Introduction

Proprotein convertase subtilisin/kexin type 9 (PCSK9) is the ninth member of the proprotein convertase family and is a key regulator of cholesterol homeostasis. PCSK9 is secreted by the liver and acts as an endogenous inhibitor of the LDL receptor (LDLR). PCSK9 loss-of-function (LOF) mutations contribute to low levels of plasma LDL-cholesterol and protection against cardiovascular diseases [1, 2]. For instance, the PCSK9 LOF low-frequent variant p.R46L (rs11591147; minor allele frequency in the HapMap-CEU (Centre d'Etude du Polymorphisme [Utah residents with northern and western European ancestry] population: 2.8%) has been shown to increase the number of cell surface LDLRs in hepatocytes in vitro [3], leading to a mean reduction in LDL-cholesterol of 9% to 16% in p.R46L carriers. Recently, several Phase 3 trials have been conducted with human monoclonal antibodies directed against PCSK9 in patients with cardiovascular diseases, or familial or primary hypercholesterolaemia. PCSK9 inhibition has led to a drastic and consistent reduction in baseline LDL-cholesterol levels of up to ~60%, independent of background lipid-lowering therapy [4].

As well as being expressed mainly in the liver, PCSK9 is also expressed in mouse and human pancreatic islets and is able to downregulate the LDLR in human isolated islets [5]. Since it has been suspected that LDLR mediated lipotoxicity can alter beta cell function [6], a potential safety issue with the use of PCSK9 inhibitors is the increased risk of type 2 diabetes. However, results obtained from Pcsk9-deficient mice are discordant. One study found no deleterious effect of Pcsk9 deficiency on glucose homeostasis [5], while another demonstrated that Pcsk9-deficient mice are glucose intolerant with reduced insulin secretion [7].

The aim of the present study was to assess the associations between the PCSK9 p.R46L variant and (1) changes in several glucose and lipid homeostasis variables in non-diabetic French participants; (2) type 2 diabetes in a French case–control study; (3) the incidence of type 2 diabetes over a 9 year follow-up period.

Methods

Study populations

Genotyping of p.R46L was performed in two European-French populations: the Data from an Epidemiological Study on the Insulin Resistance Syndrome (DESIR) and Corbeil studies.

The DESIR cohort is a 9 year longitudinal study in a French general population. We analysed 4630 DESIR participants successfully genotyped for p.R46L, of whom 4319 were non-diabetic during the follow-up period, 2207 had normal glucose homeostasis (and were used as controls in the case–control study), 127 presented with type 2 diabetes at baseline (and were used as cases in the case–control study) and 184 developed type 2 diabetes during the 9 year follow-up.

The Corbeil study included 1342 unrelated French type 2 diabetes patients successfully genotyped for p.R46L who were recruited by the Endocrinology-Diabetology Department of the Corbeil-Essonnes Hospital (Corbeil-Essonnes, France).

Glycaemic status was defined according to 1997 ADA criteria; normal glucose homeostasis was defined as a fasting plasma glucose (FPG) level <6.1 mmol/l without glucose-lowering treatment, and type 2 diabetes was defined as FPG ≥7.0 mmol/l or treatment with glucose-lowering agents. Non-diabetic participants had FPG <7 mmol/l and were not treated with glucose-lowering agents.

The study protocol was approved by the local ethics committees and all participants provided written informed consent.

Genotyping of p.R46L variant

Genotyping of the p.R46L variant (rs11591147) was performed using Metabochip DNA arrays (Illumina, San Diego, CA, USA), as previously described [8]. The individual call rate for the DESIR and Corbeil participants was >98% for the p.R46L variant and no departures from the Hardy–Weinberg equilibrium were observed in either study (p > 0.05). Cluster plots are shown in electronic supplementary material (ESM) Fig. 1.

Index calculation

HOMA-B and HOMA-IR were calculated as previously described [9].

Statistical analyses

The associations between p.R46L and quantitative traits was assessed in 4319 non-diabetic participants from DESIR through linear regression models adjusted for age, sex and BMI (except for the analysis of BMI, which was adjusted for age and sex only).

The case–control analysis included type 2 diabetes participants from DESIR (at baseline) and from Corbeil, and normal glucose controls from DESIR (during the follow-up period). The association between p.R46L and type 2 diabetes was assessed using a logistic regression model adjusted for age, sex and BMI.

The association between p.R46L and incident type 2 diabetes was assessed in DESIR using a Cox regression model adjusted for sex, age and BMI at baseline, and the timescale used was age.

All the analyses were performed under an additive model.

Given the size of our studies, we would have been able to detect a significant OR of 1.55 (type 2 diabetes case–control study) and a significant HR of 4.40 (incident type 2 diabetes study), with a power of 80% (ESM Fig. 2).

Except for statistical power analyses, we performed statistical analyses using SPSS (version 14.0), IBM, Armonk, NY, USA.

Results

In the 4,319 non-diabetic participants from the DESIR study, we assessed the association of PCSK9 p.R46L with BMI, glucose homeostasis related traits (including FPG, FSI, HOMA-B, HOMA-IR and, HbA1c), lipid homeostasis variables (including total cholesterol, HDL-cholesterol, LDL-cholesterol, triacylglycerol, apolipoprotein B [APOB] and apolipoprotein A1 [APOA1]) and liver enzyme levels (alanine aminotransferase and aspartate aminotransferase [Table 1]). When we adjusted our models for age, sex and BMI, we confirmed a marked association between the p.R46L variant and decreased total cholesterol levels (effect size [95% CI] −0.394 [−0.537, −0.251] mmol/l; p = 7.05 × 10−8), decreased LDL-cholesterol levels (effect size −0.393 [−0.525, −0.261] mmol/l; p = 5.18 × 10−9) and decreased APOB levels (effect size −0.099 [−0.136, −0.061] g/l; p = 2.42 × 10−7 [Table 1]). These association results were similar when the models were adjusted for age, sex, BMI and lipid-lowering medication (data not shown). However, we did not find any significant association between p.R46L and variations in other traits (Table 1). According to the Meta-analyses of Glucose and Insulin-Related Traits Consortium (MAGIC), in >133,000 non-diabetic participants, p.R46L was nominally (but not significantly) associated with increase fasting glucose levels (effect size 0.028 [−0.006, 0.050] mmol/l; p = 0.013), and it was not associated with fasting insulin levels (effect size 0.016 [−0.008, 0.040] pmol/l; p = 0.16) [10].

Table 1 Associations between the PCSK9 p.R46L variant and glucose homeostasis related traits, lipid homeostasis variables and liver enzyme levels in 4,319 non-diabetic participants from DESIR
Full size table

Of note, in the 4,319 non-diabetic participants from DESIR, there was no significant difference regarding statin use between p.R46L carriers (n = 4 of 171 [2.3%] participants informed for statin use) and non-carriers (n = 108 out of 4,082 [2.6%] participants informed for statin use; p = 0.81).

In the type 2 diabetes case–control study, we did not find any significant association between p.R46L and type 2 diabetes (p = 0.261) or incident type 2 diabetes over the 9 years of follow-up (p = 0.065 [Table 2]). In addition, there was no significant association between p.R46L and occurrence of impaired fasting glucose (FPG ≥6.1 mmol/l; n = 508 incident cases) over the 9 years of follow-up (HR [95% CI] 0.91 [0.58, 1.42]; p = 0.66; data not shown). According to the Diabetes Genetics Replication and Meta-analysis (DIAGRAM) consortium, p.R46L was not associated with type 2 diabetes risk in a total of 42,590 European participants (OR [95% CI] 1.09 [0.98, 1.22]; p = 0.12) [11].

Table 2 Association between the PCSK9 p.R46L variant and type 2 diabetes risk and incidence
Full size table

Discussion

The major finding of this study is that the PCSK9 p.R46L LOF variant was not associated with impaired glucose homeostasis in humans. Notably, we showed that p.R46L carriers did not have an increased incidence of type 2 diabetes during a 9 year follow-up or type 2 diabetes risk in a case–control study.

Although positive correlations were observed between plasma PCSK9 and both FPG and HOMA-IR in the Dallas Heart Study [12], we did not find any associations between p.R46L and either FPG, HbA1c, or markers of insulin resistance (HOMA-IR) or insulin secretion (HOMA-B). However, a non-clinically relevant tendency for a positive association between p.R46L and FPG was observed in MAGIC. As expected [1, 2], p.R46L remained associated with markers of LDL metabolism with reduced total cholesterol, LDL-cholesterol and APOB concentrations in people carrying the p.R46L allele. The use of statins is associated with an increased risk of type 2 diabetes and could be a potential confounding factor [13]. However, adding lipid-lowering medication status into the model did not modify the results.

The present results are in line with our previous results obtained in PCSK9-deficient mice, in which there was no alteration of glucose homeostasis, as assessed by both glucose and insulin tolerance tests [5]. While PCSK9 deficiency is associated with an increased expression of LDLR in pancreatic islets, there is no increase in intracellular cholesterol content in PCSK9-deficient islets, even under lipotoxic conditions when incubating isolated islets with high concentrations of LDL [5]. This absence of intracellular cholesterol accumulation might be due to reduced circulating LDL levels in PCSK9-deficient mice and/or to an increase in cholesterol efflux from beta cell.

Our study has several limitations that should be highlighted. First, we only focussed on the effect of the most frequent PCSK9 LOF variant, i.e. p.R46L, and we cannot exclude the possibility that other rare PCSK9 LOF mutations with stronger functional effects may alter glucose homeostasis. Second, the intensity of PCSK9 inhibition observed in heterozygous carriers of the p.R46L variant is inferior to those obtained with PCSK9 monoclonal antibodies. Indeed, the decrease of LDL-cholesterol is 10% in p.R46L carriers compared with ~60% with alirocumab or evolocumab in Phase 2 trials [4]. However, such a small decrease in LDL-cholesterol was consistently associated with a highly significant reduction of cardiovascular events in individuals carrying the p.R46L allele [1, 2], certainly because it was maintained throughout their lives.

In conclusion, a genetic PCSK9 deficiency is not associated with an increased risk of type 2 diabetes. These data are reassuring regarding the safety of PCSK9 inhibitors. Precise glucose phenotyping of individuals treated with PCSK9 monoclonal antibodies in ongoing Phase 3 trials will give some additional insights into the relationship between PCSK9 and glucose homeostasis.