Plasma cystatin C and neutrophil gelatinase-associated lipocalin in relation to coronary atherosclerosis on intravascular ultrasound and cardiovascular outcome: Impact of kidney function (ATHEROREMO-IVUS study).

BACKGROUND AND AIMS
We investigated whether plasma cystatin C (CysC) and neutrophil gelatinase-associated lipocalin (NGAL) are associated with intravascular ultrasound (IVUS)-derived characteristics of coronary atherosclerosis and 1-year adverse coronary events in patients with normal and mildly-to-moderately impaired kidney function.


METHODS
Between 2008 and 2011, virtual histology (VH)-IVUS of a non-culprit coronary artery was performed in 581 patients undergoing coronary angiography. Creatinine, CysC and NGAL were measured in pre-procedural blood samples. Presence of VH-IVUS-derived thin-cap fibroatheroma (TCFA) lesions, lesions with plaque burden (PB)≥70% and lesions with minimal luminal area (MLA)≤4 mm2 was assessed. Major adverse coronary events (MACE) comprised the composite of all-cause mortality, acute coronary syndrome, or unplanned coronary revascularization. Analyses were stratified using eGFRCr of 90 ml/min/1.73 m2 as the cut-off.


RESULTS
In patients with normal kidney function, those with higher CysC levels had fewer lesions with PB ≥ 70% and fewer VH-TCFA lesions (adjusted odds ratios (ORs) and 95% confidence intervals (CIs): 0.46 [0.30-0.69] and 0.59 [0.44-0.83], respectively, per standard deviation (SD) ln[ng/mL] CysC). Those with higher NGAL levels also had fewer lesions with PB ≥ 70% (adjusted OR [95% CI]:0.49 [0.29-0.82]) In patients with impaired kidneys, no differences in high-risk lesions were observed for CysC or NGAL. However, those with higher CysC had higher risk of MACE (hazard ratio (HR):1.4, 95% CI [1.03-1.92]). This was not the case in patients with normal kidney function. NGAL did not influence risk of MACE.


CONCLUSIONS
Mild-to-moderate kidney dysfunction modifies the relationship between CysC and high-risk coronary lesions. This has not been established before, and offers an explanation for the difference in findings between experimental and epidemiologic studies.


Introduction
Kidney impairment, as assessed by creatinine-based equations of glomerular filtration rate (eGFR Cr ), is associated with cardiovascular disease independently of established cardiovascular risk factors [1]. In persons with mild kidney dysfunction (eGFR Cr in the range of 60e89 ml/min/1.73 m 2 ), cystatin C (CysC) may outperform eGFR Cr as a predictor of adverse outcome. This is illustrated by the fact that CysC displays a linear association with mortality in patients with such mild GFR reduction, while eGFR Cr has a J-shaped association with mortality, and risk only starts to rise when eGFR Cr falls beneath 60 ml/min/1.73 m 2 [2,3]. Although some studies have shown linear associations of eGFR Cr with adverse outcome, these associations were linear only in particular ranges of eGFR Cr (specifically, eGFR Cr above 60) [4].
CysC is a cysteine protease inhibitor produced by most nucleated cells, and can be detected in serum or plasma [5]. In in vitro and animal experiments, a reduction of CysC correlated with increased activity of cysteine proteases cathepsins K and S, which led to breakdown of the elastic lamina in the blood vessel wall [6]. Altered CysC expression has been identified in diseases which progress by extracellular proteolysis, such as atherosclerosis and aortic aneurysms, and metastasis [7,8]. These experiments, pointing towards a favourable role for CysC, do not concur with the positive associations of CysC with adverse outcomes found in epidemiological studies. Studies on the in-vivo association between plasma CysC and coronary atherosclerosis may provide further insight into this discrepancy, but have not yet been performed.
Neutrophil gelatinase-associated lipocalin (NGAL) is a clinically relevant biomarker in acute kidney injury [9] due to its marked increase in plasma and urine after tubulo-interstitial kidney damage [10]. Recently, overexpression of plasma NGAL has been found in coronary plaques, where NGAL inhibits elimination of matrix metalloproteinasee9 (MMP-9) [11,12]. MMP-9 is involved in extracellular matrix degradation, herewith increasing the risk of plaque rupture [13]. NGAL and NGAL/MMP-9 complex have been shown to predict major adverse cardiovascular events in epidemiological studies [14,15].
In spite of the above-described associations that have been demonstrated between CysC, NGAL and adverse cardiac events, the presence and shape of a relationship between plasma CysC, NGAL, and coronary atherosclerosis have not yet been investigated invivo. To the best of our knowledge, we are the first to perform such an investigation, and to herewith provide a link between fundamental experiments and epidemiological studies. Specifically, our study aimed to investigate whether plasma CysC and NGAL are associated with IVUS-derived characteristics of in-vivo coronary atherosclerosis and 1-year adverse coronary events in patients with normal and mildly-to-moderately impaired kidney function.

Study population
We have previously described the design of The European Collaborative Project on Inflammation and Vascular Wall Remodeling in Atherosclerosis -Intravascular Ultrasound (ATHEROREMO-IVUS) [16]. In this study, we included 581 patients undergoing diagnostic coronary angiography or percutaneous coronary intervention (PCI) for acute coronary syndrome (ACS) or stable angina pectoris (SAP) between 2008 and 2011 in the Erasmus MC, Rotterdam, the Netherlands. Following coronary angiography, intravascular ultrasound (IVUS) of a non-culprit coronary artery was performed. The human research ethics committee of Erasmus MC, Rotterdam, the Netherlands has approved this study. All included patients have signed informed consent, and the study protocol conformed to the Declaration of Helsinki. This study is registered in ClinicalTrials.gov (number: NCT01789411).

Kidney function assessment
Estimated Glomerular Filtration Rate (eGFR Cr ) was assessed by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [17]. Patients were categorized according to eGFR by using the modified definition from the National Kidney Foundation e Kidney Disease Outcome Quality Initiative (K/DOQI) clinical practice guidelines [18]: normal (GFR!90 ml/min/1.73 m 2 ), mild (GFR 60e89 ml/min/1.73 m 2 ), moderate (GFR 30e59 ml/min/1.73 m 2 ), and severe (GFR 15e29 ml/min/1.73 m 2 ) kidney dysfunction, and kidney failure (GFR<15 ml/min/1.73 m 2 ). No patients with kidney failure were present in this study, and only one patient had eGFR Cr <30 ml/min/1.73 m 2 . The latter was excluded from further analyses. Patients were stratified into those with normal kidney function and those with mildly-to-moderately impaired kidney function, using an eGFR Cr of 90 ml/min/1.73 m 2 as the cut-off value.

Biomarkers
Arterial blood was taken before the procedure and stored at À80 C within 2 h. Samples were available in 570 patients. An immunoturbidimetric high sensitivity assay (Roche Diagnostics Ltd., Rotkreuz, Switzerland) on the Roche Cobas 8000 modular analyser platform was used in the Erasmus MC clinical laboratory to measure the level of C-reactive protein (CRP) in serum samples. The plasma EDTA samples were transported at a temperature of À80 C to Myriad RBM, Austin, Texas, USA, where cystatin C and NGAL concentrations were assessed by a validated multiplex assay (Custom Human Map, Myriad RBM, Austin, Texas, USA). As a result of the batch-wise handling of the samples, with an update of the composition of the multiplex assay by the manufacturer inbetween two batches, cystatin C was measured in the full cohort of 570 patients, and NGAL in a random subset of 473 patients. Both laboratories were blinded to clinical and imaging data.

Grayscale and radiofrequency intravascular ultrasound (IVUS)
Following coronary angiography, we performed IVUS imaging of the most proximal part of a non-culprit, non-treated coronary vessel. The non-culprit vessel was selected based on the following order: left anterior descending artery; right coronary artery; left circumflex artery. We obtained all IVUS data by the Volcano s5/s5i Imaging System (Volcano Corp., San Diego, CA, USA) using a Volcano-Eagle-Eye Gold IVUS catheter of 20 MHz with an automatic pullback system and a standard pullback speed of 0.5 mm/s. Subsequently, an independent core laboratory (Cardialysis BV, Rotterdam, the Netherlands) analysed IVUS images offline, blinded for clinical and biomarker data. The IVUS virtual histology (IVUS-VH) was assessed by pcVH 2.1 and qVH (Volcano Corp., San Diego, CA, USA) software. In each frame, the external elastic membrane and luminal borders were outlined (median interslice distance, 0.40 mm).
The degree (plaque volume and plaque burden) and composition of the atherosclerotic plaque were assessed. Plaque volume was defined as the total volume of the external elastic membrane occupied by atheroma [19]. Plaque burden was defined as the plaque and media cross-sectional area divided by the external elastic membrane cross-sectional area and is presented as a percentage ( Fig. 1). A coronary lesion was defined as a segment with a plaque burden of more than 40% in at least three consecutive frames [16]. The composition of the atherosclerotic plaque was characterized into fibrous, fibro-fatty, dense calcium and necrotic core [20]. Subsequently, three types of VH-IVUS high-risk lesions were identified: 1. Thin-cap fibroatheroma (TCFA) lesion: a lesion with the presence of >10% confluent necrotic core in direct contact with the lumen; 2. A lesion with a plaque burden of !70%; 3. a lesion with a minimal luminal area (MLA) of 4.0 mm 2 [21].

Follow-up
Clinical follow-up started at inclusion and lasted one year. The primary clinical endpoint -MACE -was the composite of all-cause mortality, ACS, or unplanned coronary revascularization. ACS was defined as the clinical diagnosis of ST-segment elevation myocardial infarction (STEMI), non-STEMI, or unstable angina pectoris using the guidelines of the European Society of Cardiology [22,23]. Unplanned coronary revascularizations were defined as unplanned coronary artery bypass grafting or repeat percutaneous coronary intervention. The secondary endpoint was the composite of allcause mortality or ACS. The endpoints were adjudicated by a clinical event committee blinded for biomarker and IVUS data.

Statistical analysis
The Kolmogorov-Smirnov test was used to test distributions of continuous variables for normality. CysC and CRP were not normally distributed and were ln-transformed for further analyses. Categorical variables are presented as numbers and percentages. Continuous variables that were normally distributed are presented as mean ± standard deviation (SD); non-normally distributed continuous variables are presented as median and interquartile range (IQR). For reasons of uniformity, all biomarkers are presented as median (IQR).
We examined the associations of plasma CysC and NGAL levels with plaque burden, plaque volume, and the presence of high-risk coronary lesions. Plaque volume was normalized for the imaged segment length. We used linear regression and logistic regression analyses with continuous ln-transformed CysC and NGAL concentrations consecutively as independent variables. To assess the effect of kidney function, we included interaction terms (ln-transformed CysC or NGAL, respectively, with dichotomized eGFR Cr (above or below 90 ml/min/1.73 m 2 )) into the logistic regression models. Subsequently, we stratified all analyses on eGFR Cr of 90 ml/min/ 1.73 m 2 . To test whether effect estimates differed between patients with ACS and patients with SAP, Z-tests for heterogeneity were performed. Cox proportional hazards regression analyses were performed to evaluate the associations between CysC and NGAL and the clinical study endpoints.
Age, gender, indication for coronary angiography, diabetes mellitus, hypertension, and CRP concentration were considered as potential confounders, and were therefore entered into the multivariable linear and logistic regression models. Multivariable adjustment of Cox proportional hazards models was constrained due to the number of clinical endpoints, and was therefore performed in two steps. For MACE, in the first step the adjustment included age, gender, and indication for angiography; in the second Plaque burden is defined as plaque and media cross-sectional area (i.e., area between yellow contour and red contour) divided by external elastic membrane cross-sectional area (contoured in red); (B) VH-IVUS derived thin-cap fibroatheroma lesion (VH-TCFA), defined as a lesion (i.e., plaque with a plaque burden >40%) with presence of confluent necrotic core >10% in direct contact with the lumen in at least three frames; (C) Odds ratio (OR) per standard deviation increase in ln-transformed cystatin C with 95% confidence interval (CI) for lesions with PB ! 70%. (D) Odds ratio (OR) per standard deviation increase in ln-transformed cystatin C with 95% confidence interval (CI) for VH-TCFA lesions. FI, fibrous; FF, fibro-fatty; NC, necrotic core, DC, dense calcium. a adjusted for age, gender. diabetes, hypertension, indication for angiography, Creactive protein.
step, diabetes mellitus, hypertension and CRP were added.
Finally, we determined the cut-off values of CysC and NGAL that carry the optimal discriminative ability with respect to presence of high-risk coronary lesions and occurrence of MACE. For this purpose, we drew receiver operating characteristic (ROC) curves and calculated the Youden index (highest sum of sensitivity and specificity À1) [24]. We considered only statistically significant associations.
All data were analysed with SPSS software (SPSS 20.0; IBM Corp., Armonk, NY). All statistical tests were two tailed, and p values < 0.05 were considered statistically significant.
Overall, no differences could be demonstrated between CysC and NGAL in either plaque burden or normalized plaque volume of the entirely imaged segment (Table 3 and Supplementary Table 2). Nevertheless, CysC showed a tendency towards lower normalized segment plaque volume (per SD increase in ln-transformed CysC: b [95% CI]:À0.43 [À1.02e0.16], p ¼ 0.16) in patients with normal kidney function; whereas no differences were observed in patients with mild-to-moderate kidney dysfunction.
There was no heterogeneity between ACS and SAP patients regarding the differences in IVUS grayscale parameters according to CysC or NGAL levels.
Both in the total population and in patients with mild-tomoderate kidney dysfunction, a CysC of 849.0 ng/ml was the optimal cut-off value to identify patients who developed MACE (CysC !849.0 ng/ml) ( Supplementary Fig. 7).
Patterns of risk of the secondary endpoint (all-cause mortality and ACS) according to CysC and NGAL levels were similar to those of MACE (Supplementary Table 3).
Finally, stratification on the indication for angiography confirmed the risk patterns which were found in the full cohort (Supplementary Table 4).

Discussion
We found that in patients with normal kidney function, those with higher CysC levels had fewer high-risk coronary lesions (VH-TCFA and lesions with PB ! 70%), while risk of MACE was not different. Conversely, when kidney function was mildly-tomoderately impaired, no differences in high-risk lesions were observed, but those with higher CysC levels had higher risk of MACE. Therefore, with regard to prediction of cardiovascular risk, CysC appears to carry potential only when eGFR Cr is below 90 ml/ min/1.73 m 2 . Furthermore, patients with higher NGAL levels had fewer lesions with PB ! 70%, but only when they had normal kidney function. No differences in MACE were found for NGAL, and thus its use for cardiovascular risk prediction could not be substantiated. Altogether, our results on CysC suggest novel pathophysiological insights, because they offer an explanation for the difference in findings observed in experimental and epidemiologic studies so far, and imply that the association between CysC and cardiovascular disease may not be solely explained through its correlation with GFR.
Higher CysC levels have been associated with occurrence of cardiovascular events in various epidemiological studies [25]. Conversely, animal experiments suggest that higher CysC may be favourable. Atherosclerotic mice deficient in CysC display increased plaque size and macrophage content, increased elastic lamina degradation and accumulation of smooth muscle cells [26,6]. Studies in humans have also found reduced CysC in atherosclerotic and aneurysmatic aortic lesions [7]. Xu et al. have demonstrated that immune cells (CD8 þ dendritic cells (DC) and macrophages), which are involved in atherosclerotic processes, are major contributors to the circulating CysC pool [27,28]. However, besides a correlation with GFR, the mechanisms that may explain the link between CysC and cardiovascular disease are still unclear. Our study provides additional insights. We found that in patients with normal kidney function, those with higher CysC levels had fewer high-risk coronary lesions, and did not have higher risk of MACE. This is in accordance with a potential 'atheroprotective' effect.
Conversely, in patients with mild-to-moderate kidney dysfunction, differences in high-risk lesions according to CysC level were not present. This could possibly be explained by the changes in CysC physiology that occur in impaired kidneys. When kidney function deteriorates, circulating plasma CysC increases and oxidative stress advances, both of which stimulate Cys to form homodimers [28,29]. When CysC forms homodimers, it cannot inhibit cysteine proteases, because the inhibitory region is hidden within the dimer interface. Thus, it may no longer be able to exhibit 'athero-protective' properties [30]. Although these hypotheses are compelling, additional clinical and experimental studies are necessary to further substantiate the effect modification by kidney function that we observed.
Our findings suggest that NGAL may act on coronary artery disease through a different mechanism than currently investigated. A potential lack of predictive precision due to a limited number of MACE may explain the difference between the current results and previous studies [15,31]. On the other hand, a recent meta-analysis that investigated NGAL as a predictor of cardiovascular disease concluded that strong evidence for independent predictive value of NGAL is still lacking [32]. Notably, we found higher plasma NGAL levels in ACS patients compared to SAP patients, independently of kidney function. This could possibly be explained by neutrophilia as a consequence of more severe cardiac damage in ACS patients compared to SAP patients [33]. However, no heterogeneity between ACS and SAP patients was observed in the relationship between NGAL and IVUS-features of coronary atherosclerosis. Hazard ratio (HR) per standard deviation increase in lntransformed cystatin C and per standard deviation increase in NGAL with 95% confidence interval (CI). a unadjusted model; b adjusted for age, gender, indication for angiography; c adjusted for age, gender, indication for angiography, diabetes mellitus, hypertension, C-reactive protein; multivariable adjustment was constrained by the limited number of clinical endpoints. Some limitations of this study merit consideration. This study is currently the largest cohort in which the associations between IVUS plaque characteristics, CysC and NGAL were investigated. Yet, we cannot exclude the possibility of a chance finding with regard to effect modification by kidney function. However, both the cut-off value (based on K/DOQI guidelines) and the study population (no kidney failure/eGFR<30) were chosen a priori. Still, our findings should be considered hypothesis-generating and warrant external validation. Second, kidney function was determined by the creatinine-based CKD-EPI formula, without direct measurement of GFR. Although the CKD-EPI formula has displayed better performance than the Modification of Diet in Renal Disease (MDRD) equation [17], it is still possible that a few patients are misclassified. Third, VH-IVUS imaging was limited to a pre-specified target segment of a non-culprit coronary artery. This study design was chosen based on the hypothesis that such a non-stenotic segment reflects coronary wall pathophysiology of the larger coronary tree [34,35]. This hypothesis, on its part, was based on ex-vivo, as well as in-vivo studies using IVUS in patients with myocardial infarction. These studies have demonstrated the presence of TCFAs in places other than the culprit lesion or even culprit artery [16,36]. In fact, we were subsequently able to confirm this hypothesis, by demonstrating that imaging characteristics of the non-culprit artery are associated with increased risk of MACE within the current study population [34]. Therefore, this study design allows us to investigate whether the patient's burden and vulnerability of atherosclerotic disease e as reflected by the phenotype of a non-culprit artery segment e is associated with blood biomarkers [16]. Finally, although the spatial resolution of IVUS-VH is formally too low to detect thin caps, we have demonstrated that VH-IVUS derived TCFA lesions strongly and independently predict the occurrence of MACE within the current study population [34].
In conclusion, this study provides new insights into the role of plasma CysC and NGAL in coronary atherosclerosis. Most importantly, it shows that in patients with normal kidney function, those with higher CysC levels have fewer high-risk coronary lesions, while in patients with impaired kidneys, those with higher CysC have higher risk of MACE. Thus, this study implies that mild-tomoderate kidney dysfunction modifies the relationship between plasma CysC and coronary artery disease. This has not been established before, and it offers an explanation for the difference in findings observed in experimental and epidemiologic studies. With regard to cardiovascular risk prediction, CysC showed predictive capacities when eGFR Cr was below 90 ml/min/1.73 m 2 , whereas NGAL levels were not predictive of MACE.