Our salient finding was an association between higher ADMA and future deterioration of glucose tolerance, irrespective of the degree of baseline IR, in men with stable CAD over an about 4.5-year follow-up.
Endothelial dysfunction as a precursor of type 2 diabetes in previous clinical reports
Our results extend previous reports suggestive of an association between incident type 2 diabetes and biochemical indices of endothelial activation [2–4] or vasodilatory endothelial dysfunction , which retained significance upon multivariate adjustment including the HOMA-IR index  or fasting insulinemia , surrogate markers of IR. This association was interpreted as a possible consequence of an impaired insulin ability to stimulate blood flow via the L-arginine–NO pathway in the skeletal muscle and adipose tissue due to endothelial dysfunction at the arteriolar and capillary level with dysfunctional endothelium as an underlying cause of IR [2, 3, 31]. That adjustment for baseline insulin sensitivity (derived from the frequently sampled intravenous glucose tolerance test) only slightly attenuated the relationship between CV risk factors and a 5-year cumulative incidence of type 2 diabetes in 872 participants of the Insulin Resistance Atherosclerosis Study , is also in keeping with the predominant role of endothelial dysfunction with regard to the risk of developing diabetes. Given anti-inflammatory properties of the normal endothelium, this concept is also concordant with a joint synergistic contribution of higher hs-CRP and peripheral microvascular endothelial dysfunction, but not IR, to the risk of new-onset type 2 diabetes . Admittedly, we observed no significant association between deteriorating glucose tolerance and hs-CRP in contrast to some previous reports [3, 5, 33]. Nevertheless, those studies were dealing with mainly untreated subjects, whereas at the time of blood collection our patients were chronically receiving statins, known to decrease CRP levels  but not ADMA , which might have obscured the discussed relationship. An analogous mechanism, in addition to the ability of ACEI, administered to all our subjects, to lower ADMA concentrations , could also be responsible for the lack of correlation between ADMA and hs-CRP. It is noteworthy that ADMA and hs-CRP were positively interrelated in the participants of the Ludwigshafen Risk and Cardiovascular Health study  but not the AtheroGene study , both of which were focused on CAD subjects presenting mainly stable angina.
Associations between the L-arginine–NO–ADMA pathway and insulin signaling
Sydow et al.  reported enhanced whole-body insulin sensitivity in transgenic mice overexpressing dimethylarginine dimethylaminohydrolase (DDAH), an enzyme responsible for the predominant pathway of ADMA degradation . Moreover, a more effective insulin signaling in the liver and a trend for increased baseline glucose uptake in an isolated fast-twitch muscle were observed. Additionally, in wild-type animals low ADMA levels (2 μmol/L) impaired insulin-mediated glucose uptake and incorporation of glucose into glycogen in an isolated fast-twitch muscle but not a slow-twitch muscle . Furthermore, a functional variant of the DDAH-2 gene, linked to depressed expression of the isoform of DDAH which predominates in the vascular endothelium , was associated with a higher IR in 1,485 non-diabetic subjects of European ancestry, out of whom 527 had insulin sensitivity assessed as insulin-stimulated glucose disposal rate during the last hour of a hyperinsulinemic euglycemic clamp, the golden standard method . According to Baron and colleagues , a superimposed intrafemoral artery infusion of N
-monomethyl-L-arginine during hyperinsulinemic euglycemic clamps not only abolished the insulin-dependent NO-mediated vasodilation [42, 43] but also reduced leg glucose uptake. Thus, besides impaired hepatic insulin signaling, the ADMA-mediated inhibition of glucose uptake by the skeletal muscles – in part dependent on the blockade of vascular endothelial-type NO synthase (eNOS) and consequently blunted muscular perfusion under hyperinsulinemic conditions [41–43] – might have contributed to the relations between DDAH-2 polymorphism and diabetes incidence or IR , providing also a mechanistic explanation for our findings.
Additionally, a modulation of neuronal-type NO synthase (nNOS), whose variant is expressed predominantly in fast-twitch skeletal muscle fibers , might be responsible for the association between IR and the L-arginine–NO pathway. It is noteworthy that although both nNOS-knockout and eNOS-deficient mice were insulin-resistant compared to wild-type animals, the former exhibited IR exclusively at the level of peripheral tissues, whereas in the latter IR was present also in the liver . Furthermore, a polymorphism of the gene encoding DDAH-1, an isoform of DDAH tending to accompany nNOS , was associated with both a decreased risk of type 2 diabetes and a tendency towards a lower HOMA-IR index in 814 Taiwanese subjects .
ADMA levels versus the magnitude of IR in clinical conditions
Although the presented experimental and genetic data suggest reciprocal and stimulatory interactions between the L-arginine-NO pathway and metabolic sensitivity to insulin [22, 39, 41, 45, 46], there are conflicting reports on the relationship between ADMA and IR in clinical conditions [13–23]. Intriguingly, we detected no correlation between ADMA and the HOMA-IR index, in accordance with the data by Mittermayer et al.  who described an independent relationship between higher ADMA levels 14–16 weeks after delivery with future deterioration of glucose tolerance in 18 out of 77 women with previous gestational diabetes mellitus who progressed to impaired glucose tolerance in 9 and developed new-onset type 2 diabetes in further 9 cases after a median follow-up of almost 3 years. The lack of association between ADMA and the magnitude of IR was also observed by other authors, including ourselves [14, 18–20], in contrast to studies reporting a positive relationship between the respective parameters [15–17, 21, 23]. Admittedly, the absence of positive correlations between ADMA and the HOMA-IR index or BMI in our subjects – in disagreement with a large study  – might have resulted from a relatively low inter-individual variability of ADMA values. Nevertheless, there are also inconsistent data on the effects of rosiglitazone, an insulin-sensitizing agent, on ADMA levels [15, 47].
Ferrannini et al.  demonstrated that deterioration of glucose tolerance over time was associated with longitudinal decreases in both insulin sensitivity and β-cell sensitivity to glucose. Thus, given IR as the primary determinant of future development of type 2 diabetes [6–9], the present study suggests that endothelial dysfunction and higher ADMA might rather predispose to further potentiation of IR without an unequivocal association with IR at baseline in cross-sectional analyses [13–23]. Additionally, we cannot exclude the possibility that ADMA might progressively impair β-cell function, another contributor to the development of type 2 diabetes [6–9].
Admittedly, a correlation does not imply a cause-and effect relationship and elevated ADMA might be only an epiphenomenon accompanying some abnormalities in the pre-diabetic state. These abnormalities can include higher glucose levels, a well-recognized cause of elevated ADMA via depressed DDAH activity due to a higher intracellular oxidative stress . Additionally, the hyperglycemia-induced depression of DDAH activity  might counteract the hyperinsulinemia-stimulated uptake of ADMA into the cells via an augmented expression of cationic amino acid transporters (CAT) . A complex interplay between these opposite effects of glucose and insulin on plasma ADMA  might be responsible for the previously mentioned discordant reports on the relationship between IR and ADMA [13–23]. Moreover, the same opposing regulatory mechanisms could also explain negative correlations between the HOMA-IR index and the SDMA/ADMA ratio  or SDMA  in Caucasians, an insignificant tendency to which was also found in our study group. Accordingly, the hyperinsulinemia-stimulated CAT-dependent uptake of ADMA and SDMA into the cells, probably within the liver and the kidney (both referred to as a “sink” for circulating ADMA ) – might hypothetically be counteracted – for ADMA but not SDMA – by the hyperglycemia-induced depression of DDAH activity . In agreement with this concept, insulin-resistant subjects are prone to hyperinsulinemia and hyperglycemia and Eid et al.  reported a fall in ADMA during acute euglycemic hyperinsulinemia in humans. It is noteworthy that these two apparently contrary effects of insulin and glucose in terms of extracellular ADMA are likely to converge into increased intracellular ADMA levels . Thus, the lack of relationship between plasma ADMA and HOMA-IR index, as in the present study, does not exclude the hypothetical continuous modulation of insulin signaling by intracellular ADMA.
Another interfering factor could be hepatic ability to clear plasma from endogenous dimethylarginines, responsible for about 30% and 20% of the combined hepatic and renal uptake of ADMA and SDMA in humans, respectively . DDAH-1 expression was reduced during diabetogenesis in the liver, kidney and adipose tissue of mice with a point mutation in leptin receptor, an experimental motel of type 2 diabetes . However, whether similar abnormalities occur in human pre-diabetic subjects, is as unclear as their speculative relevance for circulating ADMA levels or the risk of developing type 2 diabetes.
That our CAD subjects were receiving a standard therapeutic regimen in agreement with current guidelines, may point into a potential relevance of our results for the clinical reality. As progressive impairment of glucose tolerance is associated with a continuous rise in CV risk until its highest values in overt diabetes , the observed relation between elevated ADMA and a decline in glucose tolerance might contribute to the detrimental effect of ADMA on CV outcome [12, 37, 38]. Furthermore, Rijkelijkhuizen et al.  suggested conversion to type 2 diabetes as a basis of an excessive future CV mortality in Hoorn study participants exhibiting IFG. Therefore, besides the well-recognized link between ADMA and endothelial dysfunction , our finding might add to novel putative mechanisms of detrimental ADMA effects, e.g. its association with elevated arterial stiffness  or increased plasma viscosity  described in pre-diabetes and dyslipidemia, respectively. It is noteworthy that the correlation between ADMA and arterial stiffness in pre-diabetes was additive to an independent effect of low-density lipoproteins cholesterol . Thus, excessive ADMA accumulation not only accompanies atherosclerotic risk factors, such as hypercholesterolemia  and essential hypertension , but also appears to amplify coexistent abnormalities.
Additionally, it is noteworthy that ACEI, angiotensin receptor blockers and statins – despite improving NO bioavailability – exert opposite effects on the risk of new-onset type 2 diabetes [59, 60]. As ACEI and angiotensin receptor antagonists are able to lower ADMA levels  in contrast to the majority of reports on statins [35, 40], our observation could provide a possible rationale for differential effects of these drug classes on the risk of developing type 2 diabetes.
First, the lack of repeated determinations of ADMA, SDMA and IR over time constrains mechanistic conclusions based on a single measurement of these parameters in a relatively small number of study participants. Additionally, relatively high coefficients of variation of the ELISA assays for ADMA and SDMA pose another limitation. Second, we did not assess lifestyle factors or longitudinal changes in body weight, known to affect the risk of type 2 diabetes . Third, in our retrospective analysis exclusively IFG was classified as pre-diabetes to avoid a bias resulting from the fact that results of an OGTT during the follow-up were available only in a minority of the study subjects. Fourth, although no significant intergroup differences in the use of various CV drug classes during the follow-up period were detected, the type of medication was determined exclusively by the analysis of an available medical documentation. Finally, IR was quantified on the basis of an indirect measure, i.e., the log-transformed HOMA-IR index. Nevertheless, this parameter not only exhibited a close correlation with the glucose clamp-derived insulin sensitivity index , but it also was one of the best predictors of incident type 2 diabetes among surrogate estimates of IR based on assays of fasting insulin and glucose .