Introduction

Advanced glycation end products (AGEs) arise from a non-enzymatic reaction between reducing sugars and biological proteins.1 Increased AGE formation is a typical molecular consequence of diabetes, and it has been proposed to have an important role in cardiovascular complications of diabetes.2 AGEs bind to a specific cell-surface receptor (RAGE) of the immunoglobulin superfamily, which are expressed in various cells and tissues and contribute to vascular function, including the endothelium and vascular smooth muscle cells.3

AGEs also have a pathophysiological role in the progressive stiffening of large arteries as a manifestation of vascular ageing. As a result of AGE accumulation in the arterial wall, collagen fibers are crosslinked.4, 5, 6 In contrast to the physiological crosslinking of collagen (on N- and C-terminal ends only), AGE-mediated crosslinking increases the collagen content in the arterial wall (primarily because of its higher resistance to enzymatic proteolysis and decreased degradation rate), which is clinically manifested as an increased pulse wave velocity (PWV).6, 7 The community-based Baltimore Longitudinal Study of Ageing (493 participants) showed that elevated serum AGEs were independently associated with an increased PWV.8 A smaller study by McNulty et al.9 demonstrated the same association in 30 untreated hypertensive patients.

The soluble isoform of the receptor for AGEs (sRAGE) acts as a decoy for capturing circulating AGEs, preventing them from binding to the cell-surface receptor and protecting them against the RAGE–AGE axis-elicited pathophysiological processes (that is, circulating sRAGE may suppress the pathways initiated by the cell-surface RAGE and act as ‘physiological defense against AGEs’10). High plasma levels of sRAGE are associated with a lower incidence of coronary artery disease in non-diabetic males,11 while low plasma levels of sRAGE have been reported to be associated with coronary artery disease, carotid or femoral atherosclerosis and calcific aortic valve stenosis.12, 13, 14 To the best of our knowledge, the association between sRAGE and large artery stiffness has only been studied in hypertensive patients to date.15, 16 In the present paper, we investigated this association in a random, general population-based sample.

Materials and Methods

Study population

The study population consisted of a random general population sample. A survey of the risk factors was undertaken in 2008 as part of the Czech post-MONICA study.17 One percent of Pilsen residents, aged 25–75 years, were selected from the General Health Insurance Registry. A total of 1417 subjects (663 males and 754 females, mean age 55.2 years) responded to the survey (that is, >68% of the invited residents). Informed consent was obtained from all subjects, and all personal data were stored under the provisions of the Czech Data Protection Act. Data from the PWV measurements were available in 1131 of the subjects. Additionally, another 54 subjects whose sRAGE levels could not be determined (missing samples or insufficient plasma volumes and so on) were excluded from the analysis.

Examinations and materials

All study procedures were performed according to the Good Clinical Practice guidelines and were approved by the local Ethics Committee. All respondents were interviewed and examined by standardized methods using the World Health Organization MONICA Study standard manual. The interview methods have been described in detail elsewhere.18 Briefly, information on the personal and demographic characteristics, personal and family history of coronary artery disease, lifestyle and current pharmacotherapy was obtained. The following standardized examinations were performed: height and weight in light, indoor clothing without shoes using a DETECTO 20 (DETECTO, Webb City, MO, USA) scale and measuring stick. Blood pressure (BP) was measured three times in the sitting position on the right arm using standard mercury sphygmomanometers (Klinik ERKA.meter, ERKA Kallmeyer Medizintechnik GmbH & Co., Bad Tölz, Germany) with the value rounded up to the nearest 2 mm Hg; an appropriate cuff size was used. The mean arterial pressure, which was derived from the office BP measurement, was calculated as the diastolic pressure plus one-third of the pulse pressure.

Regarding methodology, large artery properties were measured as per the 2006 Expert Consensus Document.19 PWV was quantified using a Sphygmocor device (AtCor Medical Ltd, West Ryde, NSW, Australia). With the patient in the supine position, we measured the aortic PWV (aPWV) between the carotid and femoral arteries. Registrations of the pulse waves were electrocardiogram-gated so that we could calculate the time shift between the appearance of the wave at the first and second sites. The distance between the two sites was measured on the body surface. To obtain the travel distance along the aorta, we measured the distance from the jugular fossa to the pulsation of the femoral artery in the groin, which was subtracted from the distance between the jugular fossa and carotid pulsation. PWV was calculated as the ratio of the travel distance (in meters) to the transit time (in seconds).

Venous blood samples were drawn after at least 12 h of overnight fasting. All study procedures were performed on the same day at approximately the same time during the examination campaign. Frozen samples, stored at −80 °C, were used for biochemical laboratory analyses. The laboratory examinations, including assessment of the total (TCHOL) and high-density lipoprotein cholesterol (HDL), triglycerides (TG) and creatinine, were performed by the central laboratories of the Czech post-MONICA survey (Institute for Clinical and Experimental Medicine, Prague, Czech Republic) from serum samples evaluated using a Cobas Mira/ROCHE analyzer (Roche, Basel, Switzerland) and commercially available kits from the same manufacturer. The glucose levels were analyzed by enzymatic methods in the above laboratory using the standard kits (Lachema, Brno, Czech Republic). Low-density cholesterol (LDL) was calculated with the Friedewald equation, that is, LDL=TCHOL−HDL−(TG/2.22).20 The creatinine clearance was estimated using the Cockcroft–Gault formula, that is, ((140–age in years) × body weight × 1.23 (or 1.04 for women)/serum creatinine).21 Using an ACCESS 2 analyzer (Beckman-Coulter, Brea, CA, USA) and commercial kits from the same manufacturer, 25-hydroxyvitamin D (25-OH-D3) was assessed in the research laboratory of the Department of Immunodiagnostics, University Hospital, Pilsen, Czech Republic from serum samples. sRAGE was quantified by enzyme-linked immunosorbent assay methods using a Human RAGE Quantikine ELISA Kit (R&D Systems, Minneapolis, MN, USA).

Data analysis

The study had a cross-sectional design. Statistical data analyses were performed using the STATISTICA 9 (StatSoft Inc., Tulsa, OK, USA) and STATA 6 (StataCorp LP, College Station, TX, USA) software packages. Power calculations were performed using the s.ds. ascertained in our previous studies. Conventional risk factors were dichotomized by the usual cutoff points (see the relevant section of tables) with respect to the current Fifth Joint European Guidelines for Cardiovascular Prevention.18 Namely, overt diabetes was defined as a fasting plasma glucose 7.0 mmol l−1 and/or the use of antidiabetic medications, ‘raised BP’ according to systolic BP140 and/or diastolic BP90 mm Hg and hypertension according to a raised BP and/or treatment with antihypertensives. We defined a moderately decreased glomerular filtration as the calculated creatinine clearance <80 ml min−1 (none of the subjects had a clearance <30 ml min−1). Low D vitamin status was defined as 25-OH-D3 levels <40.9 ng ml−1 (that is, upper limit of the bottom quartile). sRAGE was divided into the following quartiles: <918, 918–1205, 1206–1573, and 1574 pg ml−1. The upper limit of the bottom quartile was used as the cutoff point for low sRAGE. Increased aPWV was defined as 9.3 m s−1, that is, the fourth quartile of its distribution.

For statistical analysis, we used conventional statistical methods, namely, descriptive statistics, and multivariate linear and logistic regressions to evaluate the association between sRAGE and large artery properties after adjusting for potential confounders (the methods are described in the relevant sections).

Results

The study included 1077 subjects, 509 males and 568 females, with a mean age of 54.8 years (s.d. ±13.1). Their basic characteristics are given in Table 1.

Table 1 Basic characteristics of the study sample (mean (s.d.) or factor proportion)

sRAGE and its covariates

Table 2 shows the clinical covariates of sRAGE. Using Spearman’s correlation, sRAGE was significantly positively associated with HDL cholesterol and 25-OH-D3, while it was negatively correlated with the age, body mass index, waist circumference, mean arterial pressure, TGs, fasting glycemia and aPWV. Concentrations of sRAGE are also significantly lower in males than in females (1212 vs. 1397 pg ml−1; P<0.0001 by the Mann–Whitney U-test) in subjects with a personal history of vascular disease compared with those without (1264 vs. 1312 pg ml−1; P=0.035) as well as in patients with diabetes compared with those without (1103 vs. 1330 pg ml−1; P<0.0001). They did not differ according to smoking status (1330 vs. 1299 pg ml−1 in current smokers vs. non-smokers; P=0.10); not shown in the Table. The second part of Table 2 compares the differences in the conventional risk factors between the bottom and top sRAGE quartiles using unifactorial analysis. Subjects in the bottom sRAGE quartile had a significantly higher age, body mass index, waist circumference, mean arterial pressure, TG level, fasting glycemia and aPWV, whereas they had lower HDL cholesterol and 25-OH-D3 levels.

Table 2 Clinical covariates of sRAGE in a continuous manner (by Spearman’s correlation) and differences in these factors between bottom and top quartile of sRAGE (P-value by Mann–Whitney U-test)

We also performed an analysis to exclude the potential association between sRAGE and concomitant treatment. The following drug classes were tested as independent predictors of sRAGE (adjusted for age and gender): angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, calcium antagonists, diuretics, beta-blockers, other antihypertensives, statins, and antidiabetic drugs. None of these drug classes significantly entered the multiple regression models with sRAGE as the dependent variable, either in a continuous or categorized (<918 pg ml−1) manner; not shown in the Table.

sRAGE and arterial stiffness

aPWV significantly increased across sRAGE quartiles (Figure 1). An aPWV of 1 m s−1 was associated with a 37% increase in the risk of low sRAGE (>918 pg ml−1, bottom quartile; P-value=0.018) (adjusted for age decade and raised BP), not in the Table.

Figure 1
figure 1

Aortic pulse wave velocities by sRAGE quartiles (box and whisker plots). P-value adjusted for the age decade and gender; the limits of the sRAGE quartiles are as follows: <918, 918–1205, 1206–1573 and 1574 pg ml−1.

Table 3 shows the adjusted associations between aPWV and sRAGE in a categorized manner. Using step-wise logistic regression, we identified the following variables as significant predictors of increased aPWV (9.3 m s−1, model A): age decade, body mass index 30 kg m−2, raised BP, treatment with antidiabetics, and low sRAGE (in the bottom quartile). In a separate analysis (model B) of only non-diabetic subjects, the results were virtually similar (the predictive potential of low sRAGE somewhat increased). In contrast, when only diabetic subjects were analyzed in model A, the sRAGE predictive potential disappeared (and only the age decade and raised BP significantly entered the regression), not shown in the Table. In an additional step, we analyzed the effect of co-incident hypertension. Low sRAGE levels significantly predicted increased aPWV in hypertensive, non-diabetic patients (Table 3, model C). By contrast, this association again disappeared in normotensive, non-diabetic subjects (odds ratio in the same but the non-stepwise model for low sRAGE was 0.76 and the 95% confidence interval (CI) was 0.26–2.22).

Table 3 Predictors of increased aortic pulse wave velocity

We also compared the aPWV in the sRAGE categories (dichotomized using the 918 pg ml−1 cutoff point) in hypertensive subjects as well as according to treatment with renin–angiotensin–aldosterone system (RAAS) blockers (Figure 2). Hypertensive patients with low sRAGE levels (<918 pg ml−1) showed significantly higher aPWV than those with sRAGE 918 pg ml−1, which was only when patients were untreated with RAAS blockers (Figure 2b). In contrast, no such difference in the aPWV among sRAGE categories was observed in the RAAS blocker-treated subjects (Figure 2a). Both groups (that is, treated with RAAS vs. not treated with RAAS) substantially differed in the proportion of other concomitantly used antihypertensives as follows: calcium antagonists 35.1% vs. 17.6%, diuretics 31.7% vs. 10.8%, beta-blockers 37.0% vs. 25.4 and other antihypertensives 9.8% vs. 3.4%. Therefore, we performed exploratory analysis using these drug classes as potential covariates of the core association (aPWV by sRAGE categories). None of these drugs significantly influenced this association, and the P-value remained virtually the same (0.0045).

Figure 2
figure 2

Aortic pulse wave velocities among sRAGE categories and with respect to RAAS blocker (angiotensin-converting enzyme inhibitors or angiotensin receptor blocker) treatment (means and s.d.s.). P-value adjusted for the age decade, gender and mean arterial pressure.

Discussion

In the general population, decreased concentrations of the sRAGE were independently associated with increased stiffness of the elastic central arteries, quantified as aPWV, but this association only reached statistical significance in non-diabetic patients with concomitant hypertension. These subjects, when they were in the bottom quartile of RAGE, had a greater than twofold odds ratio of increased aPWV (9.3 m s−1).

To the best of our knowledge, no previous study has addressed this question in an unselected general population, but we are in agreement with the previous studies that were performed in hypertensive patients alone.15, 16 Dimitriadis et al.15 reported that low sRAGE was associated with higher office and ambulatory BP and that sRAGE was independently and inversely correlated with the aPWV. Yoon et al.16 reported similar results in the same setting. In the present study, we sought to answer this question in an unselected general population; however, based on our results, the sRAGE–PWV association is strongest in subjects with concomitant hypertension.

In our population sample, low sRAGE was surprisingly unrelated to aortic stiffness in subgroups of patients with overt diabetes. It is evident that impaired glucose metabolism (including overt diabetes) represents an important co-factor of individually accelerated arterial stiffening.22 Additionally, in the present study, overt diabetes was the third most powerful predictor of increased aPWV (after raised BP and age). Indeed, circulating concentrations of sRAGE primary reflect the degree of tissue glycation1, 3; as a result, low concentrations are typical for patients with diabetes23 (as can also be observed in our sample). We speculate that the association with arterial stiffness ‘dissolved’ because of both the particularly suppressed sRAGE concentrations in patients with overt diabetes and the robust association between diabetes and increased PWV. Moreover, patients with overt diabetes were relatively uncommon in our series (only 9%), and this subgroup was not adequately powered to analyze the relationship between diabetes, sRAGE and aortic stiffness in more detail.

It can be assumed that the pathophysiological role of impaired glucose metabolism on arterial stiffening is already expressed in the early stages of the disease. Pietri et al.24 found a linear increase in arterial stiffness in hypertensive patients with metabolic syndrome compared with normoglycemic patients, and this trend included impaired fasting glucose and impaired glucose tolerance subjects up to overt diabetes patients. This trend in the arterial stiffness (as a macrovascular complication) was independent and was parallel with relation to glycemic status and microalbuminuria (as a microvascular complication). The authors concluded that the global cardiovascular risk in patients with metabolic syndrome largely is determined by the current glycemic status.24 In the present study, the sRAGE (which is a quantitative and more sensitive indicator of the current glycemic status) had the same macrovascular impact in euglycemic hypertensive patients.

It has been suggested that in addition to the main presumed pathophysiological mechanism of AGEs in arterial stiffening (that is, collagen crosslinking), several other more dynamic and potentially reversible mechanisms could be involved. Circulating AGEs were negatively associated with endothelium-dependent vasodilation25 as well as with the bioavailability of endothelium-derived nitric oxide.26 Moreover, because RAGE acts as a common multi-ligand receptor, it has been demonstrated to have a role in the early stages of atherosclerosis by several pathways, including expression of adhesion molecules, inflammatory mediators, growth factors, oxidative stress and so on.27, 28, 29 An important mechanism by which AGEs contribute to arterial stiffening is related to the pathophysiology of hypertension, which represents cross-talk between RAGE and RAAS, although we could not fully elucidate the details. In an animal model, Thomas et al.30 found that angiotensin II infusion accelerated the formation and accumulation of AGEs in rat glomeruli and renal tubules, while treatment with valsartan, an angiotensin receptor blocker, reduced the renal levels of AGEs in AGE-injected animals in the same model. Similarly, treatment with an angiotensin-converting enzyme inhibitor (ramipril) decreased the circulating and renal tissue levels of AGEs in rats with diabetic nephropathy.31 In humans, treatment with low-dose valsartan for diabetic patients was followed by a decrease in the serum AGEs in a BP-independent manner.32 Various antihypertensive classes have been extensively tested to determine their efficacy for BP-independent reduction of the arterial stiffness, but RAAS blockers seem to be at least moderately more potent than others. This is mainly supported by several smaller studies,33, 34, 35, 36 and it was also confirmed in the large CAFE trial.37 The most common explanation for the effect of RAAS on arterial stiffness is that they block the pro-fibrotic action of the renin–angiotensin system, that is, turnover of key components in the extracellular matrix (that is, elastin and collagen) responsible for the mechanical properties of the arterial wall.38, 39 In the present study, we observed (Figure 2) that low sRAGE (as a surrogate of high AGE expression in the vessel wall) was associated with a significantly higher aPWV in subjects who were not treated with RAAS blockers, while there was no difference in the aPWV among the sRAGE categories in those treated with this class of drugs. Our results support the hypothesis that at least part of the effect of RAAS blockers on arterial stiffness could be attributed to their potential to effect on AGE metabolism.

Specific drugs may also directly influence the pathophysiological mechanism of the RAGE–AGE axis in arterial stiffening. Aminoguanidine blocks the formation of AGEs, while alagebrium (ALT-711, Alteon Corp.) non-enzymatically breaks AGE crosslinks of collagen fibers. With the exception of one human study,40 the effect of aminoguanidine has only been tested in animal models,41, 42 and the compound was eventually withdrawn from clinical use owing to its side effects. The therapeutic potential of alagebrium was only tested in smaller clinical studies. Kass et al.43 reported that 56-day treatment with alagebrium was followed by a significant 15% increase in arterial compliance and an 8% decrease in PWV compared with the placebo. However, these results were not replicated in another similarly designed clinical study.44 Treatment with alagebrium for 16 weeks was also followed by regression of left ventricular hypertrophy and improved Doppler indices of diastolic function in patients with diastolic dysfunction (where the pathophysiology of AGEs is also presumed).45 Clinical testing of alagebrium was discontinued afterwards, which was probably for economic reasons.

Conclusions

In the present study, we confirmed the pathophysiological role of AGEs in arterial stiffening. In our random general population sample, the association between RAGE and aortic stiffness was limited to hypertensive subjects, which is probably because of the causal crosslink between the RAAS system and AGE metabolism. We speculate that therapeutic manipulation of the RAGE–AGE axis represents a promising concept for further research into options for influencing age-related arterial stiffening.