Adipocytes release a number of peptide hormones, collectively known as adipokines, which are essential in regulation of intermediate metabolism, and which may contribute to the pathogenesis of insulin resistance and related disorders [1].

Among these is resistin, a 12.5 kDa cysteine-rich protein that is secreted also by macrophages [2–4]. Several pieces of evidence suggest that this molecule antagonises insulin action and fosters inflammation [2–4]. In humans, most cross-sectional studies have reported that elevated resistin levels are associated with insulin resistance [5–7] and type 2 diabetes [8]. In most [9–11], although not all [12], prospective studies, resistin turned out to be a predictor of type 2 diabetes and cardiovascular disease. Taken together, these findings highlight the role of serum resistin as an emerging pathogenic factor and a potential therapeutic target for insulin resistance-related disorders. This latter possibility is reinforced by the observation that resistin expression in human adipose tissue is inhibited by thiazolidinediones [13], insulin-sensitising molecules that are used as oral hypoglycaemic agents in the treatment of patients with type 2 diabetes.

The mechanisms regulating resistin expression, secretion and circulating levels are still poorly understood. Recent data clearly indicate that resistin serum levels are under genetic control [6, 14]. Several single-nucleotide polymorphisms (SNPs) in the RETN gene coding for resistin, have been associated with serum resistin levels and insulin resistance traits [6, 8, 14–18].

Among others, rs1862513 at position g.−420 in the RETN 5′ flanking region has attracted much attention. This SNP has been associated with high resistin levels [8, 16], insulin resistance [17], obesity [15, 17, 18] and type 2 diabetes [8, 19]. A recent meta-analysis, including more than 3,500 individuals, has shown that those homozygous for the G allele at rs1862513 have a 30% increase in the odds of being affected by type 2 diabetes [8]. Along the same lines, the G allele was predictive of glycaemic deterioration in a Chinese population followed for 5 years [20]. These associations are supported by functional data demonstrating a stronger activation of the RETN promoter by Sp1 and Sp3 transcription factors in the presence of the G allele [8]. In agreement with these data, the same allele has been reported to be associated with higher abdominal-fat resistin mRNA levels [21]. Thus, so far, so good for SNP rs1862513.

Recently, a fine-mapping analysis of the RETN gene has been carried out in a sample of more than 2,500 individuals of European ancestry [14]. The authors found that alleles at rs1477341, rs4804765, rs1423096 and rs10401670 in the gene 3′ UTR were associated with circulating resistin levels; the latter was associated also with fasting glucose and, once SNP-by-BMI interaction was taken into account, with incidence of type 2 diabetes. Quite surprisingly, SNP rs1862513 was not associated with resistin levels [14]. When all published data on more than 5,000 individuals were meta-analysed, a significant association was observed between this SNP and resistin levels; however, it was mainly driven by data obtained in Asian populations, thus suggesting some heterogeneity caused by ethnic-specific genetic effects [14].

In this issue of Diabetologia, Asano et al. [22] have analysed the relationships between serum resistin levels, metabolic traits and RETN variants in a cross-sectional study, carried out in a large cohort comprising more than 3,000 aged Japanese. Like Hivert et al. [14] they used a fine-mapping approach, although limited to the RETN gene. With a straightforward study design and an accurate methodology, they found that two SNPs (i.e. rs34861192 in the promoter and rs3745368 in the 3′ UTR region) were significantly and independently associated with serum resistin, explaining a very large proportion (∼40%) of its variability.

Of note, while resistin levels were associated with HDL-cholesterol, triacylglycerol and obesity, no significant associations were observed between rs34861192 and rs3745368 and these metabolic traits, or the prevalence of type 2 diabetes. Finally, a significant association was observed also for the promoter SNP rs1862513, which, however, was no longer significant in a multiple SNP model, being mostly driven by rs34861192.

Asano et al. wanted to dissect cause–effect relationships among the observed associations by using Mendelian randomisation, an approach to robustly identify causal correlations [23]. However, when a genetic variant is associated with an intermediate phenotype, but not with disease outcome (i.e. resistin levels and metabolic traits, respectively, in this study), the usefulness of such an approach is highly questionable and results remain inconclusive [24].

The study by Asano et al. extends previous investigations in several important aspects. First, it further confirms the association between variability at the RETN locus and circulating resistin levels in a Japanese population.

Second, it shows that this association depends on polymorphisms in both the promoter and the 3′ UTR unique to this sample and different from those previously reported in other Asian [8, 16] and European [14] samples. Taken together, these and previous data reinforce the possibility of a population-specific genetic effect [14]. Variability of the RETN gene across different samples, including discrepancies in haplotype structure, allele frequencies (with rs3745368 being very rare in Europeans [14]) and the presence of monomorphism (with Europeans carrying only the GG genotype at rs34861192 [22]), might explain such a specific effect. Differences across populations in environmental exposures (e.g. different dietary habits) and/or in genetic background might have also played a role. Inter- and intra-ethnic different genetic associations have been recently reported for type 2 diabetes, with the effect of PPARG P12A SNP, an established genetic marker of this disease, being stronger in Asians than in Europeans and among the latter being easily appreciable in northern populations and virtually absent in the Mediterranean area [25]. Placing this issue in a broader perspective, present and previous findings strongly suggest that population-specific genetic effects may occur and, therefore, should be always considered when analysing and interpreting data on the genetics of complex traits.

Third, in this Japanese cohort, the combination of only two RETN variants may modulate a proportion of resistin levels variability that is more than an order of magnitude higher than that previously reported for resistin itself [6, 14], and definitely uncommon in the dominion of quantitative traits, including other adipokines [26, 27]. In the present understanding of the role of common genetic variants that are likely to exert minor effects, this result is totally unexpected and definitively deserves confirmation in further studies.

The genetics of adipokines is still an open and constantly growing area of interest in understanding the pathophysiology of insulin resistance traits. The paper by Asano et al. [22] has made some progress in dissecting the contribution of the RETN locus to resistin circulating levels; further studies are still needed to unravel whether this contribution is really different in different populations. To this purpose, a multicentre collaborative effort, able to recruit and analyse several large samples of different ethnicity, is desirable. In addition, functional studies are necessary to investigate whether the variants so far described, and possibly others yet to be identified, influence RETN transcriptional activity and mRNA levels and/or resistin protein synthesis and production in appropriate tissues. A final and very challenging aim is to examine the genome-wide SNPs and expression datasets in a variety of populations. This may lead to the identification of additional genes able to explain the still large proportion of serum resistin genetic variance that is not accounted for by polymorphisms at the RETN locus.