Circulating PCSK9 levels and CETP plasma activity are independently associated in patients with metabolic diseases

We studied 450 participants (median age, 60 years; 49 % women) who attended the metabolism unit because of metabolic syndrome (MetS; 78 %) according to the ATPIII criteria, atherogenic dyslipidemia (AD; 32 %), obesity (defined as BMI ≥ 30 kg/m²; 50 %), type 2 diabetes mellitus (T2DM; 72 %), or other risk factors (hypercholesterolemirefa/hypertriglyceridemia; 13 %). Subjects with chronic lung, renal or liver disease; cancer; or any other serious disease were excluded. Patients on lipid-lowering drugs (N = 391) underwent a wash-out period of at least 6 weeks (8 weeks if they were on fibrates). Anamnesis and physical examination data were recorded. The Hospital Ethical Committee approved this study, and all the patients provided written consent to participate in the study.

A blood sample was obtained after overnight fasting from each patient in the study cohort. Plasma and serum aliquots were prepared and stored at −80 °C in the BioBanc at our centre until further use. Cellular buffy coats were obtained from 425 individuals, and the cells were stored at −80 °C until DNA analyses were performed.

Biochemical parameters, lipids, and apolipoproteins were measured using colorimetric, enzymatic and immunoturbidimetric assays (Spinreact, SA, Spain; Wako Chemicals GmbH, Germany; Polymedco, NY, US; CV < 4 %) that were adapted to a Cobas Mira Plus Autoanalyser (Roche Diagnostics, Spain). The lipid profile was analysed according to Spintrol “H” CAL (Spinreact, SA, Spain) GC–MS reference methods. Spintrol “H” Normal (Spinreact, SA, Spain) was used as a quality control. All samples were subjected to the Liposcale test. This advanced lipoprotein test is based on 2D diffusion-ordered (1)H nuclear magnetic resonance (NMR) spectroscopy. This method adds diffusion coefficients to classical NMR determinations to provide a direct measure of mean particle size and number [23]. CETP activity was measured using a fluorometric assay (BioVision, CA, USA). Lecithin-cholesterol acyltransferase (LCAT) activity was assessed using a fluorometric assay (Calbiochem, CA, USA). Circulating PCSK9 levels were measured using commercial ELISA kits (R&D Systems, MN, USA).

We genotyped the rs11591147 (R46L) genetic variant of PCSK9, which was selected from the International HapMap database (http://​hapmap.​ncbi.​nlm.​nih.​gov/​). The LOF PCSK9 genetic variant rs11591147 (R46L) is associated with lower levels of plasma PCSK9 [24]. Genomic DNA was extracted from peripheral leukocytes that had been isolated from anticoagulated venous blood using a QIAamp DNA Blood Kit (Qiagen Iberia SL, Madrid, Spain) according to the manufacturer’s instructions. The DNA was genotyped for the genetic variant on the Sequenom MassARRAY platform using the iPLEX Gold protocol as specified by the manufacturer (Sequenom Inc., San Diego, CA) [25]. The genotypes of 5 of the samples were confirmed using duplicate Sequenom runs, and they showed 100 % consistency. Genotyping was performed at the Spanish National Genotyping Centre.

The results are expressed as the mean ± SD for normally distributed data, the median (interquartile range) for data that was not normally distributed, and frequencies for categorical data. The differences between groups were assessed using the Mann–Whitney U test or the Kruskal–Wallis test. Correlations were evaluated using Spearman’s test. Multiple linear regression analysis was used to test the association between CETP activity and PCSK9 levels with the following covariates: age, gender, BMI, systolic blood pressure (SBP), LDL-C, triglycerides, glucose, LCAT and MetS. All the association analyses were adjusted for age, sex and BMI (when appropriate). The Hardy–Weinberg equilibrium test was performed in the studied population using a χ² test. Statistical analyses were performed using SPSS software (IBM SPSS Statistics, version 22.0). All statistical tests were two-tailed and P < 0.05 was taken as significant.

The clinical, metabolic and biochemical parameters of the 450 recruited subjects (male/female: 231/219) are shown in Table 1. The median age of the study participants was 60 years; 78 % had MetS, 32 % had AD, 50 % had obesity, 72 % had T2DM, and 13 % had other risk factors.

Both PCSK9 levels and CETP activity were significantly elevated in patients with MetS [327.9 (261.9–417.9) vs 288.1 (233.7–360.9) ng/mL, P < 0.004, and 12.03 (10.33-14.15) vs 10.90 (10.08–12.02) nmol/L per h, P < 0.0001, respectively]. PCSK9 levels and CETP activity increased significantly as the number of MetS components increased (P = 0.002 and P < 0.0001, respectively) (Fig. 1). After adjustment PCSK9 levels and CETP activity for age and gender, the association with the MetS components remained significant in both parameters (P < 0.0001).

CETP activity stratified by quartiles of PCSK9 concentration in the whole cohort is shown in Fig. 2. A higher PCSK9 quartile corresponded with higher CETP activity (P < 0.001). The univariate analysis showed that this relation persisted after adjustment for age, gender, BMI, SBP, triglycerides, LDL-C and glucose (P = 0.003). A significant positive association between PCSK9 and CETP was observed in the whole group (r = 0.256, P < 0.001) (Table 2). After adjusting PCSK9 levels for age, gender, BMI, SBP, LDL-C, triglycerides and glucose, the correlation with CETP remained significant (r = 0.158, P = 0.003). CETP was also significantly positively associated with SBP, diastolic blood pressure, cholesterol, LDL-C, triglycerides and glucose and negatively associated with LCAT (P < 0.05) in the whole group. LDL particle number showed a positive association with CETP activity (P < 0.001), whereas there was a negative association with medium HDL particles (r = −0.193, P < 0.001) and positive associations with small and large HDL particles (P < 0.001). PCSK9 was positively associated with LDL and LDL particle number and inversely associated with LDL size. PCSK9 levels also correlated with an AD profile (Table 2).

To further explore the relationship between CETP and PCSK9, a linear regression model was generated (Table 3). The significant direct relation between CETP and PCSK9 remained robust after additional adjustment for the following covariates (model 5): age, gender, BMI, SBP, LDL-C, triglycerides, glucose, LCAT and MetS (beta 0.110; P = 0.003).

The impact of the LOF PCSK9 genetic variant rs11591147 (R46L) on CETP activity was studied in 425 patients. The frequency of PCSK9 rs11591147 (R46L) heterozygosity was 5.2 % among the whole group (Table 1 ). As expected, individuals harbouring the variant GT genotype had significant lower levels of LDL-C [2.55 (2.40–3.46) mmol/L vs 3.28 (2.55–4.08), P = 0.010] and PCSK9 [193.6 (147.8–264.9) vs 328.8 (258.9–410.2) ng/mL, P < 0.0001)] than individuals with the wild type GG genotype (Fig. 3a, b, respectively). Interestingly, individuals with the variant GT genotype had significantly lower CETP activity than individuals with the wild type GG genotype [10.15 (9.59–11.93) vs 11.65 (10.31–13.50) nmol/L per h, P = 0.009]; this finding remained significant (P = 0.044) after adjusting for age, gender and BMI (Fig. 3c).

In this study, we explored the association between circulating PCSK9 concentration and plasma CETP activity in a group of patients with increased cardiovascular risk due to metabolic alterations who were not on lipid-lowering therapies or who underwent a wash-out period. We found a moderate but significant and robust direct association between both parameters that was maintained after multiple adjustment for confounding covariates, including plasma lipid concentrations. This is the first report suggesting a direct correlation between these two biochemical parameters in humans. Moreover, our data suggest that this association is independent of intermediate factors.

Because both parameters had parallel behaviour in the presence of metabolic alterations, such as MetS components, a possible explanation could be that higher PCSK9 concentrations, which evoke higher LDL-C plasma levels, would increase CETP activity by increasing substrate concentration. Conversely, higher CETP activity would elicit higher LDL-C and PCSK9 concentrations. Abifadel et al. observed that endogenous plasma CETP activity was significantly increased in familial hypercholesterolemia patients carrying the PCSK9 Leu181Arg gain-of-function mutation. They suggest this effect could be due to an increase of plasma levels of cholesterol ester acceptors [26]. However, the association between PCSK9 and CETP activity in our study was maintained after adjusting for multiple covariates, including LDL-C plasma concentration, thereby weakening this hypothesis.

Recently, it was reported that CETP inhibitors influence PCSK9 levels as an off-target effect. In vivo data showed that CETP inhibitors decrease PCSK9 levels in humans [17] and in rhesus macaques [16]. This effect could explain why CETP inhibitors reduce (V)LDL-C. Dong et al. demonstrated in liver cells that CETP inhibitors reduced the mature form of SREBP2, leading to attenuated transcription of LDLR and its degrader PCSK9 [22]. In addition, Miyosawa et al. demonstrated that K-312, a new CETP inhibitor, decreased the occupancy of SREBP1 and SREBP2 on the sterol regulatory element within the PCSK9 promoter [21]. Consistent with these results, van der Tuin et al. demonstrated in ApoE*-Leiden.CETP mice that anacetrapib lowers hepatic PCSK9 expression by interacting with SREBP2, leading to reduced plasma PCSK9 levels and increased hepatic LDLR increasing LDLR-mediated hepatic clearance of remnant lipoprotein and decreasing atherogenicity in animal models [15]. Although these studies suggest an effect of CETP inhibitors on PCSK9 independent of CETP itself, a metabolic association between these parameters cannot be excluded. As in the PCSK9 gene, the promoter region of the CETP gene has recognition domains for SREBP1a and SREBP2, and thus a pathophysiological parallel regulatory mechanism that affects both PCSK9 and CETP protein expression must be considered. In our study, we did not measure CETP mass; however, CETP mass and ex vivo activity were highly correlated in previous studies [27]. To confirm this finding, we analysed CETP mass and activity in a group of 30 patients with T2DM. There was a significant positive association between CETP mass and activity (r = 0.543, P = 0.002) in the diabetic patient group (Additional file 1: Figure S1A). In addition, PCSK9 was also significantly positively associated with CETP mass (r = 0.521, P = 0.003) (Additional file 1: Figure S1B), supporting the hypothesis of common gene upregulation in patients with metabolic alterations.

In our cohort, 5.2 % of the patients harboured the PCSK9 R46L LOF mutation, similar to previous data in Caucasian subjects [24, 28]. This subgroup of patients had both low PCSK9 and LDL levels, and interestingly, CETP activity was also reduced. These data support the hypothesis that the correlation between PCSK9 and CETP activity is not dependent on external environmental factors that influence these parameters. A possible mechanism could be that PCSK9 LOF mutations result in increased LDLR occupancy at the cell surface, thereby increasing LDL internalization and intracellular cholesterol concentrations and leading to lower SREBP activity, which could influence PCSK9, and possibly CETP, expression.

There are several implications of this putative metabolic association that should be considered when ascertaining the impact of CETP inhibitors on PCSK9 levels that is attributed to off-target effects. PCSK9 inhibitors, in addition to their important LDL-lowering effect, induce a moderate but consistent increase in HDL cholesterol concentration. Our data could partially explain this phenomenon as a consequence of the decrease in CETP activity associated with lower PCSK9 levels.

In general, a parallel reduction of CETP activity along to lower PCSK9 levels could be seen as beneficial, however the overall effect of CETP on atherogenesis is under debate. While hypertriglyceridemia stimulates CETP enzymatic activity [29], a reduction on cholesterol ester transfer from triglyceride rich lipoprotein to HDL has been associated to coronary heart disease presence in diabetic patients [30]. If these observations are reproduced in patients on PCSK9 inhibitors treatment is not known, but it could explained the wide impact of these drugs on lipid profile affecting in a moderate way, triglycerides and HDL reinforcing the role of this treatment in patients with combined hyperlipidemias.

Our study has several limitations. It was an observational study, and thus, we cannot explain the mechanisms underlying the findings. Unfortunately, we had no access to patients on PCSK9 inhibitors who could have provided more information regarding the CETP-PCSK9 association; however, we have data on patients with genetically driven low PCSK9 levels. Our included patients with metabolic alterations, so we cannot extrapolate our findings to other populations.