Abstract
Objective. Depending on physiological context, the adipokine chemerin can reduce or enhance cardiovascular risk. We investigated whether chemerin concentrations represent cardiovascular disease risk in rheumatoid arthritis (RA).
Methods. We assessed ELISA-determined chemerin concentrations and those of 4 early endothelial activation molecules as well as angiopoietin 2, which mediates angiogenesis and thereby contributes to advanced atherosclerosis, the common carotid artery intima-media thickness (cIMT), and carotid artery plaque by ultrasound in 236 patients (114 black and 122 white) with RA. Relationships were identified in potential confounder and mediator-adjusted mixed regression models.
Results. Mean (SD) chemerin and median (interquartile range) angiopoietin 2 concentrations were 114 (35) ng/ml and 2560 (2044–3341) pg/ml, respectively; the mean (SD) cIMT was 0.708 (0.110) mm, and 40.3% of patients had plaque. Chemerin concentrations were not related to those of early endothelial activation molecules, but associated with those of angiopoietin 2 [β SE = 0.002 (0.0004), p < 0.0001] and plaque [OR 1.006 (95% CI 1.00–1.013), p = 0.05] in all patients. The presence of major conventional cardiovascular risk factors, generalized and abdominal obesity, and RA severity markers modified the independent chemerin-cardiovascular risk relations (interaction p < 0.05). Consequently, chemerin concentrations were associated with cIMT in those with but not without overweight or generalized obesity and abdominal obesity [β SE = 0.001 (0.0003), p = 0.005 and 0.001 (0.0001), p = 0.001 vs −0.001 (0.0004), p = 0.2 and −0.0002 (0.0004), p = 0.6, respectively], and with plaque in those without but not with generalized obesity [OR 1.008 (95% CI) 1.000–1.016, p = 0.03 vs 1.003 (0.990–1.017), p = 0.6, respectively]. The β (SE) for the chemerin-intima-media thickness relations in patients with overweight or generalized obesity and abdominal obesity were larger than in those without these characteristics (p < 0.0001 and = 0.04, respectively).
Conclusion. Chemerin is associated with endothelial activation and atherosclerosis in RA. Adiposity influences the chemerin-atherosclerotic phenotype relations in RA.
Rheumatoid arthritis (RA) predisposes substantially to traditional and nontraditional risk factor dependent atherosclerotic cardiovascular disease (CVD) with a prevalence and incidence as large as in diabetes1,2,3. Atherogenesis in RA remains poorly understood, and currently recommended cardiovascular risk strategies4 have potentially important shortcomings5,6. It is against this background that the need for identifying novel biomarkers of cardiovascular risk in this inflammatory disease has been emphasized5,6.
Chemerin, also known as tazarotene-induced gene 2 or retinoic acid receptor responder 2, was originally identified as a gene upregulated in psoriatic skin by the synthetic retinoid tazarotene in 19977. In 2007, chemerin was identified as a novel adipokine that regulates adipogenesis and adipocyte metabolism7. Chemerin is a natural ligand and chemotactic signal for cells expressing G protein-coupled receptor chemokine-like receptor 17,8.
A large body of evidence derived from mechanistic and clinical studies indicates that chemerin is proinflammatory and implicated in metabolic risk7,8. As an early biomarker of metabolic and inflammatory disease, chemerin concentrations predicted incident metabolic risk more strongly than measures of insulin resistance and leptin concentrations in a population without clinical symptoms of metabolic syndrome8. Chemerin also induces the production of E-selectin and intercellular adhesion molecule 1 (ICAM-1) in endothelial cells9, and further increases angiogenesis10, which contributes to advanced atherosclerosis by augmenting plaque development and destabilization11. Importantly, in the present context, chemerin participates in normal physiological processes and also exhibits protective properties in that, depending on (patho)physiological status, it can have antiinflammatory activity, and both reduced and elevated chemerin signaling may cause aberrant glucose metabolism8. Therefore, disease status and severity require consideration upon evaluating the biological effects of chemerin8.
As applies to other adipokines12, chemerin is involved in the pathophysiology of RA13,14,15. Chemerin is expressed in synovial endothelial, lining, and sublining cells, and its production is upregulated by tumor necrosis factor-α (TNF-α) and interferon-γ in RA; chemerin stimulated the production of interleukin 6 (IL-6), chemokine (C-C motif) ligand 2, Toll-like receptor 4 mRNA, and matrix metalloproteinase 3 by fibroblast-like synoviocytes13,14.
Adipokines other than chemerin are linked to metabolic risk, early endothelial activation, and inflammation in RA16,17,18,19,20,21,22,23,24. However, circulating concentrations of these molecules were reported to be unrelated to atherosclerosis25, except for retinol-binding protein 426. In our study, we examined the independent relationships between chemerin concentrations and metabolic risk, surrogate markers of early endothelial activation as well as angiogenesis, and atherosclerosis in 236 African (114 black and 122 white) patients with RA. We also determined whether the presence of major traditional risk factors, excess adiposity, and RA severity markers modify these relationships.
MATERIALS AND METHODS
Patients
We conducted our study according to the principles outlined in the Helsinki declaration. The Human Research Ethics Committee (Medical) from the University of the Witwatersrand in Johannesburg, South Africa, approved the protocol (approval number: M06-07-33). Participants gave informed, written consent. This investigation forms part of an ongoing study on cardiovascular risk in RA18,19,21,22. There were 236 consecutive African patients (114 black and 122 white) enrolled who met the 1988 American College of Rheumatology (ACR) and 2010 ACR/European League Against Rheumatism criteria for RA27,28. All invited participants agreed to participate. Data were missing in fewer than 5% of any of the recorded characteristics.
Assessments
Baseline characteristics and conventional metabolic risk factors were recorded using previously reported methods18,19,21,22. Briefly, we recorded demographic features and lifestyle factors. Height, weight, and waist and hip circumference were measured using standard approaches. The body mass index (BMI) was calculated, and abdominal obesity and fat distribution were estimated by waist circumference and waist circumference-hip ratio, respectively. We recorded disease duration and rheumatoid factor (RF) status. Disease activity was assessed by the Clinical Disease Activity Index and the Disease Activity Score in 28 joints (DAS28). Extraarticular manifestations included the current or previously recorded (hospital record review) presence of pericarditis, pleuritis, Felty’s syndrome, cutaneous vasculitis, neuropathy, scleritis or episcleritis, retinal vasculitis, glomerulonephritis, vasculitis affecting other organs, amyloidosis, keratoconjunctivitis sicca, xerostomia, Sjogren syndrome, pulmonary fibrosis, bronchiolitis obliterans organizing pneumonia, cervical myelopathy, subcutaneous nodules, and rheumatoid nodules in other locations. C-reactive protein (CRP) concentrations were determined using immunoturbidimetric methods, and those of IL-6 by ELISA. Standard laboratory blood tests of erythrocyte sedimentation rate, renal and liver function, hematological variables, lipids, and glucose were performed. The glomerular filtration rate (GFR) was estimated using the Modification of Diet in Renal Disease equation. Cardiovascular drug use was recorded.
Recorded metabolic risk factors included systolic, diastolic, and mean blood pressure, lipid concentrations and ratios, and glucose levels. Hypertension was defined as an average systolic blood pressure ≥ 140 or/and diastolic blood pressure ≥ 90 mmHg or/and current use of antihypertensive medications. Dyslipidemia was diagnosed when the atherogenic index, i.e., the cholesterol/high-density lipoprotein (HDL) cholesterol ratio was > 44. Diabetes was identified as the use of glucose lowering agents or a fasting plasma glucose ≥ 7 mmol/l.
We evaluated concentrations of 5 other adipokines, including total and high molecular weight adiponectin, leptin, and resistin as previously reported18,19,21,22, as well as retinol-binding protein 4 using a solid-phase sandwich ELISA (QuantikineHS, R&D Systems Inc.) with a lower detection limit of 0.224 µg/ml and the interassay and intraassay coefficients of variation of 7.2% and 6.9%, respectively.
We measured early endothelial activation molecule concentrations, including those of soluble E-selectin, vascular cell adhesion molecule 1 (VCAM-1), ICAM-1, and monocyte chemoattractant protein 1 (MCP-1), as well as angiopoietin 2, and using solid-phase sandwich ELISA (QuantikineHS). Their lower detection limits were 0.009 ng/l, 0.6 ng/l, 0.096 ng/l, 5.0 pg/ml, and 1.2 pg/ml, respectively; their interassay and intraassay coefficients of variation were 7.9 and 5.8, 7.0 and 3.1, 5.5 and 4.6, 5.7 and 5.8, and 8.9 and 5.9%, respectively.
Carotid artery ultrasound (US) measurements were done by 2 operators, 1 on private healthcare patients and 1 on public healthcare patients. Both resulted in images of at least 1-cm length of the distal common carotid arteries for measurement of the intima-media thickness of the far wall from an optimal angle of incidence defined as the longitudinal angle of approach where both branches of the internal and external carotid artery are visualized simultaneously29, and with high-resolution B-mode US (Image Point, Hewlett Packard and SonoCalc IMT, Sonosite Inc.) using linear array 7.5 MHz probes. The details of the methodology were reported30. The equipment used with the public healthcare patients involved the application of a unique semiautomated border detection program that was previously found to provide highly reproducible results29. The intima-media thicknesses in the left and right common carotid artery were measured, and the carotid artery intima-media thickness (cIMT) was defined as the mean of these. Carotid artery plaque was defined as a focal structure that encroaches into the arterial lumen of at least 0.5 mm or 50% of the surrounding intima-media thickness value, or demonstrates a thickness of > 1.5 mm as measured from the media-adventitia interface to the intima-lumen interface31. Both operators were blinded to the cardiovascular risk profiles of the patients. Repeat US examinations by both operators on 23 patients revealed intraobserver coefficients of variation of 5.8% and 4.1% for private healthcare patients and public healthcare patients, respectively, and an interobserver coefficient of variation of 8.0% for measurements made by the 2 operators. Both operators identified carotid artery bulb or/and internal carotid artery plaque in 11 of these 23 patients, with full agreement.
Chemerin concentrations were quantified using a solid-phase sandwich ELISA (QuantikineHS). The lower detection limit was 1.08 pg/ml and the interassay and intraassay coefficients of variation were 7.3% and 3.9%, respectively.
Data management analysis
Dichotomous variables are expressed as proportions or percentages, and continuous variables as mean (SD) or median (interquartile range) when non-normally distributed. Non-normally distributed characteristics were also logarithmically transformed prior to their inclusion in multivariable statistical analysis. An endothelial activation score was used to provide a summary measure of early endothelial activation and was calculated from SD (z) scores as follows: [z (selectin) + z (VCAM-1) + z (ICAM-1) + z (MCP-1)]19,22.
Associations of age at disease onset or at the time of the study, and sex and population grouping with chemerin concentrations were assessed by entering the respective characteristics together in single mixed regression models. Associations of other baseline characteristics with chemerin concentrations were evaluated in models with adjustment for demographic characteristics that included age at the time of the study and not age at disease onset, because the latter was not related to chemerin concentrations.
The independent relations of chemerin concentrations with metabolic risk factors were assessed in demographic characteristic, waist circumference, GFR, leflunomide (LEF) and TNF-α blockade use (potential confounders or/and determinants identified in previous analysis), and cardiovascular drug use adjusted mixed linear regression models. The independent relations of chemerin concentrations with endothelial activation, cIMT, and plaque were determined in Framingham score (calculated from age, sex, and major conventional risk factors), race, waist, GFR, CRP, and leptin concentrations, and LEF and TNF-α blockade use adjusted mixed (linear or logistic as appropriate) regression models.
Patients with RA who experience conventional risk factors or severe disease are reportedly at high risk of CVD1,2,4. For these reasons, together with the dependence of chemerin effects on physiological context8 and our recent experience with adipokine metabolism in RA18,19,20,21,22, we assessed the effect of patient characteristics on chemerin-cardiovascular risk relations by adding interaction terms to the models, and stratified analysis when indicated (significant interaction p values), that is, in subgroups with and without patient characteristics of interest in the present context. For this purpose, patients with a BMI of ≥ 30 kg/m2 and those who met the National Cholesterol Education Program for metabolic syndrome waist criterion32 were considered to sustain overall and abdominal obesity, respectively. When appropriate, patients were categorized in subgroups based on median values. In view of the small number of patients who were RF-positive or had extraarticular features, sensitivity analysis in subgroups based on the presence or absence of these characteristics was not performed.
Statistical computations were made using the GB Stat program (Dynamic Microsystems Inc.) and SAS software, version 9.1 (The SAS Institute). Significance was set at p ≤ 0.05.
RESULTS
Descriptive statistics of the recorded characteristics are given in Table 1. The mean (SD) Framingham score was 5 (7), as determined using an algorithm33. However, this characteristic was non-normally distributed with a median (interquartile range) value of only 2 (1–6). All patients were using synthetic disease-modifying agents and only 1 of them used a biologic agent other than TNF-α blockade (rituximab).
Associations of baseline characteristics with chemerin concentrations
As presented in Table 2, demographic characteristics, anthropometric measures, the use of anti-hypertensives, disease activity, and the GFR were each significantly associated with chemerin concentrations. Among the disease activity variables, CRP concentrations were most strongly related to chemerin concentrations. LEF and TNF-α blockade use were also borderline related to chemerin concentrations (p = 0.06 and 0.1, respectively), and hence additionally included as potential confounders in subsequent analyses. In separate analysis, when BMI and waist were entered into the same demographic characteristic and GFR-adjusted model, only waist circumference remained associated with chemerin concentrations, the DAS28 and waist were related to chemerin concentrations independent of one another, and the antihypertensive agent use-chemerin concentrations relation was explained by abdominal adiposity (data not shown). In view of these results, waist circumference was the anthropometric measure that was included in subsequent models. All cardiovascular drug use was consistently forced into the models in which chemerin-cardiovascular risk relations were evaluated.
Independent relationships of chemerin concentrations with metabolic risk
As also shown in Table 2, in all 236 patients, chemerin concentrations were independently related to those of leptin. Chemerin levels were not related to any of the other metabolic risk factors.
An RA duration of > 10 years affected the chemerin-HDL cholesterol concentration and chemerin-cholesterol-HDL cholesterol ratio associations (interaction p = 0.04 and 0.006, respectively). The presence of deformed joints affected the chemerin-systolic, -diastolic, and -mean blood pressure relations (interaction p = 0.01, 0.03, and 0.01, respectively). In stratified analysis, chemerin concentrations were independently associated with systolic blood pressure, diastolic blood pressure, mean blood pressure, and cholesterol-HDL-cholesterol ratio in patients with but not in those without a disease duration of > 10 years [β (SE) = 0.117 (0.056), p = 0.03, 0.068 (0.033), p = 0.04, 0.100 (0.045), p = 0.03 and 0.006 (0.005), p = 0.03 vs −0.031 (0.078), p = 0.7, −0.050 (0.043), p = 0.2, −0.038 (0.062), p = 0.5 and 0.001 (0.004), p = 0.7, respectively], and with systolic and diastolic blood pressure in patients with but not in those without deformed joints [β (SE) = 0.111 (0.056), p = 0.05 and 0.090 (0.046), p = 0.05 vs −0.089 (0.071), p = 0.2 and −0.072 (0.057), p = 0.2, respectively].
Generalized obesity affected the chemerin-leptin and chemerin-retinol binding protein 4 concentration relations (interaction p = 0.02 and 0.03, respectively); abdominal obesity affected on the chemerin-total adiponectin concentration association (interaction p = 0.03). In stratified analysis, chemerin concentrations were related to leptin levels in patients without but not in those with generalized and abdominal obesity [β (SE) = 0.003 (0.001), p = 0.0008 and 0.003 (0.001), p = 0.02 vs −0.001 (0.001), p = 0.5 and 0.001 (0.001), p = 0.2, respectively].
Independent relationships of chemerin concentrations with endothelial activation and atherosclerosis
Table 3 shows that whereas chemerin concentrations were not related to early endothelial activation in all patients, they were strongly associated with angiopoietin 2 levels. Further, chemerin concentrations were related to cIMT and plaque in univariate and multivariate analysis, respectively. In a separate analysis, the chemerin concentrations-plaque relationship became significant only once leptin levels were added to the model (data not shown).
The presence of major cardiovascular risk factors affected the chemerin-selectin concentration, chemerin-endothelial activation score, and chemerin-plaque relationship (interaction p = 0.05, 0.03, and 0.02, respectively); generalized obesity affected the chemerin-plaque association (interaction p = 0.01); and abdominal obesity affected the chemerin-ICAM-1 and chemerin-cIMT relationship (interaction p = 0.02 and 0.04, respectively).
Table 4 gives results obtained in the related stratified analysis. Chemerin concentrations were related to the endothelial activation score in patients with but not in those without major cardiovascular risk factors, with cIMT in patients that were overweight or obese, and those who sustained abdominal obesity but not in those without these risk factors, and with plaque in patients without but not in those with generalized obesity. The β (SE) for the chemerin-cIMT relations in patients who were overweight or obese and those who experienced abdominal obesity were larger than in those without these characteristics (p < 0.0001 and 0.04, respectively). Chemerin concentrations were consistently associated with those of angiopoietin 2 among subgroups.
In the models shown in Table 4, we adjusted for the Framingham score, which includes age. Nevertheless, the confounding effect of age may have been underestimated. When we replaced the Framingham score by age, sex, hypertension, dyslipidemia, smoking, and diabetes, the chemerin-cIMT relationship remained significant in patients with generalized obesity, overweight, or generalized obesity and abdominal obesity [β (SE) = 0.001 (0.0004), p = 0.01, 0.001 (0.0003), p = 0.02 and 0.001 (0.0003), p = 0.002, respectively]. This approach was not to be used for chemerin-plaque relations because it would have meant using overfitted logistic regression models, which can give false-positive or false-negative results.
DISCUSSION
In this RA study, chemerin concentrations were consistently and independently associated with those of angiopoietin 2, which contributes relevantly to angiogenesis34, and hence is an endothelial activation of advanced rather than early atherosclerosis11. Angiopoietin 2 concentrations are related to prevalent and incident CVD in RA35,36.
Chemerin concentrations were also independently associated to carotid artery plaque in all patients. Interestingly, this relationship was leptin-driven, and hence suggests a combined effect of leptin and chemerin in atherosclerosis among patients with RA. However, our most striking finding was that the presence of major conventional cardiovascular risk factors, excess abdominal and generalized adiposity, and RA severity markers consistently modified the independent chemerin-cardiovascular risk relations. This translated into associations of chemerin concentrations with cIMT in patients who were overweight or had generalized obesity and those who experienced abdominal obesity, and with carotid artery plaque in patients who were normal or overweight.
Maradit Kremers, et al37 and Escalante, et al38 reported a paradoxical protective effect of BMI against ischemic heart disease and overall mortality, respectively, in patients with RA. A modulating effect of systemic inflammation on the influence of body mass on cardiovascular risk and mortality was implicated. Abdominal adiposity also relates paradoxically to reduced endothelial activation in RA39.
Importantly in the present context, cIMT and plaque are different phenotypes of atherosclerosis, and biologically and genetically distinct40,41,42,43,44,45. Intima-media thickening reflects mostly an enlarged tunica media due to an adapted response to aging and particularly blood pressure41, and is related to stroke risk factors and prevalence44,45. By contrast, carotid plaque develops because of intimal pathology40 and represents advanced atherosclerosis that is more closely linked to coronary artery disease risk factors and prevalence41,44. We have documented that BMI or generalized adiposity associates independently with blood pressure and cIMT, whereas anthropometric measures of abdominal adiposity are related to lipids and carotid artery plaque in women with intensively managed RA46. Thus, BMI and the waist-hip ratio are associated with different metabolic risk factor profiles and reflect different aspects of atherosclerosis in RA. Also, the INTERHEART study investigators47 found that the waist-hip ratio was more strongly associated with myocardial infarction than was BMI, whereas in a large study performed in Finland and reported by Hu, et al48, BMI but not waist-hip ratio enhanced the risk of stroke in women. Notably, coronary artery but not cerebrovascular event rates were included as outcome characteristics in the Maradit Kremers, et al study37.
In our previous investigation on adiposity and atherosclerosis in RA, the relationships of BMI with cIMT and waist-hip ratio with plaque were explained by metabolic risk factors46. Conversely, in our current study, the independent relationship between chemerin and atherosclerosis in both obese and nonobese participants was not explained by metabolic risk factors, as well as several other potential mediating and confounding characteristics. Taken together, our findings indicate that chemerin represents CVD risk in both obese and nonobese patients with RA. Additionally, our results suggest that chemerin may contribute to the link between inflammation and increased risk of coronary artery disease among nonobese patients with RA37. Indeed, systemic inflammation was strongly associated with chemerin concentrations in our analysis, and suppression of inflammation with adalimumab is linked to reduced chemerin concentrations in RA15. Nevertheless, it remains uncertain whether our findings show a unique link between chemerin and CVD and that chemerin thereby contributes to the excess atherosclerosis burden in RA because the present investigation did not include patients without RA. Future studies should determine the association of chemerin concentrations with prevalent and incident cerebrovascular and coronary artery disease in obese and nonobese patients with RA, as well as in patients without RA.
In the recent study by Gonzalvo-Feo, et al49, retinoic acid-activated endothelial cells were shown to promote dendritic cell transmigration across endothelial cell monolayers through the endogenous production of chemerin, upregulation of CCRL2, and activation of dendritic cell β1 integrin affinity. Inflammatory cytokines are markedly overproduced in RA and can induce the production of retinoic acid49. The potential role of chemerin derived from endothelial cell activation in atherogenesis merits further study.
Among patients with major conventional risk factors, chemerin was associated not only with angiopoietin 2 concentrations, but also with surrogate markers of early endothelial activation in our study. These patients also had an increased OR for plaque, but this did not reach significance (p = 0.06). The presence of major conventional risk factors also modifies the potential effects of resistin on surrogate markers of early endothelial activation in RA19.
RA severity additionally affected the chemerin-conventional metabolic risk relations, but in contrast to findings on resistin in atherogenesis, disease severity modified neither the chemerin-endothelial activation associations nor the chemerin-atherosclerosis relation. Leptin is an adiposity marker in RA18, and in our study, both leptin and chemerin concentrations were higher in obese compared to nonobese patients (data not shown). Reminiscent of the paradoxical effect of adiposity on cardiovascular risk37, we found that chemerin concentrations are independently related to those of leptin in nonobese rather than obese patients. Whereas chemerin concentrations reportedly predict metabolic risk development in patients without RA8, we did not find an association of chemerin concentrations with prevalent conventional metabolic risk factors.
A recent South Korean study performed in patients with RA revealed that disease activity rather than obesity was associated with chemerin concentrations in RA50. We found that disease activity and adiposity were related to chemerin concentrations independent of one another. We also showed that chemerin concentrations are higher in African black patients compared to white patients with RA. In this regard, we previously reported that both the production and biological effects of adipokines can differ by population or ethnic origin in RA20,22.
Besides angiopoietin 2 concentrations, we assessed the production of 4 endothelial activation molecules that mediate the initial stages of atherosclerosis, and were shown to be strongly upregulated and associated with prevalent and incident atherosclerosis in RA51,52. CIMT and carotid artery plaque that were our main outcome measures independently predicted incident cardiovascular events in patients with RA and patients without RA53,54.
Our study has limitations. Conceptually, endothelial activation marker concentrations can represent disease activity. The chemerin-endothelial activation relationship identified in the present investigation were independent of disease activity variables. Our cross-sectional study design precludes drawing inferences on the direction of causality. Circulating CRP concentrations are determined not only by disease activity, but also by insulin resistance in RA55, and those of chemerin do not necessarily represent its tissue levels8. We did not include markers of insulin resistance such as the homeostasis model assessment of insulin resistance, which could have been informative in the present context. The standard z-score of endothelial activation as used in our study was previously reported by us18,19,22,39, but nevertheless awaits external validation.
Contrary to other investigated adipokines, apart from retinol binding protein 4, chemerin relates not only to surrogate markers of endothelial activation, but also atherosclerosis in RA. Chemerin represents CVD risk and adiposity influences the chemerin-atherosclerotic phenotype relations in RA. Additional longitudinal studies are needed to fully determine the role of chemerin concentrations in cardiovascular risk stratification in RA.
Acknowledgment
We thank Belinda A. Stevens for performing carotid ultrasound examinations in 104 of the study participants.
Footnotes
-
Supported by The South African Medical Research Council (grant MRC2008_DES) and National Research Foundation.
- Accepted for publication April 24, 2014.