Associations of the APOC3 rs5128 polymorphism with plasma APOC3 and lipid levels: a meta-analysis

Background Studies of the association between the apolipoprotein C3 gene (APOC3) rs5128 polymorphism and plasma levels of apolipoprotein C3 (APOC3) and lipids have reported apparently conflicting findings. This meta-analysis aimed to investigate the associations of the rs5128 polymorphism with fasting APOC3 and lipid levels. Methods The following information was abstracted for each study: ethnicity, age, sex, health condition, sample size, genotyping and lipid assay methods, mean and standard deviation or standard error by genotypes for APOC3 and lipid variables. There were 42 eligible studies with 23846 subjects included in this meta-analysis. A dominant model was used for this meta-analysis. Results The results showed that the carriers of the variant allele G had higher levels of APOC3 [standardized mean difference (SMD): 0.22, 95% confidence interval (CI): 0.12-0.31, P < 0.00001], triglycerides (TG) (SMD: 0.33, 95% CI: 0.23-0.44, P < 0.00001), total cholesterol (TC) (SMD: 0.15, 95% CI: 0.09-0.22, P < 0.00001), and low-density lipoprotein cholesterol (LDL-C) (SMD: 0.11, 95% CI: 0.04-0.17, P = 0.001) than the non-carriers. No significant association between the APOC3 rs5128 polymorphism and lower levels of high-density lipoprotein cholesterol (HDL-C) was detected under the dominant model (SMD: −0.03, 95% CI: −0.06-0.01, P = 0.156). Conclusions The results from the present meta-analysis demonstrate a significant association between the APOC3 rs5128 polymorphism and higher levels of APOC3, TG, TC and LDL-C, but further studies are needed to elucidate the underlying mechanisms. Electronic supplementary material The online version of this article (doi:10.1186/s12944-015-0027-0) contains supplementary material, which is available to authorized users.


Introduction
Cardiovascular disease (CVD) is now the leading cause of death worldwide, both in developed countries and in many developing countries [1]. After decades of study, a number of CVD risk factors have been identified. Among these risk factors, dyslipidemia, characterized by elevations of triglycerides (TG), total cholesterol (TC) and/or low-density lipoprotein cholesterol (LDL-C), or reduction of high-density lipoprotein cholesterol (HDL-C) levels, accounts for at least 50% of the populationattributable risk for CVD [2]. In recent years, much has been learned about the chromosomal regions and specific genes that influence plasma lipid or lipoprotein levels [3]. However, due to various reasons, it is difficult to identify the susceptibility genes and most associations have not been replicated across studies.
The apolipoprotein C3 gene (APOC3) is a member of the APOA1/C3/A4/A5 gene cluster and located on chromosome 11q23, an area in strong linkage with lipid metabolism [3]. Apolipoprotein C3 (APOC3) is an essential constituent of TG-rich particles including chylomicron and very low-density lipoprotein (VLDL), to a less extent of high-density lipoprotein (HDL). Evidence from gene engineering studies suggested that the main role of APOC3 could be involved in the regulation of plasma TG levels [4,5]. Additional copies of human APOC3 in transgenic pig were associated with hypertriglyceridemia (HTG) [4], whereas APOC3 deficiency prevents hyperlipidemia induced by apolipoprotein E gene overexpression [5]. In vitro, APOC3 was found to delay the catabolism of VLDL by inhibiting the lipoprotein lipase (LPL) which is the rate-limiting enzyme for TG hydrolysis [6]. Furthermore, it also interferes the apolipoprotein E-mediated remnant removal by displacement of apolipoprotein E from VLDL particles in vivo [7,8]. APOC3 also play important roles in modulating other lipid variables [9]. Given its role in plasma lipid metabolism, APOC3 is considered a candidate gene for dyslipidemia.
Of several variants within APOC3, a transversion from C to G in the 3' untranslated region (3'UTR) of exon 4 results in a rs5128 polymorphism (also known as SstI, sacI, 3238C > G or 3175C > G). A recent metaanalysis [10] demonstrated that the rs5128 polymorphism is associated with CVD, but whether this polymorphism is also associated with dyslipidemia remains to be further examined. A large body of studies has investigated the associations of this polymorphism with plasma APOC3 and lipid levels . Some studies found that this polymorphism is associated with higher plasma levels of APOC3 [14,20,27,41,42], TG [18,20,23,25,28,30,32,33,38,41,42,44,46,51], TC [18,32,51] and LDL-C [30,51], and lower levels of HDL-C [24]. However, the results from other studies were conflicting and inconclusive. In this paper, a meta-analysis was performed on previous reports to investigate the associations of the rs5128 polymorphism with APOC3 and fasting lipid levels. Our analysis results can provide the opportunity to elucidate the interrelationship among the rs5128 polymorphism, dyslipidemia and CVD.

Identification and eligibility of relevant studies
All articles published before May 2014 on the associations of the APOC3 rs5128 polymorphism with plasma APOC3 and lipid levels were identified. The language was limited to English. A comprehensive search of the literature was carried out by using PubMed and EMBASE. The keywords used for this search were "APOC3 OR APOC-III OR apolipoprotein C-III OR apolipoprotein CIII OR apolipoprotein C3 OR APOC3 OR APO C3" concatenated with "polymorphism OR variant OR SNP OR mutation". This meta-analysis is limited to APOC3 and the plasma lipid variables including TG, TC, LDL-C, and HDL-C. The studies that fulfilled the following criteria were included: (1) data reported on fasting variables; (2) data reported on APOC3 and/or at least one of the four plasma lipid variables; (3) pre-intervention baseline data were used in interventional studies; (4) the studies in which mean plasma lipid levels and standard deviations (SD) or standard errors (SE) by genotype were available. In order to find other published work which was not indexed by PubMed and EMBASE, all references cited in the included articles were reviewed. Case reports, review articles, abstracts, animal studies, reports with incomplete data, and studies based on pedigree data were excluded.

Data extraction
The irrelevant and overlapping studies were excluded after being reviewed independently by two reviewers using a structured data collection form. The results were compared, and the disagreements were resolved by discussion. Regarding the overlapping articles, only those publications that reported the most extensive information were included. For each study included in the present meta-analysis, the following information was extracted: first author, year of publication, ethnicity, country of origin, age, sex, health condition, genotyping and lipid assay methods, sample size, mean APOC3 or lipid variables and SD or SE by genotypes.

Statistical analysis
All data were presented as mean ± SD in this analysis. For those included articles in which mean ± SE was given, the value of the SD was calculated. The STATA software package v 10.0 (Stata Corporation, College Station, TX) was used for the meta-analysis. Due to the low frequencies of the GG genotype, a dominant model [(CG + GG) versus CC] was employed to ensure adequate statistical power. When data was presented for more than one subpopulation (for example, female or male subjects, the subjects with CVD or type 2 diabetes, the subjects from different ethnicity) in one article, each subpopulation was treated as a separate comparison in this meta-analysis. In addition, subgroup analyses were conducted by age, gender, ethnicity and health condition. Age subgroups were defined as adults and children (under 18 years old). Ethnic subgroups were defined as Caucasian, Asian, and the populations of other ethnic origins. The meta-analyses on the subgroup were only performed with at least four comparisons to ensure adequate statistical power.
The pooled standardized mean difference (SMD) and its 95% confidence interval (CI) were used to assess the differences of the variables between the genotypes. A random effects model was used for all analyses because both between-study and within-study heterogeneity is considered in this model; it provides a more conservative evaluation of the significance of the association than the fixed effects model [53]. Heterogeneity between studies was tested by Cochran's χ 2 -based Q-statistic at a significance level of P < 0.05. Galbraith plot was used to detect the potential sources of heterogeneity, and the pooled SMD was recalculated after removal of the outlier studies identified in the plot. The populations in the studies were tested for Hardy-Weinberg equilibrium (HWE) by χ 2 test; the significance level is defined as α < 0.05. Publication bias was assayed by Egger's linear regression test [54], and a significance level of 0.05 was used to indicate the presence of potential publication bias.

Characteristics of the included studies
Initial search of the literature yielded 646 publications. Five hundred and eighteen studies were excluded according to title and abstract. Then full text articles were retrieved and assessed on the basis of the inclusion criteria. Eighty-six papers were ineligible for the following reasons: 57 papers did not provide complete data for this meta-analysis, 24 papers presented data on other polymorphisms, 2 papers had subjects overlap with other publications, and 3 studies were based on pedigree data. In the end, 42 studies were selected for this meta-analysis.

Summary statistics
Sixty-seven comparisons were distinguished according to the categories such as age, gender, smoking status, and health condition. Of these, 18, 64, 49, 42 and 54 comparisons were included for comparing the differences in APOC3, TG, TC, LDL-C and HDL-C, respectively (Table 2). Totally, 23846 subjects were enrolled in this meta-analysis, 74% of them (17592 subjects) have the CC genotype, and 26% of them (6254 subjects) have the GC or GG genotype. 6523, 23528, 15891, 15310 and 19527 subjects were included in comparing the differences in APOC3, TG, TC, LDL-C and HDL-C, respectively (Additional file 1: Table S1).

Discussion
A genome-wide association study has suggested that the polymorphisms in or near the APOC3 gene are among the strongest genetic determinants of plasma lipid concentrations [55]. A large body of literature has investigated the associations of the APOC3 rs5128 polymorphism with plasma APOC3 and/or lipid levels .
Associations of this polymorphism with increased levels of APOC3 [14,20,27,41,42], TG [18,20,23,25,28,30,32,33,38,41,42,44,46,51], TC [18,32,51] and LDL-C [30,51], and/or decreased levels of HDL-C [24] have been reported in some, but not all studies. The lack of consistency across these studies reflects some existed limitations such as small  In the present meta-analysis, the associations of the APOC3 rs5128 polymorphism with plasma APOC3 and lipid levels were investigated to examine these discrepancies. The frequencies of the GG homozygote were very low across the populations. Most of the included studies only provided data for CC genotype and G allele carriers (GC + GG), and did not provide data separately for heterozygote CG and homozygote GG. Therefore, a dominant model (CC vs. CG + GG) was employed for this meta-analysis to ensure adequate statistical power.
The results suggested that the APOC3 rs5128 polymorphism was significantly associated with fasting plasma levels of APOC3, TG, TC, and LDL-C under the dominant model. The carriers of G allele had higher levels of APOC3, TG, TC, and LDL-C than the non-carriers. A recent meta-analysis [10] demonstrated that the rs5128 polymorphism is associated with CVD risk. CVD is recognized as a multifactorial disease, and dyslipidemia accounts for at least 50% of the population-attributable risk [2]. Taken our results together, it is possible that the association between the rs5128 polymorphism and CVD is mediated   by the dyslipidemia caused by the G allele of the rs5128 polymorphism. Subgroup analyses by age, gender, ethnicity and health condition were performed since they might be important variables in determining associative risk with dyslipidemia.
For example, the present analyses indicated that age might modulate the association between the rs5128 polymorphism and TG levels since the significant association especially exists in males and females, but not in children ( Table 2). In addition, the significant effects of the rs5128     polymorphism on TC and LDL-C exist in females and children, but not in males; the significant effect of the polymorphism on LDL-C exists in healthy subjects and diabetic patients, but not in CVD patients. More studies should be conducted to further examine the association of this polymorphism with LDL-C in CVD patients. The associations of the rs5128 polymorphism with plasma levels of APOC3, TG, TC and LDL-C were very robust, which did not vary greatly when the analyses were performed only with the available studies in HWE. However, the significant association of the APOC3 rs5128 polymorphism with plasma HDL-C levels was not detected in this metaanalysis. Significant heterogeneity was found across the analyses for APOC3, TG, TC, and LDL-C. The main sources of heterogeneity were from ethnic origin, study design, gender     and health condition of the subjects, etc. Subgroup analyses stratified by the characteristics of the subjects were performed to explore the potential source of the observed heterogeneity, and significant heterogeneity was still observed in some subgroups. Galbraith plot was employed to further evaluate the sources of heterogeneity. Outlier studies were identified by using the plot, and the heterogeneity was effectively removed or decreased after exclusion of these outlier studies. No significant changes of the SMD value were found after excluding the outlier studies. The associations of the APOC3 rs5128 polymorphism with plasma APOC3, TG, TC and LDL-C were not likely to be type I errors (false-positive results). Firstly, the results from this meta-analysis were based on the random effects model. Comparing with fixed effects model, the random effects model is a more conservative method and less likely to produce false-positive results. Secondly, 42 studies with 23846 subjects were included in this meta-analysis. Among the subjects, 26% of them were the carriers of the rs5128 G allele. Since the incidence of the G allele carriers was sufficiently high, type I error may have been prevented.
The possible mechanism under which the rs5128 polymorphism modulates plasma APOC3 has not been clarified yet. One explanation could be that the G allele enhances the transcriptional activity of APOC3 and leads to a higher plasma APOC3 level since the rs5128 polymorphism is located in the 3'UTR of exon 4 of this gene. In the present analyses, a significant higher level of APOC3 was found in G carriers comparing with the CC genotype subjects. Previous studies have shown that APOC3 can increase plasma TG levels. Three mechanisms are involved in the elevation of TG levels by APOC3. Firstly, APOC3 promotes the assembly and secretion of VLDL in liver [56][57][58]. Overexpression of APOC3 in McA-RH7777 cells by recombinant adenovirus expression vector resulted in significantly increased VLDL assembly and secretion with a dose-dependent effect [56]. APOC3 gene knockout mice showed no increase in VLDL secretion after two weeks of high-fat diet [57]. Functional analysis found that two regions on the APOC3 polypeptide chain are closely related to VLDL synthesis and secretion [57,58]. Secondly, APOC3 inhibits LPL, which is located on the inner side of capillaries and is the main enzyme to hydrolyze TG-rich particles [59]. Thirdly, APOC3 inhibits hepatic lipase. Hepatic lipase is located on the endothelial side of liver sinusoids, and its main function is to remove the remnants of chylomicron and VLDL.
The present meta-analysis also suggested the significant associations of the rs5128 polymorphism with higher levels of LDL-C and TC. There is interrelationship between TG metabolism and cholesterol metabolism in the body, so the elevation of VLDL-TG by APOC3 might have disturbed the cholesterol metabolism and increased plasma LDL-C and TC levels. For example, LDL particles are formed in the bloodstream as VLDL particles lose TG through the action of LPL. Hence, the increase of VLDL Levels can cause the elevation of LDL levels, and accordingly the LDL-C levels. Since approximately 70% of plasma cholesterol molecules are resided in LDL particles, the elevation of LDL-C can lead to higher levels of TC. In addition, it is likely that the rs5128 polymorphism is in linkage disequilibrium with other causative mutations involved in the metabolism of LDL. APOC3 is a member of the APOA1/C3/A4/A5 gene cluster which is the key component in modulating lipid metabolism [60]. Several polymorphism sites in the APOA1/ C3/A4/A5 gene cluster have been found to significantly affect plasma LDL-C levels [61].
In the current study, we did not find significant association between the APOC3 rs5128 polymorphism and lower levels of HDL-C by meta-analysis. One reason could be that HDL is metabolized independently of APOC3 [4,5]. Unlike VLDL and LDL, nascent HDL particles are mainly formed in liver, matured in bloodstream and go back to liver with full load of cholesteryl ester molecules. In this process, HDL is mainly regulated by the proteins such as apolipoprotein A1 (APOA1), ATP-binding cassette transporter A1 (ABCA1) and cholesteryl ester transfer protein (CETP) [62]. Therefore, the APOC3 rs5128 polymorphism was found to be associated with the plasma levels of APOC3, TG, LDL-C and TC, but not HDL-C.
The limitations of the present meta-analysis should be noted. Firstly, dyslipidemia is involved in a number of genes as well as some environmental factors. However, the interactions of the rs5128 polymorphism with other polymorphic loci or environmental factors on plasma APOC3 and lipid levels have not been investigated in this analysis due to the lack of the original data of the included studies. In other words, the more precise results could have been gained if more detailed individual data were available or the stratification analyses based on the environmental factors such as diet, exercise, and smoking status were performed. Secondly, a relatively small number of subjects were included for the association analysis between the rs5128 polymorphism and plasma APOC3 levels due to the limited available studies, which may reduce the statistic power and even cause the type II errors (false-negative results). More studies with larger sample size are required to further investigate the association. Thirdly, this meta-analysis only included the studies published in English because it was very difficult to get the full papers published in various languages.

Conclusions
In conclusion, the significant associations between the APOC3 rs5128 polymorphism and higher levels of APOC3, TG, TC, and LDL-C were found in the present meta-analysis.

Additional files
Additional file 1: Table S1. Presenting the plasma levels of APOC3 and lipid variables by genotypes of individual studies included in the meta-analysis.