A tagging SNP in INSIG2 is associated with obesity-related phenotypes among Samoans

Using genome-wide association analysis, Herbert et al. [1] reported a common variant, rs7566605, in the 5' region of the INSIG2 gene associated with obesity in the Framingham Heart Study population samples and also replicated this finding in four independent cohorts of European and African American ancestries. However, the association of this variant remains inconclusive with confirmation or lack thereof in several follow-up studies conducted in populations of diverse ethnicities [2–12]. Analysis of a single variant could be perceived as a limitation in assessing genetic association of a putative locus. To guard against this potential limitation we conducted a comprehensive association analysis of common tagging SNPs in the INSIG2 gene among adult Samoans, Polynesians of the Western Pacific, residing in Samoa and American Samoa.

Tagging SNPs were selected based on pairwise r ² (≥0.8) among all common SNPs with minor allele frequency (MAF) ≥0.05 using the approach of Carlson et al. [21]. We used the Chinese HapMap database http://www.hapmap.org. Previously we had shown the portability of the tagging SNPs and LD sharing between the Chinese and the Samoans [22]. Four tagSNPs spanning a region of 28 kb in INSIG2 including rs7566605 were identified (Figure 1). Genotyping was performed using the TaqMan assay (Applied Biosystems).

We used the GenABEL package [23] in R to test for association of the SNPs under the additive model. Empirical p-values maximized over the four SNPs were computed using the emp.qtscore function with 10,000 replicates. Since these empirical p-values adjust for the four SNPs, for each trait, we have carried out 5 tests on various subdivisions of the data. A stringent multiple testing correction for this would be to use the Bonferroni correction, which would require testing at the 0.05/5 = 0.01 level. We adjusted for age in all samples, for location (Samoa and American Samoa) in the combined samples and for sex in the Samoa and American Samoa samples. Of the two obesity measurements, BMI was not normally distributed; therefore, we performed logarithmic transformation of BMI to obtain normality.

All of the four SNPs are polymorphic in both locations with minor allele frequencies (MAF) ranging from 17% to 42% (Figure 2). All four SNPs are in HWE (data not shown) with the exception of marginal departures at rs1352083 in the combined male sample (p = 0.053) and rs2161829 in the Samoa female sample (p = 0.051). Since we did not observe significant allele frequency difference between the two polities, we combined the samples and performed tests of association based on the additive model (Table 2). Males and females in this combined sample did not show association of any of the four tagging SNPs either with BMI or ABDCIR. However, when males and females were combined, rs9308762 showed significant association with both BMI (p = 0.024) and ABDCIR (p = 0.009). These results are similar when the test statistic was performed under the recessive model, with both BMI (p = 0.038) and ABDCIR (p = 0.018) showing association with rs9308762. Note that the initial study used the recessive model to fit the association signal [1]. Under additive model, each copy of C allele of rs9308762 was associated with an increase of age and gender adjusted BMI = 0.775 kg/m² and ABDCIR = 2.08 cm. Then we performed tests of associations separately for samples from Samoa and American Samoa. The results showed a significant association of rs9308762 with BMI in the sample from Samoa, but not in the American Samoa sample.

Although we did not find association of rs7566605 with either BMI or ABDCIR, the INSIG2 gene itself is of considerable interest for obesity-related quantitative traits as it is implicated in lipid regulation and synthesis [24, 25]. It should be noted that rs7566605 is located 10 kb upstream of INSIG2 and its functional relevance is not yet known. The putative association of rs7566605 could have resulted from linkage disequilibrium (LD) with a functional variant in INSIG2 in the study populations, in which the association was detected. The follow-up replication studies considered only this polymorphism with inconsistent results in populations irrespective of their ethnic affiliations [2–12, 26]. Among the Asian populations with whom the Samoans share a more recent ancestry, the association of rs7566605 with obesity has been largely negative [27–30], although one study found significant association with BMI in a population from Uyghur [31]. On account of genetic diversity across populations, the extent of LD among the genetic variants is likely to vary, and this could be one of the reasons for the inconsistent findings. To guard against this potential limitation, we conducted an association analysis of common tagging SNPs in INSIG2 with two obesity-related phenotypes. It should be noted that using both genome-wide microsatellite markers at a spacing of 10 cM and over 7,000 SNPs distributed on chromosome 21, we observed a significantly higher level of LD in the Samoans compared to the European populations [18, 22] favoring the identification of a surrogate marker further away from the true variant in the relatively isolated Samoan population. We observed significant association of the most distal SNP rs9308762 with both BMI and ABDCIR in the combined Samoan sample. Although this observation is not in complete agreement with the previous report in which the most significant association was found with rs7566605 [1], our results do not necessarily indicate the existence of different functional variants. A possible explanation is that there is a single functional variant, which is in LD with both rs7566605 and rs9308762. Note that these two SNPs are in elevated LD in the Samoan as well as in the HapMap Chinese population, r² = 0.37 and 0.45, respectively. It is plausible that the putative causal variant is more strongly associated in Europeans with rs7566605, while in Samoans the stronger association is with rs9308762. Although the observations are somewhat preliminary, these results reaffirm the involvement of INSIG2 in regulation of body weight or pathophysiology of obesity related phenotypes.

Although rs7566605 was not significantly associated with obesity in the Samoan population, this study suggests that sequence variants in INSIG2 likely influence the risk for obesity related traits. Our study reiterates the importance of comprehensive assessment of genetic variants within a gene in association studies.