Figures
Abstract
Background
Adiponectin, a secretagogue exclusively produced by adipocytes, has been associated with metabolic features, but its role in the development of the metabolic syndrome remains unclear.
Objectives
We investigated the association between serum adiponectin level and metabolic traits, using both observational and genetic epidemiologic approaches in a multiethnic population assembled in Canada.
Methods
Clinical data and serum adiponectin level were collected in 1,157 participants of the SHARE/SHARE-AP studies. Participants were genotyped for the functional rs266729 and rs1260326 SNPs in ADIPOQ and GCKR genes.
Results
Adiponectin level was positively associated with HDL cholesterol and negatively associated with body mass index, waist-to-hip ratio, triglycerides, fasting glucose, fasting insulin, systolic and diastolic pressure (all P<0.002). The rs266729 minor G allele was associated with lower adiponectin and higher HOMA-IR (P = 0.004 and 0.003, respectively). The association between rs266729 SNP and HOMA-IR was no longer significant after adjustment for adiponectin concentration (P = 0.10). The rs266729 SNP was associated with HOMA-IR to an extent that exceeded its effect on adiponectin level (0.15 SD 95% C.I. [0.06, 0.24], P<0.001). There was no significant interaction between rs266729 SNP and ethnicity on adiponectin or HOMA-IR. In contrast, the SNP rs1260326 in GCKR was associated with HOMA-IR (P<0.001), but not with adiponectin level (P = 0.67).
Citation: Mente A, Meyre D, Lanktree MB, Heydarpour M, Davis AD, Miller R, et al. (2013) Causal Relationship between Adiponectin and Metabolic Traits: A Mendelian Randomization Study in a Multiethnic Population. PLoS ONE 8(6): e66808. https://doi.org/10.1371/journal.pone.0066808
Editor: Rocio I. Pereira, University of Colorado Denver, United States of America
Received: January 8, 2013; Accepted: May 13, 2013; Published: June 24, 2013
Copyright: © 2013 Mente et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: A. Mente is supported by an Early Career Award from Hamilton Health Sciences Foundation. D. Meyre is supported by a Canada Research Chair in Genetics of Obesity and Diabetes. M.B. Lanktree was supported by the Canadian Institutes of Health Research (CIHR) MD/PhD Studentship Award and the University of Western Ontario MD/PhD Program and is a CIHR Fellow in Vascular Research. R.A. Hegele is a Career Investigator of the Heart and Stroke Foundation of Ontario and holds the Edith Schulich Vinet Canada Research Chair (Tier I) in Human Genetics, the Martha G. Blackburn Chair in Cardiovascular Research, and the Jacob J. Wolfe Distinguished Medical Research Chair at the University of Western Ontario. H. Gerstein holds a Aventis Population Health Institute Chair in Diabetes Research, McMaster University. S. Yusuf holds an endowed chair in Cardiovascular Research from the Heart and Stroke Foundation of Ontario. S.S. Anand holds the Michael G. DeGroote/Heart and Stroke Foundation of Ontario endowed Chair in Population Health Research, Canada Research Chair in Ethnicity and Cardiovascular Disease, and the Eli Lilly-May Cohen Chair in Women’s Health Research. This work was supported by CIHR (MOP-13430, MOP-79523, CTP-79853), the Heart and Stroke Foundation of Ontario (NA-6059, T-6018, PRG-4854, NA-03785), Genome Canada through the Ontario Genomics Institute, and the Pfizer Jean Davignon Distinguished Cardiovascular and Metabolic Research Award. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors received funding from a commercial source (Jean Davignon Distinguished Cardiovascular and Metabolic Research Award, sponsored by Pfizer). This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
Introduction
Adiponectin is a secretagogue exclusively produced by adipocytes and abundantly secreted into the bloodstream where it accounts for 0.01% of total plasma protein [1]. Extensive study of adiponectin function in animal models led to the conclusion that adiponectin promotes insulin sensitivity in the muscle and liver [2], [3]. Cross-sectional epidemiological studies in humans have shown that low serum adiponectin is associated with insulin resistance, obesity, dyslipidemia, coronary artery disease, hypertension or type 2 diabetes (T2D) [4]–[8]. In longitudinal studies, hypoadiponectionemia was shown to predict the development of insulin resistance [9], dyslipidemia [10], T2D [11], hypertension [12], coronary artery disease [13] but was not a predictor of subsequent weight gain [14]. Heritability studies provided strong evidence that plasma adiponectin level is genetically determined and that an overlap exists between the genetic architecture of adiponectin and features of the metabolic syndrome such as fasting insulin or HDL-cholesterol [15], [16]. The gene encoding adiponectin (ADIPOQ) is located on chromosome 3q27 and comprises three exons [17]. Candidate gene studies [18], and more recently genome-wide association (GWA) studies [19], fine-mapping [20] or large-scale deep resequencing experiments [21] have identified numerous common and rare variants at the ADIPOQ locus associated with serum adiponectin level. Even though ADIPOQ is the major genetic determinant of serum adiponectin level, 12 additional contributing loci have been identified through large-scale GWA meta-analyses [22]–[25]. Datsani and colleagues recently showed convincing evidence of association between a multiple SNP adiponectin-decreasing gene score and decreased body mass index (BMI), increased waist to hip ratio (WHR), higher triglyceride (TG) levels, lower HDL-C concentrations, higher post oral glucose tolerance test (OGTT) 2-hr glucose level, higher homeostatic model assessment-insulin resistance (HOMA-IR) and increased T2D risk using large datasets from international consortia [25]. Overall, these data suggest that adiponectin may be a key hormone in the development of metabolic syndrome.
The common SNP -11377 C>G (rs266729) is located in the promoter region of the gene ADIPOQ and has functional consequences on ADIPOQ gene expression [26]. The rs266729 G variant allele has been consistently associated with lower serum adiponectin concentration in diverse ethnic groups [18], [27] and with increased risk of T2D [28] and coronary artery disease [29]. In this study we investigated the causal association between serum adiponectin level and the different components of the metabolic syndrome, using both observational and genetic epidemiology approaches in a multiethnic population randomly assembled in Canada.
Subjects and Methods
Ethics Statement
This study was approved by the Hamilton Health Sciences/Faculty of Health Science Research Ethics Board and written informed consent was obtained from each participant including consent to analyze genetic specimens. The investigation conforms to the principles outlined in the Declaration of Helsinki.
Study Population
The study population was comprised of Canadians of European, South Asian, Chinese, and Aboriginal origin who participated in the Study of Health Assessment and Risk in Ethnic groups (SHARE) and the Study of Health and Research Evaluation in Aboriginal Peoples (SHARE-AP), two cross-sectional studies of cardiovascular disease (CVD) risk factors conducted between 1996 and 1998. Individuals between age 35–75 years were randomly selected from three cities (Toronto, Hamilton, Edmonton), and from the Six Nations Reservation (Ohsweken, Ontario) as previously described [30], [31].
Clinical Measurements and Biochemical Analyses
We recorded each participant’s health behaviours and medical history using standardized questionnaires as previously described [30], [31]. Blood pressure, height, weight, waist and hip circumference were measured using a standardized protocol [32]. Fasting blood samples were collected in the morning from all participants. Blood samples were collected and processed according to a standard protocol and were shipped to the core laboratory in Hamilton for analysis. All subjects underwent an eight-hour fast before blood was drawn for glucose, insulin, lipids and adiponectin plasma level measurements. Plasma levels of total cholesterol, triglycerides, and glucose were measured using enzymatic methods [33]–[35] and plasma HDL cholesterol level was measured after precipitating the VLDL and LDL with phosphotungstic acid and magnesium chloride [36]. Plasma LDL cholesterol level was calculated according to the method developed by Friedewald et al. [37]. Plasma insulin level was measured by manual radioimmunoassay assay (Diagnostic Products Corporation, Los Angeles USA). Plasma adiponectin level was measured in the laboratory of Dr. Stephan Blankenberg at the University of Mainz, Germany using a commercially available Human adiponectin ELISA assay (RD195023100), produced by Biovendor Research Products. The intra-assay imprecision is 6.4–7.0% and the interassay imprecision is 7.3–8.2%. Basal insulin resistance was calculated using the previously validated HOMA-IR model [38].
Genetic Analysis
The genotypic information of the rs266729 SNP in ADIPOQ was extracted from the gene-centric 50 K single nucleotide polymorphism (SNP) array described elsewhere [39]. Genomic DNA was extracted from leukocytes using the Illumina Human CVD beadchip scanned on the Illumina BeadStation 500 G at the Centre for Applied Genomics for SHARE and from whole blood using the Puregene DNA purification kit (Gentra Systems, Minneapolis MN, USA) at the Clinical Trials and Clinical Research Laboratory (CTCRL) for SHARE-AP. Genotyping of the gene-centric 50 K array was carried out in the Centre for Applied Genomics (Hospital for Sick Children, Toronto, Ontario, Canada; www.tcag.ca) for SHARE and in the Genome Quebec Innovation Center (McGill University, Montreal, QC; www.genomequebec.com) for SHARE-AP. In selecting SNP(s) for analysis, we considered functional rather than surrogate SNPs, since the linkage disequilibrium structure in participants may vary according to the ethnic background. The only available functional SNP in ADIPOQ in the gene-centric 50 K array was rs266729. The rs266729 SNP was in Hardy-Weinberg equilibrium within each ethnic group and the call rate for the rs266729 SNP was 98.9%. The polymorphism is common in each ethnic group (G minor allele frequency comprised between 22 and 29% in the four ethnic groups). The G variant is significantly less frequent in Chinese compared to other ethnic groups (Table S1).
We also searched for SNPs for use in a reciprocal Mendelian randomization approach [40] to help further decipher the causal relationships between adiponectin level and HOMA-IR. We first searched for a biologically relevant genetic variant conclusively associated with HOMA-IR in the literature and genotyped in the 50 K array. We identified the rs1260326 SNP in GCKR as a relevant SNP for this approach. The P446L coding non-synonymous SNP rs1260326 is biologically relevant [41], [42] and has been conclusively associated with HOMA-IR in a French population (P<5×10−8) [43]. The rs1260326 SNP was in Hardy-Weinberg equilibrium within each ethnic group and the call rate for the rs1260326 SNP was 98.9%. The polymorphism is common in each ethnic group (A minor allele frequency comprised between 5% and 21% in the four ethnic groups). The A variant is significantly less frequent in South Asians (5%) and Aboriginals (6%) compared to other ethnic groups (16% in Europeans and 21% in Chinese).
Statistical Analysis
The statistical analyses were conducted using SAS version 9.1 (Cary, South Carolina), the R package, or PLINK software [44]. Two-tailed P-values <0.05 after Bonferroni correction were considered statistically significant. Hardy-Weinberg equilibrium for each SNP within ethnic groups was assessed by square goodness-of-fit test (P>0.001). The chi-square contingency test was used to compare allele and genotype frequency distributions between ethnic groups and sexes. Non-normally distributed variables were natural log-transformed prior to statistical analyses. Multiple linear regressions were used to assess associations of the rs266729 SNP with quantitative traits, while adjusting for covariates of age, sex and ethnicity coded as indicator variables.
To estimate the genetic effect on metabolic traits as a function of genetically lowered adiponectin levels, we performed a Mendelian randomization analysis. First, we used an instrumental variable analysis in which we determined the quantitative trait change per 1 SD of higher genetically predicted adiponectin concentration. We fit the data to a generalized least squares regression model with the rs266729 SNP as a predictor variable, adjusting for covariates (age, sex, ethnicity), and then we regressed the genetically predicted adiponectin values against the metabolic trait. Second, we calculated a predicted increase in the metabolic parameter based on the effect on adiponectin levels. We converted log adiponectin and log HOMA-IR into z-scores and then used linear regression to model the effect of the rs266729 SNP on log adiponectin z-scores, adjusting for age, sex, and ethnicity. This slope value was then multiplied by the amount that log HOMA-IR z-score changes for every 1 z-score change in log adiopnectin z-score (as determined from a linear regression equation of log adiponectin predicting log HOMA-IR, adjusting for covariates). The resulting value was the predicted increase in the metabolic parameter based on the effect on adiponectin. Third, linear regression was used to assess the effect on log HOMA-IR z-score for the modeled allele. We compared this observed effect with the predicted effect to test whether HOMA-IR may be causally related to rs266729 SNP determined higher adiponectin level. Finally we used a reciprocal Mendelian randomization approach [40] to help further decipher the causal relationships between adiponectin level and HOMA-IR. The functional rs1260326 SNP in GCKR is a relevant SNP for a bi-directional Mendelian randomization analysis.
An additive model was used in all the analyses including the rs266729 and the rs1260326 SNP. All of the results, with the exception of the socio-demographic data, are reported based on multivariable analyses. In a sub-analysis, we assessed the association of the metabolic traits with the rs266729 SNP by ethnicity. To further assess the possible heterogeneity in the effects by ethnicity, we conducted a meta-analysis of the parameter estimates obtained for each ethnic group using a random effects model. With a sample size of 1,157, we had 80% power to detect a standard deviation difference in log adiponectin of 0.17 (equal to a difference of 0.12 log ug/mL) and a standard deviation difference in log HOMA-IR as small as 0.17 (absolute difference in log HOMA-IR of 0.105) for a MAF of 20%.
Results
Participant Characteristics
Genotype and phenotype data were available from 1,186 participants from the SHARE and SHARE-AP population. From this sample, 29 individuals who had implausible levels of adiponectin (>100 µg/mL) were excluded, leaving a final sample of 1,157 participants (286 Aboriginals, 258 Europeans, 319 South Asians, and 294 Chinese).
The characteristics of participants are displayed in Table S2. The mean age of the overall population is 50.3 years. Women are less represented among the South Asians compared to Aboriginals (P = 0.02). Significant differences in age, health behaviours including current smoking, measures of adiposity, glycemic index and load, adiponectin and leptin concentration, and insulin resistance are present among ethnic groups (Table S2). For example, Aboriginal people have substantially higher BMI and abdominal obesity relative to other ethnic groups. Conversely people of Chinese origin have the lowest BMI and abdominal obesity. Europeans have a higher BMI, yet less abdominal obesity compared to South Asians. Despite these differences, South Asians and Aboriginal people are more insulin resistant compared to the Europeans (all P<0.05). Adiponectin levels are significantly higher in Europeans and Aboriginals compared to the other ethnic groups (overall P<0.001). (Table S2).
Epidemiologic Assessment of Adiponectin Level and Metabolic Parameters
To investigate the association between adiponectin level and metabolic parameters in this multiethnic sample, multivariable linear regression analyses were performed, adjusting for age, sex and ethnic origin. As shown in Table 1, adiponectin level is positively associated with HDL cholesterol (standardized B coefficient = +0.322; P<0.0001) and negatively associated with body mass index (B = –0.193; P<0.0001), waist-to-hip ratio (B = –0.196; P<0.0001), triglycerides (B = –0.247; P<0.0001), fasting glucose level (B = –0.158; P<0.0001), fasting insulin level (B = –0.297; P<0.0001), HOMA-IR (B = –0.297; P<0.0001), systolic (B = –0.098; P = 0.002) and diastolic (B = –0.123; P<0.0001) blood pressure. LDL cholesterol is not associated with adiponectin level (P = 0.42) (Table 1). All of the associations remained statistically significant after Bonferoni adjustment for multiple testing (10 tests, P adjusted <0.005).
Association of rs266729 SNP with Adiponectin Level and Related Metabolic Parameters
We assessed the association of the rs266729 polymorphism, under an additive model, with adiponectin level and metabolic parameters found to be associated with adiponectin in our study. The SNP is significantly associated with serum adiponectin level (Table 2). The minor G allele is associated with lower mean adiponectin concentration (G/G: 2.16, C/G: 2.32 and C/C: 2.38 mmol/L log units, respectively) (P = 0.004), adjusting for age, sex, and ethnic origin. Each additional copy of the minor G allele predicts lower adjusted adiponectin by 0.09 mmol/L log units (95% CI: 0.03, 0.15; P = 0.003) (Table 3).
The rs266729 polymorphism is also significantly associated with HDL cholesterol and HOMA-IR, after adjusting for age, sex, and ethnic origin (Table 2). The minor allele (G) is associated with lower mean HDL concentration (G/G: 0.95 mmol/L, C/G: 1.07 and C/C: 1.08 mmol/L, respectively) (P = 0.004). The G allele is also associated with higher mean HOMA-IR (C/G and C/C) (G/G: 1.28, C/G: 1.08 and C/C: 1.06 log units, respectively) (P = 0.003). The associations remain statistically significant after accounting for multiple testing (11 tests, P adjusted <0.0045).
All other metabolic characteristics which are associated with adiponectin concentration (body mass index, waist-to-hip ratio, fasting triglycerides, glucose, insulin, and systolic and diastolic blood pressure) are not significantly associated with the rs266729 polymorphism.
The association between the SNP rs266729 and HOMA-IR [45], [46], has been reported previously. Conversely, since the association between the SNP rs266729 and HDL cholesterol was novel, we sought replication of this finding in a publically available independent sample of >100,000 subjects of European ancestry [47]. As the association was not confirmed in this large sample (P = 0.87), we considered the initial significant association between the SNP rs266729 and HDL cholesterol as a random false-positive result, and we did not assess the causal association of adiponectin level and HDL cholesterol in the Mendelian randomization analysis.
Mendelian Randomization Study for Serum Adiponectin Using SNP rs266729
In order to examine whether a higher HOMA-IR (as observed in carriers of the G allele) may be causally related to the rs266729 SNP determined lower adiponectin level, we performed a Mendelian randomization analysis. If serum adiponectin is causally related to increased insulin resistance, then we would expect that genetically reduced adiponectin (e.g., in individuals carrying the rs266729 G allele) would increase insulin resistance to the same extent as observed in epidemiological studies. We present three main results: i) an estimated causal change in HOMA-IR for a genetically induced change in adiponectin concentration (instrumental variable analysis); ii) a predicted increase in HOMA-IR based on the effect on adiponectin levels (column 6 in Table 4); and iii) an observed increase in HOMA-IR (column 7 in Table 4).(see Methods section).
Using instrumental variable analyses, we estimated a causal change in log HOMA-IR of –0.61 (95% CI: –0.70, –0.52; p<0.001) z-score units for a genetically induced change of one log adiponectin z-score unit. Table 4 shows the association of HOMA-IR with the rs266729 SNP. The relationship is also shown pictorially in Figure 1. The minor G allele that correlates with lower adponectin is also associated with increased insulin resistance (P<0.001, column 7 in Table 4). For this SNP, the observed effect exceeds the effect predicted based on the serum adiponectin change. Specifically, the G allele decreases log adiponectin by 0.14 SD units (95% CI: 0.05, 0.24; P = 0.002) and is expected to increase log HOMA-IR by 0.038 SD units; however we observed a 0.15 SD unit (95% CI: 0.06, 0.24; P<0.001) increase in HOMA-IR. The 95% CI for the estimated causal effect does not include the point estimate for the observed association between adiponectin and HOMA-IR, suggesting that the two estimates are truly different. In order to strengthen confidence in our observation that rs266729 may be causally related to HOMA-IR we reciprocally assessed another biologically relevant SNP known to be associated with HOMA-IR. The SNP rs1260326 in GCKR was strongly associated with HOMA-IR (P<0.001), but not with adiponectin level (P = 0.67). Consequently, we did not further explore reciprocal Mendelian randomization analyses using this SNP.
To examine whether the effect of the rs266729 SNP on HOMA-IR is mediated by the variation of adiponectin level, we added log adiponectin to the multivariable model shown in Column 7 of Table 4. With the addition of serum adiponectin into the model, the association of the SNP rs266729 with log HOMA-IR (B = 0.09; 95% CI: −0.01, 0.18; P = 0.10) is attenuated and no longer statistically significant.
There is no significant heterogeneity in the gene effect on adiponectin or HOMA-IR variation across ethnic groups (P for interaction = 0.52, and 0.71 respectively) (Table S3). To further test if the SNP effects are influenced by ethnicity, we conducted a meta-analysis of the parameter estimates obtained for each ethnic group. The association of adiponectin, HOMA-IR) and the rs266729 SNP were unaltered (Table S4).
Discussion
By using both observational, genetic epidemiology and Mendelian randomization approaches in the multi-ethnic SHARE/SHARE-AP study we found that the adipocyte-secreted hormone adiponectin may have a causal role in the variation of insulin resistance (as measured by the HOMA-IR index). This study is the first to our knowledge to explore the relationships between adiponectin and metabolic traits using Mendelian randomization. Mendelian randomization is thought to overcome the limitations of classical observational epidemiology and provides useful evidence to support or reject causal hypotheses linking intermediary biomarkers and complex diseases [48]. Our findings suggest a possible causal relation between a low serum adiponectin level and a higher level of insulin resistance. More specifically, we show that 1) serum adiponectin level is negatively associated with HOMA-IR, 2) the adiponectin-increasing C allele of the promoter ADIPOQ SNP rs266729 is significantly associated with lower HOMA-IR index, 3) the association of the rs266729 SNP with HOMA-IR is attenuated after adjusting for serum adiponectin level in the regression model, 4) the rs266729 SNP is associated with HOMA-IR to an extent that exceeds its effect on adiponectin level, and 5) the validity of this observation is strengthened because the functional SNP rs1260326 in GCKR while strongly associated with HOMA-IR, is not associated with adiponectin. Thus our findings do not support a reciprocal causal association of HOMA-IR on adiponectin level.
A strong body of evidence in the literature supports the view that adiponectin may have a causal role in insulin resistance. Serum adiponectin level has been associated with insulin resistance in cross-sectional and longitudinal studies conducted in European [10], [49] and non-European populations [9], [50]–[52]. A recent family-based study showed a heritability (the proportion of total phenotypic variability caused by genetic variance in a population) of 55% for plasma adiponectin and genetic correlations ranging from 18.6 to 20% between plasma adiponectin, HOMA-IR and plasma insulin [15]. These data suggest that the genetic architecture of plasma adiponectin overlaps with the genetics of insulin resistance. Disruption of adiponectin causes a phenotype of insulin resistance in mice [53] and over-expression of adiponectin improves insulin sensitivity wild-type and leptin deficient obese mice [54]–[57]. The physiological mechanisms supporting a link between adiponectin and insulin sensitivity include increased glucose utilization and fatty-acid oxidation in skeletal muscle, suppression of glucose production in the liver, increased storage of triglyceride in the adipocyte and increased synthesis and turn-over of triglycerides in brown adipose tissue [2], [3], [56]–[59]. The significant associations we found between the promoter ADIPOQ SNP rs266729, serum adiponectin level and insulin resistance are in line with previous reports [18], [27], [28], [45], [46], [60]–[62]. It is noteworthy that the SNP rs266729 effect on insulin resistance significantly exceeds its predicted effect in the Mendelian randomization experiment. This result is further supported by our finding that adiponectin attenuates rather than negates the effect of the SNP rs266729 on insulin resistance. These data suggest that a low level of adiponectin may alter insulin sensitivity via direct and indirect physiological mechanisms. For instance, insulin resistance state may have effects on gene expression- i.e. negative or positive feedback and amplify the effect of the SNP rs266729 on insulin resistance. Further mechanistic studies are needed to understand the complex relationships between the hormone adiponectin and insulin resistance.
We found no evidence for an interaction between the ADIPOQ SNP rs266729 and ethnicity on variation of adiponectin or HOMA-IR. Most of the SNPs identified through GWAS are thought to be proxies in linkage disequilibrium with causal variants and may therefore not been informative in diverse ethnic populations [20], [63], [64]. As an illustration, Croteau-Chonka et al. recently demonstrated that a “synthetic” GWAS association for serum adiponectin level previously reported in the Filipino population [23] was induced by a rare coding variant (R221S) at the ADIPOQ locus in linkage disequilibrium with the common SNPs. The R221S variant was unique to the Filipino population and was not found in sequences of 12,514 individuals of European ancestry [20]. Given the similar pattern of genetic effects identified across ethnic groups, our findings illustrate the relevance of using functional instead of tag SNPs in assessing the role of gene variants on metabolic phenotypes in a multi-ethnic Mendelian randomization context [48]. It should be noted that the power of our study remains modest to investigate interactions between the SNP and ethnicity in relation to metabolic traits and further work with this SNP is needed in larger multiethnic samples.
The data suggest that a consistent association between adiponectin and insulin sensitivity is observed in populations of European, South Asian, East Asian or Native Canadian ancestries. This result is in line with previous reports in SHARE/SHARE-AP and in other populations, since they did not provide consistent evidence of varying associations between adiponectin and metabolic traits like insulin resistance according to the ethnic background [65]–[67]. This study is the first to our knowledge to apply Mendelian randomization in a multi-ethnic sample, an important prerequisite to evidence valid causal associations representative of the worldwide human diversity. An important limitation of using Mendelian randomization in a multi-ethnic context therefore resides in its sensitivity to population stratification [48]. We therefore recommend including genetic markers that are robust to inter-ethnic MAF variations in multi-ethnic Mendelian randomization experiments.
Despite the fact that we found a significant association between adiponectin level and HDL cholesterol, HOMA-IR, BMI, WHR, TG, fasting glucose level, fasting insulin level, HOMA-IR, systolic and diastolic blood pressure in our multi-ethnic observational study, we only found a convincing genetic association between the ADIPOQ promoter SNP rs266729 and HOMA-IR. One possible explanation is that the association found between adiponectin level and metabolic traits is not causal and explained by confounding factors for all traits but HOMA-IR. However, the modest sample size of the current study (N = 1,157) leads us to not totally exclude the presence of subtle genetic effects of the SNP rs266729 on other metabolic traits that we were unable to detect. For instance, the absence of association between the SNP rs266729 and BMI may be unexpected, since insulin resistance is a key feature of obesity [68]. However, the SNP rs266729 was not associated with BMI in the large-size GIANT study (N >120,000 European subjects, P = 0.30), in line with our results [69]. Another limitation of this study is we have used total adiponectin as opposed to the active high–molecular weight multimer to estimate the circulating levels of adiponectin. Despite the fact that total adiponectin is strongly associated with the HOMA-IR in our study, this may introduce bias the Mendelian randomization model. We are also aware that insulin resistance may have been more precisely estimated using oral glucose tolerance test-derived indexes.
Conclusion
In conclusion, our data suggest a possible causal association between serum adiponectin level and insulin resistance as measured by HOMA-IR, irrespective of the ethnic background. More Mendelian randomization experiments are needed in well-powered multi-ethnic studies to decipher the complex pleiotropic role of the adipocyte-secreted hormone adiponectin on metabolic traits.
Supporting Information
Table S1.
Characteristics of the rs266729 SNP genotyped in Aboriginal, South Asian, Chinese and European participants.
https://doi.org/10.1371/journal.pone.0066808.s001
(DOC)
Table S2.
Characteristics of study participants by ethnicity.
https://doi.org/10.1371/journal.pone.0066808.s002
(DOC)
Table S3.
Association of metabolic traits with the rs266729 polymorphism. Values are beta-coefficient (95% CI). †
https://doi.org/10.1371/journal.pone.0066808.s003
(DOC)
Table S4.
Association of metabolic traits with the rs266729 polymorphism, based on meta-analysis of adjusted estimates obtained for each ethnic group. Values are based on fixed effect model. †
https://doi.org/10.1371/journal.pone.0066808.s004
(DOC)
Author Contributions
Conceived and designed the experiments: HG SY SSA. Performed the experiments: ADD RM. Analyzed the data: AM MH. Contributed reagents/materials/analysis tools: RH. Wrote the paper: AM DM SSA. Critically revision of the article: AM DM ML MH HG RH SY SSA.
References
- 1. Scherer PE, Williams S, Fogliano M, Baldini G, Lodish HF (1995) A novel serum protein similar to C1q, produced exclusively in adipocytes. J Biol Chem 270: 26746–26749.
- 2. Fruebis J, Tsao TS, Javorschi S, Ebbets-Reed D, Erickson MR, et al. (2001) Proteolytic cleavage product of 30-kDa adipocyte complement-related protein increases fatty acid oxidation in muscle and causes weight loss in mice. Proc Natl Acad Sci U S A 98: 2005–2010.
- 3. Yamauchi T, Kamon J, Waki H, Terauchi Y, Kubota N, et al. (2001) The fat-derived hormone adiponectin reverses insulin resistance associated with both lipoatrophy and obesity. Nat Med 7: 941–946.
- 4. Ouchi N, Kihara S, Arita Y, Maeda K, Kuriyama H, et al. (1999) Novel modulator for endothelial adhesion molecules: adipocyte-derived plasma protein adiponectin. Circulation 100: 2473–2476.
- 5. Weyer C, Funahashi T, Tanaka S, Hotta K, Matsuzawa Y, et al. (2001) Hypoadiponectinemia in obesity and type 2 diabetes: close association with insulin resistance and hyperinsulinemia. J Clin Endocrinol Metab 86: 1930–1935.
- 6. Iwashima Y, Katsuya T, Ishikawa K, Ouchi N, Ohishi M, et al. (2004) Hypoadiponectinemia is an independent risk factor for hypertension. Hypertension 43: 1318–1323.
- 7. Matsubara M, Maruoka S, Katayose S (2002) Decreased plasma adiponectin concentrations in women with dyslipidemia. J Clin Endocrinol Metab 87: 2764–2769.
- 8. Kumada M, Kihara S, Sumitsuji S, Kawamoto T, Matsumoto S, et al. (2003) Association of hypoadiponectinemia with coronary artery disease in men. Arterioscler Thromb Vasc Biol 23: 85–89.
- 9. Yamamoto Y, Hirose H, Saito I, Nishikai K, Saruta T (2004) Adiponectin, an adipocyte-derived protein, predicts future insulin resistance: two-year follow-up study in Japanese population. J Clin Endocrinol Metab 89: 87–90.
- 10. Wildman RP, Mancuso P, Wang C, Kim M, Scherer PE, et al. (2008) Adipocytokine and ghrelin levels in relation to cardiovascular disease risk factors in women at midlife: longitudinal associations. Int J Obes (Lond) 32: 740–748.
- 11. Lindsay RS, Funahashi T, Hanson RL, Matsuzawa Y, Tanaka S, et al. (2002) Adiponectin and development of type 2 diabetes in the Pima Indian population. Lancet 360: 57–58.
- 12. Chow WS, Cheung BM, Tso AW, Xu A, Wat NM, et al. (2007) Hypoadiponectinemia as a predictor for the development of hypertension: a 5-year prospective study. Hypertension 49: 1455–1461.
- 13. Laughlin GA, Barrett-Connor E, May S, Langenberg C (2007) Association of adiponectin with coronary heart disease and mortality: the Rancho Bernardo study. Am J Epidemiol 165: 164–174.
- 14. Vozarova B, Stefan N, Lindsay RS, Krakoff J, Knowler WC, et al. (2002) Low plasma adiponectin concentrations do not predict weight gain in humans. Diabetes 51: 2964–2967.
- 15. Henneman P, Aulchenko YS, Frants RR, Zorkoltseva IV, Zillikens MC, et al. (2010) Genetic architecture of plasma adiponectin overlaps with the genetics of metabolic syndrome-related traits. Diabetes Care 33: 908–913.
- 16. Povel CM, Boer JM, Feskens EJ (2011) Shared genetic variance between the features of the metabolic syndrome: heritability studies. Mol Genet Metab 104: 666–669.
- 17. Saito K, Tobe T, Minoshima S, Asakawa S, Sumiya J, et al. (1999) Organization of the gene for gelatin-binding protein (GBP28). Gene 229: 67–73.
- 18. Vasseur F, Helbecque N, Dina C, Lobbens S, Delannoy V, et al. (2002) Single-nucleotide polymorphism haplotypes in the both proximal promoter and exon 3 of the APM1 gene modulate adipocyte-secreted adiponectin hormone levels and contribute to the genetic risk for type 2 diabetes in French Caucasians. Hum Mol Genet 11: 2607–26014.
- 19. Heid IM, Henneman P, Hicks A, Coassin S, Winkler T, et al. (2010) Clear detection of ADIPOQ locus as the major gene for plasma adiponectin: results of genome-wide association analyses including 4659 European individuals. Atherosclerosis 208: 412–420.
- 20. Croteau-Chonka DC, Wu Y, Li Y, Fogarty MP, Lange LA, et al. (2012) Population-specific coding variant underlies genome-wide association with adiponectin level. Hum Mol Genet 21: 463–471.
- 21. Warren LL, Li L, Nelson MR, Ehm MG, Shen J, et al. (2012) Deep resequencing unveils genetic architecture of ADIPOQ and identifies a novel low-frequency variant strongly associated with adiponectin variation. Diabetes 61: 1297–1301.
- 22. Jee SH, Sull JW, Lee JE, Shin C, Park J, et al. (2010) Adiponectin concentrations: a genome-wide association study. Am J Hum Genet 87: 545–552.
- 23. Wu Y, Li Y, Lange EM, Croteau-Chonka DC, Kuzawa CW, et al. (2010) Genome-wide association study for adiponectin levels in Filipino women identifies CDH13 and a novel uncommon haplotype at KNG1-ADIPOQ. Hum Mol Genet 19: 4955–4964.
- 24. Richards JB, Waterworth D, O'Rahilly S, Hivert MF, Loos RJ, et al. (2009) A genome-wide association study reveals variants in ARL15 that influence adiponectin levels. PLoS Genet 5: e1000768.
- 25. Dastani Z, Hivert MF, Timpson N, Perry JR, Yuan X, et al. (2012) Novel loci for adiponectin levels and their influence on type 2 diabetes and metabolic traits: a multi-ethnic meta-analysis of 45,891 individuals. PLoS Genet 8: e1002607.
- 26. Laumen H, Saningong AD, Heid IM, Hess J, Herder C, et al. (2009) Functional characterization of promoter variants of the adiponectin gene complemented by epidemiological data. Diabetes 58: 984–991.
- 27. Ong KL, Li M, Tso AW, Xu A, Cherny SS, et al. (2010) Association of genetic variants in the adiponectin gene with adiponectin level and hypertension in Hong Kong Chinese. Eur J Endocrinol 163: 251–257.
- 28. Han LY, Wu QH, Jiao ML, Hao YH, Liang LB, et al. (2011) Associations between single-nucleotide polymorphisms (+45T>G, +276G>T, −11377C>G, −11391G>A) of adiponectin gene and type 2 diabetes mellitus: a systematic review and meta-analysis. Diabetologia 54: 2303–2314.
- 29. Yang Y, Zhang F, Ding R, Wang Y, Lei H, et al. (2012) Association of ADIPOQ gene polymorphisms and coronary artery disease risk: A meta-analysis based on 12 465 subjects. Thromb Res 130: 58–64.
- 30. Anand SS, Yusuf S, Vuksan V, Devanesen S, Teo KK, et al. (2000) Differences in risk factors, atherosclerosis, and cardiovascular disease between ethnic groups in Canada: the Study of Health Assessment and Risk in Ethnic groups (SHARE). Lancet 356: 279–284.
- 31. Anand SS, Yusuf S, Jacobs R, Davis AD, Yi Q, et al. (2001) Risk factors, atherosclerosis, and cardiovascular disease among Aboriginal people in Canada: the Study of Health Assessment and Risk Evaluation in Aboriginal Peoples (SHARE-AP). Lancet 358: 1147–1153.
- 32. Anand SS, Yusuf S, Vuksan V, Devanesen S, Montague P, et al. (1998) The Study of Health Assessment and Risk in Ethnic groups (SHARE): rationale and design. The SHARE Investigators. Can J Cardiol 14: 1349–57.
- 33. Allain CC, Poon LS, Chan CS, Richmond W, Fu PC (1974) Enzymatic determination of total serum cholesterol. Clin Chem 20: 470–475.
- 34. McGowan MW, Artiss JD, Strandbergh DR, Zak B (1983) A peroxidase-coupled method for the colorimetric determination of serum triglycerides. Clin Chem 29: 538–542.
- 35. Neeley WE (1972) Simple automated determination of serum or plasma glucose by a hexokinase-glucose-6 -phosphate dehydrogenase method. Clin Chem 18: 509–515.
- 36. Warnick GR, Nguyen T, Albers AA (1985) Comparison of improved precipitation methods for quantification of high-density lipoprotein cholesterol. Clin Chem 31: 217–222.
- 37. Friedewald WT, Levy RI, Fredrickson DS (1972) Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18: 499–502.
- 38. Wallace TM, Matthews DR (2002) The assessment of insulin resistance in man. Diabet Med 19: 527–534.
- 39. Keating BJ, Tischfield S, Murray SS, Bhangale T, Price TS, et al. (2008) Concept, design and implementation of a cardiovascular gene-centric 50 k SNP array for large-scale genomic association studies. PLoS One 3: e3583.
- 40. Timpson NJ, Nordestgaard BG, Harbord RM, Zacho J, Frayling TM, et al. (2011) C-reactive protein levels and body mass index: elucidating direction of causation through reciprocal Mendelian randomization. Int J Obes (Lond) 35: 300–308.
- 41. Beer NL, Tribble ND, McCulloch LJ, Roos C, Johnson PR, et al. (2009) The P446L variant in GCKR associated with fasting plasma glucose and triglyceride levels exerts its effect through increased glucokinase activity in liver. Hum Mol Genet 18: 4081–4088.
- 42. Rees MG, Wincovitch S, Schultz J, Waterstradt R, Beer NL, et al. (2012) Cellular characterisation of the GCKR P446L variant associated with type 2 diabetes risk. Diabetologia 55: 114–122.
- 43. Vaxillaire M, Cavalcanti-Proença C, Dechaume A, Tichet J, Marre M, et al. (2008) The common P446L polymorphism in GCKR inversely modulates fasting glucose and triglyceride levels and reduces type 2 diabetes risk in the DESIR prospective general French population. Diabetes 57: 2253–2257.
- 44. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, et al. (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81: 559–575.
- 45. Petrone A, Zavarella S, Caiazzo A, Leto G, Spoletini M, et al. (2006) The promoter region of the adiponectin gene is a determinant in modulating insulin sensitivity in childhood obesity. Obesity (Silver Spring) 14: 1498–1504.
- 46. Ferguson JF, Phillips CM, Tierney AC, Pérez-Martínez P, Defoort C, et al. (2010) Gene-nutrient interactions in the metabolic syndrome: single nucleotide polymorphisms in ADIPOQ and ADIPOR1 interact with plasma saturated fatty acids to modulate insulin resistance. Am J Clin Nutr 91: 794–801.
- 47. Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, et al. (2010) Biological, clinical and population relevance of 95 loci for blood lipids. Nature 466: 707–713.
- 48. Ebrahim S, Davey Smith G (2008) Mendelian randomization: can genetic epidemiology help redress the failures of observational epidemiology? Hum Genet 123: 15–33.
- 49. Wagner A, Simon C, Oujaa M, Platat C, Schweitzer B, et al. (2008) Adiponectin is associated with lipid profile and insulin sensitivity in French adolescents. Diabetes Metab 34: 465–471.
- 50. Yamamoto Y, Hirose H, Saito I, Tomita M, Taniyama M, et al. (2002) Correlation of the adipocyte-derived protein adiponectin with insulin resistance index and serum high-density lipoprotein-cholesterol, independent of body mass index, in the Japanese population. Clin Sci (Lond) 103: 137–145.
- 51. Hanley AJ, Connelly PW, Harris SB, Zinman B (2003) Adiponectin in a native Canadian population experiencing rapid epidemiological transition. Diabetes Care 26: 3219–3225.
- 52. Goropashnaya AV, Herron J, Sexton M, Havel PJ, Stanhope KL, et al. (2009) Relationships between plasma adiponectin and body fat distribution, insulin sensitivity, and plasma lipoproteins in Alaskan Yup'ik Eskimos: the Center for Alaska Native Health Research study. Metabolism 58: 22–29.
- 53. Kubota N, Terauchi Y, Yamauchi T, Kubota T, Moroi M, et al. (2002) Disruption of adiponectin causes insulin resistance and neointimal formation. J Biol Chem 277: 25863–25866.
- 54. Combs TP, Pajvani UB, Berg AH, Lin Y, Jelicks LA, et al. (2004) A transgenic mouse with a deletion in the collagenous domain of adiponectin displays elevated circulating adiponectin and improved insulin sensitivity. Endocrinology 145: 367–383.
- 55. Yamauchi T, Kamon J, Waki H, Imai Y, Shimozawa N, et al. (2003) Globular adiponectin protected ob/ob mice from diabetes and ApoE-deficient mice from atherosclerosis. J Biol Chem 278: 2461–2468.
- 56. Kim JY, van de Wall E, Laplante M, Azzara A, Trujillo ME, et al. (2007) Obesity-associated improvements in metabolic profile through expansion of adipose tissue. J Clin Invest 117: 2621–2637.
- 57. Shetty S, Ramos-Roman MA, Cho YR, Brown J, Plutzky J, et al. (2012) Enhanced fatty acid flux triggered by adiponectin overexpression. Endocrinology 153: 113–122.
- 58. Yamauchi T, Kamon J, Minokoshi Y, Ito Y, Waki H, et al. (2002) Adiponectin stimulates glucose utilization and fatty-acid oxidation by activating AMP-activated protein kinase. Nat Med 8: 1288–95.
- 59. Berg AH, Combs TP, Du X, Brownlee M, Scherer PE (2001) The adipocyte-secreted protein Acrp30 enhances hepatic insulin action. Nat Med 7: 947–953.
- 60. Buzzetti R, Petrone A, Zavarella S, Zampetti S, Spoletini M, et al. (2007) The glucose clamp reveals an association between adiponectin gene polymorphisms and insulin sensitivity in obese subjects. Int J Obes (Lond) 31: 424–428.
- 61. Perez-Martinez P, Lopez-Miranda J, Cruz-Teno C, Delgado-Lista J, Jiménez-Gómez Y, et al. (2008) Adiponectin gene variants are associated with insulin sensitivity in response to dietary fat consumption in Caucasian men. J Nutr 138: 1609–1614.
- 62. Karmelić I, Lovrić J, Božina T, Ljubić H, Vogrinc Ž, et al. (2012) Adiponectin level and gene variability are obesity and metabolic syndrome markers in a young population. Arch Med Res 43: 145–153.
- 63. Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, et al. (2009) Finding the missing heritability of complex diseases. Nature 461: 747–753.
- 64. Lin PI, Vance JM, Pericak-Vance MA, Martin ER (2007) No gene is an island: the flip-flop phenomenon. Am J Hum Genet 80: 531–538.
- 65. Mente A, Razak F, Blankenberg S, Vuksan V, Davis AD, et al. (2010) Ethnic variation in adiponectin and leptin levels and their association with adiposity and insulin resistance. Diabetes Care 33: 1629–1634.
- 66. Ferris WF, Naran NH, Crowther NJ, Rheeder P, van der Merwe L, et al. (2005) The relationship between insulin sensitivity and serum adiponectin levels in three population groups. Horm Metab Res 37: 695–701.
- 67. Khoo CM, Sairazi S, Taslim S, Gardner D, Wu Y, et al. (2011) Ethnicity modifies the relationships of insulin resistance, inflammation, and adiponectin with obesity in a multiethnic Asian population. Diabetes Care 34: 1120–1126.
- 68.
Robiou-du-Pont S, Bonnefond A, Yengo L, Vaillant E, Lobbens S, et al.. (2012) Contribution of 24 obesity-associated genetic variants to insulin resistance, pancreatic beta-cell function and type 2 diabetes risk in the French population. Int J Obes (Lond). In press.
- 69. Speliotes EK, Willer CJ, Berndt SI, Monda KL, Thorleifsson G, et al. (2010) Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index. Nat Genet 42: 937–948.