SEARCH

SEARCH BY CITATION

Keywords:

  • HMGB1;
  • Esrage;
  • Crage;
  • Chronic heart failure;
  • Diabetes

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

Aims

High-mobility group box-1 (HMGB1) is a ligand for the receptor for advanced glycation endproducts (RAGE). An HMGB1–RAGE interaction has been implicated in cardiac dysfunction. We assessed the association of HMGB1 and RAGE isoforms with heart failure (HF) in diabetic and non-diabetic patients.

Methods and results

We assayed serum levels of HMGB1, cleaved RAGE (cRAGE), endogenous secretory RAGE (esRAGE), high-sensitivity C-reactive protein (hsCRP), and N-terminal pro-brain natriuretic peptide (NT-proBNP) in parallel with assessment of left ventricular volumes and function in 125 diabetic and 222 non-diabetic Chinese patients with chronic HF. Of the total, 79 diabetic patients without HF and 220 normal subjects served as diabetic and normal controls, respectively. Serum HMGB1, cRAGE, hsCRP, and NT-proBNP levels were higher and, in contrast, esRAGE levels lower in HF patients than in subjects without HF (for all; P < 0.01), with higher levels of cRAGE and hsCRP in diabetic HF vs. non-diabetic HF patients (P < 0.01). For HF patients—with or without diabetes—HMGB1 levels correlated positively with left ventricular end-diastolic and end-systolic volumes (r = 0.267 and r = 0.321, respectively) and NT-proBNP values (r = 0.497), and were inversely related to ejection fraction (r = −0.461; all P < 0.001). Serum cRAGE levels correlated with NT-proBNP values (r = 0.451) and New York Heart Association functional class (r = 0.402; both P < 0.001). Multivariable regression analysis revealed that HMGB1, cRAGE, and esRAGE were consistently associated with HF in diabetic and non-diabetic patients.

Conclusion

Heart failure patients have increased serum HMGB1 and cRAGE and decreased esRAGE levels, and these are related to the severity of HF in both diabetic and non-diabetic patients. Such associations are worth further investigation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

Interaction of advanced glycation endproducts (AGEs) with the main receptor (RAGE), and subsequent intracellular signalling, has been implicated in diabetic complications.1 Advanced glycation endproducts, acting as pro-inflammatory triggers, are associated with endothelial dysfunction and coronary artery disease, in both diabetic 24 and non-diabetic patients.5 Interestingly, AGEs have also been recently implicated in cardiac dysfunction,69 thus emerging as a risk factor for poor clinical outcomes in chronic heart failure (HF).10,11

Besides full-length membrane-bound RAGE, two major additional isoforms have been recently identified, namely the endogenous secretory RAGE (esRAGE) and a cleaved form of RAGE (cRAGE).12 Endogenous secretory RAGE, a quantitatively minor isoform, is generated through alternative splicing of pre-mRNA; cRAGE is proteolytically cleaved from the cell surface by matrix metalloproteinases, and then shed into the bloodstream. Both variants may act as decoy ligands for AGEs and several inflammatory cytokines. Recent studies have, however, suggested a reciprocal relationship between these two ligands, whereby decreased esRAGE and/or increased cRAGE levels are biomarkers of heightened ligand–RAGE interaction in diabetes, atherosclerosis, and other inflammatory diseases, possibly underscoring an inadequate endogenous protective response.13 This suggestion has been recently extended to HF, where esRAGE levels are considered a negative prognostic factor.14

High-mobility group box-1 (HMGB1) is a nuclear DNA-binding protein, passively released from necrotic cells and actively secreted by activated immune cells.15 Recently, HMGB1 has been demonstrated to be a novel ligand for RAGE, cRAGE, and esRAGE.16,17 The engagement of HMGB1 with RAGE promotes the shedding of the receptor.18 Its signalling via RAGE activates inflammatory pathways and intensifies cellular oxidative stress, which results in profuse production of inflammatory cytokines and elevated expression of adhesion molecules.17 Increased HMGB1 levels are associated with ischaemia–reperfusion injury in mice, and with coronary artery disease in diabetic patients,16,19 and are also involved in post-infarction inflammatory response and left ventricular remodelling.20 On the other hand, low levels of HMGB1 have also distinctly favourable biological properties, being capable of attracting stem cells,21 facilitating myocardial cell regeneration and differentiation,22 enhancing angiogenesis, and consequently improving myocardial function and survival after myocardial infarction.23 Indeed, low-grade inflammation elicited by intra-myocardial injection of HMGB1 can favour the recovery of chronic post-infarction cardiac remodelling and limit HF.24

Research over the last two decades has provided some clues about the association of myocardial dysfunction with inflammation, increased cytokine production, and fibrous tissue deposition. Inflammatory activation and increased cytokine production, thus, apparently play some role in the clinical development and progression of HF.25 Based on these findings, we hypothesized that HMGB1 concentrations may mark—and perhaps contribute to—the development of HF, with accompanying esRAGE and cRAGE changes. We predicted a graded relationship between HF severity and changes in plasma levels of HMGB1 and related proteins. We therefore evaluated the association of serum concentrations of HMGB1 (primary objective), as well as esRAGE and cRAGE, together with levels of high-sensitivity C-reactive protein (hsCRP) as a marker of inflammation and N-terminal precursor of brain natriuretic peptide (NT-proBNP) (secondary objectives) as a HF biomarker, in non-diabetic and type 2 diabetic patients with chronic HF.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

The study protocol was approved by the local hospital Ethics Committee, and written informed consent was obtained from all participants.

Study population

We restricted the inclusion to Chinese patients of Han nationality with systolic chronic HF due to ischaemic or idiopathic dilated cardiomyopathy. The study, therefore, included 149 consecutive patients with ischaemic HF (mean age: 68.3 ± 9.9 years; duration of disease: 1.5 ± 2.1 years), and 198 patients with HF caused by idiopathic dilated cardiomyopathy (mean age: 57.0 ± 13.5 years; duration of disease: 2.9 ± 3.3 years) recruited between January 2000 and May 2006 from three teaching hospitals affiliated to the Shanghai Jiaotong University School of Medicine in China. Systolic HF was diagnosed according to the European Society of Cardiology guidelines, including patients with symptoms or signs of HF and left ventricular ejection fraction (LVEF) <45%, assessed by echocardiography. For patients with idiopathic dilated cardiomyopathy and cardiovascular risk factors (e.g. with associated type 2 diabetes), the presence of significant coronary artery disease was excluded by coronary angiography. Of the 347 patients with HF, 222 were non-diabetic and 125 had type 2 diabetes. Diabetes was defined according to the American Diabetes Association criteria as: two fasting plasma glucose levels ≥7.0 mmol/L, or symptoms of diabetes plus a casual post-prandial plasma glucose reading ≥11.1 mmol/L, or a 2 h glucose reading ≥11.1 mmol/L after a 75 g glucose load, or taking oral hypoglycaemic drugs or parenteral insulin. To avoid confounding variables, we excluded patients with a history of viral myocarditis, hypertrophic cardiomyopathy, primary valvular disease, or pulmonary heart disease. We also excluded patients with type 1 diabetes, chronic viral or bacterial infections, tumours, or immune disorders. Detailed information was obtained on demographics, clinical manifestation, and medications, as well as New York Heart Association (NYHA) functional class, and echocardiographic measurements were used to evaluate HF severity.

Two hundred and twenty normal Chinese subjects (132 men and 88 women; mean age: 60 ± 13 years) and 79 Chinese patients with type 2 diabetes only (37 men and 42 women; mean age: 64 ± 10 years) served as normal and diabetic controls, respectively. Their gender distribution was the same as in the general population; age was purposely matched between controls and the HF population. Detailed medical and family history was taken, and fasting blood samples were collected during an annual physical check-up. In the normal controls, serum levels of glucose, lipid profiles, liver and renal function tests, and the electrocardiogram were normal in all subjects, and none had a history of cardiovascular diseases (including past history of angina/myocardial infarction). In the diabetic control group, we excluded patients with prior coronary heart disease, stroke, or severe renal dysfunction.

Biochemical investigations

Peripheral venous blood samples were collected after an overnight fast. Serum glucose, blood urea nitrogen, creatinine, uric acid, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, lipoprotein (a), apolipoprotein A, apolipoprotein B, and triglycerides were measured using standard laboratory techniques on a Hitachi 912 Analyser (Roche Diagnostics, Mannheim, Germany). Serum NT-proBNP was determined using a commercially available electrochemiluminescence immunoassay kit (Roche Diagnostics). Serum HMGB1 levels were assessed with an enzyme-linked immunosorbent assay (ELISA) kit (HMGB1 ELISA kit II; Shino-Test Corporation, Tokyo, Japan) according to the manufacturer's instructions. The detection limit for HMGB1 was 0.2 ng/mL, with an inter-assay coefficient of variation <10%. Levels of cRAGE and esRAGE were also measured using ELISA kits (Quantikine; R&D Systems, MN, USA; B-Bridge International, Mountain View, CA, USA, respectively).

Echocardiographic assessment

Transthoracic two-dimensional echocardiography was performed using a Hewlett-Packard Sonos 2500 (Hewlett-Packard, San Diego, CA, USA) or a GE Vivid-7 system (General Electric Vingmed Sound AS, Horton, Norway) equipped with 2.5 or 1.7/3.4 MHz transducers, respectively. Images were obtained at rest with the patient lying in the left lateral decubitus position at end-expiration. Left ventricular end-diastolic and end-systolic volumes were measured according to the biplane Simpson's method based on the American Society of Echocardiography recommendations, and LVEF was calculated. An average of three consecutive cardiac cycles was used for each patient.

Follow-up

All HF patients were prescribed standard HF treatments and were seen every 1–3 months in a dedicated HF clinic. During each visit, heart rate, blood pressure, and new clinical manifestations were recorded; echocardiography was performed every 6 months. Adverse events (hospitalization or death from HF) were recorded during each visit or by telephone with patients or family members. Hospitalization for HF was defined as that due to progressive fluid retention and the need to increase or change medications. Two trained physicians independently reviewed all medical notes, including forms from visits to the emergency department and hospital medical records.

Statistical analysis

Continuous variables are presented as mean ± standard deviation (SD). Differences between groups were compared with two-factor (HF, diabetes) analysis of variance followed by Dunnett post hoc between-group analysis or by the non-parametric Kruskal–Wallis test. Categorical data were summarized as frequencies or percentages, and differences between groups were evaluated by the χ2 test. The sample size (347 patients with HF and 299 subjects without HF, with or without diabetes) was such to yield >80% power to detect differences between HF and non-HF patients for HMGB1, NT-proBNP, cRAGE, and hsCRP (for all, 99% power both in non-diabetic and diabetic patients), as well as esRAGE levels (99% in non-diabetic and 87% in diabetic patients, respectively) under a type I error probability of 0.05 (1−α) for a two-sided test.

Correlations of serum HMGB1 (on a logarithmic scale) with other biomarkers or echocardiographic measurements (left atrial diameters and ventricular volumes) were assessed by Pearson's test, and associations of these biomarkers with LVEF (not normally distributed) or NYHA functional class (categorical variable) by Spearman's test. The linear correlation of LVEF with HMGB1 was still represented with the Pearson's method to represent their crude distributions. We used two models in multivariable logistic regression analysis for the presence of HF in non-diabetic and diabetic subjects, respectively. In Model I, multivariable adjustment was made for conventional risk factors measured at baseline examination, and included age, gender, smoking, hypertension, systolic/diastolic blood pressure, triglycerides, total cholesterol, blood urea nitrogen, creatinine, uric acid, fasting glucose, and glycated haemoglobin (by a conditional logistic regression method). In Model II, the multivariable-adjusted odd ratios (ORs) and their 95% confidence intervals (CI) for HF associated with the biomarkers of interest were compared with the respective normal- and diabetic-matched controls, synchronously estimated together with significantly independent conventional risk factors established in Model I (by the backward conditional logistic regression method). In addition, the ORs were given for a 1- or 1/2-SD increase of each biomarker, blood pressure, creatinine, and uric acid in control group.

All statistical analyses were done using the SPSS version 13.0 software (SPSS, Inc., Chicago, IL, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

Clinical characteristics

The diabetic and non-diabetic patients with HF were more frequently male and cigarette smokers than the normal or diabetic controls. Compared with non-diabetic patients with HF, those with both HF and diabetes were older and had a higher ratio of ischaemic vs. non-ischaemic aetiology, a poorer renal function, and more abnormal lipid profiles. No significant differences between HF patients with and without diabetes were observed for left ventricular end-diastolic or end-systolic volumes or LVEF, with the exception of a larger left atrial size in diabetic patients with HF. During a 2-year follow-up, 22 out of 125 (17.6%) diabetic and 23 out of 222 (10.4%) non-diabetic patients with HF died of sudden cardiac death or refractory HF (Table 1).

Table 1. Baseline clinical characteristics and biochemical assessments of patients studied
 Non-diabetic patientsDiabetic patients
 No HF (n = 220)HF (n = 222)P-valueNo HF (n = 79)HF (n = 125)P-value
  • a

    Values are given as mean (standard deviation) or number (percentage).

  • b

    HF, heart failure.

  • **

    P < 0.01, diabetic patients with no HF vs. normal controls [non-diabetic subjects without HF (no HF)].

  • P < 0.05.

  • P < 0.01, diabetic patients with HF vs. normal controls.

  • P < 0.05.

  • §

    P < 0.01, diabetic patients with HF vs. non-diabetic patients with HF.

Male/female (n)132/88167/55 37/4294/31 
Age (years)59.5 (13.2)60.4 (14.5)0.5564.1 (10.4)**64.4 (10.4),§0.137
Aetiology (ischaemic/dilated cardiomyopathy, n) 68/154  81/44§ 
Hypertension [n (%)]25 (11.4)94 (42.3)<0.00153 (67.4)94 (75.4)0.208
Cigarette smoking [n (%)]19 (14.4)101 (45.4)<0.00116 (20.3)48 (38.6)0.007
NYHA functional class II/III/IV (n) 109/107/6  29/85/11 
Systolic blood pressure (mmHg)123.1 (18.5)124.5 (20.4)0.53136.0 (19.5)129.1 (20.4),0.080
Diastolic blood pressure (mmHg)75.2 (10.2)77.1 (11.9)0.21580.1 (10.6)79.0 (11.2)0.342
Heart rate (b.p.m.)71.6 (4.7)77.8 (16.1)0.18477.1 (9.7)77.8 (13.1)0.954
NT-proBNP (pg/mL)157.9 (195.3)3297.4 (5409.8)<0.001172.1 (361.8)2659.0 (3302.7)<0.001
Left atrial diameter (mm)37.5 (4.8)45.0 (6.3)0.00740.0 (6.8)46.9 (7.1),§<0.001
Left ventricular end-diastolic volume (mL)107.3 (27.0)294.3 (122.9)<0.001104.3 (16.8)276.0 (134.6)<0.001
Left ventricular end-systolic volume (mL)37.4 (12.7)158.1 (84.8)<0.00135.5 (10.8)151.4 (99.1)<0.001
Ejection fraction (%)63.3 (5.9)36.0 (6.9)<0.00165.1 (6.1)36.6 (7.6)<0.001
Fasting glucose (mmol/L)4.73 (0.50)5.27 (1.16)0.0016.91 (2.38)**6.66 (2.32),§0.036
Glycated haemoglobin (HBA1c, %)5.85 (0.35)5.99 (0.52)0.8637.88 (1.27)**7.30 (1.24),§0.001
Blood urea nitrogen (mmol/L)5.04 (1.22)6.93 (3.18)<0.0015.81 (1.78)7.60 (2.72)0.003
Creatinine (μmol/L)79.1 (17.3)104.6 (70.5)<0.00184.2 (21.9)104.8 (42.5)<0.001
Uric acid (μmol/L)310.5 (65.0)395.3 (120.5)<0.001324.9 (72.9)394.8 (134.2)<0.001
Triglycerides (mmol/L)1.09 (0.50)1.53 (0.96)<0.0012.09 (1.33)**1.64 (1.02)<0.001
Total cholesterol (mmol/L)4.62 (0.65)4.31 (0.98)<0.0014.82 (1.12)4.21 (0.98)<0.001
High-density lipoprotein cholesterol (mmol/L)1.42 (0.28)1.21 (0.45)<0.0011.20 (0.29)1.07 (0.30),0.020
Low-density lipoprotein cholesterol (mmol/L)2.79 (0.58)2.50 (0.78)0.0112.80 (0.86)2.52 (0.81)0.025
Apoprotein A (g/L)1.24 (0.17)1.15 (0.21)<0.0011.26 (0.19)1.09 (0.23)0.001
Apoprotein B (g/L)0.78 (0.15)0.85 (0.19)0.0410.95 (0.25)0.85 (0.20)0.020
Lipoprotein (a) (g/L)0.19 (0.14)0.24 (0.21)0.1120.19 (0.16)0.19 (0.18)0.314
Angiotensin-converting enzyme-inhibitors [n (%)] 162 (72.8)  73 (58.8)0.022
Angiotensin receptor blockers [n (%)] 52 (23.4)  43 (34.7)0.035
β-Blockers [n (%)] 209 (94.2)  113 (90.1)0.193
Nitrates [n (%)] 107 (48.1)  93 (74.3)<0.001
Statins [n (%)] 61 (27.7)  67 (53.5)<0.001
Diuretics [n (%)] 167 (75.2)  88 (70.7)0.400
Aspirin [n (%)] 183 (80.7)  117 (90.8)0.018
Digoxin [n (%)] 160 (72.3)  83 (66.3)0.280
Hospitalization during the 2-year follow-up [n (%)] 99 (44.6)  67 (53.6)0.107
Death during the 2-year follow-up [n (%)] 23 (10.4)  22 (17.6)0.054

Influence of heart failure and its aetiology on biological analytes

Serum levels of HMGB1(7.57 ± 8.70 vs. 2.66 ± 4.22 ng/mL), cRAGE (1005.7 ± 1176.3 vs. 594.5 ± 401.5 pg/mL), hsCRP (16.80 ± 17.08 vs. 7.83 ± 6.67 mg/L), and NT-proBNP (3509.9 ± 5206.8 vs. 201.9 ± 567.5 pg/mL) were higher, but esRAGE levels (241.5 ± 243.0 vs. 345.5 ± 146.3 pg/mL) were lower in the overall population of patients with HF than in those without HF (all P < 0.01), while elevated levels of cRAGE (643.5 ± 1176.3 vs. 594.5 ± 401.5 pg/mL in diabetic and non-diabetic subjects, respectively), hsCRP (11.39 ± 9.31 vs. 5.15 ± 4.38 mg/L; both P < 0.05), and reduced esRAGE levels (253.5 ± 109.9 vs. 415.3 ± 131.3 pg/mL; P < 0.001) were present in association with diabetes in patients without HF. In other words, levels of HMGB1 and NT-proBNP, in contrast to the other analytes assayed, were selectively influenced by the presence of HF. Higher levels of cRAGE and hsCRP, but not of HMGB1 or esRAGE, were observed in diabetic vs. non-diabetic patients with HF (for all; P < 0.01; Figure 1 and Table 2). In addition, esRAGE levels were selectively related to the aetiology of HF, with higher levels in dilated vs. ischaemic cardiomyopathy (260.5 ± 290.3 vs. 222.3 ± 182.0 pg/mL; P = 0.015); whereas cRAGE levels were influenced by both the aetiology of HF (P = 0.015) and diabetes (P = 0.003), with the highest levels seen in diabetic patients with dilated cardiomyopathy (1743.8 ± 1838.0 pg/mL), and the lowest levels in non-diabetic patients with ischaemic cardiomyopathy (752.5 ± 804.6 pg/mL). However, HMGB1 levels were not influenced by aetiology (5.38 ± 0.69 vs. 6.91 ± 0.72 ng/mL in ischaemic and dilated cardiomyopathy, respectively; P = 0.066) or the presence of diabetes (6.14 ± 0.79 vs. 6.10 ± 0.60 ng/mL in diabetic and non-diabetic patients, respectively; P = 0.123) in HF (Table 3).

Table 2. Serum levels of high-mobility group box-1 and other biological analytes in diabetic and non-diabetic patients with or without heart failure
 Non-diabetic patientsDiabetic patientsDiabetesHFDiabetes × HF
 No HF (n = 220)HF (n = 222)P-valueNo HF (n = 79)HF (n = 125)P-valueP1-valueP2-valueP3-value
  • a

    Values are given as mean ± standard deviation. P1, P2, and P3 values stand for the contributions of diabetes, HF, and HF aetiology-diabetes interaction to the differences of biological analytes among the four groups by two-way analysis of variance, respectively.

  • *

    P < 0.05.

  • **

    P < 0.01, diabetic patients with no HF vs. normal controls (non-diabetic subjects with no HF).

  • P < 0.01, diabetic patients with HF vs. normal controls.

  • §

    P< 0.05.

  • §§

    P < 0.01, diabetic patients with HF vs. non-diabetic patients with HF.

HMGB1 (ng/mL)1.84 ± 2.276.31 ± 7.58<0.0012.41 ± 3.465.99 ± 8.24<0.0010.423<0.0010.446
cRAGE (pg/mL)550.5 ± 397.1962.91 ± 1155.50.001666.1 ± 355.1*1159.6 ± 1309.7, §§<0.0010.035<0.0010.044
esRAGE (pg/mL)467.0 ± 128.5245.5 ± 281.7<0.001285.3 ± 109.0**237.2 ± 180.00.001<0.001<0.001<0.001
hsCRP (mg/L)3.32 ± 2.2320.68 ± 32.04<0.0017.08 ± 4.03*33.40 ± 23.20, §<0.0010.025<0.0010.033
NT-proBNP (pg/mL)157.9 ± 195.33297.4 ± 5409.8<0.001172.1 ± 361.82659.0 ± 3302.7<0.0010.411<0.0010.531
Table 3. Serum levels of high-mobility group box-1 and other biological analytes in diabetic and non-diabetic patients according to different heart failure aetiology
 Ischaemic cardiomyopathy (ICM)Dilated cardiomyopathy (DCM)DiabetesAetiologyDiabetes × aetiology
 No diabetes (n = 68)Diabetes (n = 81)No diabetes (n = 154)Diabetes (n = 44)P1-valueP2-valueP3-value
  • a

    Values are given as mean ± standard deviation. P1, P2, and P3 values stand for the contributions of diabetes, aetiology, and diabetes–aetiology interaction to the differences of biological analytes among the four groups by two-way analysis of variance, respectively.

  • *

    P < 0.05.

  • **

    P < 0.01, diabetic patients of different aetiology vs. non-diabetic patients of the same aetiology.

  • P < 0.01, diabetic patients of DCM vs. non-diabetic patients of ICM.

  • §

    P < 0.05.

  • §§

    P < 0.01, diabetic patients of DCM vs. diabetic patients of ICM.

  • P < 0.05, non-diabetic patients of DCM vs. non-diabetic patients of ICM.

HMGB1 (ng/mL)5.31 ± 4.875.60 ± 7.946.75 ± 8.486.87 ± 8.920.1470.0660.431
cRAGE (pg/mL)752.5 ± 804.6946.3 ± 1102.5*1001.3 ± 1196.11743.8 ± 1838.0**,, §§0.0150.0030.072
esRAGE (pg/mL)194.3 ± 114.6183.9 ± 123.0286.2 ± 188.4241.2 ± 294.1§0.0780.0150.248
hsCRP (mg/L)10.68 ± 15.7417.19 ± 25.5220.14 ± 34.1123.28 ± 39.840.0370.2610.352
NT-proBNP (pg/mL)1612.5 ± 1824.82987.1 ± 3376.93357.1 ± 3032.63944.4 ± 4204.00.1620.3820.309
image

Figure 1. Comparison of serum levels of high-mobility group box-1 (HMGB1), cleaved receptor for advanced glycation endproducts (cRAGE), endogenous secretory receptor for advanced glycation endproducts (esRAGE), high-sensitivity C-reactive protein (hsCRP), and N-terminal pro-brain natriuretic peptide (NT-proBNP) among diabetic and non-diabetic patients with or without HF

Download figure to PowerPoint

Serum HMGB1 levels correlated positively with hsCRP (Pearson's r = 0.208; P < 0.001) and cRAGE values (Pearson's r = 0.086; P = 0.042), and were inversely related to esRAGE (Pearson's r = −0.203; P < 0.001). Serum esRAGE correlated negatively with hsCRP levels (Pearson's r = −0.249; P < 0.001).

Association of biological analytes with heart failure severity

The relationship between serum measurements of HMGB1 levels and HF severity is depicted in Figure 2. High-mobility group box-1 was positively related to NT-proBNP (Pearson's r = 0.497; P < 0.001), and left ventricular end-diastolic and end-systolic volumes (Pearson's r = 0.267 and r = 0.321, respectively; both P < 0.001), but was inversely related to LVEF (Spearman's r = −0.461; P < 0.001) in both diabetic and non-diabetic HF patients. Moreover, HMGB1 levels showed a positive correlation with NYHA functional class in HF diabetic patients (Spearman's r = 0.184; P = 0.039), but not in HF non-diabetic patients (Spearman's r = 0.074; P = 0.291).

image

Figure 2. Correlation of serum high-mobility group box-1 (HMGB1) levels (on a natural logarithmic scale) with left ventricular volumes and ejection fraction (LVEF), and N-terminal pro-brain natriuretic peptide (NT-proBNP) levels (on a logarithmic scale) in heart failure (HF) patients. LVEDV, left ventricular end-diastolic volume; LVESV, left ventricular end-systolic volume. Due to positively skewed distribution of left ventricular ejection fraction values in heart failure patients, the correlation of high-mobility group box-1 levels with left ventricular ejection fraction was analysed by the Spearman's method, but displayed as Pearson's linear correlation to show the crude distributions of values

Download figure to PowerPoint

Similarly, serum levels of cRAGE were positively related to NT-proBNP levels independent of diabetes (Spearman's r = 0.451; P < 0.001), but were positively associated with left atrial diameters (Pearson's r = 0.360; P < 0.001), left ventricular end-diastolic and end-systolic volumes (Pearson's r = 0.358 and r = 0.328, respectively; both P < 0.001), and inversely related to LVEF (Spearman's r = −0.356; P < 0.001) only in diabetic patients with HF. In addition, in both non-diabetic and diabetic patients with HF, serum cRAGE levels correlated significantly with NYHA functional class (Spearman's r = 0.326 and r = 0.567, respectively; both P < 0.001). There was a stepwise elevation of cRAGE levels with worsening NYHA functional classes [median and 25–75% percentiles: 260 (83–663) vs. 381 (155–1160) pg/mL in class II; 995 (519–2297) vs. 1085 (428–2733) pg/mL in class III; 5838 (3194–6489) vs. 2220 (802–5175) pg/mL in class IV, in non-diabetic vs. diabetic patients, respectively].

However, serum esRAGE levels correlated inversely with NT-proBNP levels (Pearson's r = −0.303; P < 0.001) only in non-diabetic patients, and with left atrial diameters (Pearson's r = −0.259; P = 0.003) and left ventricular end-systolic volumes (Pearson's r = −0.175; P = 0.048) only in diabetic patients with HF. No significant correlation between esRAGE and LVEF was found in either group (both P > 0.05). In addition, esRAGE levels were negatively related to NYHA functional class in diabetic and non-diabetic patients with HF (Spearman's r = −0.292 (P = 0.001) and r = −0.161 (P = 0.011), respectively). The detailed association between these biological analytes and HF severity is shown in Table 4.

Table 4. Association of biological analytes with disease severity in heart failure patients
Variables correlatedTotal HF patientsNon-diabetic patients with HFDiabetic patients with HF
 Correlation coefficientP-valueCorrelation coefficientP-valueCorrelation coefficientP-value
  • a

    LAD, left atrial diameter (mm); LVEDV, left ventricular end-diastolic volume (mL); LVESV, left ventricular end-systolic volume (mL).

  • a,b

    Correlation coefficients were calculated by the Spearman's method (NYHA class being categorical or LVEF non-normally distributed); the others were calculated by the Pearson's method (both variables being continuous).

HMGB1—NT-proBNP0.378<0.0010.347<0.0010.2900.002
HMGB1—LAD0.1760.0200.1280.0230.2130.012
HMGB1—LVEDV0.1340.0260.1530.0210.1240.031
HMGB1—LVESV0.1530.0110.1780.0090.1380.021
HMGB1—LVEFa,b−0.1270.021−0.1810.011−0.1030.024
HMGB1—NYHA classa,b0.0850.1310.0740.2910.1840.039
cRAGE—NT-proBNP0.451<0.0010.446<0.0010.4770.001
cRAGE—LAD0.221<0.0010.1100.1160.360<0.001
cRAGE—LVEDV0.0960.0760.0930.1820.328<0.001
cRAGE—LVESV0.1340.0130.0500.4820.358<0.001
cRAGE—LVEFa,b−0.1790.001−0.0150.663−0.356<0.001
cRAGE—NYHA classa,b0.402<0.0010.567<0.0010.326<0.001
esRAGE—NT-proBNP−0.227<0.001−0.303<0.001−0.1320.124
esRAGE—LAD−0.1350.011−0.0550.213−0.2590.003
esRAGE—LVEDV−0.0650.396−0.0170.423−0.1720.134
esRAGE—LVESV−0.1520.258−0.1380.321−0.1750.048
esRAGE—LVEFa,b−0.0490.341−0.0190.316−0.0740.423
esRAGE—NYHA classa,b−0.223<0.001−0.1610.011−0.2920.001

Multivariable logistic regression analysis for the presence of heart failure

Multivariable stepwise logistic regression analysis run in all subjects, with or without HF, revealed that, adjusting for traditional cardiovascular risk factors (Model I), smoking (OR: 2.235, 95%: CI 1.455–4.582; P = 0.009), blood urea nitrogen (OR: 1.596, 95%: CI 1.162–2.193; P = 0.004), and systolic blood pressure (OR: 0.386, 95%: CI 0.194–0.771; P = 0.007, for a 1-SD increase) were independent determinants for HF in non-diabetic subjects, while blood urea nitrogen (OR: 1.314, 95%: CI 1.068–1.613; P = 0.012), systolic blood pressure (OR: 0.386, 95%: CI 0.194–0.771; P = 0.007, for 1-SD increase), and creatinine (OR: 2.364, 95%: CI 1.228–4.213; P = 0.018, for a 1-SD increase) were independent risk factors for HF in diabetic subjects. When NT-proBNP, HMGB1, and RAGE isoform measurements together with the above risk factors were included in the multivariable analysis (Model II), HMGB1, esRAGE, cRAGE, and NT-proBNP resulted in being independently associated with the presence of HF in both non-diabetic and diabetic patients (Table 5).

Table 5. Multivariable stepwise logistic regression model for the presence of heart failure
VariablesNon-diabetic subjectsDiabetic subjects
 OR (95% confidence interval)P-valueOR (95% confidence interval)P-value
  1. a

    OR, odd ratios.

Univariable conditional logistic regression adjusting for conventional risk factors (Model I)
 Female gender0.639 (0.254–1.606)0.3410.712 (0.270–1.877)0.493
 Age [years (SD)]0.862 (0.574–1.295)0.4741.289 (0.860–1.931)0.219
 Hypertension2.877 (0.999–6.351)0.0501.621 (0.636–4.134)0.312
 Smoking2.235 (1.455–4.582)0.0091.174 (0.445–3.094)0.746
 Systolic blood pressure [mmHg (SD)]0.386 (0.194–0.771)0.0070.543 (0.305–0.966)0.038
 Diastolic blood pressure [mmHg (SD)]1.503 (0.883–2.559)0.1331.054 (0.688–1.613)0.811
 Blood urea nitrogen (mmol/L)1.596 (1.162–2.193)0.0041.344 (1.068–1.690)0.012
 Creatinine [µmol/L (SD)]1.230 (0.512–2.956)0.6022.364 (1.228–4.213)0.018
 Uric acid [µmol/L (SD)]3.861 (0.556–8.797)0.2721.191 (0.875–1.622)0.266
 Fasting glucose (mmol/L)1.938 (1.039–3.616)0.0380.877 (0.571–1.097)0.154
 Glycated haemoglobin (%)1.625 (1.124–1.989)0.0230.932 (0.418–4.135)0.236
 Triglyceride (mmol/L)2.011 (0.962–4.201)0.0630.962 (0.696–1.328)0.435
 Total cholesterol (mmol/L)0.893 (0.557–1.430)0.5400.740 (0.507–1.078)0.117
Backward stepwise regression adjusting for independent conventional risk factors and all biomarkers (Model II)
 Smoking1.325 (0.643–1.681)0.219
 Systolic blood pressure (mmHg, SD)2.512 (0.816–5.752)0.1080.754 (0.561–1.014)0.062
 Blood urea nitrogen (mmol/L)1.237 (0.842–1.816)0.2781.555 (1.001–2.417)0.050
 Creatinine [µmol/L (SD)]1.429 (0.852–2.293)0.176
 Fasting glucose (mmol/L)1.282 (0.926–1.775)0.135
 Glycated haemoglobin (%)1.425 (1.013–1.938)0.011
 NT-proBNP [pg/mL (SD/2)]1.512 (1.043–1.886)0.0011.602 (1.261–2.036)0.001
 HMGB1 [ng/mL (SD)]1.681 (1.164–1.696)0.0021.357 (1.068–1.722)0.008
 cRAGE [pg/mL (SD/2)]1.402 (1.055–2.069)0.0431.256 (1.150–1.971)0.032
 esRAGE [pg/mL (SD)]0.441 (0.283–0.689)0.0080.394 (0.196–0.878)0.023

Association of biological analytes with mortality

During a 2-year follow-up in HF patients, non-survivors had higher serum HMGB1 levels than survivors (12.41 ± 17.39 vs. 6.01 ± 7.02 ng/mL; P = 0.002), and patients with one or more hospitalization for HF had higher serum cRAGE and NT-proBNP levels than those without (1279.5 ± 1405.1 vs. 840.7 ± 1053.3 pg/mL and 4584.45 ± 5726.94 vs. 1025.4 ± 1501.3 pg/mL, for serum cRAGE and NT-proBNP levels, respectively; for both P < 0.001).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

Although AGEs and several RAGE-related proteins are suggested to be closely involved in atherosclerotic vascular disease in both diabetic and non-diabetic patients, their relation to HF has been much less explored. Our study is the first to show that elevated serum levels of HMGB1 and cRAGE, and decreased esRAGE levels, are associated with HF in both diabetic and non-diabetic patients.

High-mobility group box-1, an intracellular regulator of gene expression and also a RAGE ligand, has been shown to contribute to inflammatory reactions, sepsis, atherosclerosis, and re-perfusion injury.15,2628 However, the in vivo pathological effects of HMGB1 remain controversial. Andrassy et al.16 observed that administration of HMGB1 led to a pro-inflammatory response and worsened cardiac function in mice subjected to ischaemia–re-perfusion injury, while HMGB1 antagonists significantly reduced such injury. In contrast, the experiments by Takahashi et al.24 demonstrated a clear-cut attenuation of local inflammation and fibrosis in rats with myocardial infarction given intra-myocardial HMGB1 injection, with an accompanying improvement of cardiac function. The existence of such beneficial effects of HMGB1 is also supported by the experiments of Limana et al.22 and Kitahara et al 23; the former showed myocardial regeneration induced by exogenous HMGB1 administration through measurements of cardiac c-kit+ cell proliferation and differentiation;22 and the latter showed improvements of angiogenesis, cardiac function, and survival after myocardial infarction.23

Our results, obtained in the largest population studied thus far in HF, evidenced a positive relationship of HMGB1 with HF, consistent with deleterious effects of HMGB1. In our study, the HMGB1 value paralleled the clinical severity and worse outcome of HF independent of the presence of diabetes. The reasons for the above-mentioned discrepancies in the effects of HMGB1 in the setting of HF remain unclear at present, but the variety of experimental conditions used in the different studies probably contribute to the differences in results. In our study, we recruited HF patients with different aetiologies, whose disease course was mostly longer than 6 months, which is at variance with the acute effects observed in well-controlled experimental models or cell lines. In conditions of ischaemia–reperfusion, a diffuse myocardial injury provoked by the abundant production of oxygen-derived free radicals is the main pathophysiological mechanism, while persistent myocardial ischaemia is a dominant stimulus in mice with myocardial infarction.24 Infection-associated myocarditis, autoimmune inflammation, and mutation-induced loss/dysfunction of matrix protein primarily initiate the development of dilated cardiomyopathy.29 Thus, the final results may vary between studies, depending on the combined pathogenetic mechanisms in which HMGB1 and multiple other factors are involved with differing roles. Of note, the in vivo concentrations and distribution of HMGB1 are quite different in previously reported studies, contributing to the disparity in findings.

Similar to the reports of Koyama et al.,30 we observed an elevation of cRAGE levels in HF patients, which was more remarkable in diabetes and in dilated cardiomyopathy. Since RAGE cleavage by metalloproteinase can be induced by HMGB1 and other inflammatory cytokines,18 the increase in cRAGE levels may represent a distinctive pathophysiological correlate of severe inflammatory reactions in these diseases, whereby HMGB1 might actually be the trigger for increased cRAGE concentrations. While cRAGE levels might become elevated in response to a variety of other stimuli, including AGEs,12 levels of HMGB1 would be more specific for HF, independent of the presence or absence of diabetes. This would explain the interaction of HF and diabetes in determining the levels of cRAGE, but not of HMGB1.

Endogenous secretory RAGE is an endogenous protein thought to neutralize AGEs and some inflammatory cytokines, and acts as a protective, anti-atherogenic factor. In the present study, we observed that esRAGE levels were significantly lower in non-diabetic patients with HF than in normal controls, similar to levels found in diabetes. These results indicate that protective systems against AGEs and inflammatory cytokines may be severely impaired in HF, and such impairment seems to be more prominent in ischaemic HF, where esRAGE levels were lower.

Taken together, the present findings show that increased levels of HMGB1 and cRAGE and decreased esRAGE levels are associated with the development and the severity of HF. Levels of cRAGE were particularly elevated in HF patients with diabetes or dilated cardiomyopathy, as well as in patients experiencing a re-hospitalization for HF during the follow-up. Moreover, HMGB1 levels were significantly higher in non-survivors vs. survivors within patients with HF. These findings support the notion that RAGE, RAGE variants, and their ligands are closely—but each of them differentially—marking the development or severity of HF, and possibly, because of their biological actions, involved in the pathogenetic mechanism of HF. Here, HMGB1 was related to the development and progression of HF in both diabetic and non-diabetic subjects, while cRAGE and esRAGE possibly play a selective role in diabetes-related worsening of heart function. In our study, lower enrolment or the loss to follow-up of more severe cases with the worst prognosis may have contributed to the low mortality observed compared with that reported in previous trials and registries as well as epidemiological studies.

Study limitations

We recognize a few limitations in our study. First, this study was mainly cross-sectional with little significance of the follow-up data due to the low numbers of events accrued, thereby only allowing us to detect associations. Due to the study design, predictions and causal inferences are impossible. Second, although differences in magnitude were identified with the current number of subjects, and fulfilling the original hypotheses, the sample size in our study is still relatively small, and larger prospective studies are now warranted to confirm both the association and the predictive role of HMGB1 and related proteins with HF. Third, despite being statistically significant, the magnitude of the associations found (the slope of the relationship in regression analyses) was small, and interactions of variables not measured here may have occurred; importantly, information on diastolic function was not available. Fourth, all HF patients studied were Chinese; thus, it remains uncertain whether these results are fully applicable to other ethnicities. However, the identification of a new marker, with additional possible pathogenetic relevance, has to be regarded as a novelty in the area.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

The present study demonstrates higher serum levels of HMGB1 and cRAGE, and lower levels of esRAGE in diabetic and non-diabetic patients with HF, associated with the severity of changes in cardiac function, symptoms, and clinical outcome. Such associations are now worth further exploration.

Funding

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References

This study was supported by a key grant from the Science and Technology Commission of Shanghai Municipality-‘Optimal Therapy of Myocardial Infarction with Diabetes' (05DZ19503).

Conflict of interest: none declared.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Funding
  9. References
  • 1
    Goldin A Beckman JA Schmidt AM Creager MA Advanced glycation end products: sparking the development of diabetic vascular injury Circulation 2006 114 597 605
  • 2
    Kilhovd BK Berg TJ Birkeland KI Thorsby P Hanssen KF Serum levels of advanced glycation end products are increased in patients with type 2 diabetes and coronary heart disease Diabetes Care 1999 22 1543 1548
  • 3
    Kiuchi K Nejima J Takano T Ohta M Hashimoto H Increased serum concentrations of advanced glycation end products: a marker of coronary artery disease activity in type 2 diabetic patients Heart 2001 85 87 91
  • 4
    Tan KC Chow WS Ai VH Metz C Bucala R Lam KS Advanced glycation end products and endothelial dysfunction in type 2 diabetes Diabetes Care 2002 25 1055 1059
  • 5
    Kanauchi M Tsujimoto N Hashimoto T Advanced glycation end products in nondiabetic patients with coronary artery disease Diabetes Care 2001 24 1620 1623
  • 6
    Avendano GF Agarwal RK Bashey RI Lyons MM Soni BJ Jyothirmayi GN Regan TJ Effects of glucose intolerance on myocardial function and collagen-linked glycation Diabetes 1999 48 1443 1447
  • 7
    Asif M Egan J Vasan S Jyothirmayi GN Masurekar MR Lopez S Williams C Torres RL Wagle D Ulrich P Cerami A Brines M Regan TJ An advanced glycation endproduct cross-link breaker can reverse age-related increases in myocardial stiffness Proc Natl Acad Sci USA 2000 97 2809 2813
  • 8
    Aronson D Cross-linking of glycated collagen in the pathogenesis of arterial and myocardial stiffening of aging and diabetes J Hypertens 2003 21 3 12
  • 9
    Hartog JW van de Wal RM Schalkwijk CG Miyata T Jaarsma W Plokker HW van Wijk LM Smit AJ van Veldhuisen DJ Voors AA Advanced glycation end-products, anti-hypertensive treatment and diastolic function in patients with hypertension and diastolic dysfunction Eur J Heart Fail 2010 12 397 403
  • 10
    Hartog JW Voors AA Schalkwijk CG Scheijen J Smilde TD Damman K Bakker SJ Smit AJ van Veldhuisen DJ Clinical and prognostic value of advanced glycation end-products in chronic heart failure Eur Heart J 2007 28 2879 2885
  • 11
    Koyama Y Takeishi Y Arimoto T Niizeki T Shishido T Takahashi H Nozaki N Hirono O Tsunoda Y Nitobe J Watanabe T Kubota I High serum level of pentosidine, an advanced glycation end product (AGE), is a risk factor of patients with heart failure J Card Fail 2007 13 199 206
  • 12
    Basta G Receptor for advanced glycation endproducts and atherosclerosis: from basic mechanisms to clinical implications Atherosclerosis 2008 196 9 21
  • 13
    Lu L Pu LJ Zhang Q Wang LJ Kang S Zhang RY Chen QJ Wang JG De Caterina R Shen WF Increased glycated albumin and decreased esRAGE levels are related to angiographic severity and extent of coronary artery disease in patients with type 2 diabetes Atherosclerosis 2009 206 540 545
  • 14
    Chen Y Pat B Zheng J Cain L Powell P Shi K Sabri A Husain A Dell'italia LJ Tumor necrosis factor-alpha produced in cardiomyocytes mediates a predominant myocardial inflammatory response to stretch in early volume overload J Mol Cell Cardiol 2010 49 70 78
  • 15
    Scaffidi P Misteli T Bianchi ME Release of chromatin protein HMGB1 by necrotic cells triggers inflammation Nature 2002 418 191 195
  • 16
    Andrassy M Volz HC Igwe JC Funke B Eichberger SN Kaya Z Buss S Autschbach F Pleger ST Lukic IK Bea F Hardt SE Humpert PM Bianchi ME Mairbaurl H Nawroth PP Remppis A Katus HA Bierhaus A High-mobility group box-1 in ischemia-reperfusion injury of the heart Circulation 2008 117 3216 3226
  • 17
    Sparvero LJ Asafu-Adjei D Kang R Tang D Amin N Im J Rutledge R Lin B Amoscato AA Zeh HJ Lotze MT RAGE (Receptor for Advanced Glycation Endproducts), RAGE ligands, and their role in cancer and inflammation J Transl Med 2009 7 17
  • 18
    Raucci A Cugusi S Antonelli A Barabino SM Monti L Bierhaus A Reiss K Saftig P Bianchi ME A soluble form of the receptor for advanced glycation endproducts (RAGE) is produced by proteolytic cleavage of the membrane-bound form by the sheddase a disintegrin and metalloprotease 10 (ADAM10) FASEB J 2008 22 3716 3727
  • 19
    Yan XX Lu L Peng WH Wang LJ Zhang Q Zhang RY Chen QJ Shen WF Increased serum HMGB1 level is associated with coronary artery disease in nondiabetic and type 2 diabetic patients Atherosclerosis 2009 205 544 548
  • 20
    Kohno T Anzai T Naito K Miyasho T Okamoto M Yokota H Yamada S Maekawa Y Takahashi T Yoshikawa T Ishizaka A Ogawa S Role of high-mobility group box 1 protein in post-infarction healing process and left ventricular remodelling Cardiovasc Res 2009 81 565 573
  • 21
    Palumbo R Sampaolesi M De Marchis F Tonlorenzi R Colombetti S Mondino A Cossu G Bianchi ME Extracellular HMGB1, a signal of tissue damage, induces mesoangioblast migration and proliferation J Cell Biol 2004 164 441 449
  • 22
    Limana F Germani A Zacheo A Kajstura J Di Carlo A Borsellino G Leoni O Palumbo R Battistini L Rastaldo R Muller S Pompilio G Anversa P Bianchi ME Capogrossi MC Exogenous high-mobility group box 1 protein induces myocardial regeneration after infarction via enhanced cardiac C-kit+ cell proliferation and differentiation Circ Res 2005 97 e73 e83
  • 23
    Kitahara T Takeishi Y Harada M Niizeki T Suzuki S Sasaki T Ishino M Bilim O Nakajima O Kubota I High-mobility group box 1 restores cardiac function after myocardial infarction in transgenic mice Cardiovasc Res 2008 80 40 46
  • 24
    Takahashi K Fukushima S Yamahara K Yashiro K Shintani Y Coppen SR Salem HK Brouilette SW Yacoub MH Suzuki K Modulated inflammation by injection of high-mobility group box 1 recovers post-infarction chronically failing heart Circulation 2008 11814 Suppl. S106 S114
  • 25
    Heymans S Hirsch E Anker SD Aukrust P Balligand JL Cohen-Tervaert JW Drexler H Filippatos G Felix SB Gullestad L Hilfiker-Kleiner D Janssens S Latini R Neubauer G Paulus WJ Pieske B Ponikowski P Schroen B Schultheiss HP Tschope C Van Bilsen M Zannad F McMurray J Shah AM Inflammation as a therapeutic target in heart failure? A scientific statement from the Translational Research Committee of the Heart Failure Association of the European Society of Cardiology Eur J Heart Fail 2009 11 119 129
  • 26
    Andersson U Wang H Palmblad K Aveberger AC Bloom O Erlandsson-Harris H Janson A Kokkola R Zhang M Yang H Tracey KJ High mobility group 1 protein (HMG-1) stimulates proinflammatory cytokine synthesis in human monocytes J Exp Med 2000 192 565 570
  • 27
    Kalinina N Agrotis A Antropova Y DiVitto G Kanellakis P Kostolias G Ilyinskaya O Tararak E Bobik A Increased expression of the DNA-binding cytokine HMGB1 in human atherosclerotic lesions: role of activated macrophages and cytokines Arterioscler Thromb Vasc Biol 2004 24 2320 2325
  • 28
    Porto A Palumbo R Pieroni M Aprigliano G Chiesa R Sanvito F Maseri A Bianchi ME Smooth muscle cells in human atherosclerotic plaques secrete and proliferate in response to high mobility group box 1 protein FASEB J 2006 20 2565 2566
  • 29
    Maisch B Richter A Sandmoller A Portig I Pankuweit S Inflammatory dilated cardiomyopathy (DCMI) Herz 2005 30 535 544
  • 30
    Koyama Y Takeishi Y Niizeki T Suzuki S Kitahara T Sasaki T Kubota I Soluble receptor for advanced glycation end products (RAGE) is a prognostic factor for heart failure J Card Fail 2008 14 133 139