The associations between proprotein convertase subtilisin/kexin type 9 E670G polymorphism and the risk of coronary artery disease and serum lipid levels: a meta-analysis

Background Studies had investigated the associations between proprotein convertase subtilisin/kexin type 9 (PCSK9) E670G polymorphism and coronary artery disease (CAD) and lipid levels, but the results were controversial. Thus, we performed this meta-analysis to investigate the association between PCSK9 E670G polymorphism and lipid levels and the susceptibility to CAD. Methods All relevant articles according to the inclusion criteria were retrieved and included in the present meta-analysis. Odds ratios (ORs) with 95 % confidence interval (CI) were used to analyze the strength of the association between PCSK9 E670G polymorphism and the susceptibility to CAD. At the same time, the pooled standardized mean difference (SMD) with 95 % CI was used for the meta-analysis of PCSK9 E670G polymorphism and lipid levels. The publication bias was examined by using Begg’s funnel plots and Egger’s test. Results A total of seventeen studies met the inclusion criteria. For CAD association, the pooled effects indicated that the G allele carriers had higher risk of CAD than non-carriers in dominant genetic model (OR:1.601, 95 % CI: 1.314–1.951, P < 0.001), as well as in allelic genetic model (OR: 1.546, 95 % CI: 1.301–1.838, P < 0.001). When the subgroup analysis stratified by ethnicity and HWE was performed, the positive result existed in most of the subgroups. For lipid levels association, the pooled effects indicated that the G allele carriers had higher TC and LDL-C levels than the non-carriers (for TC, SMD: 0.126, 95 % CI: 0.023–0.229, P = 0.016; for LDL-C, SMD: 0.170, 95 % CI: 0.053–0.287, P = 0.004, respectively). There was no difference in the levels of TG and HDL-C between the G carriers and the non-carriers in the whole population (SMD: 0.031, 95 % CI: −0.048–0.110, P = 0.440; SMD: −0.123, 95 % CI: −0.251–0.006, P = 0.061, respectively). When the studies were stratified by ethnicity and type of study, the G carriers had higher TC levels than the non-carriers (SMD: 0.126, 95 % CI: 0.014–0.238, P = 0.027) in the non-Asian subgroup. The similar results existed in cohort subgroup. The association between PCSK9 E670G polymorphism and LDL-C levels was significant in all subgroups. Meanwhile, the G carriers had higher TG levels than the non-carriers (SMD: 0.113, 95 % CI: 0.012–0.214, P = 0.028) in the case–control subgroup. AG + GG genotypes had lower HDL-C levels than AA genotype in Asian subgroup (SMD: −0.224, 95 % CI: −0.423– −0.025, P = 0.027) and in case–control subgroup (SMD: −0.257, 95 % CI: −0.467–−0.048, P = 0.016). Conclusions The present meta-analysis concluded that PCSK9 E670G polymorphism was associated with CAD risk and lipid levels. Electronic supplementary material The online version of this article (doi:10.1186/s12944-015-0154-7) contains supplementary material, which is available to authorized users.


(Continued from previous page)
Conclusions: The present meta-analysis concluded that PCSK9 E670G polymorphism was associated with CAD risk and lipid levels.
Keywords: Proprotein convertase subtilisin/kexin type 9, Dyslipidemia, Polymorphism, Gene, Coronary artery disease, Meta-analysis Background Coronary artery disease (CAD) is the major manifestations of atherosclerotic process, which is expected to remain one of the leading causes of mortality until at least 2030 [1]. Epidemiological studies demonstrated that elevated serum low-density lipoprotein cholesterol (LDL-C) levels had consistently been shown to be the risk factors for the occurrence and development of CAD. In addition, genetic and other environmental factors are involved in the pathogenesis of CAD [2].
Proprotein convertase subtilisin/kexin type 9 (PCSK9), originally named neural apoptosis regulated convertase-1 (NARC-1), is the ninth member of the proprotein convertase family [3]. PCSK9 is mainly expressed in liver, kidney and intestine. The pro-PCSK9 undergoes autocatalytic intramolecular cleavage at FAQ15↓2SIP residue in endoplasmic reticulum (ER) and Golgi apparatus. After the auto-catalytic procession, the mature and active PCSK9 protein is generated and secreted. By the catalytic domain, PCSK9 binds to the epidermal growth factor-like repeat A domain (EGF-A) of LDL receptor (LDLR) and enhances the degradation of LDLR, which increases the serum lipid levels and accelerates the procession of atherosclerosis. Recently, numerous studies revealed that PCSK9 was closely associated with lipid levels and the risk of CAD.
The coding gene for human PCSK9 protein is located on chromosome 1p32.3 and it encompasses 12 exons and encodes a 692 amino acid glycoprotein. It contains a signal peptide (1-30), a pro-domain (31-152) and a catalytic domain (153-451) followed by a cysteinehistidine rich domain (526-692). The pro-domain acts as an intramolecular chaperone, which ensures the catalytic domain to fold correctly in the ER.
The mutations of PCSK9 were identified as the cause of a third monogenetic form of autosomal-dominant hypercholesterolemia (ADH), except LDLR and apolipoprotein B (apoB) genes [4]. Recently, there is an increasing interest in the role of the gene polymorphisms of PCSK9 in the serum lipids homeostasis and the pathogenesis of CAD.
Up to date, a total of 163 mutations in PCSK9 gene have been found, including 153 substitutions, eight insertions and two deletion mutations (www.ucl.ac.uk/ldlr/ LOVDv.1.1.0/, the last update February 10, 2011). These mutations are divided into gain-of-function (GOF) and loss-of-function (LOF) mutations. LOF mutations in PCSK9 gene increase the concentration of LDLR on the cell surface, which promotes the uptake of serum LDL-C and prevents the procession of CAD.GOF mutations enhance the degradation of LDLR, which reduces the uptakes of LDL-C and ultimately increases the circulating LDL-C levels. In addition, it may also regulate the production and secretion of apoB and enhance the levels of very low-density lipoprotein (VLDL) [5].
PCSK9 E670G variant (rs505151), a common variant in exon 12, is located within the cysteine-rich Cterminal domain, which participates in the regulation of auto-catalysis [6]. The guanine is substituted for adenine at nucleotide 23,968, leading to a change from E to G at the position 670 of the PCSK9 protein, and this change may increase the affinity of PCSK9 for the LDLR. The frequency of 670G was varied greatly in different ethnic populations. Previous studies have revealed that E670G polymorphism was associated with high levels of total cholesterol (TC) and LDL-C and the risk of CAD [7,8], but the results was inconsistent [9,10]. For instance, Slimani A et al. found that the plasma TC and LDL-C levels were significantly higher in 670G carriers than in non-carriers and 670G increased the risk and the severity of CAD [8]. On the contrary, in a Chinese study, this variant was associated neither with elevated LDL-C levels, nor with the CAD susceptibility [10].
To enhance statistical power to produce a more precise result, we performed this meta-analysis based on available data aimed to derive a more precise association of the PCSK9 gene polymorphism and lipid levels and the risk of CAD.

Studies selection
The meta-analysis followed the Perferred Reporting Items for Systematic Reviews and Meta-analysis (Additional file 1: PRISMA) criteria [11]. The electronic databases, such as PubMed, Foreign Medical Journal Service (FMJS) and Embase, were searched for relevant papers on the association between PCSK9 E670G polymorphism and serum lipid levels and/or the risk of CAD. We also searched the Chinese electronic databases, such as Wanfang Data and China National Knowledge Infrastructure (CNKI) databases for Chinese literatures. The following keywords were used in the search: "PCSK", "proprotein convertase subtilisin/kexin type 9", "neural apoptosis regulated convertase-1", "NARC-1", "polymorphism", "gene", "rs505151", "mutation", "variant", "lipid", "dyslipidemia", "coronary heart disease","myocardial infarction", "coronary artery disease", "ischemic heart disease", "acute coronary syndrome", "CAD" and "CHD". The reference lists of the relevant papers that we identified were also checked.

Inclusion and exclusion criteria
The eligible study must met the following criteria: 1) Case-control study or cohort study investigating the relationships between PCSK9 E670G polymorphism and serum lipid levels and/or the risk of CAD; 2) For the lipid levels, the data were presented as mean ± standard deviations (SD) and the study had one of the serum lipid levels, at least including TC, triglyceride (TG), LDL-C or high density lipoprotein cholesterol (HDL-C); 3) For risk of CAD, the frequencies of genotypes in controls and CAD groups were clear.
Exclusion criteria: 1) The study was not conducted in human; 2) Family-based study. If there are several multiple publications from the same population, the most recent literature with the largest sample size was adopted.

Data extraction
A special form was prepared for recording the available information. Two reviewers (Cai and Zhang) independently extracted the following data from each eligible study: first author, year of publication, country, ethnicity, genotype and allele distributions, geontyping method, lipid levels, type of study, Hardy-Weinberg equilibrium (HWE). The disagreements were resolved by consulting with the third author (Shi).

Data analysis
The deviation from the HWE for the PCSK9 E670G genotype distributions was assessed by Fisher's exact test. The odds ratios (ORs) with 95 % confidence interval (CI) were applied to evaluate the strength of the association between the PCSK9 E670G polymorphism and the susceptibility to CAD. Because the frequency of genotype GG in most of studies was low, the pooled ORs were only performed for allelic model (G vs. A) and dominant model (AG + GG vs. AA). The pooled standardized mean difference (SMD) with 95 % CI was used for the meta-analysis of PCSK9 E670G polymorphism and lipid levels. We defined PCSK9 670G carriers as having the AG and GG genotypes. If the unit of the lipids is presented as mg/dl, we converted it to mmol/l. In several studies, the subjects were divided into different subgroups (e.g. male or female, drinker or nondrinker, CAD or ischemic stroke), so we treated each subpopulation as a separate comparison in our metaanalysis. Between-study heterogeneity was investigated and measured by using Cochran's Q test. It was also detected by using the I 2 statistic. If the between-study heterogeneity was significant (I 2 > 50 %, P ≤ 0.05), a random-effect model (a Dersimonian-Laird method) was used to calculate the results. Otherwise, the fixed effect model (a Mantel-Haenszel method) was adopted [12].
Because of the significant heterogeneity among studies, we carried out sensitivity analysis while evaluating the results again by omitting one single study each time. We also performed the subgroup analysis stratified by ethnicity ("Asian" or "non-Asian") and type of study ("case-control" or "cohort") to explore the sources of heterogeneity. The publication bias of literature was examined by Begg's funnel plots, which was verified by Egger's linear regression test.
The STATA version 12.0 (StataCorp LP, College Station, Texas 77845 USA) was used for the meta-analysis.

Characteristics of included studies
Through the initial retrieval, a total of 147 records were selected. 123 records were excluded according to abstract. Full-text articles were retrieved and seven literatures were excluded for the following reasons: (1) data were repetitive or overlapping (n = 2); (2) data were obviously wrong (n = 2); (3) data were not obtained (n = 2); (4) family-based study (n = 1). In the end, 17 literatures met all the inclusion criteria and are involved in the present meta-analysis [7,8,10,[13][14][15][16][17][18][19][20][21][22][23][24][25][26]. The flow diagram of the study selection process is presented in Fig 1. All of the eligible studies included in this study had been approved by the Ethics Committee of their affiliations and our study was approved by the Ethics Committee of Wujin hospital, affiliated to Jiangsu University.
The characteristics of the included studies were listed in Tables 1 and 2. For CAD, nine comparisons from eight articles including 1517 cases and 1795 controls dealt with the association between PCSK9 E670G polymorphism and risk of CAD. All of them were case-control studies and the minor allelic frequencies (MAF) varied greatly from 2.3 to 48.3 %. None of the distribution of genotypes in controls deviated from HWE, except Gao et al's study. Among them, seven studies were performed among Asians, one among Caucasians and one among Africans, respectively. The genotyping method in most studies was PCR-RFLP, except for Mo et al's study.
For lipid levels, ten papers separately provided data for more than one subpopulation (such as male and female groups, drinker and nondrinker groups, and so on), and we treated each subpopulation as a separate comparison. Therefore, 25 comparisons from 15 papers were used to evaluate the relationship between PCSK9 E670G polymorphism and serum lipid levels, including 14,558 participants. Of these, 22 comparisons from 14 papers, 20 comparisons from 13 papers, 21 comparisons from 14 papers and 23 comparisons from 14 papers presented the data on TC, TG, HDL-C and LDL-C, respectively. 11 comparisons were conducted in Asians, 11 in Caucasians, and 3 in other ethnic populations. Genotype distributions in 19 populations or subpopulations were consistent with HWE. Seven papers were case-control studies and eight papers were cohort studies.

Meta-analysis results
Association of the PCSK9 E670G polymorphism with the risk of CAD Due to the rare GG genotype, the dominant and allelic contrast genetic models were used to evaluate the association of the PCSK9 E670G polymorphism with the risk     Fig. 2a, b). Then the subgroup analyses stratified by ethnicity and HWE were performed. In the subgroup analyses stratified by ethnicity, significant associations were found among both Asians (for dominant model Association of the PCSK9 E670G polymorphism with serum lipid levels Table 4, in the whole population, the pooled effects indicated that the G allele carriers had higher TC and LDL-C levels than the non-carriers (for TC, SMD: 0.126, 95 % CI: 0.023-0.229, P = 0.016; for LDL-C, SMD: 0.170, 95 % CI: 0.053-0.287, P = 0.004, respectively) (Fig. 3a, b). There was no difference in the levels of TG and HDL-C between the G carriers and the non-carriers in the whole population (SMD: 0.031, 95 % CI: -0.048-0.110, P = 0.440; SMD: -0.123, 95 % CI: -0.251-0.006, P = 0.061, respectively) (Fig. 3c, d).

Heterogeneity analysis
For CAD, the between-study heterogeneity was not significant (for dominant model, I 2 = 30.4 %, P = 0.175; for allelic model, I 2 = 36.3 %, P = 0.128). For lipid levels, there was significant heterogeneity among the total comparisons for TC, HDL-C and LDL-C (for TC, I 2 = 57.4 %, P heterogeneity < 0.001; for HDL-C, I 2 = 63.7 %, P heterogeneity < 0.001; for LDL-C, I 2 = 68.2 %, P heterogeneity < 0.001). In order to explore the possible sources of heterogeneity, subgroup analyses stratified by ethnicity (Asian or non-Asian) and type of study (cohort or case-control) were performed. But the between-study heterogeneity was not obviously reduced under the subgroup analyses.

Sensitivity analysis
To evaluate the influence of single study on the whole result, the sensitivity analysis was carried out by calculating pooled estimates again when omitting a single study each time. The results suggested that no single study could influence the stability of the overall pooled estimates not only for CAD, but also for lipid levels (Additional file 2: Figure S1A-E).

Publication bias
The Begg's funnel plot was performed to assess the publication bias and no visual publication bias was found in all the comparisons, which was confirmed by Egger's test (for CAD: P = 0.095; for TC: P = 0.143; for TG: P = 0.585; for HDL-C: P = 0.295; for LDL-C: P = 0.464, respectively) (Additional file 3: Figure S2A-E).

Discussion
To our knowledge, the current meta-analysis was the first study to investigate the relationships between PCSK9 gene E670G polymorphism and lipid levels and the risk of CAD. Our study concluded that PCSK9 E670G polymorphism might be associated with lipid levels and the susceptibility to CAD. A growing body of evidence indicated that PCSK9 might influence the serum lipid levels and the progression of atherosclerosis in vitro and in vivo. Through a post-transcriptional pathway, PCSK9 regulates the level of LDLR and then plays a major role in cholesterol homeostasis. In addition to the degradation of LDLR, PCSK9 may also have a potential effect on the apoB100, which may be another factor to atherosclerosis. The monoclonal antibodies of PCSK9 are going to the phase III clinical trials, which are expected to be the new hope for the patients with ADH.
In 2003, Abifadel et al. [27] firstly identified the mutation in the PCSK9 gene. Since then, a large number of LOF and GOF mutations were reported. Among them, E670G, as the GOF mutation, is one of the most commonly investigated PCSK9 gene polymorphisms. The association of PCSK9 genetic polymorphism with the disorders of lipid profile and risk of CAD has been intensively studied, but the results are inconclusive [9, 13-15, 18, 26, 28]. Even in the same ethnic population, the results were also inconsistent [7,10,21]. The discrepancies of the associations between PCSK9 E670G polymorphism and lipid levels and CAD among various populations may be due to the different characteristics of the study population, such as age, gender, ethnicity and/or environmental factors [29]. In addition, a small sample size may also be one cause. To diminish these influences, we retrieved all the relevant articles and analyzed the potential associations between PCSK9 E670G polymorphism and lipid levels and susceptibility to CAD. The present meta-analysis showed that 670G carriers had higher levels of TC and LDL-C than non-carriers and higher risk of CAD in overall populations (SMD: 0.126, 95 % CI: 0.023-0.229, P = 0.016; SMD: 0.170, 95 % CI: 0.053-0.287, P = 0.004, respectively). The dominant model suggested that 670G carriers were at a 1.601-fold higher risk of CAD than non-carriers. In the whole population, there was no difference in the levels of TG and HDL-C between the 670G carriers and non-carriers.
For the lipid study, there was a significant betweenstudy heterogeneity, which may affect the interpretation of the results. To explore the heterogeneity, the subgroup analyses stratified by ethnicity and type of study were performed. But, we still could not explain the source of heterogeneity entirely. In the subgroup analyses, we found that 670G carriers had high TC levels in non-Asian populations (SMD: 0.126, 95 % CI: 0.014-0.238, P = 0.027) and in cohort studies (SMD: 0.133, 95 % CI: 0.010-0.256, P = 0.034). The association between this mutation and LDL-C levels was significant in all subgroups. Interestingly, the 670G carriers had higher TG levels in case-control subgroup (SMD: 0.113, 95 % CI: 0.012-0.214, P = 0.028). Moreover, the 670G carriers had lower HDL-C levels in Asian population (SMD: -0.224, 95 % CI: -0.423--0.025, P = 0.027) and case-control subgroup (SMD: -0.257, 95 % CI: -0.467--0.048, P = 0.016). The sensitivity analysis found that the pooled effects did not change after excluding single study each time, which indicated that the results were stable.
As other meta-analyses, there were several inherent limitations in this study. Firstly and mainly, the between-study heterogeneity is significant, which was a potential problem that may affect the interpretation of the results. As we all know, the between-study heterogeneity may be influenced by age, sex, sample size, type of study, and so on. Although the subgroup analyses were performed, the heterogeneity was not explained entirely. Secondly, the relationships between this SNP and CAD risk and lipid levels did not consider the confounding factors, such as age, sex, smoking, drinking and other lifestyle factors. In the study conducted by Evans et al, E670G polymorphism was associated with increased LDL-C levels in man but not in women. Thirdly, the number of studies included in the meta-analysis was small for CAD risk. Only one study was conducted in Caucasian population and one study was in African population. Fourthly, although the Egger's tests indicated no remarkable publication bias in our meta-analysis, the inevitable publication bias may remain in the results because the papers having negative result were probably more difficult to be accepted for publication. Finally, the present meta-analysis was not able to assess gene-gene and gene-environment interactions.