology for the GOLDN study have been described previously (10). Written informed consent was obtained from each participant. The protocol was approved by the Institutional Review Boards at the University of Alabama at Birmingham, the University of Minnesota, the University of Utah, and Tufts University.
Anthropometric measurements were collected using standard methods. Fasting blood samples were drawn by a certified phlebotomist. Aliquots were saved and stored at -80°C until processed. Using the American Diabetes Association (ADA) criteria, subjects were classified as having type 2 diabetes when fasting plasma glucose concentration was ≥126 mg/dl or use of insulin or diabetes medication was reported (12). Overweight (BMI≥25) and obesity (BMI≥30) were classified based on international standards (13). Abdominal obesity was defined as a condition in which a subject has a waist circumference ≥102 cm in men, ≥88 cm in women (14). Physical activity was estimated as a physical activity score based on the Paffenbarger questionnaire of the Harvard Alumni Activity Survey (15).
For the GOLDN population, dietary intake was estimated using the Dietary History Questionnaire (DHQ), a cognitively-based food frequency questionnaire, developed by the National Cancer Institute (available online at http://riskfactor.cancer.gov/DHQ/). For the BPRHS population, the food frequency questionnaire was developed specifically for this population and has been validated (16). The food list for the FFQ was developed using the format of the National Cancer Institute/Block food frequency, but with data from the HHANES dietary recalls for Puerto Rican adults and tested in Puerto Rican subjects, aged 60 years and older, in Massachusetts. Because the Puerto Rican population has a typical diet and portion size that differs considerably from the general US population, we expanded the food groups and portion sizes. Comparison of our instrument with the Block FFQ showed that the our questionnaire captured the intakes reported in 24 hour recalls more accurately for total nutrients and in ranking of individuals (16). In both studies, nutrient intake profiles are calculated using the USDA National Nutrient Database for Standard Reference (17). Intake of total fat, saturated fatty acids (SFA), monounsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA), including n-3 and n-6 PUFA, were presented as percentage of total daily energy intake.
The WDTC1 gene maps to 1p36.1 on chromosome 1, and encompasses about 73.6 kbp. Because there was no known report of WDTC1 genotype-phenotype association, we selected SNPs for genotyping based primarily on identification of SNPs that served as proxies for major haplotype blocks using HapMap genotypes (www.hapmap.org) known in White (European-American) and African populations. Seven SNPs were selected for initial genotyping but two SNPs (rs3813791 and rs4469729) with a minor allele frequency lower than 0.05 were not further analyzed. Thus, only five SNPs, rs11247626, rs4460661, rs11589265, rs3813790, and rs4970518 (Table 1) were genotyped in both populations, with HUGO names given based on nomenclature recommendations of the Human Genome Variation Society (http://www.hgvs.org/rec.html).
Genomic DNA was isolated from buffy coats of peripheral blood using QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) according to the vender's recommended protocol. SNPs were genotyped with Applied Biosystems TaqMan SNP genotyping system (18).
Statistical analyses were performed using SAS 9.1. (Cary, NC, USA). We assessed the relationship between WDTC1 variants and obesity-related traits (BMI, overweight, and obesity) by covariance analysis. For dichotomous characters, such as overweight, obesity, and abdominal obesity, we employed logistic regression. In these analyses, the dependent variables were BMI, overweight or obesity status, or abdominal obesity. Genotypes of the individual WDTC1 SNPs served as independent variables. Analyses were adjusted for potential confounders (age, sex, smoking, alcohol intake, physical activity, population admixture (see below), and medication for hypertension and dyslipidemia) using a linear or logistic regression model. Men and women were analyzed together, as well as separately to examine sex specific effects. A P-value ≤0.05 was considered statistically significant. When examining genotype by dietary interaction, we categorized subjects into two subgroups based on the population mean of the dietary intake.
Pair-wise linkage disequilibria (LD) among SNPs were estimated as correlation coefficients (i.e., r2) using the HelixTree program (Golden Helix, Bozeman, MT). In the GOLDN population, LD was estimated based on unrelated subjects only. For haplotype analysis, we estimated haplotype frequencies using the Expectation-Maximization (EM) algorithm (19). To determine the association between haplotypes and phenotypes, we used haplotype trend regression analysis with the option of composite haplotype estimation implemented in HelixTree (20, 21). P-values were further adjusted for multiple tests by a permutation test. In the GOLDN population, the haplotype trend regression analysis was conducted for overweight and obesity using GENMOD method in SAS and haplotypes estimated by HelixTree using the Expectation-Maximization (EM) algorithm while adjusting for family relationship.
In the BPRHS participants, population admixture was calculated using STRUCTURE 2.2, an extension of an earlier method, based on Bayesian clustering, using a Markov chain Monte Carlo (MCMC) algorithm (22). This new method enables estimation of population admixture using linked markers. We estimated population admixture based on 96 SNPs genotyped in this population, with minor allele frequencies greater than 0.05 or less than 0.95, representing 35 genes located on 15 different chromosomes (11).
In the BPRHS population (Table 2), the total energy intake, total fat intake, and the percentages of individuals who reported smoking or drinking alcohol, were significantly higher in men than in women. In contrast, the carbohydrate energy intake (% of total energy), the mean BMI, the percentages of participants who were overweight (BMI≥25) or obese (BMI≥30) were significantly higher in women than in men. Other demographic characteristics did not differ significantly by sex. In the GOLDN population, the total energy intake and the percentage of participants with cardiovascular diseases or overweight were significantly higher in men than in women. Conversely, women had significantly higher carbohydrate intake (% of total energy) than men. A comparison of these two populations finds that the mean age of the BPRHS was about eight years older than that of GOLDN for men and women combined. The rates of obesity, T2DM, CVD, the percentage of smokers, and carbohydrate intake (% of total energy) were significantly higher in BPRHS than in GOLDN for men and women. On the other hand, the GOLDN population had significantly higher physical activity score than the BPRHS. While total fat intake was similar for men and women in both populations, GOLDN participants had significantly higher saturated fatty acid and MUFA intake than BPRHS subjects for both men and women.
Minor allele frequencies of all five SNPs at the WDTC1 locus, except m12502C>T, were significantly different (P<0.001) between the two populations (Table 1). Notably, minor allele frequencies of SNPs i22835A>G and i61970A>G were 0.21 and 0.29 in the BPRHS population, each of which is almost two-fold higher than those in the GOLDN (0.12 and 0.18, P<0.001 and P<0.001, respectively). No significant differences in allelic frequency were observed between men and women within each population (data not shown). All SNPs were in Hardy-Weinberg equilibrium in both populations, and SNP i22835A>G and i61970A>G were in strong linkage disequilibrium (r2=0.803 and 0.782, respectively in the BPRHS and GOLDN populations), whereas pair-wise LD measures for other SNPs were weak (r2 ≤ 0.4, data not shown).
In the BPRHS population, we found a significant association between SNP i22835A>G and BMI (Table 3, P=0.030). Carriers of the major allele i22835A had higher mean BMI (31.5 and 31.0 kg/m2 for AA and GA, respectively) than GG subjects (28.6 kg/m2). As expected the i61970A>G SNP, in significant LD with i22835A>G, showed a similar association with BMI. Carriers of the major allele i61970A showed higher BMI (31.8 and 31.0 kg/m2, for AA and AG, respectively) as compared with GG subjects (30.2 kg/m2); however, this association did not reach statistical significance (P=0.075). Conversely, the other three SNPs examined in this study showed no significant associations with BMI. Moreover, in the GOLDN population, there were no statistically significant associations detected between any of the WDTC1 variants and BMI.
To determine if other factors confound the association between WDTC1 variants and BMI in the GOLDN population, we examined the association according to gender and by combining minor allele carriers (i.e., GG+GA vs AA) in this population. Our analyses identified a significant sex by genotype interaction (P=0.016) on BMI kg/m2 for SNP i22835A>G (Fig. 1). Male carriers of the minor allele i22835G (GG + GA, n=112) exhibited a trend towards lower BMI than AA subjects (n=427), the same direction seen in BPRHS, although in GOLDN this correlation did not reach statistical significance (30.4 vs 31.1 kg/m2, P=0.160). Conversely, female carriers of the minor G allele at the i22835A>G SNP (n=135) had a higher BMI than AA homozygotes (n=446) (31.4 vs 30.3 kg/m2; P=0.054). A similar non-statistically significant association was observed for SNP i61970A>G. However, we did not observe a sex-genotype interaction in BPRHS. Since the BPRHS population has a higher frequency of T2DM, we also examined if T2DM interacts with genotypes influencing BMI. Again, we did not observe a T2DM by genotype interaction on BMI.
We next examined the risk of being overweight or obese in relation to WDTC1 genotypes (Table 4). In the BPRHS population, the minor allele carriers (GG and GA) of the i22835A>G SNP were 74% and 29% less likely to be overweight (BMI≥25 kg/m2) or obese (BMI≥30 kg/m2) (OR=0.26 and 0.71 for GG and GA, respectively, P=0.003) than non-carriers (AA). Consistently, these subjects presented a reduced risk of being obese (P=0.075 for BMI≥30 kg/m2, P=0.006 for abdominal obesity). Conversely, the major allele carriers of the i61970A>G SNP, which is in strong LD with the former, had increased risk of overweight (BMI≥25 kg/m2, P=0.011), obesity (BMI≥30 kg/m2, P=0.128), or abdominal obesity (P=0.015) when compared to non-carriers (GG).
In the GOLDN population, the risk of being overweight (BMI≥25 kg/m2) for the minor carriers of i22835A>G or i61970A>G did not reach statistical significance (Table 4, P=0.847, 0.167 for i22835A>G, and P=0.325, 0.657 for i61970A>G, respectively for men and women). However, female carriers of the i22835G minor allele have significant increased risk of obesity (BMI≥30 kg/m2, OR=1.54 P=0.044). Male carriers of the major allele i61970A showed increased risk of obesity (BMI≥30, OR=1.41), but not statistically significant (P=0.073). Such increased risk is more apparent for abdominal obesity (OR=1.72, P=0.007 for i22835A>G in women, OR=1.55 P=0.028 for i61970A>G in men).
To explore the combined effects of WDTC1 variants on the likelihood of being overweight or obese, we conducted haplotype analysis using two SNPs i22835A>G and i61970A>G. In the BPRHS population, four haplotypes A-A, A-G, G-G, and G-A, were identified with frequencies of 0.63, 0.16, 0.13, and 0.08, respectively. WDTC1 haplotypes showed strong association with being overweight or obese (BMI>25 kg/m2) at a global significance (P=0.024) after permutation correction. In particular, carriers of haplotype G-G were 48% less likely to be overweight or obese (OR=0.52, P=0.004), whereas carriers of the haplotype A-A (OR=1.62, P=0.005) were 62% more likely to be overweight or obese compared to non-carriers. A similar significant association (P=0.011 at a global significance after permutation correction) was found between WDTC1 haplotype and abdominal obesity. However, the association between WDTC1 haplotyes and obesity (BMI≥30 kg/m2) did not reach statistical significance at a global level (P=0.157) after permutation correction.
In the GOLDN study, the haplotype analysis was conducted separately for men and women because of genotype by sex interaction. Three major haplotypes: A-A, G-G, A-G were identified in this population with frequencies of 0.82, 0.11, and 0.06 respectively in men, 0.81, 0.13, and 0.06 respectively in women. In women (N=579), WDTC1 haplotypes were significantly associated with overweight (BMI≥25 kg/m2), obesity (BMI≥30 kg/m2), or abdominal obesity at a global level (P=0.046, 0.012, and 0.011, respectively) after permutation correction. In particular, female carriers of the G-G haplotype had significantly higher risk of being overweight (BMI≥25 kg/m2, OR=3.94, P=0.096), obese (BMI≥30 kg/m2, OR=6.44 P=0.012), and abdominal obesity (OR=6.44, P=0.011) than non-carriers, whereas such risk for the female carriers of the A-A haplotype did not reach statistical significance (data not shown). However, in men (N=536) no association between WDCT1 haplotypes and obesity-related traits (overweight, obesity, or abdominal obesity) reached a global significance (data not shown).
We examined whether dietary fat intake modulates the association between WDTC1 genotypes and BMI by categorizing subjects into two subgroups according to the population mean of dietary fat intake expressed as the percentage of total energy. In the BPRHS population, we found that dietary MUFA intake displayed a strong interaction (P=0.012) with i22835A>G genotypes influencing BMI. As depicted in Figure 2, when MUFA intake as a continuous variable was plotted against the predicted BMI, the carriers (GG+GA) of the minor allele i22835G exhibited increased BMI, whereas the non-carriers (AA) had decreased BMI, as MUFA intake was increasing. However, this interaction was not observed for intakes of saturated fatty acid, polyunsaturated fatty acid, or total fatty acids (data not shown). In the GOLDN population, no significant interaction between fatty acid and genotype was found either with men and women separately, or both combined (data not shown).
adp, first described as a Drosophila obesity gene (6), was recently shown to be conserved in the mouse and to function similarly in fat storage (7). Thus, WDTC1, the ortholog of adp, was postulated to function as a “skinny” gene in humans. Our report provides the first supporting evidence for this hypothesis and demonstrates that WDTC1 variants are associated with obesity in two US adult populations of diverse ethnicity. This observation was further confirmed by risk and haplotype analyses. Homozygous subjects of the minor G allele at the i22835A>G SNP, representing about 4% of the BPRHS population, were indeed 74% less likely to be overweight or obese compared to homozygous subjects (AA). In contrast, carriers of the major allele i22835A at this SNP, accounting for about 96% of the population, were more likely to be overweight or obese (BMI>25). Furthermore, WDTC1 variants showed consistent influence on abdominal obesity in both populations. These data constitute the first report whereby an obesity role for a gene identified in Drosophila was extended via genetic association to a parallel function in humans.
adp was first described as a “thrifty” gene in Drosophila (23-25). adp mutants with obese phenotypes might have been selected for survival during periods when natural food sources were scarce. Our observations that WDTC1 variants associated with obesity in the BPRHS and GOLDN populations prompts the question whether WDTC1 has been subject to selection during human evolution. We therefore investigated possible natural selection of WDTC1 in human populations. Based on genotyping data from the HapMap Phase II project (26), natural selection has been estimated for the WDTC1 variants (27, http://hg-wen.uchicago.edu/selection/haplotter.htm). We found that WDTC1 SNP i22835A>G (rs4460661) has been subject to strong positive selection in all three analyzed populations (European, Asian, and African) (Fay and Wu's H = -57.14, -84.48, and -4.62, respectively). Fay and Wu's H is a powerful statistic to detect positive selection when the selected variants are in high frequencies (27, 28). A high negative H value suggests selective sweep (28). However, while measured by integrated haplotype score (iHS), an alternative method to detect recent positive selection (27), positive selection was not statistically significant (iHS = -0.701, 0.646, 0.556, for European, Asian, and African, respectively). While this finding requires confirmation, its combination with our observation that the high frequencies of the major allele associating with elevated BMI suggests that WDTC1 might have experienced strong positive selection at some point(s) in human history, likely before diversification of the three populations analyzed by Voight et al. (27). Moreover, WDTC1 could act as a thrifty gene in humans, providing carriers of the major allele (i22835A) with an advantage in resistance to starvation when ancestors to modern humans lived through times of limited food resources and/or challenging environments. Conversely, in contemporary civilization with abundant food supplies, such variants could predispose carriers of certain alleles to an increased risk of being overweight or obese. Therefore, higher frequencies of those major alleles at WDTC1 which associated with increased BMI may contribute to the health disparities currently observed in the two diverse populations studied here.
While WDTC1 variants show strong association with BMI in both populations, the pattern of association is different. The minor allele i22835G associated with low BMI in the BPRHS population, whereas this allele correlated with a high BMI in women of the GOLDN population. In addition, WDTC1 variants exhibited strong genotype by MUFA intake interaction in BPRHS, whereas such interaction was not found in the GOLDN population. These discrepancies may be attributed to genetic and environmental divergence between these two populations. The GOLDN population was recruited from Minneapolis, MN and Salt Lake City, UT and comprised European Americans whereas the BPRHS population comprised Puerto Rican Hispanics living in the Boston area, who had admixture ancestry from African, European and Native Americans (11, 29). The frequency difference in the minor alleles for four SNPs between the BPRHS and GOLDN populations (Table 1), especially for SNP i22835A>G and i61970A>G, further supports the genetic divergence between these two populations. LD and haplotype structures are also slightly different between the two populations. The genetic background difference could contribute to factors influencing the gender-specific effect of the minor allele i22835G on BMI in the GOLDN population. On the other hand, the dietary habits and mean age were also different between both populations. As indicated in Table 2, the BPRHS population is about eight years older than the GOLDN on average. For dietary habits, BPRHS women had significantly higher total energy intake than GOLDN women (Table 2). In addition, while total fat intake is similar in both populations, MUFA intake is significantly higher (P<0.001 for both men and women) in the GOLDN than in the BPRHS population. The female carriers of the minor allele i22835G in the GOLDN had a high BMI, instead of a low BMI as in the BPRHS, and this could be explained by the interaction between MUFA intake and i22835A>G genotypes. As shown in Figure 2, as dietary MUFA intake increased, the rank of BMI between i22835A>G genotypes switched, i.e., after MUFA intake is greater than about 12% of the total energy intake. Minor allele carriers (GG+GA) have a high BMI, instead of a low BMI, whereas non-carriers (AA) have a lower BMI. As the mean MUFA intake in the GOLDN population is 13%, it is anticipated that the female carriers (GG+GA) of the minor allele i22835G have a higher BMI, instead of a lower BMI, than the non-carriers (AA).
The question arises why male carriers of i22835G in the GOLDN population did not respond to a high MUFA intake as female carriers did. This could be explained by the fact that as the BPRHS population was mainly represented by women (72%), the observation of MUFA intake by WDTC1 genotype interaction primarily reflected the high proportion of women in the BPRHS population. Therefore, that GOLDN female carriers of i22835G allele had a higher BMI than the non-carriers in response to a high MUFA intake is consistent with the interaction between MUFA intake and WDTC1 genotypes observed in the BPRHS population but not in GOLDN which had a lower proportion of women (52 %). In addition, the frequency of the minor allele i22835G in the GOLDN population is almost half that in the BPRHS (0.21 vs 0.12). Thus, the combination of a low frequency of the minor allele i22835G and the high MUFA intake in the GOLDN population could be the primary contributor to the reduced power to detect the WDTC1 genotype by MUFA intake interaction.
Alternatively, the food sources of MUFA or those sources in combination with other non-MUFA-containing components of the diet could be different between the two populations as their dietary habits differed from each other (Table 2). Thus, the dietary source of MUFA could contribute the differential pattern of association and interaction between two populations.
We have observed a strong association between WDTC1 variants and obesity in two US populations; however, the association between this locus and obesity across other ethnic groups remains to be determined. As indicated, because WDTC1 might have been subject to natural selection, other genetic and environmental factors in addition to fatty acid intake could modulate the association between WDTC1 variation and obesity.
The research presented here was supported by the National Institutes of Health, National Institute on Aging, Grant Number 5P01AG023394-02 and NIH/NHLBI grant number HL54776 and NIH/NIDDK DK075030 and contracts 53-K06–5-10 and 58–1950-9–001 from the U.S. Department of Agriculture, Agriculture Research Service.
Added fact: WDTC1 SNP i22835A>G (rs4460661) was also found to be associated with obesity in Asian populations in the same manner as observed in the Puerto Ricans.