The INSIG2 rs7566605 genetic variant does not play a major role in obesity in a sample of 24,722 individuals from four cohorts

Obesity results from an imbalance between food intake and energy expenditure, and is a major risk factor for many of the chronic diseases of adulthood including hypertension, type 2 diabetes, coronary heart disease and stroke [1, 2]. Although the results of twin and adoption studies provide strong support for a genetic influence on body weight [3‐5], mutations causing monogenic forms of obesity are rare in the population [6], and major genes contributing to common obesity have not been described. In addition, interaction between genes and environment is suggested by an age-adjusted prevalence of 32.9% reported for American adults with a BMI ≥ 30 kg/m² in 2002–2004, while only 15% met this criterion in the 1976–1980 National Health Examination Survey [7, 8]. Recent genetic studies have focused on unbiased genome-wide associations studies (GWAS) of the relationship between large numbers of common genetic polymorphisms measured simultaneously and disease risk. The development of high-density genotyping technology and the completion of the HapMap resource [9] have facilitated this approach, and enabled the successful detection of SNPs with moderate effects that contribute to common variation in obesity [10‐18].

Herbert et al. identified a genetic variant near the INSIG2 locus associated with obesity as assessed by a BMI ≥ 30 kg/m² in a GWAS carried out in participants in the Framingham Heart Study [19]. This finding was subsequently replicated in four of five additional populations. Since the homozygous CC genotype conferring susceptibility was found to be present in approximately 10% of the individuals studied, the authors speculated that although the rs7566605 polymorphism has a moderate influence on the risk for obesity (pooled odds ratio (OR) = 1.22, 95% confidence interval (CI) = 1.05–1.42, p = 0.0080), there may be a considerable impact of this gene on public health due to the high frequency of the allele in the general population. Similarly, when the rs7566605 variant was subsequently genotyped in nine additional cohorts from multiple ethnicities the association with BMI was reproduced in only five of the study samples suggesting that the effect of the variant might differ between populations [20]. INSIG2 was also independently identified as a genetic determinant of obesity in a genome-wide association study carried out in 1,000 unrelated Caucasian adults of Northern European origin [15]. We, therefore, assessed the effect of rs7566605 on obesity risk in individuals enrolled in the Atherosclerosis Risk in Communities (ARIC) study [21], the Coronary Artery Risk Development in Young Adults (CARDIA) study [22, 23], the Genetic Epidemiology Network of Arteriopathy (GENOA) study [24], and two cohorts of extremely obese and lean white individuals ascertained at the University of Ottawa [25].

INSIG2 maps to chromosome 2q14, and encodes a 225-amino acid protein that participates in two pathways implicated in the regulated synthesis of cholesterol, fatty acids, and triglycerides in the liver and other organs. The first of these entails the sterol-dependent cleavage of two transcription factors, sterol regulatory element-binding proteins 1 and 2, found in the endoplasmic reticulum. If sterols are present, INSIG2 blocks lipid synthesis by preventing the translocation of the transcription factors to the nucleus and transcription of target genes declines [26‐31]. INSIG2 can also bind to the enzyme 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMG CoA reductase) which catalyzes the rate-limiting step in cholesterol biosynthesis. Following interaction with INSIG2, HMG CoA reductase is ubiquitinated and rapidly degraded in the proteasome so that cholesterol synthesis is inhibited [32‐35]. A role for Insig-2 in vivo was later established using an animal model when overexpression of the Insig-2 cDNA in Zucker diabetic fatty rats was shown to decrease the levels of triacylglycerols in the liver and plasma when compared to uninfected littermate controls [36].

A description of the study sample divided into case and comparison groups on the basis of obesity (BMI ≥ 30 kg/m²) and stratified by racial group is shown in Table 1. The allele and genotype frequencies for the INSIG2 polymorphism were in agreement with Hardy-Weinberg equilibrium expectations in each of our study populations. In African-American, white, and Hispanic obese individuals, there was no discernable difference in the frequencies of the three INSIG2 SNP genotypes when compared to non-obese study subjects (Table 1). When the relationship between the rs7566605 variant and BMI considered either as a categorical variable or a continuous variable was studied, no significant association with obesity was found for participants in any of the four cohorts or in a combined analysis of all of the study groups (p = 0.38; p = 0.88 for Mantel-Haenszel test of homogeneity) (Table 2). Weight, waist circumference, and waist-to-hip ratio were also examined in three of the cohorts after stratification by race (Tables 2 and 3). There were no significant mean differences observed among INSIG2 rs7566605 genotypes for these anthropometric measurements with the exception of a significant association for reduced waist-to-hip ratio for white ARIC study participants with the INSIG2 CC genotype, and for an increased waist-to-hip ratio for African-Americans in the ARIC cohort carrying the same allele. After combining all of the study participants and adjusting for age, gender, race, and study, there was no significant association between rs7566605 and any of the obesity-related quantitative traits (Tables 2 and 3).

The role of INSIG2 rs7566605 as a determinant of obesity risk was investigated in a sample of 24,722 individuals belonging to four different cohorts and a significant association was not found under a recessive genetic model. Herbert et al. identified a genetic variant (rs7566605) 10 kb upstream of INSIG2 that was associated with obesity as assessed by a BMI ≥ 30 kg/m² in participants in the Framingham Heart Study. This finding was then replicated in four of five additional populations including individuals of Western European ancestry, African-Americans, and children [19]. However, the absence of an association between rs7566605 and BMI levels when DNA samples from the Nurses Health Study cohort were genotyped suggested that the SNP may have variable effects in different populations. The initial association with measures of body size has subsequently been reproduced in some [15, 20, 40‐43] but not all cohorts [44‐56]. In one of three of the earliest studies designed to replicate this finding, Dina et al. reported a lack of association between rs7566605 and either BMI or obesity under a recessive model in 4,998 unrelated middle-aged French participants in the Donne Epidemiologiques sur le Syndrome d'Insulino-Resistance (DESIR) cohort [45]. There was also no evidence of association between the sequence variant and obesity in two other large studies of unrelated European adults when 4,916 individuals in the European Prospective Investigation of Cancer (EPIC) Norfolk study or 1,683 subjects in the Medical Research Council (MRC) Ely study were genotyped [46]. Although Rosskopf et al. did not observe any association between the INSIG2 variant and either obesity or BMI when the entire Study of Health in Pomerania (SHIP) cohort consisting of 4,310 unrelated German individuals was assessed, a subgroup analysis of 2,701 overweight individuals in SHIP showed that both the mean BMI and risk of being obese was increased for rs7566605 CC homozygotes when compared to those with the GC and GG genotypes. Based on these results, the authors speculated that since INSIG2 is a regulator of insulin-mediated fatty acid synthesis [57], the effect of the rs7566605 variant might be stronger under conditions of increased insulin stimulation, and might explain the failure to find an association in the Nurses' Health Study where the mean BMI was lower than in the other study samples [19, 47]. A similar suggestion was made by Hall et al. to explain the lack of replication of an association between the INSIG2 variant and obesity-related phenotypes in 1,428 members of 248 British families where about half of the cohort was overweight [44]. However, no difference in the distribution of rs7566605 genotype frequencies, or association with measures of body composition including BMI, waist circumference, waist-to-hip ratio, and percentage body fat was found when 1,026 severely obese patients with a mean BMI of 46.0 kg/m² were compared with 818 population-based controls from Utah [53]. Similarly, no association with overweight, obesity, or obesity-related quantitative traits was reported in a combined sample of 18,014 Danish subjects from four cohorts, or in an obese subgroup of 3,878 subjects with a mean BMI of 32.8 kg/m²[52].

There was also no consistent evidence that the INSIG2 polymorphism is a determinant of obesity risk in studies including individuals of non-European descent. No association between rs7566605 and BMI or measures of obesity was found in two separate cohorts in India [48], a group of 747 patients with type 2 diabetes that included Afro-Caribbean and Indian subjects [49], or in three Japanese cohorts [50, 51, 54], while a significant association with obesity was found for 908 Japanese patients with a mean BMI ≥ 30 kg/m² when compared to 1,495 controls whose weight was in the normal range [40]. In the study reported here, there was no significant association with obesity in Mexican-American participants in the GENOA cohort who were ascertained as members of families with at least two siblings with diabetes. The high percentage of obese individuals (49.1%) (Table 1) and an average BMI of 30.82 kg/m² are consistent with expectation since obesity is a well-established risk factor for diabetes in this population [58]. An association between the INSIG2 polymorphism and obesity both in unrelated and family based samples was described by Herbert et al. for African-Americans from Maywood, Illinois [19], but was not confirmed in a later study in the same population [20]. There was no replicated association for the INSIG2 variant and either obesity or BMI observed for African-American participants in the ARIC, CARDIA, or GENOA studies.

The influence of INSIG2 rs7566605 on variation in other anthropometric measures including weight, waist circumference, and waist-to-hip ratio was also analyzed. There were no significant mean differences observed among INSIG2 rs7566605 genotypes except for an association with waist-to-hip ratio for white and African-American ARIC study participants. Although BMI is a widely used surrogate measure of adiposity, waist-to-hip ratio is an index of abdominal obesity and has been shown to be a better predictor of both myocardial infarction and stroke [59, 60]. However, the difference in the direction of effect observed for the two racial groups, and the marginal p-values given the large number of statistical tests conducted make this result difficult to interpret, particularly in the absence of direct measures of body composition. In addition, a combined analysis of all study participants adjusted for age, gender, race, and study did not show a significant association between the INSIG2 rs7566605 variant and waist-to-hip ratio (Table 3).

There are a variety of possible reasons to explain the different associations detected between the INSIG2 variant and obesity in different populations. These include differences in ascertainment or study design, population substructure, genotype call rate, degree of LD between rs7566605 and a true causative variant, and heterogeneity between the populations due to unknown genetic, lifestyle, or environmental factors. In this context, a potential interaction between the INSIG2 polymorphism and level of physical activity was observed in the Danish Inter99 cohort. Study participants who were carriers of either the rs7566605 C allele or G allele and were physically inactive had a BMI that was 1.00 kg/m² or 0.54 kg/m² higher, respectively, than their physically active counterparts with the same genotype [52]. An additional reason for the failure to replicate an association between the INSIG2 genetic variant and measures of obesity could be low statistical power. In this study, there was adequate power (≥ 80%) to detect an association similar to that reported in earlier studies for the risk of obesity under a recessive genetic model (OR = 1.29–1.75) [19, 20, 40, 42] for each of the racial groups within the ARIC and GENOA cohorts, while the ORs detectable for white and African-American participants in the CARDIA study were 1.96 and 1.87, respectively. For all of the obesity-related continuous traits, the power reached 95% to observe a small effect of the INSIG2 sequence variant (R² ≤ 1%) given the INSIG2 rs7566605 minor allele frequencies found for each study population after stratification by race.

In the initial report describing the association between INSIG2 and obesity, the lack of replication in the Nurses' Health Study was attributed to the presence of fewer individuals with a high BMI when compared to the Framingham Heart Study or KORA samples in which an association between the rs7566605 variant and obesity was observed [19]. In the case of the white ARIC participants, the BMI distribution appears to be comparable to that of the KORA cohort. It was later suggested that the effect of INSIG2 variation may be more prominent in cohorts of young individuals [20]. Therefore, it is noteworthy that we did not observe an association between rs7566605 and BMI in the young CARDIA cohort. In summary, the results of an analysis of a sample of 24,722 individuals belonging to four cohorts do not support a major role for the INSIG2 rs7566605 variant as a determinant of obesity risk.