Trp64Arg β3‐Adrenoceptor: When Does a Candidate Gene Become a Disease‐Susceptibility Gene?

Since the initial discovery ((1)) of the Trp64Arg variant in the β3-adrenoceptor gene (Trp64Arg ADRB3) in 1995 and reports of its association with obesity and related traits ((1)) ((2)) ((3)), there have been more than 100 published articles that have investigated the relationship of this common variant with various phenotypes ((4)) ((5)). Many of these reports have confirmed and extended the initial findings of modest, but statistically significant, associations of Trp64Arg ADRB3 with one or more traits related to obesity, but many others have not. In this month's issue of Obesity Research, Kurokawa et al. ((6)) report a meta-analysis of the relationship between the Trp64Arg ADRB3 variant and body mass index (BMI). Evaluating 35 subgroups that included over 6500 Japanese subjects, they concluded that the Trp64Arg ADRB3 variant accounts for a small but significant increase in BMI (0.26 kg/m2; 95% confidence interval: 0.18 to 0.42; p < 0.01). How should these results be placed in context with the current literature? The answer to this question is not straightforward but depends on the criteria that we choose to define an obesity-susceptibility gene variant. The following considerations are worth noting in defining these criteria not only for Trp64Arg ADRB3, but also more generally for candidate gene variants. It is possible that type 1 error could contribute to false-positive results, i.e., even in the absence of a true genotype-phenotype association, 1 in 20 studies will show a statistically significant result at the p < 0.05 level of statistical significance by chance alone. Moreover, as the number of comparisons within any particular study increases, the probability of a significant result on any one test increases by chance alone. To minimize this possibility, some studies use the Bonferroni adjustment or similar corrections, although such procedures may be an overcorrection and lead to inflated type 2 error rates because the tested phenotypes tend to be correlated with each other. Others still would argue that Bonferroni's correction is too conservative for hypothesis-driven investigations; for example, when there is a priori evidence for a genotype–phenotype association. Another source of potential confounding is publication bias. Positive studies are more likely to be published than negative studies, which might amplify the proportion of positive studies, both true-positives and false-positives as the result of type 1 error. It seems unlikely that all of the reported associations of Trp64Arg ADRB3 can be attributed to type 1 error. Under the null hypothesis of no association, one would expect inflated type 1 error rates to result in statistically significant results being reported in both directions, i.e., for the Trp64Arg allele to be significantly associated in some studies with increased obesity and in others with decreased obesity. However, to our knowledge, there is not a single report (from more than 100 studies published to date) showing a statistically significant association of the Trp64Arg ADRB3 variant with decreased BMI or leanness, as would be expected if all associations were the result of type 1 error. There are of course many factors that contribute to a failure to detect statistically significant associations, the most common being that most studies are underpowered to detect small differences between groups, thus resulting in false-negative results. Other factors that can lead to inconsistent results among studies include differences in ascertainment scheme and study design, ethnic (genetic) background, phenotyping, and statistical analyses. Furthermore, the presence of underlying population admixture can lead to inflated type 1 or type 2 error rates. With respect to the Trp64Arg ADRB3 variant, there are a few positive studies ((2)) ((7)) that have used family-based collections, which minimize potential problems with stratification bias due to admixture. Others, as well as ourselves, have used both candidate-gene and genome-wide approaches to identify obesity-susceptibility genes. The candidate-gene approach has resulted in the identification of several gene variants that seem to influence obesity, whereas linkage studies have identified some chromosomal loci that are likely to harbor obesity-susceptibility genes ((5)). None of these gene variants identified to date would be considered to have a major effect on the common forms of obesity or related phenotypes. Although the obesity genetics research community hopes that there are a few gene variants with major effects, it is also possible that typical obesity is so heterogeneous and polygenic that there will be no major genes. Instead, there may be 20 or more common gene variants, each contributing a genetic burden of a fraction of a BMI unit. To date, there have been three published meta-analyses of the Trp64Arg ADRB3 ((6)) ((8)) ((9)). Each of the studies differed subtly by the criteria used for inclusion of studies and by differing statistical methods. Although at first glance the results of these studies might seem contradictory [two providing evidence for a modest gene effect ((6)) ((9)) and one not ((8))], the results of these studies are actually remarkably consistent with each other (Table 1). In fact, the absolute magnitude of the effect was found to be very similar among the three studies (0.19 to 0.30 kg/m2); only the degree of statistical significance of the effect differed. One important distinguishing characteristic of the most recent study by Kurokawa et al. ((6)) is that only reports in which Japanese subjects were studied were included. It is possible that Japanese may have a genetic background and environment conducive for phenotypic expression of the variant. Only Allison et al. ((8)) examined the effect of gene dose on phenotype. They found that there was a significant effect on the number of Arg alleles on BMI, with an increment of 0.81 kg/m2 from zero to two Arg alleles (p = 0.016). An important limitation of all three meta-analyses is that traits more proximal to the physiology of the β3-adrenoceptor (e.g., lipolysis and resting metabolic rate) were not examined. One clear finding is that Trp64Arg ADRB3 is common in diverse populations, with more than one-half of Pima Indians and Native Eskimos, one-third of Japanese, one-quarter of Mexican Americans and African Americans, and one-fifth of whites harboring at least one copy of the variant allele. Thus, even if the effect of the Trp64Arg ADRB3 variant on obesity-related phenotypes is small, it still may account for significant population-based attributable risk of obesity. A weakness of all of the association studies described above is that they were performed retrospectively; therefore, it was difficult to control for potential confounding variables. In recently performed studies in which subjects were specifically recruited by ADRB3 genotype, matched for potential confounders, and specific hypotheses were tested, subjects carrying the Trp64Arg ADRB3 variant were significantly more insulin resistant ((10)) and had lower resting metabolic rates ((11)), independent of BMI. Similarly, in prospective studies, Walston et al. ((12)) found that subjects homozygous for the Trp64Arg ADRB3 variant had decreased acute insulin release. These findings are potentially consistent with studies showing an earlier onset of type 2 diabetes in subjects who harbor the Trp64Arg ADRB3 variant, and also with recent studies that suggest that β3-adrenoceptors are present in islets and that expression of mutant (Arg64) β3-adrenoceptors attenuate insulin secretion in transfected β-cells ((13)). Several studies have directly or indirectly investigated the functional consequences of this single nonconservative missense mutation, which resides at the junction between the first transmembrane domain and the first intracellular loop of this seven-transmembrane-spanning receptor. Overexpression of the wild-type and mutant receptors in vitro has been performed by at least four groups ((14)) ((15)) ((16)) ((17)); one group showing no changes in function, and three groups showing markedly decreased signaling capabilities (despite normal receptor binding). Indeed, recent studies performed by McLenithan et al. ((17)) show that the Arg64-substituted receptor does not associate with G-proteins and acts as a dominant negative receptor. Furthermore, there is good concordance in studies of omental fat from humans with and without the Trp64Arg ADRB3 variant; two independent studies ((18)) ((19)) show decreased lipolytic activity. A third study ((20)) reportedly did not detect differences in lipolytic activity between genotypes but was a substudy of one of the subsequently reported studies ((18)) that did show significant differences between genotypes. A study of subcutaneous adipose tissue ((21)) did not detect any differences in lipolytic rates between genotypes, which is not surprising because β3-adrenoceptors are expressed in visceral and not subcutaneous fat. If obesity is a heterogeneous and polygenic disease (as almost every obesity genetics article dogmatically states), then why would we expect studies of single gene variants to be conclusive or even consistent with one another? Indeed, some of the most intriguing evidence for the contribution of Trp64Arg ADRB3 to obesity and related phenotypes are recent gene–gene interaction studies. We have shown that there are significant additive or synergistic effects on BMI of Trp64Arg ADRB3 with a number of other candidate gene variants that are in the same or converging biochemical pathways, including Pro12Ala PPARγ2 ((22)), Glu12/Glu9 α-2b-adrenoceptor ((23)), intron 2 variant of PPARγ co-activator-1 ((24)), and Thr92Ala type 2 deiodinase ((25)). Others have independently found some of these interactions ((26)) and have reported additional interactions of Trp64Arg ADRB3 with A-3826G uncoupling protein-1 ((27)) ((28)) and Gly971Arg insulin receptor substrate-1 ((29)). In summary, when considering all of the variability among studies (i.e., ascertainment scheme and study design, ethnic/genetic background, phenotyping, and analyses) and the statistical considerations described above, consistencies among a large proportion of published Trp64Arg ADRB3 association studies seem unprecedented. Meta-analyses are consistent with a small effect of the Trp64Arg ADRB3 variant on BMI. Several functional studies showing defective signaling properties are quite convincing. Recent evidence suggests that Trp64Arg ADRB3 has additive and interactive effects with a number of other candidate gene variants. Given the heterogeneous and polygenic nature of human obesity, would we have expected anything different of an obesity-susceptibility gene?