Genetic and Environmental Determinants of Dimethylarginines and Association With Cardiovascular Disease in Patients With Type 2 Diabetes

Asymmetric dimethylarginine (ADMA) is considered an independent risk factor for cardiovascular (CV) events and mortality in patients with a range of CV risk (1). Although ADMA plasma levels were reported to be elevated in patients with type 2 diabetes (2) and insulin resistance (3), there are discrepancies concerning the association of ADMA with glycemic control and its prognostic value regarding CV disease (CVD) in diabetes patients (4–8). Results from prospective studies reporting that high levels of ADMA predicted CV events in patients with diabetes mellitus (9,10) could not be confirmed in population-based prospective cohorts (7). A possible explanation for these conflicting results may be that activity of the enzyme responsible for degradation of ADMA, dimethylarginine dimethylaminohydrolase (DDAH), is influenced by diabetes and diabetes therapy (11,12). Variation in the DDAH1 and/or DDAH2 genes were associated with differences in ADMA levels in diabetes subjects (13) and with insulin sensitivity in nondiabetic subjects (14) and predicted major CV events in diabetes subjects (15).

Recent evidence indicates that symmetric dimethylarginine (SDMA) might also play a role in diabetes mellitus and CVD. In patients with type 2 diabetes mellitus, SDMA was inversely associated with glycemic control and with insulin resistance in nondiabetic Caucasians (5,16). Furthermore, prospective studies found that SDMA was an independent predictor for mortality in patients with CVD (17,18). While the majority of ADMA is metabolized by DDAH, SDMA is mainly cleared via the kidneys. However, both dimethylarginines can also be degraded by the enzyme alanine–glyoxylate aminotransferase (AGXT2) (19,20).

There appears to be a complex interaction between dimethylarginines, the genes involved in their metabolism, glycemic control, and CVD in patients with type 2 diabetes mellitus. Therefore, the aims of this study were to 1) investigate determinants of dimethylarginine plasma levels in a representative sample of older men and women with type 2 diabetes mellitus; 2) explore the influence of single nucleotide polymorphisms (SNPs) in the DDAH1, DDAH2, and AGXT2 genes on plasma levels of dimethylarginines; and 3) determine whether plasma dimethylarginines and/or dimethylarginine-related SNPs are associated with pre-existing or incident CVD in type 2 diabetes.

The Edinburgh Type 2 Diabetes Study (ET2DS) is a prospective, population-based cohort study of 1,066 men and women aged 60–75 years with established type 2 diabetes mellitus recruited between 2006 and 2007 in the Lothian region of Scotland, U.K. The design of the ET2DS and details of the physical examination undertaken at baseline and 1 year after recruitment have been described previously (21). The ET2DS population has been shown to be representative of patients aged 60–75 years with type 2 diabetes living in Lothian (22), and subjects returning after 1 year (n = 939) are similar to the original study population (23). Genotyping for the current analysis was performed in participants attending the 1-year examination using DNA extracted from whole blood samples taken at baseline (n = 935), and plasma L-arginine and dimethylarginines were measured in 783 of these subjects (data missing at random due to insufficient plasma in stored aliquots). All ET2DS participants were followed up after 4 years for CV events (in 2010–2011). Use of routine data sources (record linkage) and general practitioner (GP)/hospital notes as well as direct patient contact ensured that follow-up for CVD events included all ET2DS participants.

Mass spectrometric determination of L-arginine, ADMA, and SDMA plasma samples was performed as described by Schwedhelm et al. using a validated high-throughput liquid chromatography–mass spectrometry/mass spectrometry assay (24).

Prevalent CVD at baseline was recorded according to predefined criteria using a combination of subject report of a doctor diagnosis of disease, positive World Health Organization (WHO) chest pain questionnaire, and positive Edinburgh Claudication questionnaire and electrocardiogram (ECG) findings (Minnesota coded) together with hospital discharge codes from record linkage at the Information and Statistics Division of the National Health Service for Scotland, details of which have been described previously (22). Four years after recruitment, participants were followed up for new CV events using a combination of repeat self-completion (or GP) questionnaire, repeat Information and Statistics Division record linkage for hospital discharge and death certificate data (both ICD-10 coded), and review of clinical case notes as required. Criteria for fatal and nonfatal events were as follows: myocardial infarction (MI), 1) ICD-10 code for new MI on hospital discharge/death record, dated after baseline, plus either subject report of a doctor diagnosis of MI, positive WHO chest pain questionnaire for MI, report of MI on GP questionnaire (all with date consistent with ICD-10-coded event), or ECG codes for MI that were not present at baseline or 2) clinical criteria for MI met following scrutiny of hospital and/or GP notes; new angina (in subjects without a diagnosis of angina at baseline), 1) ICD-10 code for angina as primary diagnosis on hospital discharge record, dated after baseline, 2) at least two of (a) subject report of a doctor diagnosis (self-report) of angina or of starting angina medication since baseline, (b) ECG codes for ischemia that were not present at baseline, and (c) positive WHO chest pain questionnaire, or 3) clinical diagnosis of angina on scrutiny of hospital notes; stroke, 1) ICD-10 code for stroke as primary diagnosis on hospital discharge/death record, dated after baseline, or 2) clinical criteria for stroke met on scrutiny of clinical notes in subjects with either self-report of stroke or with nonprimary ICD-10 hospital discharge/death code for stroke; transient ischemic attack (TIA), 1) ICD-10 code for TIA as primary diagnosis on hospital discharge record or 2) clinical criteria for TIA met on scrutiny of clinical notes in subjects with either self-report of stroke or with nonprimary ICD-10 hospital discharge code for stroke or TIA.

Genotyping was performed at the Genetics Core of the Welcome Trust Clinical Research Facility in Edinburgh. SNPs of the DDAH1 gene, the DDAH2 gene, and the AGXT2 gene were determined using the TaqMan OpenArray Genotyping System (Applied Biosystems Inc., Carlsbad, CA). The sample success rate and SNP success rate were 96.5% and 100%, respectively. In total, 96.5% of the genotypes could be determined. Using the tagger algorithm implemented in Haploview 4 (25), tag SNPs across DDAH1, DDAH2, and AGXT2 genes, including the known or expected promoter regions, were selected. The selection was based on the linkage disequilibrium patterns observed in Caucasian samples genotyped within the International HapMap Project. From each linkage disequilibrium block identified, 2–3 tag SNPs with an r² of at least 0.8 were chosen to cover the whole locus, including SNPs reported to be associated with diabetes (rs2474123 and rs2935) (13). Linkage equilibrium was assessed using the locus zoom software. Haplotypes were generated using the THESIAS software and analyzed for possible effects. For further analyses, only those haplotypes with an estimated effect were considered.

Normality of L-arginine and dimethylarginines was assessed using the Kolmogorov–Smirnov test. L-arginine, SDMA, and ADMA were naturally log transformed for parametric tests. One L-arginine value of 252.7 µmol/L was excluded as an outlier. Differences between groups were calculated using the χ² test for categorical data and the t test, Mann–Whitney test, or Kruskal–Wallis test for continuous data. Nonparametric Spearman’s correlation coefficients and multiple linear regression analysis with stepwise backward selection were used to assess determinants of dimethylarginines. To determine the associations of clinical variables and SNPs with prevalent and incident CVD, logistic regression analyses were performed. To correct for multiple testing in the SNP analyses, Bonferroni correction was used, which set the P value for statistical significance at P ≤ 0.0021. A P value of <0.05 was considered statistically significant for all other analyses. Statistical analyses were performed using SPSS 17.0/19.0 for Windows (SPSS Inc., Chicago, IL).

Clinical characteristics of participants attending the year 1 research clinic (the total study population used in the current analysis; n = 939) and of the study population with plasma dimethylarginine available (n = 783) together with their associations with dimethylarginines are summarized in Table 1.

Median (interquartile range [IQR]) plasma ADMA levels were statistically significantly higher in subjects with history of stroke (0.50 [0.40–0.57] vs. 0.45 [0.39–0.51] µmol/L; P = 0.01), but no further associations with CVD were apparent. Significantly higher median (IQR) plasma SDMA levels were found in subjects with MI (0.55 [0.46–0.66] vs. 0.48 [0.41–0.57] µmol/L; P < 0.001), angina (0.52 [0.43–0.64] vs. 0.48 [0.41–0.56] µmol/L; P < 0.001), stroke (0.57 [0.48–0.74] vs. 0.49 [0.42–0.58] µmol/L; P < 0.001), and intermittent claudication (0.56 [0.45–0.70] vs. 0.49 [0.42–0.58] µmol/L; P = 0.004), with a maximum difference of 0.08 µmol/L between subjects with and without stroke. Neither of the dimethylarginines was significantly associated with categories of diabetes treatment (diet-only treatment, oral agents only, or insulin therapy), although a trend toward slightly higher SDMA levels in subjects with insulin treatment was apparent (median [IQR] SDMA 0.52 [0.42–0.66] vs. 0.49 [0.42–0.58]; P = 0.051). Associations of ADMA and SDMA with prevalent CVD lost statistical significance after adjusting for age, sex, and renal function. In the prospective analysis, neither of the dimethylarginines was significantly associated with incident coronary artery disease or cerebrovascular disease (Table 2).

Plasma ADMA was related to SNPs in the DDAH1 gene but not to SNPs in the DDAH2 or AGXT2 genes, whereas plasma SDMA was related to SNPs in the AGXT2 gene only (Table 3). SNPs that were significantly associated with dimethylarginines showed an increase in plasma levels of the respective dimethylarginine from the major to the minor allele, with a maximum increase in ADMA of 0.06 µmol/L for the DDAH1 rs1554597 SNP and a maximum increase of SDMA of 0.17 µmol/L for the AGXT2 rs28305 SNP.

In univariate regression analysis, SDMA was significantly associated with age, sex, creatinine, L-arginine, diastolic blood pressure, waist–hip ratio, diabetes duration, and prevalent CVD, and ADMA was significantly associated with age, sex, creatinine, fasting glucose, L-arginine, and triglycerides. SNPs were entered into the regression models individually according to P value and as common haplotypes restricted to SNPs under linkage equilibrium (DDAH1: rs1554597, rs11161614, rs1146381, and the TTG haplotype; AGXT2: rs28305, rs16899974, rs344156, and the GCG haplotype). Table 4 displays the final models for the multivariate analyses. Addition of further SNPs or analysis by haplotypes did not improve the overall models.

The distribution of genotypes did not differ significantly between subjects with or without prevalent CVD for the SNPs we investigated (Table 3). Only weak associations between SNPs in the DDAH1 and AGXT2 genes and CVD were apparent in the logistic regression analysis, and these did not reach statistical significance after correction for multiple testing. In subgroup analyses, carriers of the AGXT2 rs28305 C allele more frequently had a history of intermittent claudication. In the analysis of this SNP according to genotype, the frequency of prevalent intermittent claudication was 4.6% in the GG genotype, 11.9% in the GC genotype, and 33.3% in the CC genotype (P < 0.001). In the allelic analysis, intermittent claudication was present in 12.5% of the carriers of the C allele as compared with 5.4% in carriers of the G allele (P < 0.001). The age-, sex-, and creatinine-adjusted odds ratio (95% CI) for carriers of the C allele was 2.50 (1.56–4.01) (P < 0.001) and the odds ratio for the GC genotype versus GG genotype was 2.80 (1.60–4.92) (P < 0.001) and the CC versus GG genotype was 12.86 (1.10–150.05) (P = 0.04).

In our cohort of older adults with type 2 diabetes, we observed a lack of association of dimethylarginines or SNPs in the genes of their metabolizing enzymes with prevalent or incident CVD after adjustment for relevant covariates. However, subgroup analysis revealed an interaction between plasma SDMA levels, the AGXT2 rs28305 SNP, and intermittent claudication. Furthermore, we observed strong association of SNPs in the DDAH1 gene with plasma levels of ADMA and of SNPs in the AGXT2 gene with SDMA.

The lack of association of ADMA with CVD in our cohort is in contrast to a number of previous studies. Krzyzanowska et al. reported that elevated ADMA levels were associated with macrovascular disease in diabetic patients (26). Furthermore, ADMA was predictive of future CV events and all-cause mortality in two different cohorts of diabetes patients (9,10). However, data from two larger study collectives recently published showed ADMA levels to be associated with CV events and all-cause mortality only in nondiabetic patients (7,15). In our large cohort of diabetes patients, only weak associations of ADMA with clinical features of diabetes and no association with medication was apparent, making it less likely that glycemic control or medication directly affects ADMA and its associated risk. However, there is some evidence that early stages of diabetic nephropathy with renal vasodilatation and hyperfiltration could lower circulating ADMA, while in later stages of renal insufficiency, ADMA levels rise (8,27). The predictive value of ADMA could therefore be highly dependent on the selection of study subjects. In this context, another possible bias is that we investigated older, obese subjects, and results might be different in younger cohorts with lower BMI.

The association between SDMA, rs28305, and CVD in our study adds to increasing evidence that links SDMA with macrovascular disease. In previous studies, SDMA was an independent predictor of mortality or CVD events in patients with prevalent CVD (17,18,28). To date, no data are available on the associations of SDMA or SNPs in AGXT2 and CVD in patients with diabetes. Therefore, it is interesting to note that in our cohort, SDMA was more strongly related to CVD than ADMA in unadjusted analyses. Furthermore, the independent association of AGXT2 rs28305 with intermittent claudication strengthens the hypothesis that SDMA might be both a marker of increased risk and involved in the mechanisms leading to vascular disease, which are most likely nitric oxide synthase independent. There is some evidence linking SDMA to an increased release of reactive oxygen species as well as proinflammatory cytokines (29,30). Because the development of macroangiopathy in type 2 diabetes is associated with chronic subclinical inflammation (31), it would be important to investigate the molecular mechanisms that connect SDMA, inflammation, and macrovascular disease in diabetes patients.

Consistent with our results, Abhary et al. reported that several tag SNPs in the DDAH1 gene were associated with ADMA levels in an elderly cohort of patients with type 2 diabetes (13). Moreover, they found a significant, but weaker, association of one tag SNP in the DDAH2 gene (rs313183) and ADMA levels. We cannot rule out that the lack of association between DDAH2 and dimethylarginine levels in our group results from the selection of the SNPs, which did not include rs31318. However, subgroup analysis by retinopathy status among the participants of the above-mentioned study revealed that the majority of the SNPs were associated with ADMA levels in patients without documented diabetic retinopathy. Therefore genetic differences or differences in the clinical manifestation of the disease between the cohorts could have contributed to the discrepancies between the observed associations. Nonetheless, despite the hypothesis that DDAH2 might be the enzyme responsible for degrading plasma ADMA owing to its high expression in vascular tissue (32), the clinical data indicate a greater contribution of DDAH1 in regulating circulating ADMA levels in patients with type 2 diabetes. This observation is in line with our previous genome-wide association study in a population-based cohort, in which DDAH1 emerged as the only gene locus significantly associated with ADMA in multivariate analysis (33).

Another enzyme capable of degrading dimethylarginines in vivo is AGXT2 (19,20). In our study, SNPs in the AGXT2 gene were not related to ADMA but only to SDMA plasma concentrations. AGXT2 rs28305 was significantly associated with SDMA plasma concentrations after adjustment for clinical variables, including renal function. This is an interesting observation, as SDMA is mainly eliminated via renal glomerular excretion, and our results suggest that enzymatic degradation by AGXT2 contributes to SDMA clearance. Preliminary data from our group support this finding, as we observed an association of AGXT2 and SDMA in genome-wide association studies (33).This observation remains consistent with the close relation between SDMA and renal function, as AGXT2 is predominantly expressed in the kidneys (20).

Discrepancies between previous studies regarding ADMA plasma levels in diabetic patients raise questions about determinants of dimethylarginines in diabetes and their associations with glycemic state and diabetes medication. In our cohort, there were nominally significant, but weak, positive correlations of ADMA with fasting glucose levels and of SDMA with BMI, but neither of the dimethylarginines were significantly related to HbA_1c or diabetes treatment. In line with our results, previous studies found that ADMA was related to markers of renal function and L-arginine plasma concentrations but not to diabetes duration, glucose concentrations, HbA_1c, or insulin therapy in patients with type 2 diabetes (9,26). In a population-based study, including diabetic and nondiabetic subjects, ADMA levels were associated with age, BMI, and smoking status, but these clinical variables only explained approximately 3.5% of interindividual variation in ADMA (7). In contrast, the multivariable model in our study, including rs7554597 and the precursor molecule L-arginine, was able to explain 21.6% of the variation of ADMA. In another sample of the same population-based cohort, SDMA plasma levels were related to age, creatinine, homocysteine, BMI, and diastolic blood pressure, very similar to our results (34). However, while the multivariate model including these parameters was only able to account for 8% of variation in SDMA levels, the multivariate model we used explained 44.5% of the interindividual variation in circulating SDMA. Therefore, it is important to consider the regulatory pathways of dimethylarginines, including genetic variation in the metabolizing enzymes, when assessing the determinants of ADMA or SDMA in diabetic patients, whereas the contribution of glycemic control or antihyperglycemic agents seems to be only minor.

We conclude that dimethylarginines are only weakly related to glycemic state, antihyperglycemic medication, or CVD in older patients with type 2 diabetes mellitus. Furthermore, their predictive value for incident CVD appears to be limited. The strong associations between DDAH1 SNPs and ADMA levels, as well as AGXT2 SNPs and SDMA levels, demonstrate that genetic variations in the metabolizing genes should be taken into consideration when determining dimethylarginines.

Acknowledgments. The authors thank the participants, staff, investigators, and collaborators on the ET2DS. Furthermore, the authors thank the Wellcome Trust Clinical Research Facility for their collaboration (research clinic and genotyping) and Tanja Zeller and Christian Müller from the Cardiology Department of the University Medical Center Hamburg-Eppendorf for helping with the genetic analyses. The excellent technical assistance of Mariola Kastner and Anna Steenpass, both of the Department of Clinical Pharmacology and Toxicology, University Medical Center Hamburg-Eppendorf, is gratefully acknowledged.

Funding. This study was supported by the European Association for the Study of Diabetes, the U.K. Medical Research Council, and the European Commission (EU-FP7 grant number 278397, EuRhythDia).

Duality of Interest. Additional support was provided by Pfizer. E.S. and R.H.B. are named as inventors on patents relating to analytical assays for methylarginines and receive royalties from them. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.A. and S.M. researched data and wrote the manuscript. N.L. and E.S. contributed to the data analysis and discussion. C.R., R.M.W., and M.W.J.S. researched data and reviewed and edited the manuscript. R.A. reviewed and edited the manuscript. P.J.G., R.H.B., and J.F.P. designed the study, contributed to the discussion, and reviewed and edited the manuscript. R.H.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.