Association of PCSK9 plasma levels with metabolic patterns and coronary atherosclerosis in patients with stable angina

Design and primary results of the EVINCI study have been previously published [21]. In brief, EVINCI was a multicenter, multinational, prospective, observational study that included 697 patients with stable chest pain or equivalent symptoms and intermediate probability of CAD from 14 centres in 7 European countries (http://​www.​clinicaltrials.​gov, NCT00979199). The recruitment period was 39 months (March 2009 to June 2012). According to the protocol, each patient underwent coronary CTA, stress imaging by myocardial perfusion imaging or wall motion imaging. If at least one of non-invasive tests was positive, invasive coronary angiography was performed. Blood samples were collected from patients in fasting state during the first enrolment visit, before non-invasive imaging, and plasma aliquots were stored in the EVINCI Bio-Bank. Ethics Committee approval was provided by each participating center, and all subjects provided written informed consent.

For the purposes of the present analysis, plasma levels of PCSK9 were evaluated in patients from the EVINCI study who had a completed clinical profile and available blood samples stored in the BioBank (clinical population). Further analyses were performed in the subgroup of patients with available coronary CTA core laboratory analyses (CTA population). The study flow diagram is detailed in Fig. 1.

Height and body weight were measured by standard methods with the participants in light clothing. Body mass index (BMI) was calculated as body weight (in kg) divided by the square of the height (in m). Homeostatic model assessment of insulin resistance index (HOMA-IR) was calculated as fasting glucose (mg/dL) × fasting insulin (pmol/L)/8.66 [29]. Presence and extent of obesity was defined according to body mass index (BMI) categories [30]. Diabetes was defined as fasting plasma glucose (FPG) > 125 mg/dL or treatment. The presence of metabolic syndrome was diagnosed according to the National Cholesterol Education Program Adult Treatment Panel III [31] criteria, when three or more of the following criteria were present:

Since data on waist circumference were not collected during the EVINCI study, central obesity was defined as BMI over 30 kg/m², based on the assumption that in this condition waist circumference does not need to be measured [32].

PCSK9 plasma levels were measured in blood samples stored in the EVINCI biological bank at the Consiglio Nazionale delle Ricerche (CNR)-Institute of Clinical Physiology, Pisa, Italy) [20, 22] by a dedicated ELISA (Quantikine ELISA, R&D Systems) (0.07 ± 0.007 ng/L LoD; 3.96% intra-run variation; 8.33% inter-run variation). Additional specific metabolic and inflammatory biomarkers were measured using standard methods [24]. LDL cholesterol was calculated according to Friedewald formula [33].

The methodology for coronary CTA acquisition and analysis in the EVINCI study has been previously reported in detail [21, 23, 24]. Briefly, each segment of the AHA 17-coronary segment model was assessed for interpretability, and interpretable segments were evaluated for the degree of stenosis of the coronary artery. If a plaque was present, plaque composition was visually determined (calcified, non calcified, and mixed). Only one type of plaque composition could be assigned to a single segment. A previously validated CTA score, used as an indicator of the global coronary atherosclerotic burden and risk, was derived in each patient by integration of all data on the location, severity and composition of CAD [27, 28].

Categorical variables are presented as numbers (percentage), continuous variables as mean ± SD. The logarithmic transformation of continuous variables was used in the analyses. To better discriminate patients with the highest and the lowest PCSK9 plasma levels, clinical, biohumoral and imaging features were compared among 4 groups according to PCSK9 quartiles (Additional file 1: Tables S1, S2, S3) or among 3 groups (I vs II–III vs IV quartiles) using analysis of variance (ANOVA) with post hoc tests for multiple comparisons or the Chi square test, as appropriate. Comparison between two groups was performed using the Student’s t test. Pearson correlation was used to assess the relation between bio-humural variables in specific patients subgroups.

Multivariate linear regression was used to estimate the effect of clinical and bio-humoral variables, including PCSK9 plasma levels, on the CTA score. A multivariable model was developed, considering variables with a P value < 0.1 at univariable analysis, and then using backward and forward stepwise selections to build-up the final model. All analyses were performed using the SPSS 23 software. A 2-sided value of P < 0.05 was considered statistically significant. There is no multiplicity adjustment implemented in statistical testing.

The clinical population consisted of 539 EVINCI patients with a completed clinical and bio-humoral profile, and in whom PCSK9 plasma levels were determined (Fig. 1). The mean value of PCSK9 was 212.0 ng/mL (SD 104.9 ng/mL), and the median value was 183.8 ng/mL (95% CI 203.2–220.9 ng/mL).

Clinical characteristics among different PCSK9 groups are detailed in Table 1. Patients in the highest PCSK9 quartile had a more frequent family history of CAD and a lower BMI. On the other hand, metabolic syndrome was more prevalent in the lowest Quartile and was progressively less frequent from Quartile I to IV. Among medications, the use of anti-diabetic drugs and aspirin, but not of statins, varied significantly among groups.

The bio-humoral profile comparison among the various PCSK9 groups are reported in Table 2. There was a significant increasing trend in circulating transaminases, alkaline phosphatases, total cholesterol, LDL cholesterol, Apo B and Lipoprotein (a) passing from the lowest to the highest PCSK9 quartiles. Conversely, HDL cholesterol, total/HDL cholesterol ratio, Apo A1 and adiponectin levels were lower, and insulin levels, HOMA-IR and HO-1 higher, in the lowest PCSK9 quartile. No differences were observed for inflammatory biomarkers among groups. In the overall population, PCSK9 levels were significantly lower in patients with the metabolic syndrome, or diabetes, or high BMI, or low HDL cholesterol (Fig. 2). The correlation of PCSK9 with LDL cholesterol plasma levels was lost in these specific patients groups (Fig. 3).

Statin treatment did not significantly affect PCSK9 plasma levels (222 ± 1 ng/mL vs 202 ± 9 ng/mL, no statin users vs statin users, P = ns). Moreover, statin treatment abolished the significant correlation between PCSK9 and LDL cholesterol, while had no effect on the tight correlation of PCSK9 levels with HDL cholesterol (Fig. 4).

PCSK9 plasma levels were not associated with obstructive CAD at coronary CTA (Table 3). In fact, the diagnosis of obstructive CAD (i.e., at least one plaque with either > 50% or > 70% diameter stenosis) and the number of obstructive plaques per patient were similar across PCSK9 groups. On the other hand, PCSK9 plasma levels were inversely related with non obstructive CAD and the extent of global coronary atherosclerotic burden. In fact, the number of non-obstructive plaques and in particular of mixed plaques, as well as the global CTA score were significantly higher in patients with the lowest PCSK9 levels (Quartile I) as compared with those with the highest levels (Quartile IV). In the whole population, PCSK9 plasma levels showed a significant trend to reduction among CTA score classes, in particular in patients with CTA score > 5 (Fig. 5).

At univariable and multivariable analyses, including all clinical (Additional file 1: Table S4) and bio-humoral variables (Additional file 1: Table S5), low PCSK9 levels were an independent predictor of the CTA score together with age, gender, interleukin-6 and leptin (Table 4). Interestingly, when HDL cholesterol plasma levels were forced into the final multivariable model, the association of PCSK9 with the CTA score lost statistical significance (Table 4).

The relationship between circulating PCSK9 levels and CAD has been extensively explored. PCSK9 plasma levels have been associated with obstructive coronary lesions in patients with acute coronary syndrome and myocardial infarction [6‐10] as well as with subclinical coronary atherosclerosis and imaging markers of atherosclerotic risk even if in some reports this relationship is not so clear [11‐18]. The association of PCSK9 with coronary atherosclerosis in contemporary populations may be influenced by a number of factors. There is a shift in the clinical presentation of atherosclerosis from acute coronary syndromes to chronic atherosclerotic disease, which implies a possible change in the underlying mechanisms and prevents a simple extension of the results obtained by association studies in high risk patients to lower risk populations [19]. The diffusion of treatments with lipid lowering drugs have altered the relationship between lipid profiles and coronary disease [19, 20] and may uncover other emerging disease determinants, such as obesity, metabolic syndrome and diabetes. The present study was specifically designed to explore the relationships between circulating levels of PCSK9, comprehensive circulating determinants of the atherosclerotic risk and precise characterization of coronary artery disease in contemporary patient populations with suspected CAD. We did not find an association of PCSK9 levels with obstructive CAD but we documented a clear inverse correlation between PCSK9 and global coronary atherosclerotic burden as described by CTA score. These somewhat surprising results could be explained tacking into account some peculiarity of the EVINCI study. Firstly, the population was a contemporary European multicentric group of patients with stable angina, receiving optimal medical treatment and with a lower prevalence of obstructive CAD (35% of patients with CTA stenosis > 50% and 11% with stenosis > 70%) than usually found either in patients with acute coronary syndromes [7, 8] or referred to coronary catheterization in whom circulating PCSK9 had been previously measured [9, 10]. Secondly, using a comprehensive CTA score, we were able to assess the global coronary atherosclerotic burden at the patient level [27, 28]. In fact, this score, incorporating number and characteristics of all coronary plaques, is a better marker of the presence and extent of coronary atherosclerosis than estimates of CAD presence and severity only based on “hemodynamically significant” lesions. Thus, the CTA score takes into account the possible effects of multiple atherogenic processes on the final coronary phenotype and is a comprehensive predictor of coronary outcome [28]. This was particularly relevant considering that the present study group had a peculiar metabolic profile linked to diffuse coronary atherosclerosis in the absence of high LDL cholesterol levels usually associated with obstructive CAD.

In our population circulating PCSK9 was still correlated with total and LDL cholesterol [together with Apo B and Lipoprotein (a)] but showed also a strong correlation with HDL cholesterol levels (together with Apo A1). The relationship of PCSK9, total or LDL cholesterol and Lipoprotein (a) has been demonstrated in many different populations and in large study cohorts [11, 34‐37]. On the other hand, the relationship of PCSK9 levels and HDL cholesterol is less well established, and a clear involvement of PCSK9 in HDL metabolism is indeed uncertain [12, 38‐42]. Interestingly, we found that the relationships of PCSK9 with LDL cholesterol was affected and, as matter of fact, obscured by statin treatment, while that with HDL cholesterol was not (Fig. 4). This is consistent with previous observations studying the effects of statin treatment on PCSK9 levels [43‐45] showing that the direct correlation between PCSK9 and LDL cholesterol is lost after treatment [43]. Current results add evidence that statins do not affect the strong relationship of PCSK9 with HDL cholesterol.

We found a relationship of low PCSK9 levels with insulin resistance or diabetes, obesity or metabolic syndrome. The association of PCSK9 with glucose metabolism has been extensively studied at epidemiological, genetic, clinical and animal levels [46‐50]. The possible interaction of PCSK9, glucose metabolism and treatment has also been investigated. Some antidiabetic drugs have been shown to decrease PCSK9 levels [51] while treatment with PCSK9 inhibitors was not associated with increased incidence of diabetes [52]. Even if most findings suggest a direct correlation of PCSK9 with abnormal glucose metabolism and insulin resistance, some studies suggest an opposite relation similar to that found in the present work.

In particular, patients with specific loss-of-function PCSK9 mutations or polymorphisms had increased prevalence in pre-diabetes and diabetes [53] as well as higher HOMA-IR index and insulin levels [54]. Interestingly, the direct relationship between PCSK9 and Lipoprotein (a) observed in the present study is in keeping with the association of low PCSK9 levels with insulin resistance and diabetes. In fact, in a large population study and metanalysis Paige et al. demonstrated the association of low Lipoprotein (a) concentrations and higher risk of type-2 diabetes [55].

Similarly, the inverse correlation of circulating PCSK9 with BMI, consistent with the direct correlation with adiponectin, observed in the present study, was not a completely expected result [46, 56]. Actually, an inverse association between PCSK9 and BMI had been already observed in some low cardiovascular risk populations such as healthy subjects [11], patients with an altered metabolic liver function, such as those with an increased hepatic fat [57], children and adolescents after adjustment for age, glucose, insulin, and adiponectin [58]. However, the mechanism of these associations is still unknown.

Similarly, clinical, genetic, and experimental evidence linked PCSK9 with metabolic syndrome [46]. In the present study, we found no association between PCSK9 and some components of the metabolic syndrome such as triglycerides and hypertension, while we found a very strong association between low PCSK9 levels and HDL cholesterol levels, independent of statin use. The association between lower PCSK9 levels and lower levels of adiponectin further underlines the link of PCSK9 with a specific metabolic phenotype [56]. As matter of fact, these results add some complexity to the known relation between PCSK9 and cholesterol. In fact, at least in this population, diabetes, obesity and the metabolic syndrome seem to override the influence of PCSK9 on LDL metabolism, as demonstrated by the loss of the correlation between PCSK9 and LDL cholesterol in patients with altered metabolic phenotype (Fig. 3).

Moreover, we also observed a tight relationship between circulating PCSK9 and liver enzymes, classically considered as markers of liver cellular damage. More recently liver enzymes have been also considered as surrogate biomarkers of liver metabolic function [59]. In the present population liver enzymes where in the normal range and the observed correlation with PCSK9 might express variable liver metabolic function (Additional file 1: Tables S6, S7). This observation needs further validation and the molecular mechanisms could be explored.

Given the peculiar metabolic patterns associated with low PCSK9 levels and the inverse relationship observed between PCSK9 and the CTA score, the present results suggest a possible pathophysiologic link between low circulating PCSK9, a metabolic phenotype characterized in particular by low HDL cholesterol and diffuse coronary atherosclerosis in a population at low prevalence of obstructive CAD. In a previous study performed in this population, we demonstrated that low plasma HDL cholesterol was an independent predictor of the CTA score together with age, gender, leptin and IL6 [23]. In the present study, low PCSK9 was an independent predictor of the CTA score in a multivariable model including the same variables and outperforming a wide variety of risk factors and bio-humoral markers. However, when HDL cholesterol levels were added to the analysis, the association of PCSK9 with the CTA score lost significance, strongly suggesting a contribution of HDL cholesterol to PCSK9-associated atherosclerotic risk. It has been very recently reported that individuals who reach low levels of LDL cholesterol may be persistently at high risk of atherosclerotic disease when total/HDL cholesterol ratio is discordantly higher [60]. Patients with such discordance had a greater atherogenic clinical profile with high levels of triglycerides, high BMI and a high prevalence of risk factors, including diabetes [60]. Interestingly, a similar metabolic pattern was recognized in the present study in patients with low PCSK9 levels who also showed low HDL and high total/HDL cholesterol ratio (Table 2).