Advanced Glycation End Products and Antioxidant Status in Type 2 Diabetic Patients With and Without Peripheral Artery Disease

For our study, we evaluated blood levels of AGEs, pentosidine, MDA, TRAP, and vitamin E in a group of type 2 diabetic patients with and without PAD. Ninety-nine consecutive type 2 diabetic patients (57 men and 42 women) who regularly attended our outpatient clinic were enrolled in the study. Coronary heart disease was evaluated by clinical history (specified from charts) of myocardial infarction, angina pectoris, coronary artery surgery, angioplasty, and/or definite myocardial infarction on an electrocardiogram, interpreted according to the Minnesota code (18). Cerebral vasculopathy was investigated by history of symptoms of transient ischemic attack and/or stroke. Hypertension was diagnosed if blood pressure was >130/80 mmHg and/or if antihypertensive drugs were being taken (19). A physical examination and ophthalmoscopy were also performed.

The control group was composed of 20 subjects (10 men and 10 women) who were selected from an initial group of 32 subjects with normal glucose tolerance according to the American Diabetes Data Group criteria (20). Compared with the diabetic patients, the control subjects had similar distributions of sex, age, and diet, particularly as regards consumption of fruit and fresh vegetables. Informed written consent was obtained from all subjects participating in the study.

No diabetic patients were receiving insulin therapy, but >90% of them took both hypoglycemic and antihypertensive drugs. Thirty percent of patients with PAD and 27% without PAD took statins; 91% of patients with PAD and 45% without (P < 0.001) were receiving antiplatelet therapy (aspirin 100 mg/day).

On the day of the study, fasting blood samples were taken to determine fasting plasma glucose (FPG), A1C, pentosidine, AGEs, TRAP, vitamin E, MDA, cholesterol, and triglycerides. In the morning, after a 12-h overnight fast, blood was collected and immediately centrifuged at 1700g for 20 min at 4°C. Glucose was measured immediately, and the remaining samples were frozen at −80°C until assayed. The assays were performed within 3 months of sample collection. Under these conditions, previous experience had proven that no alteration occurs.

The ABI was estimated at rest by strain-gauge plethysmography, by means of a compressor connected to three cuffs, placed around one arm (that with higher pressure) and both ankles; distally, mercury strain gauges were placed, respectively, around the thumb of the above-mentioned arm and both terminal digits (toes) of the foot. By inflating the cuff to a pressure sufficient to stop blood flow, the plethysmographic transducer reveals a flat line. During slow deflation of the cuff, the appearance of the wave pulse indicates the level of systolic pressure of that area. The ABI is the ratio between the systolic pressures of ankles and arm, measured at the same time (21). A value of <0.9 was considered to indicate PAD, according to the conventional cutoff (15,16). The lowest mean ABI of three consecutive measurements in both legs was used in analysis. All subjects were also evaluated by echo color Doppler ultrasound of the peripheral arterial hypoaortic tree.

Pentosidine was measured by a liquid chromatography (LC) method (22). Plasma samples (350 μl), after the addition of 100 μl of NaBH₄ (0.2 mol/l), were precipitated on ice with an equal volume of 10% cold trichloroacetic acid. Samples were then hydrolyzed (6 N HCl at 110°C for 18 h), and, after acid removal (centrifugation with a Speed Vac SC 110; Savant), pellets were suspended in 500 μl of 0.01% heptafluorobutyric acid (Sigma-Aldrich, Milan, Italy) and then filtered. Samples were analyzed by LC (ProStar detector; Varian, Turin, Italy). The column was a 3.9 × 300 mm C₁₈ μBondapack (Waters Italia, Milan, Italy). The LC equipment was programmed with a linear gradient of acetonitrile in water and 0.1% heptafluorobutyric acid. Pentosidine was detected by fluorescence excitation at 335 nm and emission at 385 nm. The pentosidine standard was a kind gift of Professor V.M. Monnier (Case Western Reserve University, Cleveland, OH).

AGEs were measured by an enzyme-linked immunosorbent assay method (23), using AGE-BSA as the adsorbed antigen. AGE-albumin was prepared by incubating albumin (50 mg) with 0.5 mmol/l glucose in 0.2 mol/l NaPO₄ buffer (pH 7.4) under sterile condition. After incubation, unbound material was removed by extensive dialysis against PBS. Microliter plates were coated with AGE-BSA by adding 100 μl of a solution of AGE-BSA (30 μl/ml dissolved in PBS) to each well and incubating for 2 h at room temperature. Wells were washed three times with 0.15 ml of a solution containing PBS, 0.05% polyoxyethylene sorbitol ester (Tween 20) (PBS-Tween; Sigma Chemical, St. Louis, MO), and 1 mmol/l NaN₃. Wells were then blocked by incubation for 1 h with 0.1 ml of a solution of blocking buffer (a proprietary protein in PBS containing Kathon Antimicrobial Agent; Pierce Biotechnology, Rockford, IL). After washing with PBS-Tween, 50 μl of competing antigen was added, followed by 50 μl of antiserum. Plates were incubated for 3 h at room temperature. Wells were then washed with PBS-Tween and developed with alkaline phosphatase-linked anti-rabbit IgG (an antibody directed against rabbit IgG, purified from serum by chromatography after immunization of sheep with rabbit IgG; Roche Diagnostics, Mannheim, Germany), with p-nitrophenyl phosphate as the colorimetric substrate.

TRAP was evaluated according to the method of Ghiselli et al. (24). In this method, the production of peroxyl radicals obtained by thermal decomposition of 2,2′-azobis-(2-amidinopropane) dihydrochloride leads to a linear decrease in R-phycoerythrin (R-PE) fluorescence emission over 1 h. When plasma is added to the reaction mixture, a period of complete protection of R-PE is observed. The length of this lag phase (T) is here taken to be directly related to total plasma antioxidant capacity. To quantify TRAP, the T produced by plasma is compared with the T produced by a known amount of Trolox. By comparing the T of plasma with the T of Trolox, taking into account the concentration of Trolox, the TRAP value of a plasma sample is obtained according to the following proportion: concentration Trolox:T Trolox = X:T plasma.

The resulting value of X is then multiplied by 2.0 (the stoichiometric factor of Trolox) and by the dilution factor of plasma (250); values are expressed as micromoles per liter. 2,2′-Azobis-(2-amidinopropane) dihydrochloride, R-PE, Trolox, and all other chemicals were purchased from Sigma Chemical.

Vitamin E levels were measured by high-precision chromatography on a reverse-phase column by a diode array spectrophotometric detector, according to the method of Tsan et al. (25). MDA was measured by the highly sensitive fluorometric method (high-performance LC) after deproteination of the sample and its reaction with thiobarbituric acid and extraction of the adduct with n-butylic alcohol (26). Plasma glucose was determined by a glucose oxidase method (27). A1C was measured by an LC method (28) (Bio-Rad, Milan, Italy). Total cholesterol and LDL and HDL cholesterol were measured by enzymatic analytical chemistry (CHOD-PAP method, Roche Diagnostics, Milan, Italy) (29,30) as plasma triglycerides (GPO-PAP colorimetric enzyme test, Roche) (31).

Data are expressed as means ± SD. For parameters occurring as frequency, the statistical difference between the two groups of patients was determined by means of the χ² test with Yates’ correction (32); when a frequency was <5, Fisher’s exact test was applied.

For statistical multiple comparison of continuous variables, a one-way ANOVA followed by the Bonferroni post hoc test was used. Statistical difference was accepted when P < 0.05.

The presence of any relationship between pair variables was evaluated by least-squares linear regression. Pearson’s correlation coefficient r was used to quantify the strength of the relationship.

Multiple regression was used to further explore the linear relationship among the variables. The equation was of the form: y = a + b₁x₁ + b₂x₂ + b₃x₃.

Regression parameters were estimated as well as correlation coefficient r. ANOVA statistics were used to assess the significance of the regression and accepted for P < 0.05.

Principal component analysis (PCA) is useful in reducing the dimensionality of the dataset, may help to identify new meaningful underlying variables, and can also indicate the presence of clusters within multivariate data. PCA transforms a number of possibly correlated variables into fewer uncorrelated variables defined as principal components, which are linear combinations of the original variables. After principal components were obtained, they were plotted to observe any groupings in the dataset. PCA computation was performed according to a correlation matrix and standardized principal component score using AMADA software (33) or BiPlot software Excel macros (34). The first three components were considered for classification of data. A BiPlot graphic display was used to present the behavior of variables (columns) to examine their correlation (35) on the same chart. Both length and directions of vectors (rays) may be important in interpreting the data (35). In this case, the most useful parameter is the cosine of the eigenvectors, which suggested correlations among different variables. When the angle between eigenvectors is close to 0°, the variables are positively correlated, the angle for negative correlations approaches 180°, and angles of 90° indicate no correlation.

On the basis of the ABI values, the 99 type 2 diabetic patients were divided into one subgroup of 33 individuals with PAD (ABI ≤0.9) and another subgroup of the remaining 66 patients without PAD (ABI >0.9). All of the diabetic patients with PAD showed the characteristic below-the-knee macroangiopathy, without media calcification, confirming ABI data obtained by plethysmography. No evidence of macroangiopathy above the knee was found.

The clinical characteristics and laboratory and instrument data of patients and control subjects are listed in Table 1. No significant differences were found in age or BMI between type 2 diabetic patients with or without PAD nor was there any difference in diabetes duration or smoking habits.

There were no significant differences in serum levels of A1C, FPG, or lipid profiles. Serum creatinine values were normal in all subjects, and none of the diabetic patients had retinopathy. Serum concentrations of AGEs, pentosidine, and MDA were significantly higher in patients with vasculopathy than in patients without (P < 0.001 for each parameter). Vitamin E and TRAP were significantly lower in patients with PAD (P < 0.001).

When attempting to fit data pairwise using linear regression analysis and considering all diabetic patients, we found no significant correlations between FPG or A1C against any glycolipid oxidation parameter evaluated (AGEs, pentosidine, and MDA) and ABI. Furthermore, no correlations were found between either cholesterol (total, LDL, and HDL) or triglyceride levels and ABI.

After data were fitted using the linear regression model, ABI was inversely correlated with serum AGE concentrations in all patients (slope = −0.0308, r = −0.6685, P < 0.001), even when patients with PAD (slope = −0.0201, r = −0.8162, P < 0.001) (Fig. 1A) and without PAD (slope = −0.0166, r = −0.5366, P < 0.001) were considered. ABI was also found to be correlated with pentosidine in all patients (slope = −0.0065, r = −0.7490, P < 0.001) and when the groups with PAD (slope = −0.0050, r = −0.9154, P < 0.001) (Fig. 1B) and without PAD were evaluated separately (slope = −0.0031, r = −0.5716, P < 0.001). A negative relationship was also found between ABI and MDA but was significant only in patients with PAD (slope = −0.2292, r = −0.6714, P < 0.001). In patients with PAD, both TRAP and vitamin E were directly associated with ABI (slope = 0.0005, r = 0.4318, P < 0.05 and slope = 0.0365, r = 0.5445, P < 0.01, respectively).

A multiple regression model applied to AGEs and pentosidine, as independent variables, and ABI, as a dependent variable, revealed a significant correlation in both patients with and without PAD (r = 0.9198, P < 0.001 and r = 0.5764, P < 0.001, respectively). However, no significance was found when this model was applied to healthy control subjects. When individual regression coefficients (slopes) were tested for significance by ANOVA, the component related to pentosidine was confirmed to be significant (slope = −0.00426, P < 0.001 for patients with PAD; slope = −0.00239, P < 0.05 for those without PAD). The AGE component did not reach a statistically significant level, indicating that the model is not closely dependent on the latter variable.

Multiple regression was also performed for TRAP, vitamin E, and MDA to verify any link with the dependent variable ABI. In this case, for patients with PAD, a significant regression was obtained (r = 0.6913, P < 0.001), but the regression coefficient due to TRAP and vitamin E was not significant, indicating that the model is best explained by a single linear regression between MDA and ABI.

A further attempt to explain the links among the variables was to use the multivariate technique of PCA applied to the most important variables considered previously. Both positive and negative correlations among variables after PCA are easily observed when the results are presented as a BiPlot graph (Fig. 2), as indicated under research design and methods. The scores obtained for the two groups of patients, although closely related, show different behavior, revealing the distinct pattern due to PAD. Considering the relationships among variables, the BiPlot graph obtained with the first PCA components shows that there is close collinear behavior between AGEs and pentosidine and that they are inversely correlated with ABI. TRAP and MDA, respectively, also appear to be inversely correlated. MDA is negatively correlated with vitamin E and TRAP. Conversely, no apparent relation was found for either AGEs or pentosidine against vitamin E.

In 99 consecutive type 2 diabetic patients, highly significant increased levels of AGEs, pentosidine, and MDA were found in patients with PAD more than in patients without PAD and control subjects. AGEs and pentosidine were correlated with ABI in all diabetic patients, not only in patients with PAD; no correlation was found in the control group. These results indicate that glyco-oxidation contributes to the development of atherosclerosis in the below-the-knee peripheral artery tree in type 2 diabetes. More precisely, among AGE components, pentosidine appears to be strongly associated with the peripheral artery status of diabetic patients. This study indicates that pentosidine may be a predictor of PAD in diabetes, but it is necessary to carry out an appropriately designed longitudinal study to confirm this hypothesis. In type 2 diabetes, previous studies have demonstrated both increased serum levels of AGEs in patients with coronary artery disease (9) and increased accumulation of AGEs in atherosclerotic plaques in coronary arteries (11). Yoshida et al. (12) also showed that accumulation of pentosidine in the vessel walls of diabetic patients increases arterial stiffness and/or carotid intima-media thickness, contributing to their increased cardiovascular risk. For the first time, our results also show that the lower extremity arteries may be damaged by AGEs, particularly pentosidine, in type 2 diabetes. In addition, lipid oxidation, in terms of serum levels of MDA, was associated with peripheral diabetic angiopathy, in agreement with one of our previous studies (36).

We found that both TRAP and vitamin E levels, as expressions of a defense mechanism against glycolipid oxidation, were lower in type 2 diabetic patients with PAD than in those without PAD and control subjects. Low TRAP levels have been reported in both type 1 and type 2 diabetes (7,24). Recent epidemiological studies suggested that decreased levels of antioxidants favor cardiovascular disease in nondiabetic subjects (37–39). Several studies later addressed the issue of antioxidant status in type 2 diabetes, generating conflicting results (40,41). In our study, both these parameters were positively correlated with ABI in patients with PAD (lower values of ABI corresponded to lower values of TRAP and vitamin E). Moreover, these parameters were not correlated with AGEs or pentosidine. Only TRAP was inversely correlated with MDA. This suggests that the defensive role of antioxidants fails in the presence of the already developed vascular damage in type 2 diabetic patients. These findings may help to explain the inefficacy of treatment with antioxidants used in secondary prevention when the atherosclerotic process is already manifest. Several randomized trials have failed to show any benefit deriving from the use of vitamin E in preventing cardiovascular events in various high-risk groups, including diabetic patients (42,43). The Primary Prevention Project (PPP) trial (44) also showed the substantial failure of antioxidant vitamin E supplementation in primary prevention of major cardiovascular events in patients at risk. In that study, a marginal reduction in the risk of PAD was documented only in nondiabetic subjects.

It is possible that vitamin E supplementation had no effect at all. Alternatively, the lack of effect in those diabetic groups may be due to the fact that the study did not specify the duration of diabetes, although it is known (45) that PAD may occur even in patients with newly diagnosed diabetes and that it increases with the duration of disease. In addition, PAD is frequently asymptomatic in diabetic subjects.

In summary, our results show that, in type 2 diabetic patients, glyco-oxidation, lipid oxidation, and specifically serum levels of pentosidine and MDA are strongly associated with below-the-knee diabetic macroangiopathy. Serum antioxidant capacity, represented by TRAP, is associated with MDA but not with AGEs, showing that it cannot prevent the development of PAD caused by AGEs.