Genetic Variants of FTO Influence Adiposity, Insulin Sensitivity, Leptin Levels, and Resting Metabolic Rate in the Quebec Family Study

RESEARCH DESIGN AND METHODS

We genotyped 908 individuals from 223 French-Canadian families from the Quebec City area that participated in the Quebec Family Study (QFS), a long-term obesity study that has recruited and extensively phenotyped individuals in several different phases (11). For this study, some families were from randomly ascertained probands and others were recruited based on obese (BMI >32 kg/m²) probands. All subjects provided written informed consent, and the QFS was approved by the Laval University Medical Ethics Committee.

Details of the acquisition of the phenotypes have been reported (12–16). Briefly, percentage body fat was obtained from the Siri equation after measuring body density from underwater weighing. Fat mass (FM) was calculated by percentage body fat multiplied by weight, and fat-free mass (FFM) was calculated by weight minus FM. The six skinfold thicknesses measured were the suprailiac, subscapular, abdominal, medial calf, biceps, and triceps. RMR and respiratory quotient were measured using an open-circuit indirect calorimeter and a ventilated hood. Dietary intake parameters were derived from a quantitative food questionnaire administered by a trained dietitian. Metabolic clearance rate (MCR) and an insulin sensitivity index (ISI) were derived from an oral glucose tolerance test (OGTT) based on the methodology of Stumvoll et al. (17). The formulas for these measures are MCR-OGTT = 18.8 − (0.271 × BMI) − (0.0052 × insul120) − (0.27 × glyc90) and ISI-OGTT = 0.226 − (0.0032 × BMI) − (0.0000645 × insul120) − (0.00375 × glyc90), where insul120 is the insulin level at 120 min and glyc90 is the glucose level at 90 min.

Two SNPs (rs17817449 and rs1421085) were selected in the FTO gene for genotyping. These SNPs were chosen based on the study by Dina et al. (2), who demonstrated that these SNPs had the strongest associations with obesity compared with other SNPs in the region. The SNPs were located in highly conserved regions and were therefore putatively functional (2).

Genotyping was performed using the Sequenom iPLEX Gold Assay (Sequenom, Cambridge, MA). Locus-specific PCR primers and allele-specific detection primers were designed using MassARRAY Assay Design 3.1 software. DNA was amplified in a multiplex PCR and labeled using a locus-specific single base extension reaction. The products were desalted and transferred to a 384-element SpectroCHIP array. Allele detection was performed using Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (compact MALDI-TOF MS). Mass spectrograms and clusters were analyzed by the TYPER 3.4 software package. Ehrich et al. (18) have previously provided details of the procedure.

Regenotyping of rs17817449 was performed using a restriction fragment–length polymorphism (RFLP) assay. PCR conditions were established according to the methodology of Bailey et al. (19). Primers were designed using Primer3 (http://frodo.wi.mit.edu/) (20). The primers used were: 5′AGGACCTCCTATTTGGGACA3′ and 5′AGCTTCCATGGCTAGCATTA3′, and the digestion was incubated with 2 units of AlwN I for 16 h and heat inactivated at 65°C for 20 min.

Deviation from Hardy-Weinberg equilibrium (HWE) was evaluated using the exact test implemented in PEDSTATS (21). All phenotypes except for waist-to-hip ratio, FFM, hip circumference, percentage body fat, and MCR-OGTT were natural log transformed. Association tests for adiposity, insulin sensitivity, and energy phenotypes were performed using a sandwich estimator in proc mixed in SAS (version 8.2; SAS Institute, Cary, NC). The sandwich estimator provided an estimate of the variance-covariance matrix of parameters in our regression-based method. It accounted for correlated data due to familial relationships by correcting for the standard errors for the dependencies of subjects within families. All phenotypes were adjusted for age and sex. We tested both an additive and recessive model because previous work had demonstrated that FTO SNPs in strong linkage disequilibrium (LD) with the two SNPs that we tested (r² = 0.92–1) were associated with type 2 diabetes in these models (5). Significance was set at P < 0.05.

RESULTS

Clinical characteristics of the 908 QFS participants in this study are shown in Table 1. The mean BMI was 27.6 kg/m². Genotyping was performed for two FTO SNPs (rs17817449 and rs1421085) on 908 samples. High call rates were obtained for both SNPs (99.9%), and rs1421085 was in HWE (P = 0.1 among 363 unrelated individuals). We noted a HWE deviation for rs17817449 (P = 0.048 among 363 unrelated individuals). To ensure that this was not due to technical error, we regenotyped rs17817449 using a different technology (RFLP) in 16% of the samples (n = 145). Perfect concordance was found, and therefore the SNP was included in our analyses.

We tested the two SNPs for association with several measures of adiposity. The SNPs were in high LD (r² = 0.93). Results are shown for rs17817449 (Table 2) and rs1421085 (supplementary Table 1 [available in an online appendix at http://dx.doi.org/10.2337/db07-1267]). rs17817449 was associated with eight anthropometric indexes of adiposity (weight, BMI, FM, waist and hip circumferences, waist-to-hip ratio, percentage body fat, and skinfold thickness). The strongest associations were with BMI (P = 0.000081) and waist circumference (P = 0.00021) under a recessive model. The effect on BMI was significant in both men and women separately (data not shown). The SNP also influenced skinfold thickness (P = 0.023 additive, P = 0.0013 recessive), weight (P = 0.0059 additive, P = 0.00055 recessive), FM (P = 0.030 additive, P = 0.0014 recessive), and FFM (P = 0.036 additive, P = 0.0087 recessive). rs17817449 was associated with plasma leptin under both models (P = 0.036 additive, P = 0.0013 recessive), but this was abolished after adjusting for BMI. No association was found between either SNP and type 2 diabetes (data not shown); however, our power was low, as there were only 52 cases.

We investigated whether rs17817449 influenced glucose homeostasis by testing for association with fasting glucose and insulin and the following ISIs: glucose area under the curve (AUC), insulin AUC, 1/homeostasis model assessment of insulin resistance (HOMA-IR), MCR-OGTT, and ISI-OGTT (Table 2). rs17817449 was associated with fasting insulin under both the additive (P = 0.011) and recessive (P = 0.029) models, but an association was not observed with fasting glucose. Many ISIs were associated under both models, including those derived from the OGTT, MCR-OGTT (P = 0.0074 additive, P = 0.0039 recessive), and ISI-OGTT (P = 0.0091 additive, P = 0.0072 recessive). No associations with insulin sensitivity measures remained significant after adjusting for BMI (data not shown).

We hypothesized that the FTO SNPs contribute to body weight regulation through energy intake or energy expenditure. A nonsignificant trend toward increased total energy, lipid, and carbohydrate intake was evident for the obesigenic genotypes (Table 2). Since increased consumption might be expected in obese individuals, association analyses were also performed with energy intake per weight. In addition, we also adjusted for percentage FM, which was negatively correlated with energy intake (r = −0.64, P < 0.0001). After adjustment for these parameters, no significant associations were observed. We next tested for possible associations with energy expenditure (RMR and respiratory quotient). rs17817449 was associated with RMR under both models (P = 0.042 additive, P = 0.0034 recessive) but not after dividing by FFM. No association was observed with respiratory quotient (Table 2).

DISCUSSION

SNPs in the FTO gene have recently been associated with obesity-related phenotypes (1–3) and type 2 diabetes (1,5). The SNPs with the strongest results (rs9939609, rs17817449, rs3751812, rs1421085, and rs8050136) are all in strong LD (r² from 0.92 to 1) based on HapMap European data. In the present study, the minor allele frequencies (MAFs) for rs17817449 and rs1421085 were comparable with those originally reported by Dina et al. (2). rs17817449 had an MAF of 0.39 in all individuals in this study compared with an MAF of 0.40 in control subjects of the adult French sample. An MAF of 0.41 was found for rs1421085 in both samples. In our study, rs17817449 was not in HWE (P = 0.048). Technical error is unlikely since we found perfect concordance in a regenotyped sample using RFLP. Perhaps the ascertainment of some families by an obese proband (see research design and methods) led to the HWE result, possibly due to an overrepresentation of individuals homozygous for the FTO obesogenic genotype (data not shown). This would be consistent with the association results of adiposity traits, which are mostly stronger under a recessive model in this study (Table 2).

In the present study, we confirmed the association of rs17817449 and rs1421085 with several measures of adiposity including BMI, weight, FM, waist circumference, hip girth, percentage body fat, and the sum of six skinfold thicknesses. For rs17817449, the mean BMI for carriers of one G-allele is 0.6 units higher and for two G-alleles is 3.25 units higher when examining all individuals in our QFS sample.

Impaired insulin sensitivity frequently leads to type 2 diabetes (22). In the present study, the obesity risk alleles of rs17817449 and rs1421085 were positively associated with fasting insulin and negatively associated with the insulin sensitivity traits, 1/HOMA-IR, ISI-OGTT, and MCR-OGTT (Table 2 and Supplementary Table 1). The influence of these SNPs on insulin sensitivity appears to be mediated through adiposity, since adjusting for BMI completely eliminated the associations (data not shown). This is consistent with the results of the Wellcome Trust Case Control Consortium/U.K. Type 2 Diabetes study (1).

Excessive energy intake and/or low energy expenditure can lead to the development of obesity; hence, the effect of FTO SNPs could be mediated by either of these changes. Our results did not demonstrate an association with total energy intake after appropriate adjustment; however, a lack of power may explain the nonsignificant results observed with energy intake or utilization. An excess of body fat is accompanied by increases in FFM (23). In the present study, alleles that increased adiposity also increased FFM, a finding that is consistent with those of Frayling et al. (1) who showed similar results with lean mass (P < 0.03). FFM has been shown to be highly correlated with RMR (24), which is in agreement with our data (r = 0.84, P < 0.0001). We observed a significant association with RMR; however, this did not survive adjustment for FFM. Cross-sectional observations, such as those presented here, may be limited in their ability to disentangle the causes of obesity from its effects.

We have demonstrated that SNPs within the FTO gene are strongly associated with several measures of adiposity and are also associated with insulin sensitivity, fat-free mass, plasma leptin, and RMR. More research is needed to determine the specific role of FTO in body weight regulation and obesity. This should include sequencing to identify the specific causal variant(s) and functional studies to further examine the mechanism by which FTO influences adiposity.