The Relationship Between Adrenomedullin, Metabolic Factors, and Vascular Function in Individuals With Type 2 Diabetes

RESEARCH DESIGN AND METHODS—

We recruited four groups of Chinese subjects: 100 healthy subjects with normal glucose tolerance (fasting plasma glucose [FPG] ≤5.5 mmol/l), 60 with impaired fasting glucose (IFG) (FPG between 5.6 and 6.9 mmol/l), and 200 with type 2 diabetes (FPG ≥7.0 mmol/l) (15), half of whom did not have any evidence of nephropathy (diabetic group, n = 100) whereas the remaining 100 subjects had established diabetic nephropathy. Selection of these subjects was based on strict exclusion criteria. The healthy and IFG groups were recruited among working adults from the general population. Subjects from the healthy group were not taking long-term medications or have any known history of chronic medical illness such as diabetes or hypertension. The IFG subjects did not have any known history of chronic medical illness except that 12 self-reported having hypertension, of whom 7 were taking antihypertensive agents. These seven subjects were not excluded, since the number was small and the relationship between commonly used anitihypertensive agents and plasma adrenomedullin is still unclear based on existing literature. The type 2 diabetic subjects were recruited from the ambulatory care diabetes center of a secondary hospital. Type 2 diabetic subjects with normal renal function (the diabetic group) were strictly defined as early morning spot urinary albumin-to-creatinine ratio (ACR) ≤3.3 mg per mmol/l (i.e., 30 mg/g) and consistently normal serum creatinine. The diabetic nephropathy group (n = 100) was defined according to the presence of proteinuria ≥1.0 g/day (equivalent to spot urinary ACR ≥113 mg per mmol/l [i.e., 1,000 mg/g]) or persistently elevated serum creatinine with a mean MDRD (Modified Diet in Renal Disease) formula–estimated glomerular filtration rate (16) of ∼43 ml/min per 1.73 m² (2). Individuals were excluded from the diabetic nephropathy group when renal diseases attributable to other causes were suspected. These exclusion criteria included the presence of hematuria, renal insufficiency of unexplained origin, urinary tract infection, and history of rapidly progressive renal failure, glomerulonephritis, and polycystic kidney disease. Such strict criteria were employed so as to better understand the effect of diabetic kidney disease per se (and not other forms of renal impairment) on plasma adrenomedullin. To avoid misclassification, we decided to only include subjects with well-establish nephropathy in the diabetic nephropathy group because recent data suggested that early forms of diabetic nephropathy might remit spontaneously (17). As expected, more subjects from the diabetic nephropathy group suffered from retinopathy (44% nonproliferative and 31% proliferative) compared with subjects from the diabetic group (20% nonproliferative and 12% proliferative) (P < 0.01), since diabetic nephropathy is strongly associated with retinopathy.

We focused on the nondiabetic individuals (i.e., the healthy and IFG groups, n = 160) to explore the relationship between major metabolic and inflammatory biomarkers and MR-proADM. This was to avoid the influence of therapeutic agents such as glucose-, lipid-, and blood pressure–lowering agents, which were used extensively among the diabetic and diabetic nephropathy groups. To explore the relationship between vascular function and adrenomedullin, we randomly sampled 50 subjects from those individuals with type 2 diabetes to measure their resting forearm cutaneous microcirculatory perfusion (RCMP) using 2-dimensional laser Doppler flowmetry. We made this decision based on two considerations. First, we were interested in the relationship between plasma MR-proADM and vascular function in subjects with diabetes. Pooling heterogeneous subjects from all four study groups was therefore undesirable. Second, we observed from our initial results (Fig. 1) that subjects from the healthy and IFG groups had little variation in plasma MR-proADM concentrations. Therefore, to explore the relationship between RCMP and MR-proADM in these nondiabetic individuals would be statistically inefficient since the range of exposure (in this case, MR-proADM) was limited.

Major metabolic indicators

Anthropometric data were measured for all individuals. Blood pressure was measured using a sphygmomanometer according to standard procedures (18). Briefly, the subjects were rested for at least 15 min. Blood pressure was measured twice over the right arm 5 minutes apart in a sitting position. Should the two readings differ by >10 mmHg (either systolic or diastolic), a third reading will be taken and the average of the closest two readings will be recorded. Venous blood samples (taken after a 10-h fast) with EDTA as anticoagulant were kept in an icebox immediately after collection, and the plasma was separated from erythrocytes by centrifuging at 1,500g for 10 min at 4°C. The plasma, if not analyzed, was frozen at −80°C within 30 min of collection. Together with the fasting blood specimen, an early morning urine sample was collected for the measurement of urinary albumin and creatinine using commercial assay (Immulite; DPC UK) with a lower detection limit of 6 mg/l. Plasma total adiponectin concentration was measured using a commercial enzyme-linked immunosorbent assay kit (R&D Systems, Minneapolis, MN) with a maximum intra-assay coefficient of variance (CV) of 7.4% and interassay CV of 8.4%. Glucose measurements were carried out using the glucose oxidase method using the Vitros 700 Chemistry Analyser (Rochester, NY). Blood lipids (total cholesterol, triglycerides, and HDL cholesterol) were measured by enzymatic methods using Kodak Ektachem chemistry slides, which were then read on a Vitros 700 Chemistry Analyser. HDL cholesterol was measured after precipitation with dextran sulfate and magnesium chloride. LDL cholesterol was calculated using Friedewald's formula. Detection of MR-proADM was performed in duplicates in blinded samples using a novel sandwich immunoassay (B.R.A.H.M.S. Sevadil LIA; B.R.A.H.M.S., AG, Hennigsdorf/Berlin, Germany) as described in detail elsewhere (14). The assay has an analytical detection limit of 0.08 nmol/l, and the interassay CV is <20% for values >0.12 nmol/l. The assay is linear on dilution with undisturbed recovery of the analyte. EDTA, heparin, and citrate plasma samples are stable (<20% loss of analyte) for at least 3 days at room temperature, 14 days at 4°C, and 1 year at −20°C.

Measurement of forearm cutaneous microcirculatory function

All vascular reactivity measurements were performed on the same morning as the clinical evaluation while the subjects were still fasting. A single investigator, who performed all the measurements (S.K.G.), was blinded to the medical history of the subjects.

Detailed methods for the measurement of cutaneous microcirculatory function have been previously reported by our collaborators (19). Briefly, the skin over the extensor surface of the forearm was tested by performing laser Doppler perfusion imaging measurements at baseline and after the iontophoresis of acetylcholine (endothelium-dependent vasodilation) and sodium nitroprusside ([NaNP] endothelium-independent vasodilatation) using a laser Doppler perfusion imager (Lisca PIM 1.0; Lisca Development AB, Linkoping, Sweden). Baseline RCMP (i.e., erythrocyte flux, in volts) was first measured. Perfusion over the same area was again quantified after iontophoresis. The percentage of increase (δ) in perfusion over baseline as a reflection of magnitude of vasodilation (i.e., vascular reactivity) in response to acetylcholine and NaNP was then calculated. The reproducibility of the technique has been previously reported (20). The CV of the baseline measurement was 14.1% and during maximal hyperemic response after the iontophoresis 13.7%.

The study was approved by the relevant ethics committee and institution review board. Written informed consent was obtained from all the participants.

Statistical analysis

SPSS for Windows (version 11.5; SPSS, Chicago, IL) was used for statistical analyses. Comparison of proportions was carried out using χ² test for independence. BMI was calculated as weight in kilograms divided by the square of height in meters. A formula was used to derive the homeostasis model assessment of insulin resistance (HOMA-IR): (fasting insulin × glucose)/22.5 (21).

Analysis was not stratified by sex, as MR-proADM was unaffected by sex (14). ANOVA with post hoc Tukey's honestly significant difference test was used for comparison of MR-proADM between the four groups. Subsequently, ANCOVA was employed to adjust for potential confounders (sex, age, BMI, and blood pressure) to determine whether MR-proADM levels remained significantly different between groups. Since the conclusions were unaffected by adjustment for potential confounders, the unadjusted MR-proADM levels were reported.

Among subjects without diabetes, Pearson correlation was employed to explore the relationship between plasma MR-proADM concentrations and multiple metabolic factors (age, BMI, waist circumference, blood pressure, fasting glucose, full lipids profile, HOMA-IR, early morning spot urinary ACR, highly sensitive C-reactive protein [hsCRP], and total adiponectin). Results were ranked according to P values (Table 2). Metabolic factors (i.e., HOMA-IR, hsCRP, triglycerides, LDL cholesterol, adiponectin), which showed a potentially important correlation (P ≤ 0.1) with plasma MR-proADM (dependent variable), were incorporated as independent variables in subsequent linear regression analysis. Waist circumference was not included in the multivariate analysis due to strong colinearity with HOMA-IR (β = 0.80). In fact, waist circumference is often used as a surrogate measurement of insulin resistance in epidemiological studies (22).

Among the subpopulation of subjects with diabetes who had RCMP measured, Pearson correlation was employed to explore the relationship between plasma MR-proADM concentrations, RCMP, and postchallenge change (δ) in microcirculatory perfusion. To further quantify the strength of association between MR-proADM and RCMP, linear regression was performed using MR-proADM as the independent variable and RCMP as the dependent variable. The impact of nephropathy status (i.e., as defined in diabetic and diabetic nephropathy groups) on the relationship between MR-proADM and RCMP was also examined in multivariate analysis using a general linear model (GLM). In the GLM, MR-proADM and nephropathy status (1 = diabetic nephropathy group, 0 = diabetic with normal renal status group) were independent variables, whereas RCMP was the dependent variable. A P value of <0.05 was considered statistically significant.

RESULTS—

The clinical characteristics and plasma MR-proADM concentrations of the four groups of subjects studied are shown in Table 1. The difference in unadjusted plasma MR-proADM is also shown in Fig. 1. MR-proADM increased progressively from healthy individuals to type 2 diabetic patients with nephropathy (P < 0.001 for trend, Table 1). Post hoc pairwise comparisons revealed diabetic versus healthy and IFG, P = 0.04; diabetic nephropathy versus all, P < 0.01 (Fig. 1).

Correlations between MR-proADM and multiple metabolic factors ranked according to P values are also shown in Table 2. Multiple linear regressions (dependent variable: MR-proADM; independent variables: hsCRP, HOMA-IR, triglycerides, LDL cholesterol, and adiponecntin) revealed that hsCRP, HOMA-IR, LDL cholesterol, and adiponectin were significant predictors of plasma MR-proADM concentration with standardized CVs (β) of 0.11 (P = 0.01), 0.20 (P = 0.001), 0.31 (P < 0.001) and 0.33 (P < 0.001), respectively. Collectively, hsCRP, HOMA-IR, LDL cholesterol, and adiponectin predicted ∼84% of changes in plasma MR-proADM concentration.

Detailed results of RCMP among a random sample of 50 subjects taken from individuals with diabetes (i.e., diabetic [n = 39] and diabetic nephropathy [n = 11] groups) were as follows. The resting forearm skin temperature was 30.5 ± 0.5°C. The RCMP, measured using 2-dimensional laser Doppler flowmetry, was 0.80 ± 0.25 V. Pearson correlation between RCMP and MR-proADM among these 50 subjects with diabetes revealed moderately strong correlation (r = 0.43, P = 0.002) (Fig. 2). Linear regression (dependent variable, MR-proADM; independent variable, RCMP) revealed a CV (β) of 0.80 (95% CI 0.30–1.30; P = 0.002). In other words, a unit change in plasma MR-proADM concentration will, on average, result in a 0.8 unit increase in cutaneous blood flow.

GLM analysis (to study the impact of nephropathy status on the relationship between MR-proADM and RCMP) revealed that nephropathy status had no significant influence on the observed relationship (β = 0.004, 95% CI −0.16 to 0.15; F statistics = 0.003, P = 0.96). Therefore, analysis of RCMP was not stratified by renal status.

Correlations between MR-proADM and indexes of endothelial-dependent and -independent vascular reactivity (i.e., percentage of increase [δ] in perfusion after iontophoretic transcutanoues delivery of vasoactive substances acetylcholine and NaNP) were unremarkable: acetylcholine challenge δ perfusion (r = −0.19, P = 0.17), NaNP challenge δ perfusion (r = −0.20, P = 0.16).

CONCLUSIONS—

There were three main findings in our study. First, plasma MR-proADM concentration was elevated in type 2 diabetic subjects with preserved renal function. This was further accentuated in the presence of established diabetic nephropathy. Second, metabolic and inflammatory factors, namely insulin resistance, LDL cholesterol, adiponectin, and hsCRP, appeared to be significant determinants of plasma MR-proADM concentrations. Third, MR-proADM was well correlated with measurement of resting, unprovoked microcirculatory blood flow, suggesting that it may be one of the determinants of basal vascular perfusion in subjects with type 2 diabetes.

The first finding was interesting because it clarified the relationship between plasma adrenomedullin concentrations and type 2 diabetes. Our data suggested that MR-proADM was mildly elevated in subjects with uncomplicated type 2 diabetes (Fig. 1). However, in the presence of diabetic nephropathy, plasma MR-proADM became markedly deranged (consistent with Kinoshita et al. [13]). The strict criteria that we adopted for the recruitment of subjects in the diabetic (normal renal function) and diabetic nephropathy (well established nephropathy) groups was an advantage in helping to define the relationship between diabetes and adrenomedullin; i.e., adrenomedullin level increased with increasing severity of diabetes. Based on sequence homology, adrenomedullin was thought to belong to the calcitonin gene–related peptide superfamily (23). It was shown to be secreted from all three types of cultured vascular cells: endothelium, vascular smooth muscle cell, and adventitial fibroblast. Previous elegant studies suggested kidney as an important source of adrenomedullin (24), whereas pulmonary vascular bed could be the main site of receptor-dependent clearance of circulating adrenomedullin (25). In contrast, the exact site of clearance for MR-proADM is not known at the moment. Therefore, one possible explanation for the elevated plasma MR-proADM concentration among type 2 diabetes with nephropathy could be reduced MR-proADM clearance. It is known that adrenomedullin could exert a wide range of vascular actions (mostly protective). These include endothelium-dependent and -independent vasodilation, antioxidative stress, stimulation of endothelial nitric oxide production, antiproliferation of vascular smooth muscle cell, and adventitial fibroblast (26). Taken together, the elevation of plasma MR-proADM concentration in type 2 diabetes (especially in the presence of nephropathy) could be an appropriate physiological response to ongoing vascular injury (27).

Factors that upregulate adrenomedullin production are incompletely understood. The role of hyperglycemia is controversial. In vitro data suggested that hyperglycemia might increase vascular adrenomedullin expression (10). However, this notion could not be substantiated in vivo (7,8). Other postulated mechanisms included acute hyperinsulinemia (28), increased oxidative stress (based on in vivo studies) (29), and proatherogenic/inflammatory factors such as angiotensin II, endothelin-1 (30), interleukin-1β, and tumor necrosis factor-α (31) (based on cell culture studies). Our analysis revealed that among the multitude of metabolic factors examined (Table 2), insulin resistance, hsCRP, LDL cholesterol, and adiponecitn were significant determinants of plasma MR-proADM concentrations. Our observation was noteworthy given that insulin resistance had been associated with vascular injury (32) and diabetic nephropathy (33,34). In addition, insulin resistance was found to be associated with low-grade endothelial inflammation (manifesting as elevated hsCRP) (35), which has become increasingly recognized as a determinant of vasculopathy (36). The relationship between LDL cholesterol and adrenomedullin is poorly understood. Limited in vitro data from endothelial cells in rats suggested that oxidized LDL might stimulate the secretion of adrenomedullin (37). Adiponectin has emerged as one of the most important adipocytokines at the crossroad of energy homeostasis, inflammation, and vascular injury (38). To our knowledge, the relationship between adiponectin and adrenomedullin has not been well studied. Nevertheless, very recent studies suggested that adiponectin might be associated with reduced odds of renal dysfunction in subjects with type 2 diabetes (39). Therefore, the observed relationship between adiponectin and MR-porADM in our study is potentially novel and may require further investigations. Taken together, our data suggested that vasculopathic metabolic and inflammatory factors were associated with upregulation of MR-proADM, probably a response to injury. To identify the inciting factors is important as this could lead to the discovery of novel therapeutic interventions. For instance, insulin sensitizer had been found to ameliorate renal dysfunction in individuals with type 2 diabetes (40). Lipid-lowering therapy may retard the progression of renal impairment (41). Nevertheless, our study revealed that HOMA-IR, hsCRP, LDL cholesterol, and adiponectin collectively accounted for ∼84% of variation in plasma MR-proADM concentrations. Therefore, other determinants of plasma MR-proADM unidentified in our study should be investigated in future studies.

We observed that plasma MR-proADM concentration correlated moderately well with basal cutaneous microcirculatory blood flow among subjects with type 2 diabetes. This suggested that adrenomedullin-mediated vasodilation could probably be one of the determinants of basal microcirculatory perfusion. A growing body of animal and human pilot studies corroborated our present observation. These pilot studies showed that adrenomedullin induced vasodilation in isolated canine (42) and bovine (43) retinal arteries (another microcirculatory bed) in vitro. Similarly, adrenomedullin also induced vasodilation of retinal arteries of diabetic male Wistar rats in vivo (44). In a small number of healthy human subjects, Dorner et al. (45) reported that adrenomedullin dose dependently increased choroidal blood flow and flow velocity in the ophthalmic artery. Therefore, the novelty in our present observation was intriguing. One mechanism identified in the pathogenesis of diabetic nephropathy was hemodynamic-mediated vascular injury (46). Sustained increase in glomerular capillary pressure driven by increase in plasma flow had been observed, especially in early stages of nephropathy. The elevation in glomerular capillary pressure might be damaging to glomerular endothelial, epithelial, and mesangial cells, thereby initiating and contributing to the progression of nephropathy (47). Although numerous mediators of diabetic hyperfiltration had been proposed, the exact mechanism remained unclear (48). It is therefore tempting to speculate that endothelial-derived vasodilatory substances like adrenomedullin could be involved, since MR-proADM was increasingly elevated from healthy to renal-impaired subjects (Fig. 1) and well correlated to magnitude of microcirculatory perfusion (Fig. 2). Should this be the case, modulating adrenomedullin action would have therapeutic potential in the prevention of diabetic nephropathy (5).

There are a few limitations in our study. It would be ideal to only recruit individuals (healthy or diabetic) not receiving any form of pharmacological interventions (so as to observe the unbiased and unconfounded relationship between metabolic factors, adrenomedullin levels, and different disease state). This, however, is unrealistic since individuals with diabetes and diabetic nephropathy are expected to receive intensive treatment according to current standards of care. Secondly, the safe and noninvasive measurement of microcirculatory perfusion was based on cutaneous vasculature (instead of direct measurement of renal hemodynamics). Although experience from our group suggests that cutaneous microcirculatory function correlates well with measurement of severity of diabetic nephropathy (49), it was only a surrogate and could not help us gain direct insights into actual changes in renal hemodynamics. Third, we did not measure the serum creatinine of all the study subjects. Although we observed incremental plasma MR-proADM concentration according to categorical diabetic and renal function status, we were unable to directly investigate the relationship between plasma MR-proADM and renal function (as estimated by serum creatinine). This said, the limitation of serum creatinine as an estimate of renal function (i.e., glomerular filtration rate) has been well recognized (16).

In conclusion, our study revealed that plasma MR-proADM concentration was elevated among subjects with type 2 diabetes, which was further accentuated when nephropathy set in. MR-proADM was related to multiple metabolic factors and basal microcirculatory perfusion. Therefore, adrenomedullin might play a role in the pathogenesis of diabetic vasculopathy.