Disease risk alleles with large frequency differences between populations may account for large differences in the prevalence of diseases between populations . In some cases, large risk allele frequency differences between populations may be the result of local positive selection [41, 42], as predicted by the thrifty gene hypothesis . We previously identified the T2DM risk allele of the Gly482Ser variant in the PPARGC1A gene as a potential thrifty allele because of its unusually high frequency in Polynesians . In the present study, we further test the thrifty gene hypothesis in Polynesians by evaluating the association between the 482Ser risk allele and a T2DM-related phenotype, BMI, in Tongans and Maori. We find that the 482Ser risk allele is associated with age and sex adjusted BMI in Tongans assuming a dominance (P = 0.014) or a co-dominance model (P = 0.037), but not a recessive model of inheritance (P = 0.46). In the Maori, however, we found no significant associations between Gly482Ser genotypes and age and sex adjusted BMI (Table 1 and Figure 1). When considering the populations together and including population as a covariate, we also find no evidence of association (P = 0.56). Despite having a lower BMI, we find that the Maori have a higher PPARGC1A risk allele frequency than the Tongans. Although this observation does not support a causal relationship between the PPARGC1A risk allele and BMI, it can be explained by several confounding factors including sampling biases and environmental differences between populations.
Previous association studies between Gly482Ser and T2DM-related phenotypes have produced conflicting results, with some studies detecting significant effects [22–28] and others failing to detect effects [29–33]. These conflicting results and the differences in effect between populations observed in the present study may be the result of genetic by environment interactions or population-specific epistatic effects. It is worth noting that genome-wide association (GWA) studies have failed to detect associations between SNPs in or near PPARGC1A and T2DM [43–48] or BMI . To date, however, these GWA studies have been restricted to populations relatively distantly related to Pacific populations. Moreover, in the present study, the lack of an association in Maori could be the result of reduced statistical power compared to Tongans due to the smaller sample size (N = 110) and the high risk allele frequency (0.833) in Maori. Our results provide some support for the involvement of the PPARGC1A Gly482Ser SNP in body weight or the pathophysiology of obesity related phenotypes in Pacific populations, but large-scale GWA studies in Pacific populations are desired to test this association more robustly.
If the 482Ser risk allele was driven to high frequency in Polynesians by positive selection, we expect its frequency to be relatively high in Polynesians compared to other populations. We previously showed that the 482Ser risk allele is found at unusually high frequency in a sample of Polynesians compared to a sample of highland New Guineans . Here we extend the genotyping of this SNP to a worldwide sample and find that the highest frequencies of the 482Ser risk allele worldwide are indeed observed in Polynesian populations (Figure 2). One striking feature of the worldwide risk allele frequency distribution is the almost complete absence of the risk allele in Africa: we observe only three Gly482Ser heterozygotes out of 106 African individuals resulting in an allele frequency of only 0.01 in Africa (Figure 2 and Additional File 1, Table S1). Outside of Africa, the risk allele is generally found at intermediate frequencies: 0.36 in Europe; 0.33 in the Middle East; 0.27 in Central and South Asia; 0.39 in East Asia; and 0.25 in the Americas. The exception to this pattern is in Oceania where only a single heterozygote from Bougainville was observed among 53 Melanesians while the risk allele is found at a frequency of 0.69 among Polynesians (Figure 2 and Additional File 1, Table S1). The highest frequency of the risk allele worldwide is 0.85 and is found in Western Samoan and Niue Islanders. Such a striking difference in frequency between Melanesians and Polynesians is consistent with our previous observations  and suggests that local positive selection may be responsible for the increase in frequency of the risk allele in Polynesians. Alternatively, the non-risk allele may have gone to fixation due to selection in Melanesians. To determine whether the observed pattern is in fact due to positive selection, or whether it is the result of demographic effects such as a population bottleneck , will require more formal tests of selection at this locus.
It is worth noting that the search for thrifty alleles in Polynesians has not been restricted to the PPARGC1A gene. For example, it was recently shown that the Q223R (rs1137101) variant in the leptin receptor gene (LEPR) is associated with BMI and obesity in Pacific Islanders and thus represents a candidate thrifty locus . However, the frequency of the 223Q risk allele in Polynesians (0.10) is relatively low: 40 of 54 worldwide populations have higher frequencies of the risk allele than Polynesians . This observation suggests that the 223Q risk allele in LEPR did not rise in frequency in the Pacific due to selection and is thus unlikely to account for the unusually high prevalence of T2DM and obesity-related phenotypes in Polynesians.
The FTO gene was also evaluated as a candidate thrifty gene in Polynesians. Recent GWA studies conducted in populations of European origin have identified several SNPs associated with BMI in the FTO gene [52–54]. In Pacific populations, however, no association was detected between FTO risk alleles and BMI . In addition, the FTO risk alleles are not at unusually high frequency in the Pacific . These observations suggest that FTO is also unlikely to be a thrifty gene in Pacific populations.