Introduction

Type 2 diabetes mellitus is a chronic metabolic disorder characterised by hyperglycaemia, insulin resistance, impaired beta cell function [1, 2] and a variety of cardiovascular risk factors, including hypertension, elevated triglycerides, reduced HDL cholesterol, increased plasminogen activator inhibitor-1 (PAI-1) levels, and endothelial dysfunction [3, 4]. Collectively, these metabolic and cardiovascular risk factors have been referred to as the insulin resistance syndrome (IRS) or metabolic syndrome. Insulin resistance in type 2 diabetes results from impaired insulin action in skeletal muscle and liver [5, 6]. Also, adipocytes are resistant to the anti-lipolytic action of insulin [5, 7], and the resultant increase in plasma NEFA exacerbates the insulin resistance in muscle and liver [5–9].

The prevalence of type 2 diabetes and the IRS is rapidly and relentlessly increasing and, although major risk factors for these disorders have been identified [1], the genetic factors responsible for type 2 diabetes mellitus and the IRS remain unclear. Genome-wide scans have demonstrated linkage between a number of chromosomal regions and diabetes or diabetes-related traits [10, 11]. Candidate gene studies and a number of single-nucleotide polymorphisms (SNPs) have also been associated with type 2 diabetes and/or related diseases, including obesity and insulin resistance [12, 13].

Adiponectin (encoded by the gene ADIPOQ) is a hormone that is expressed only by adipocytes [14] and regulates energy homeostasis and glucose and lipid metabolism [15]. Rodent and human studies have demonstrated that adiponectin levels are decreased in insulin-resistant states, such as type 2 diabetes and obesity [16–19]. Administration of adiponectin to rodents with obesity and diabetes had insulin-sensitising effects [18] and reduced triglyceride accumulation in muscle [19]. Adiponectin acts through its receptors, ADIPOR1 and ADIPOR2. It was initially thought that ADIPOR1 was primarily expressed in skeletal muscle, whereas ADIPOR2 was predominantly expressed in liver [20]. However, we recently showed that ADIPOR2 is highly expressed in human muscle and may be the predominant isoform through which adiponectin exerts its insulin-sensitizing effects in skeletal muscle [21]. Adiponectin receptors mediate enhanced fatty acid oxidation and glucose uptake, increase adenosine 5′-monophosphate-activated protein kinase (AMPK) activity, and interact with peroxisome proliferator-activated receptor (PPAR) pathways [22, 23].

Common polymorphisms in the ADIPOQ gene have been studied in several populations, including Japanese [24] and Europeans [25]. These studies suggest that genetic variation in the ADIPOQ gene may predispose humans to insulin resistance and/or components of the IRS. Additionally, SNP +45 and SNP +276 polymorphisms of the ADIPOQ gene have been shown to predict conversion from impaired glucose tolerance to type 2 diabetes in the Study to Prevent Non-Insulin-Dependent Diabetes Mellitus (STOP-NIDDM) [26]. In contrast, common ADIPOQ polymorphisms were not associated with type 2 diabetes in Pima Indians [27]. Studies of SNPs in the ADIPOR1 and ADIPOR2 genes failed to find any association with type 2 diabetes in northern Europeans or African Americans [28], or in the Japanese population [29]. However, genetic variants in both ADIPOR1 and ADIPOR2 genes were found to be associated with type 2 diabetes in the Old Order Amish population [30]. More recently, it was reported that variation in ADIPOR1 may affect insulin sensitivity and liver fat content in Europeans [31].

To date, no data have been published on the effects of genetic variation in the ADIPOQ, ADIPOR1 and ADIPOR2 genes on type 2 diabetes and the IRS in Mexican Americans, a group at high risk of these diseases [32]. In this study, we conducted a preliminary survey of these three candidate genes, selecting SNPs from the public SNP database (dbSNP) and from a review of the literature. We chose a subset of SNPs in the ADIPOQ gene and all available SNPs in the recently discovered ADIPOR1 and ADIPOR2 genes. We genotyped these common polymorphisms in a Mexican American population sample, and performed association analyses with IRS-related phenotypes.

Subjects and methods

The San Antonio Family Diabetes Study

All study participants were Mexican Americans from the San Antonio Family Diabetes Study (SAFDS) [10, 33]. Briefly, probands for the SAFDS were identified by diabetes status in an earlier epidemiological survey, the San Antonio Heart Study, which has been extensively described elsewhere [34]. All first, second and third degree relatives of the probands, aged from 18–98 years, were invited to participate in the study. Metabolic, anthropometric, demographic and medical history information was obtained on 439 individuals (116 with type 2 diabetes) distributed across 27 low-income Mexican American pedigrees [10, 33]. All procedures were approved by the Institutional Review Board of the University of Texas Health Science Centre at San Antonio, and all subjects gave informed consent prior to their participation.

For this study, we used phenotypic information from those subjects without diabetes (n=323 individuals, depending on availability of data), to avoid metabolic derangements secondary to type 2 diabetes or treatment for type 2 diabetes. The following ten IRS-related phenotypes, which have been described in detail elsewhere [10], were considered for the present study: fasting plasma glucose (FPG), fasting specific insulin (FSI), homeostasis model assessment of β-cell function (HOMA-%β) and insulin sensitivity (HOMA-%S, an inverse measure of insulin resistance), fasting HDL cholesterol, fasting triglyceride concentrations, fasting leptin (LEP), BMI, waist circumference (WC) and sum of skin fold thickness (SS). The sum of eight skinfold measures was used as a measure of overall subcutaneous adiposity. Phenotype-specific distinct outliers were excluded from the analyses (e.g. triglyceride values >9 mmol/l). Where indicated, the data were log-transformed to normalise distributions.

SNP genotyping

Genomic DNA was isolated from whole blood (Qiagen, Chatsworth, CA, USA). SNPs were selected from the NCBI dbSNP database and from previous studies that reported association with diabetes, obesity and/or IRS-related traits. SNPs were obtained by either the Assay-on-Demand or Assay-by-Design service and genotyped according to manufacturer’s instructions (Applied Biosystems, Foster City, CA, USA). Alleles were scored using the allelic discrimination software Sequence Detection System v2.1 (Applied Biosystems). All SNPs were tested for Mendelian pedigree inconsistencies. For all SNPs genotyped, our mean rate of success for genotyping was >98%. Primers and probes used for the SNP genotyping are available on request.

Statistical genetic analysis

We performed association analysis in our complex pedigree-based data using the measured genotype approach (MGA) within the variance components (VC) analytical framework implemented in the program SOLAR (http://www.sfbr.org/solar/, last accessed in July 2006) [35]. The VC-based approach accounts for the non-independence among family members. In this analytical technique, VCs are modelled as random effects (e.g. additive genetic effects and random environmental effects), whereas the effects of measured covariates such as age and sex are modelled as fixed effects on the trait mean. The VCs, the association parameters, and the other covariate effects (e.g. age and sex terms) were estimated, simultaneously, by maximum likelihood techniques. A likelihood function based on multivariate normal density was numerically maximised to obtain parameter estimates.

Prior to performing MGA, the quantitative transmission disequilibrium test [36], as implemented in SOLAR, was used to examine hidden population stratification, using the notation of within (w) and between (b) family components. In MGA, generally, the marker genotypes were incorporated in the mean effects model as a measured covariate, assuming additivity of allelic effects [37, 38]. Using the notation of within (w) and between (b) family components, a significant test of b=w vs b=0 and w=0 reflects a significant difference between the genotypic means (Table 1). All analyses included age and sex terms as covariates, if found to be significant. The tests of association and population stratification and the related hypothesis testing are detailed in Table 1. The nested models were compared using the likelihood ratio test. Twice the difference between the log-likelihoods of these models yields a test statistic that is asymptotically distributed, approximating a χ 2 distribution with 1 df. A p value ≤0.05 is considered significant. Using SOLAR, linkage disequilibrium (LD) between SNP pairs was estimated using the absolute value of the correlation coefficient∣ρ∣. However, for the purpose of discussion, r 2 values are reported to describe the pairwise LD patterns.

Table 1 Association/linkage disequilibrium and population stratification analyses and hypothesis testing
Full size table

The Bayesian quantitative trait nucleotide model

The Bayesian quantitative trait nucleotide (BQTN) analytical technique is employed to analyse SNPs to find the responsible nucleotide variants (the QTNs [39, 40]) influencing a given phenotype. Given complete SNP data for a gene, this statistical technique can be used to identify the sequence variants that are either potentially functional or that exhibit the highest disequilibria with such potential functional sites. The BQTN model is a simple extension of the classical variance component model, which aims to disentangle the genetic architecture of a quantitative trait [41]. This technique has been described in detail elsewhere [41].

Bayesian model selection/model averaging

Since a candidate gene may contain a number of SNPs that could generate several possible competing models of QTN action, we employed a Bayesian model selection/model averaging approach to analyse the SNP data simultaneously in order to estimate the probability that each SNP has a direct effect on the phenotype. This technique has been described in detail elsewhere [41, 42].

The BQTN analysis can be extended to analyse the haplotype data. For a given individual, the most probable haplotypes were generated using the program MERLIN [36], as implemented in SOLAR. The BQTN analytical procedures are implemented in the computer program SOLAR [41].

Conditional linkage/QTN analysis

To examine whether the associated SNP(s) found by the QTN analysis can account for our reported linkage of triglyceride concentrations at chromosome 12p13.31, which is close to the ADIPOR2 locus (chromosome 12p13.33), we combined the QTN analysis with our identity-by-descent (IBD)-based variance component linkage analysis. If a variant, or set of variants, in the ADIPOR2 gene is responsible for the observed linkage signal, linkage analysis conditional on a fixed-effect MGA of the polymorphism will yield an expected logarithm of the odds ratio in favour of linkage (LOD) score near zero. Alternatively, if the associated polymorphism is in less than complete linkage disequilibrium with the true functional site, linkage analysis will generally yield a non-zero LOD score. This method has been described in detail elsewhere [43].

Results

The clinical characteristics of the subjects are shown in Table 2. We genotyped 59 SNPs (ADIPOQ=9, ADIPOR1=12 and ADIPOR2=38), and, of these, 38 (ADIPOQ=8, ADIPOR1=6, and ADIPOR2=24) were polymorphic and in Hardy–Weinberg equilibrium (HWE) in our dataset. SNPs that were not in HWE were discarded from further analysis. Schema of genomic structure and variants of the ADIPOQ, ADIPOR1 and ADIPOR2 genes are shown in Fig. 1. The minor allele frequencies (MAFs) of the SNPs, categorised by diabetes status, are shown in Table 3. The frequency of the minor alleles varied from less than 1 to 49%. There were no significant associations with type 2 diabetes (Table 3).

Table 2 Characteristics of the SAFDS subjects without diabetes distributed across 27 families
Full size table
Fig. 1
figure 1

Genomic structure, and location of SNPs in the ADIPOQ (a), ADIPOR1 (b) and ADIPOR2 (c) genes. Exons are shown as boxes, and introns as lines connecting the boxes. Numbers in parenthesis indicate location relative to the A of the ATG of the start codon. The asterisks (*) denote SNPs that were polymorphic in this population (MAF >0.01). The SNPs that were not in HWE are denoted by a dagger (†)

Full size image
Table 3 Minor allele frequencies of SNPs by diabetes status
Full size table

We performed association analysis using MGA, after finding no statistical evidence of hidden population stratification. The significant results (p≤0.05) of the association analyses are summarised in Table 4. Of the eight SNPs examined in the ADIPOQ gene, four exhibited significant associations with several IRS-related phenotypes. The minor allele for SNPs rs4632532 and rs182052 were significantly associated with an increase in BMI (p=0.029 and p=0.032), FSI (p=0.023 and p=0.026) and SS (p=0.0089 and p=0.0084), and a decrease in HOMA-%S (p=0.015 and p=0.016). Two other SNPs, rs266729 and rs2241767, were significantly associated with SS (p=0.036 and p=0.013). For all SNPs genotyped in the ADIPOQ gene, the pairwise LD (r 2) ranged from 0 to 0.98. The highest pairwise LD found among the eight ADIPOQ SNPs was between SNP pair rs4632532–rs182052 (r 2=0.98), followed by SNP pair rs2241766–rs2241767 (r 2=0.92), as shown in Fig. 2a. The average pairwise LD among the ADIPOQ SNPs was 0.37.

Table 4 Significant associations between genetic polymorphisms in ADIPOQ, ADIPOR1 and ADIPOR2 genes and IRS-related phenotypes
Full size table
Fig. 2
figure 2

Linkage disequilibrium (LD) between SNP pairs within ADIPOQ (a), ADIPOR1 (b) and ADIPOR2 (c) genes. SNPs are labelled on the y-axis, and the locations (bp) within the gene are shown on the x-axis. Pairwise LD is estimated using the correlation coefficient ∣ρ∣, and depicted in the figure by the colour intensity of the shaded box, as shown in the legend. The diagonal represents a comparison of each SNP against itself (i.e. ∣ρ∣=1.0)

Full size image

SNP rs7539542 of ADIPOR1 was significantly associated with BMI, SS and WC (p=0.025, p=0.047 and p=0.0062, respectively). The minor allele was associated with an increase in BMI, SS and WC, as shown in Table 4. Correlations between SNP pairs were estimated and are shown in Fig. 2b. The pairwise LD in the ADIPOR1 gene (r 2) ranged from 0 to 0.87. The highest LD was between SNP pair rs2275737–rs2275738 (r 2=0.87), as shown in Fig. 2b. The average pairwise LD among the ADIPOR1 SNPs was 0.39.

Of note, 14 of the 24 ADIPOR2 SNPs were significantly (p<0.05) associated with fasting plasma triglyceride concentrations. The mean triglyceride concentrations per genotype are depicted in Table 4 for the significantly associated SNPs. The majority of the minor alleles, with the exception of rs1029629 and rs12582624, were associated with decreased triglyceride levels. Figure 2c shows the overall pattern of LD in the ADIPOR2 gene, and Table 4 highlights SNPs that were in strong LD with each other. The average proportion of shared variation between SNPs, measured by the pairwise correlation, was 0.31, with SNP-specific values ranging from 0 to 0.99. Of particular note, four of these SNPs (rs10848569, rs929434, rs3809266 and rs12342) were in high pairwise LD (r 2=0.99) and were strongly associated with triglyceride levels (p=0.00029, p=0.00016, p=0.00027 and p=0.00021).

Given our previous findings of common genetic influences on IRS-related phenotypes, including obesity, insulin resistance and triglyceride concentrations [10], we verified whether adjustment for measures of obesity and insulin resistance may affect the observed patterns of association between variants in ADIPOR2 and triglyceride concentrations. Thus, we repeated the association analysis of triglyceride and ADIPOR2 SNP data after accounting for the effects of BMI and HOMA-%S. All of the initial findings were observed again, but with increased levels of significance, most notably for SNP rs929434 (p=0.000060), with the exception of SNP rs1468491, where the p value changed from 0.026 to 0.032.

Following the findings of the association analysis of the SNPs in ADIPOR2 using MGA (Table 4), we performed the BQTN analysis using ten of the 24 SNPs in ADIPOR2 to obtain a more plausible model, given the data. The 14 SNPs that exhibited redundancy owing to relatively high pairwise LDs (r 2 >0.80) were excluded from the BQTN analysis. The remaining ten non-redundant SNPs resulted in the examination of 1,024 possible additive gene action models. The model with only SNP rs929434 provided the strongest evidence for the observed associations, with estimated posterior probability of >0.82. The BQTN analysis was extended to analyse the predicted ADIPOR2 haplotypes. Of the 35 observed haplotypes, 13 haplotypes had frequencies >1%, with a cumulative frequency of 89%. The 35 haplotypes generated 59,536 possible models to be examined by BQTN analysis; however, haplotypic information did not improve the association.

To determine whether rs929434 has relevance to our reported linkage of fasting triglyceride concentrations at chromosome 12p13.31 [11], which is close to the ADIPOR2 locus (chromosome 12p13.33), we re-analysed the original triglyceride data using information from our recent genome scan data generated by the Centre for Inherited Disease Research (CIDR). Using the total fasting triglyceride dataset, for which SNP rs929434 data were also available, we found modest evidence for linkage (LOD=1.85) of plasma triglyceride levels to a genetic location between markers D12S374 and GATA49D12 at 12p13.31. When we re-analysed the data, conditional on SNP rs929434, the LOD score was reduced to 0.98. Thus, about 47% of the evidence for linkage at 12p13.31 was explained by the sequence variation at rs929434.

Discussion

We have previously shown that the Mexican American population is at relatively high risk of obesity, type 2 diabetes and the IRS. The genetic variation in candidate genes ADIPOQ, ADIPOR1 and ADIPOR2 could influence variation in such disease conditions and related traits. In the present study we provide evidence that ADIPOR2 polymorphisms are strongly associated with decreased plasma triglyceride levels. Of the polymorphisms that we genotyped in the ADIPOR2 gene, our strongest association was with fasting triglyceride levels, most notably SNPs rs10848569, rs929434, rs3809266 and rs12342, where we observed high pairwise correlation between all SNPs (r 2=0.99). These four SNPs with relatively high MAF of 44.3–45.1% appear to have substantial influence on variation in circulating triglyceride levels, and the evidence for association continues to be significant after Bonferroni’s correction for multiple testing. Given that there is some correlation between phenotypes and between SNPs, the Bonferroni correction is rather conservative. To further investigate the potential functional relevance of the variants in the ADIPOR2 gene, we performed a BQTN analysis, in which we compared all possible models of gene action. This analysis provided strong statistical support for a single SNP model (i.e. rs929434) for the observed association with triglyceride concentrations.

Genetic variants in the ADIPOR2 gene were found to be associated with type 2 diabetes in the Old Order Amish population [30]. Of the eight SNPs typed (rs10773980, rs1029629, rs11061971, rs10735003, rs9805042, rs12342, rs1044471 and rs2286380), three (rs1029629, rs12342 and rs1044471) in the Amish population were significantly associated with combined type 2 diabetes and impaired glucose tolerance, while in the present study, these SNPs were associated with fasting triglyceride concentrations. Interestingly, the minor allele for rs12342 in the present study, which was significantly associated with decreased fasting triglyceride, was found to be the major allele in the Amish population. In contrast, a study in a Japanese population demonstrated no association between variants (including five that we examined: rs2058033, rs12582624, rs10848569, rs12342 and rs1044471) in ADIPOR2 and type 2 diabetes mellitus [29]. Likewise, no association between variants (including five that we examined: rs10773980, rs767870, rs12578449, rs9805042 and rs730031) in ADIPOR2 and glucose and lipid metabolic parameters were found in a European population [31].

We previously reported modest evidence for linkage of fasting triglyceride concentrations to a genetic location on chromosome 12p, which is close to the ADIPOR2 locus (chromosome 12p13.33) [11]. To determine whether rs929434 could account for our linkage signal, we re-evaluated linkage on chromosome 12p, conditional on the measured genotype effect. By including rs929434 genotypes as a covariate in the model, the LOD score for fasting triglyceride concentrations was reduced from 1.85 to 0.98. Thus, about 47% of the evidence for linkage at 12p13.31 was explained by the sequence variation at rs929434, in turn suggesting that this may be one of several functional variants or in strong LD with the true functional variant(s). The majority of the minor alleles for the polymorphisms in ADIPOR2 were associated with decreased fasting triglyceride and, thus, may provide some protection against developing the IRS and atherosclerosis. Two ADIPOR2 SNPs, rs1029629 and rs12582624, that were in strong LD with each other (r 2=0.93), were associated with increased fasting triglyceride. This seemingly paradoxical finding is not unexpected, given that independent variants in the same gene may have independent effects on traits.

The molecular mechanisms by which adiponectin exerts its insulin-sensitising effects on liver and muscle appear to be related to an increase in fatty acid oxidation, an effect that is mediated, at least in part, by its activation of AMPK [23]. In skeletal muscle, adiponectin acts through its receptors (ADIPOR1 and ADIPOR2) to phosphorylate AMPK, which, in turn, phosphorylates acetyl CoA carboxylase. This leads to an increase in fatty acid oxidation [22, 23] and a reduction in toxic intracellular lipid metabolites that inhibit insulin signalling. Adiponectin also decreases hepatic glucose output by reducing the expression of enzymes involved in gluconeogenesis in the liver [44, 45], thereby contributing to the regulation of whole-body glucose homeostasis. As a result of these processes, a reduction in tissue triglycerides and toxic lipid metabolites is observed in both skeletal muscle and liver [19, 22, 23], and this contributes to improved insulin signal transduction. The effects of adiponectin on muscle and liver were recently reviewed elsewhere [46], and it was suggested that liver is the primary site of adiponectin bioactivity. ADIPOR2 is abundantly expressed in human liver [20], and an important question raised by our results is whether adiponectin acts through the hepatic ADIPOR2 to influence triglyceride biosynthesis and homeostasis.

There has been considerable interest in the role of kinetic defects in triglyceride-rich lipoprotein (TRL) metabolism in the development of atherosclerosis and dyslipidaemia (e.g. increased fasting triglycerides and low HDL cholesterol), as it relates to type 2 diabetes and the IRS [47, 48]. It has been suggested that adiponectin concentration may play an independent role in regulating TRL metabolism, aside from its link with abdominal fat, insulin resistance and dyslipidaemia [47]. These observations have relevance to a recent finding where a significant correlation was demonstrated between the expression of ADIPOR1 and ADIPOR2 mRNA in human skeletal muscle and in vivo parameters of glucose and lipid metabolism [49]. ADIPOR2 mRNA levels correlated positively and independently only with fasting triglyceride concentrations [49], whereas ADIPOR1 mRNA expression was positively correlated with serum insulin and C-peptide concentrations, first-phase insulin secretion, and elevated plasma triglyceride and cholesterol concentrations [49]. Given this independent association of ADIPOR2 gene expression with fasting triglyceride concentrations, but not with BMI, insulin or glucose, and our previous findings of common genetic influences on IRS-related phenotypes (e.g. obesity, insulin resistance and fasting triglyceride) [10], we examined whether adjustment for measures of obesity and insulin resistance may affect the observed patterns of association between variants in ADIPOR2 and fasting triglyceride. The sample size was reduced (N=273) because of the requirement of common covariate information. All the initial findings were observed again, but with increased levels of significance (with the exception of SNP rs1468491, as noted above). For example, the p value for association of SNP rs929434 with fasting triglyceride concentrations was 0.000060. These results, together with previous findings [49], indicate that adiponectin and its receptors, especially ADIPOR2, play an important role in regulating fasting plasma triglyceride levels, and that they could have a potential role in VLDL metabolism.

Several studies have shown that the ADIPOQ gene is associated with IRS-related phenotypes [24–27]. Of particular note, a haplotype including two SNPs at positions −11,391 (rs17300539) and −11,377 (rs266729), both located in the promoter sequence of the ADIPOQ gene, was shown to be strongly associated with adiponectin concentrations and type 2 diabetes in the French population [50]. Our results provide additional evidence that variants of the ADIPOQ gene are significantly associated with IRS-related traits.

We genotyped 12 SNPs in the ADIPOR1 gene (Fig. 1), and six of these were polymorphic and in HWE in our dataset. Of the six polymorphic SNPs, one was significantly associated with IRS-related traits. Interestingly, SNP rs7539542 of ADIPOR1 was significantly associated with three obesity traits: SS, BMI and waist circumference. Similar to our findings, variants in ADIPOR1 were found to be associated with type 2 diabetes in the Old Order Amish population [30]. The following five SNPs were evaluated in both studies: rs6666089, rs2275738, rs2275737, rs1342387 and rs7539542. In the Amish population, allele/genotype frequencies of SNPs rs2275737 and rs1342387 differed significantly between subjects with type 2 diabetes vs. those with normal glucose tolerance. However, there were no significant associations between ADIPOR1 SNPs and type 2 diabetes, insulin sensitivity or insulin secretion in northern Europeans, African Americans [28] or a Japanese population [29]. In contrast, a recent study in a European population demonstrated significant association between several SNPs (including the two we examined, rs666089 and rs1342387) and insulin sensitivity/liver fat [31].

In conclusion, this study demonstrates that polymorphisms in ADIPOR1, ADIPOR2 and ADIPOQ are associated with several IRS-related phenotypes in Mexican Americans, including plasma triglyceride levels, SS, BMI and waist circumference. Of particular note, multiple genetic variants in ADIPOR2 were strongly associated with decreased triglyceride levels. Our findings of multiple strong associations between ADIPOR2 polymorphisms and plasma triglyceride concentrations may have important implications for atherogenesis and/or dyslipidaemia, owing to the potential influence of ADIPOR2 genetic variation on triglyceride-rich lipoprotein metabolism.