OUP user menu

Association between the adiponectin promoter rs266729 gene variant and oxidative stress in patients with diabetes mellitus

Sarah L. Prior, David R. Gable, Jackie A. Cooper, Stephen C. Bain, Steven J. Hurel, Steve E. Humphries, Jeffrey W. Stephens
DOI: http://dx.doi.org/10.1093/eurheartj/ehp090 1263-1269 First published online: 26 March 2009

Abstract

Aims Low levels of adiponectin are associated with type 2 diabetes and coronary heart disease (CHD). Recent evidence also suggests that low levels of adiponectin are associated with increased oxidative stress. Our aim was to examine the association between the rs266729 promoter gene variant (−11377C > G) and plasma markers of oxidative stress in diabetes subjects.

Methods and results Seven hundred and sixty-seven Caucasian subjects with diabetes were successfully genotyped (CC/CG/GG). Genotype data were analysed in relation to plasma total antioxidant status (TAOS) and Oxidized-LDL (Ox-LDL). Plasma adiponectin measurements were available in 206 samples. There was a significant association between genotype and plasma TAOS (CC: 42.1 ± 13.4% vs. CG: 42.0 ± 12.0% vs. GG: 47.9 ± 12.0%, P = 0.02; for CC/CG vs. GG, P = 0.006). With respect to Ox-LDL, CC subjects had 8% higher plasma Ox-LDL compared with CG/GG [CC vs. CG vs. GG: 48.5 (36.3–60.2) U/L vs. 44.8 (35.6–54.1) U/L vs. 44.9 (41.2–49.1) U/L, for CC vs. CG/GG P = 0.03]. For plasma adiponectin, GG subjects had the highest levels [CC vs. CG vs. GG: 8.18 (5.69–15.38) µg/mL vs. 7.12 (5.34–12.97) µg/mL vs. 11.84 (6.98–25.25) µg/mL, P = 0.09; for CC/CG vs. GG, P = 0.05].

Conclusion This study shows an association between a promoter variant in the adiponectin gene and plasma markers of oxidative stress. In line with previous studies, this work supports an antioxidant role for adiponectin which may explain its cardioprotective effect. Further prospective study is necessary to explore the effect of this gene variant in diabetes in relation to CHD risk and oxidative stress.

Keywords
  • Adiponectin
  • Oxidative stress
  • rs266729
  • Gene
  • Coronary heart disease
  • Diabetes

Introduction

Adiponectin is a true adipokine, produced solely by adipocytes in white adipose tissue exhibiting anti-diabetic, anti-inflammatory, and anti-atherogenic properties.1 Serum adiponectin levels are lower in patients with coronary heart disease (CHD)2 and prospective studies have shown low levels to be predictive of the future risk of type 2 diabetes (T2DM) and myocardial infarction.3 Hypoadiponectinaemia is therefore associated with insulin resistance,2 endothelial dysfunction,4 obesity,5 T2DM,6 CHD,3 and hypertension.7 Investigations suggest that decreased plasma adiponectin levels in the context of obesity and diabetes may contribute to vascular endothelial dysfunction, playing a pivotal role in the pathogenesis of atherosclerosis and enhancing the risk of future cardiovascular events.8,9 Therefore, the established relationship between insulin resistance and vascular endothelial cell dysfunction might be partly explained by decreased levels of adiponectin.10 Considerable interest has developed on the role that increased oxidative burden may play in many major disorders, in particular CHD, diabetes, and hypertension.11 Recent evidence suggests that low levels of adiponectin are associated with increased oxidative stress10 and in vitro studies suggest that adiponectin possesses antioxidant properties.12,13 Previously, an inverse correlation has been observed between urinary levels of 8-epi-prostaglandin F, a marker of oxidative stress, and blood levels of adiponectin in humans.14,15 Furthermore, the secretion of adiponectin is suppressed by an increased oxidative burden.16

The ADIPOQ gene encoding the adiponectin protein is located on chromosome 3q27.17 In line with plasma levels of adiponectin, the ADIPOQ gene has been identified as a susceptibility locus for the metabolic syndrome, T2DM, and cardiovascular disease.18,19 Several gene variants have been identified in ADIPOQ.20,21 The rs266729 promoter variant (−11377C > G) has recently been studied in relation to cardiovascular risk. With respect to stroke risk, Hegener and colleagues22 found a protective effect associated with homozygosity for the G allele in a nested case–control study. Within the prospective Northwick Park Heart Study II (NPHSII), the CG genotype was associated with increased CHD risk compared with homozygosity for either allele.23 Within the European multicentre HIFMECH study an increase in acute myocardial infarction risk was observed with the CG genotype.23 Within these previously published studies the association between genotype and intermediate biochemical risk-phenotypes has not been studied and thus the mechanism of associated risk remains unclear.

To date, no studies have examined the rs266729 promoter variant within a sample of subjects with diabetes. Since diabetes per se is associated with variation in plasma adiponectin levels, examining the effect of this gene in a sample of diabetes subjects (a group at high-risk for vascular disease) is desirable. Furthermore, diabetes is associated with an increase in oxidative burden,2427 which is likely to contribute to vascular disease risk. Lower serum levels of adiponectin have been associated with higher urinary isoprostanes in subjects with normal glucose tolerance, although not in a cohort with abnormal glucose tolerance.15 Therefore, variation within the adiponectin gene might have differential effects in different subject groups. The aim of this study was to explore the relationship between the rs266729 promoter gene variant and plasma markers of oxidative stress. As oxidative stress is increased in the presence of atherosclerosis,28 we reasoned that it would also be essential to study the association after grouping the subjects by CHD status as previously described.29,30

Methods

Subjects

Ethical approval was granted by the institutional ethics committee and all subjects gave written informed consent before recruitment. Patients were recruited from the University College London Diabetes and Cardiovascular Study (UDACS) described elsewhere.29,30 Briefly, this comprises 1011 consecutive subjects recruited from the diabetes clinic at University College London Hospitals (UCLH) NHS Trust between the years 2001 and 2002. All patients had diabetes according to WHO criteria.31 Within this study we chose to focus only on Caucasian subjects because of potential differences in genotype distribution between different ethnic backgrounds. Of these, genotype data were available for 767 (98.3% of 780 subjects). Of the 767 subjects, a genotype CHD status was available on 764 subjects. Of these subjects plasma TAOS was measured successfully in 739 and Ox-LDL in 494 subjects. This was because of limited availability of plasma and assay failure. The presence of CHD was recorded if any patient had positive coronary angiography/angioplasty, coronary artery bypass, cardiac thallium scan, exercise tolerance test, myocardial infarction, or symptomatic/treated angina. Any individual who was asymptomatic or had negative investigations was categorized as ‘no CHD’. Smoking status was defined as current, never, or ex-smokers. The latter comprised of those who had stopped smoking for more than 12 months. Current smokers included subjects who had stopped smoking within a 12-month period. Therefore this is a case–control study. None of the subjects were knowingly taking any form of vitamin supplementation. Waist circumference was not recorded on this sample. Plasma samples were collected within a 12-month period and stored immediately at −80°C. Samples were collected during routine diabetes clinic visit. None of the samples were fasting at the time of collection.

Genotyping for the rs266729 gene variant

Genomic DNA was extracted from 5 mL ethylenediaminetetraacetic acid (EDTA) blood samples, and variants were genotyped using the method described by Gable et al.23 Briefly, analysis was conducted using polymerase chain reaction amplification and restriction fragment length polymorphism analysis. Primer sequences were: forward-CATCAGAATGTGTGGCTTGC and reverse-AGAAGCAGCCTGGAGAACTG followed by digestion with HhaI. For all variants, genotype was confirmed by two independent technicians and any discrepancies were resolved by repeat genotyping.

Measurement of plasma total antioxidant status

Plasma total anti-oxidant status (TAOS), which is inversely related to oxidative stress, was measured by Sampson's modification of Laight's photometric microassay,32 using 2.5 µL citrated plasma samples in 96-well enzyme-linked immunosorbent assay (ELISA) plates. The TAOS of plasma was determined by its capacity to inhibit the peroxidase-mediated formation of the 2,2-azino-bis-3-ethylbensthiazoline-6-sulfonic acid (ABTS+) radical. There are two arms to the assay, a control arm and a test arm. In the control arm phosphate-buffered saline is used instead of plasma. The difference in absorbance (control absorbance − test absorbance) divided by the control absorbance (expressed as a percentage) was used to represent the percentage inhibition of the reaction. The inter- and intra-assay coefficients of variation were 10.1% and 4.3%, respectively. Previously, we have shown that baseline plasma TAOS is associated with prospective risk and has a good correlation with plasma F2-isoprostanes.29

Measurement of plasma oxidized-LDL

Plasma oxidized-LDL (Ox-LDL) was measured using a commercially available ELISA kit supplied by Mercodia (Uppsala, Sweden). In this assay a monoclonal antibody is directed against antigenic determinants in the Ox-LDL molecule (mAB-4E6).

Measurement of adiponectin

Plasma total adiponectin was measured using a commercially available kit. This was the Quantikine Human Adiponectin/Acrp30 immunoassay supplied by R&D Systems (Abingdon, UK).

Statistical analysis

Analysis was performed using SPSS (SPSS Inc., Chicago, IL, USA) and STATA (STATA Corporation, TX, USA). Data are reported for those individuals among whom high-throughput genotyping was available. Results are presented as mean ± standard deviation. For data that had a normal distribution after log transformation, the geometric mean and approximate standard deviation is shown. This included systolic and diastolic blood pressure, body mass index, C-reactive protein (CRP), HbA1c, and triglyceride. For duration of diabetes and Ox-LDL, the data could not be transformed to a normal distribution and so the data are expressed as median and interquartile range. Deviations from Hardy–Weinberg equilibrium were considered using χ2 tests. Allele frequencies are shown with the 95% confidence interval (CI). Analysis of variance (ANOVA) was used to assess the association between genotypes and baseline characteristics for data with a normal distribution or after log-transformation. For duration of diabetes, Ox-LDL and adiponectin, data were analysed by the Kruskal–Wallis or Mann–Whitney U tests for data that did not have a normal distribution. Stepwise regression was initially performed to look for the final factors influencing TAOS (glucose) or Ox-LDL (LDL, triglyceride). Models used forward selection with a significance level of 5%. One hundred bootstrap samples using 18 candidate variables were performed to test stability of the models and ensure that the important confounders were identified. Analysis of covariance (ANCOVA) was then performed to test the association between genotype and TAOS or Ox-LDL after adjustment for these confounders. In addition, recessive and dominant genotype effects were included in the stepwise models to assess which genetic model best fitted the data. Model assumptions were checked by normal probability plots, plotting residuals against fitted values, and testing for nonlinearity using the ‘nlcheck’ command in Stata. For TAOS Box–Cox transformation improved non-normality and non-constant variance and so analysis was based on the transformed data. For Ox-LDL there was some evidence of non-normality which was not corrected by transformation, and the analyses have therefore been repeated using an ordinal logistic regression model which confirmed the results. χ2 Tests were used to compare differences in categorical variables by genotype. In all cases a P-value of <0.05 was considered statistically significant. Two-sided statistical testing was performed. The interaction between genotype and CHD in determining plasma TAOS/Ox-LDL was assessed using linear regression where the interaction was compared with the individual effects of these variables (likelihood ratio test). This analysis has been run with and without the inclusion of potential confounders. Correction for multiple comparisons was not applied to the results, because the study design was predominantly ‘hypothesis testing’. While making such an adjustment reduces the type I error, it leads to increase in the type II error, and fewer errors of interpretation occur when no adjustment is made.33

Results

Of the 780 Caucasian subjects, genotype data were available for 767 (98.3%). Of these 78% (n = 598) had T2DM. The genotype distribution for the ADIPO −11377C>G variant was in Hardy–Weinberg equilibrium (CC/CG/GG: 427/298/42, χ2 = 1.15, P = 0.29) with a C allele frequency of 0.75 (0.73–0.77) and a G allele frequency of 0.249 (0.23–0.27). As shown in Table 1, there was no statistically significant association between genotype and routine baseline clinical and biochemical markers. Importantly, as observed in Table 2, no statistically significant gender difference or differences by CHD status and medication use were observed. There was no CHD risk associated with this genotype if a dominant (P = 0.59), heterozygous (P = 0.41), or recessive (P=0.56) disease module was considered.

View this table:
Table 1

Baseline differences in subjects by the rs266729 promoter variant (−11377C > G)

TraitAdiponectin −11377C>GProbability three-way
CC (n = 427)CG (n = 298)GG (n = 42)
Age (years)62.6 (13.5)62.6 (13.7)62.0 (13.6)0.96
Duration (years)*11 (4–17)11 (4–14)11 (4–17)0.20
Systolic blood pressure (mmHg)140 (21)137 (20)142 (19)0.08
Diastolic blood pressure (mmHg)80 (11)78 (11)79 (11)0.11
Body mass index (kg/m2)28.8 (5.6)28.3 (5.5)28.2 (5.0)0.46
HbA1c (%)7.8 (1.6)7.8 (1.6)7.7 (1.4)0.68
Random plasma glucose (mmol/L)9.7 (4.8)9.7 (5.6)9.8 (4.6)0.99
LDL cholesterol2.9 (0.9)2.8 (0.9)2.7 (1.0)0.48
Triglycerides1.7 (1.1)1.6 (0.9)1.5 (0.7)0.21
CRP (mg/L)1.56 (1.39)1.61 (1.32)1.66 (1.48)0.86
TAOS (%)42.0 (13.4)42.0 (12.0)47.9 (12.0)0.02
Ox-LDL (U/L)*48.5 (36.3–60.2)44.8 (35.6–54.1)44.9 (41.2–49.1)0.07§
Males, % (n)61.8 (264)60.9 (181)45.2 (19)0.15
Type 2 diabetes, % (n)78.9 (337)75.5 (225)83.3 (35)0.40
Current smokers, % (n)16.2 (69)18.1 (54)14.3 (6)0.75
CHD, % (n)21 (90)18.8 (56)23.8 (10)0.65
ACE-inhibitor, % (n)43.8 (187)47 (140)38.1 (16)0.47
Aspirin, % (n)45.4 (194)43.3 (129)52.4 (22)0.53
Insulin, % (n)40.5 (173)48.3 (144)40.5 (17)0.12
Statin, % (n)28.6 (122)23.2 (69)31 (13)0.21
  • Ox-LDL, Oxidized-LDL; TAOS, total antioxidant status; CHD, coronary heart disease; ACE-inhibitor, angiotensin-converting enzyme-inhibitor; CRP, C-reactive protein; HbA1c, Haemoglobin A1c.

  • *Median and interquartile range shown for duration of diabetes and Ox-LDL. Analysis performed by ANOVA after appropriate transformation of non-normally distributed data and by Kruskal–Wallis for duration of diabetes and Ox-LDL. χ2 Test was used to compare groups.

  • Log-transformed data. Mean and standard deviation shown or geometric mean and approximate standard deviation for log-transformed data.

  • P = 0.01 after adjustment of confounders.

  • §P = 0.11 after adjustment of confounders.

View this table:
Table 2

Baseline differences in subjects by coronary heart disease (CHD) status

TraitNo CHD (n = 609)CHD (n = 155)P-value
Age (years)61.1 (14.0)68.6 (10.2)<0.001
Duration (years)*11 (5–21)11 (6–18)0.75
Systolic blood pressure (mmHg)139 (20)138 (22)0.91
Diastolic blood pressure (mmHg)70 (9)77 (12)0.002
Body mass index (kg/m2)28.3 (5.5)29.5 (5.4)0.02
HbA1c (%)7.8 (1.6)7.6 (1.5)0.15
Random plasma glucose (mmol/L)9.7 (4.9)9.6 (4.3)0.84
LDL cholesterol2.9 (0.9)2.4 (0.9)<0.001
Triglycerides1.6 (1.4)1.9 (1.0)0.007
CRP (mg/L)1.53 (1.32)1.80 (1.57)0.04
TAOS (%)42.6 (13.2)41.5 (13.2)0.35
Ox-LDL (U/L)*47.2 (36.7–57.8)45.8 (35.6–57.2)0.58
Males, % (n)59.2 (361)67.5 (104)0.05
Type 2 diabetes, % (n)73.2 (446)95.5 (148)<0.001
Current smokers, % (n)17.0 (103)12.3 (19)0.09
ACE-inhibitor, % (n)42.3 (258)56.0 (86)0.003
Aspirin, % (n)37.8 (230)75.0 (116)<0.001
Insulin, % (n)45.8 (279)35.0 (54)0.02
Statin, % (n)18.2 (111)60.6 (94)<0.001
Adiponectin −11377C > G, % (n)55.0/39.7/5.3 (335/242/32)57.4/36.1/6.5 (89/56/10)0.65
  • Ox-LDL, Oxidized-LDL; TAOS, total antioxidant status; CHD, coronary heart disease; ACE-inhibitor, angiotensin-converting enzyme-inhibitor; CRP, C-reactive protein; HbA1c, Haemoglobin A1c.

  • CHD status was available for 764 of the subjects with genotype data.

  • *Median and interquartile range shown for duration of diabetes and Ox-LDL. Analysis performed by ANOVA after appropriate transformation of non-normally distributed data and by Kruskal–Wallis for duration of diabetes and Ox-LDL. χ2 Test was used to compare groups. For CC vs. CG/GG, the OR for CHD was 1.1 (95% CI 0.77–1.57), P = 0.59. For CC/CG vs. GG, the OR for CHD was 0.8 (95% CI 0.39–1.67), P = 0.56. For a heterozygous model, CG vs. CC/GG, the odds ratio for CHD was 1.17 (95% CI 0.81–1.68), P = 0.41.

  • Log-transformed data. Mean and standard deviation shown or geometric mean and approximate standard deviation for log-transformed data.

Association between the −11371C > G variant and plasma total antioxidant status

As described, plasma TAOS was measured successfully in 739 of the above samples (CC/CG/GG: 404/294/41). As described previously,29,30 plasma TAOS correlated positively with plasma HDL-cholesterol, and negatively with triglyceride and glucose (correlation coefficient r = 0.13, −0.14, and −0.12; all P < 0.05). Glucose was found to be the most important confounder, being selected 89 times from the 100 bootstrap samples. No other variable was selected consistently. In the overall group, there was a significant association between genotype and plasma TAOS (CC: 42.1 ± 13.4% vs. CG: 42.0 ± 12.0% vs. GG: 47.9 ± 12.0%; P = 0.02). The recessive genotype effect was selected 73 times in the 100 bootstrap samples compared with just four selections which included the dominant effect. The GG genotype was associated with a mean 14% increase in plasma TAOS compared with C-allele carriers implying that this genotype may have a protective role (for CC/CG vs. GG, P = 0.006). After adjustment of plasma TAOS for glucose, the association between genotype and plasma TAOS remained unchanged (for CC vs. CG vs. GG, P = 0.01; for CC/CG vs. GG, P = 0.002).

As oxidative stress is increased in the presence of atherosclerosis,28 we reasoned that it would be important to study the association after grouping the subjects by CHD status as initially planned. Of note plasma TAOS was not statistically different between those with and without CHD (42.6 ± 13.2% vs. 41.5 ± 13.2%, P = 0.35) in the whole group (as previously described, an association with CHD risk in males in this cross-sectional sample and a prospective sample has been observed29). Table 2 shows the baseline differences in the sample by CHD status. After stratifying by CHD status, the association between genotype and plasma TAOS remained significant in those subjects without CHD. For CC vs. CG vs. GG: 42.4 ± 13.3% vs. 41.8 ± 13.0% vs. 50.9 ± 10.4%, P = 0.0003 (CC/CG vs. GG, P = 0.0003). Within those subjects with CHD, no association was observed (CC vs. CG vs. GG: 40.8 ± 13.9% vs. 42.6 ± 12.3% vs. 38.4 ± 12.2%, P = 0.56). We observed a significant interaction between genotype (CC/CG vs. GG) and CHD status in determining plasma TAOS (P = 0.02) as shown in Figure 1.

Figure 1

Plasma total antioxidant status (TAOS) grouped by the rs266729 promoter variant (−11377C > G) and coronary heart disease (CHD) status. Mean and standard deviations shown. Numbers of subjects are shown at the base of each column.

Association between the −11371C > G variant and plasma oxidized-LDL

As previously described, within this sample,29 plasma Ox-LDL was not significantly associated with pharmacotherapy but was correlated with LDL, triglyceride, and HDL (correlation coefficient r = 0.32, 0.26, −0.13, respectively, P < 0.05). LDL and triglyceride were confirmed as important covariates for ox-LDL occurring in 100% and 97% of the bootstrap samples, respectively. HDL occurred only in 20% of the bootstrap samples and was not used as a covariate in the ANCOVA. The dominant genotype effect was selected 54 times in the 100 bootstrap samples compared with 19 selections which included the recessive effect. In the overall group, after adjusting for the above correlates of Ox-LDL, there was an association between genotype and plasma Ox-LDL [CC vs. CG vs. GG: 48.5 (36.3–60.2) U/L vs. 44.8 (35.6–54.1) U/L vs. 44.9 (41.2–49.1) U/L, P = 0.07; after adjustment for LDL and triglyceride, P = 0.11]. As observed, CC subjects had higher plasma Ox-LDL compared with CG/GG (P = 0.03; after adjustment P = 0.07). Within the group there was no statistical difference in Ox-LDL between those with and without CHD (P = 0.58, Table 2). In the subjects without CHD, the CC genotype was associated with highest mean plasma Ox-LDL but this did not reach statistical significance [CC vs. CG vs. GG: 48.5 (37.6–60.1) U/L vs. 44.6 (35.0–54.9) U/L vs. 45.8 (42.6–50.0) U/L, P = 0.07; for CC vs. CG/GG, P = 0.04]. Using the ordinal logistic regression model gave similar results to the ANCOVA (adjusted for LDL and triglyceride: CC vs. CG vs. GG, P = 0.08; CC vs. CG/GG, P = 0.04). In those with CHD, no association was observed with plasma Ox-LDL [CC vs. CG vs. GG: 48.0 (32.9–60.2) U/L vs. 44.9 (37.1–52.2) U/L vs. 42.5 (35.8–45.5) U/L; P = 0.55]. No significant interaction was observed by genotype (CC vs. CG/GG) and CHD in determining plasma Ox-LDL (P = 0.51).

Association of genotype with plasma adiponectin

As a result of the associations observed between genotype and plasma markers of oxidative stress in subjects without CHD, we went on to study the association between genotype and plasma adiponectin in this group of subjects. Within UDACS, plasma was available for the measurement of adiponectin in 206 samples drawn from those without CHD. These measurements were performed as part of another unpublished study. No further samples were available to measure in the reminder of the cohort. Within this smaller sample with adiponectin measurements, no consistent covariates were found by the stepwise models among the baseline variables in relation to adiponectin. Of note, the mean age, plasma CRP, and triglyceride was higher and duration of diabetes lower in those with plasma adiponectin measurements compared with those without (data not shown). Furthermore, the group with measurements contained more patients with T2DM and a higher proportion of subjects receiving treatment with angiotensin-converting enzyme-inhibitors and aspirin. The recessive genotype effect was selected 26 times in the 100 bootstrap samples compared with 15 selections which included the dominant effect. GG subjects had the highest adiponectin level [CC vs. CG vs. GG: 8.18 (5.69–15.38) µg/mL vs. 7.12 (5.34–12.97) µg/mL vs. 11.84 (6.98–25.25) µg/mL; P = 0.09; for CC/CG vs. GG P = 0.05].

Discussion

Within this manuscript we describe for the first time the association between a variant within the adiponectin gene and a plasma marker of oxidative stress in vivo. Homozygosity for the G allele of the rs266729 variant was associated with increased antioxidant capacity. Furthermore, this effect was most apparent in those subjects without clinically manifest CHD, with the GG genotype being associated with an approximately 20% higher mean plasma TAOS. Within those subjects with CHD, no genotype association with plasma TAOS was observed. Therefore, in the presence of CHD with its increased oxidative burden and its cluster of associated risk factors it may be that the modest genotype association is overwhelmed.28 As described, we also observed a modest association between the G allele with lower plasma Ox-LDL (after adjustment for confounding factors). This is in keeping with the data for plasma TAOS. Following stratifying by CHD status, the association was of borderline significance in those without CHD and not significant in those with CHD. This lack of association with plasma Ox-LDL may be related to the observation that many of these subjects were taking statin therapy, which is associated with a reduction in Ox-LDL.34,35

There are limitations to our study. We have chosen from the outset to examine one gene variant in relation to intermediate biochemical markers of oxidative stress. We have focussed on plasma TAOS and Ox-LDL as intermediate biochemical phenotypes. There are limitations to these measurements as discussed elsewhere.29,36 Further work should be performed to look at the other variants within the promoter region of the ADIPOQ gene to allow haplotype-based analysis. This would also allow linkage disequilibrium across the region to be examined. At first sight, the lack of any significant difference in genotype distribution between those with and without CHD might appear to conflict with the genotype–phenotype association observed. However, there are several possible explanations for this. First, prospective gene-association studies are more powerful than case–control studies,23 as in our case. Secondly, within the diabetes sample increased obesity, oxidative burden, inflammation, and hyperglycaemia might all overwhelm the adiponectin genotype ‘strength of signal’ and hence the genotype effect on CHD risk might be lower than in a non-diabetic sample. Thirdly, case–control cross-sectional studies are prone to intrinsic bias, for example because of possible altered rates of disease progression, subsequent progression of secondary phenotypes, or genotype associations with death or treatment changes. Indeed, the presence of the C allele might be associated with both earlier disease presentation and earlier death in some, subsequently balanced by more aggressive secondary prevention strategies. Such influences are well-recognized confounders.23,37,38 Another limitation within this study is the limited number of samples available where adiponectin could be measured. This should be investigated in other cohorts if available.

The association between the rs266729 variant and plasma TAOS was independent of other risk factors for CHD including CRP. This suggests that this may be independent of any direct anti-inflammatory effects associated with adiponectin.1 We have not measured adiponectin for all the subjects within UDACS, however in a subset of the subjects without CHD there was a non-significant trend (P-value of 0.05) for the adiponectin levels to be highest in the GG subjects. We therefore speculate that the observed gene association with plasma TAOS may be via increased plasma adiponectin but cannot confirm this within our current study sample. Of interest, adiponectin is known to enhance the transcription of other genes which are important within diabetes. These include those involved in fatty acid metabolism, most notably peroxisome proliferator-activated receptor-α (PPAR-α).39 Activation of PPAR-α leads to an increase in levels of molecules involved in free fatty acid transport, such as CD36, and energy dissipation such as uncoupling protein-2 (UCP2),40 which also increases fatty acid oxidation. Furthermore, UCP2 is a regulator of mitochondrial reactive oxygen species generation29,30 and increased UCP2 expression is associated with a reduction in oxidative stress.41 Studies have also shown that the cardiac tissue from adiponectin knockout mice have higher levels of superoxide and peroxynitrite following myocardial ischaemia compared with wild-type.12 Following the administration of the globular domain of adiponectin prior to reperfusion, the knockout mice showed reduced myocardial ischaemia-induced superoxide production and peroxynitrite formation, and reversed proapoptotic and infarct enlargement.12 Similarly, in adult rats fed a high-fat diet, studies using isolated aortic segments demonstrate that adiponectin protects the endothelium against hyperlipidaemic injury by multiple mechanisms, including promoting endothelial nitric oxide synthase activity, inhibiting inducible nitric oxide synthase activity, preserving bioactive nitric oxide, and attenuating oxidative/nitrative stress.13 These effects may therefore be directly related to lower levels of adiponectin. Of importance, in vitro, increased mRNA levels of plasminogen activator inhibitor-1,42 leptin,43 and tumour necrosis factor-α44 inhibit expression of adiponectin.45 Thus it may be that under proinflammatory conditions, adiponectin expression and plasma levels reduce resulting in loss of antioxidant potential and increased oxidative stress.

In summary, this study shows an association between a promoter variant in the adiponectin gene and plasma markers of oxidative stress. In line with previous studies, the current work supports an antioxidant role for adiponectin which may explain its cardioprotective effect. Further prospective study is necessary to explore the effect of this gene variant in diabetes in relation to CHD risk and oxidative stress.

Funding

Diabetes UK (BDA: RD01/0001357 to J.W.S.) and the creation of UDACS, the British Heart Foundation (RG2005 014 to S.E.H.).

Conflict of interest: none declared.

References

View Abstract