Association of rs2230806 in ABCA1 with coronary artery disease: An updated meta-analysis based on 43 research studies.

BACKGROUND: As a key gene in the reverse transport pathway of cholesterol, ABCA1 (ATP-binding cassette transporter A1) plays an important role in the pathogenesis of coronary artery disease (CAD). In the ABCA1, rs2230806 is the most widely studied polymorphism and its role has been controversial.
METHODS: We performed an updated meta-analysis by searching online electronic databases using the PubMed, Web of Science, Embase, Cochrane Library, EMBASE, Google Scholar, China National Knowledge Infrastructure, and Wan Fang databases before June 28, 2019. STATA12.0 software was used to perform a series of analyses on the data, including genetic effect model, heterogeneity, sensitivity, and publication bias analysis.
RESULTS: Based on our inclusion and exclusion criteria, finally 43 articles including a total of 34,348 subjects (14,085 CAD cases and 20,263 healthy controls) were investigated. Results showed that carrying the K allele in rs223086 in the overall population significantly reduced the risk of CAD (OR = 0.745, 95% CI = 0.687-0.809, P < .001). After the ethnicity stratification analysis, the above phenomenon was found to be significant in Asian populations (OR = 0.686, 95% CI = 0.633-0.744, P < .001), marginally significant in Caucasians (OR = 0.887, 95% CI = 0.786-1.001, P = .051), and not significant in other populations (OR = 0.851, 95% CI = 0.558-1.297, P = .452). Further stratified according to the sample size in the Asian and Caucasian populations, in the Asian the K allele is more protective in small samples than large samples; however, in the Caucasian small samples carrying the K allele play a protective role while large samples are negative. In addition, according to the source of the control population and the geographical location in China, the results showed that rs2230806 was significantly associated with CAD in any group. Five genetic models (allelic, recessive, dominant, homozygote, and heterozygote) were analyzed in the above analysis.
CONCLUSION: The K allele of rs2230806 was significantly associated with decreased risk of CAD, especially in Asian populations and small sample Caucasians.

Introduction

Coronary artery disease (CAD) is currently being the main cause of death and disability in developed and developing countries.[ Epidemiological studies have shown that CAD is a multifactor and multigene regulated disease, mainly affected by genetic variation, environmental factors, blood lipid level imbalance, and the interaction between them. Atherosclerosis (AS) is generally regarded as the pathological basis of CAD. Studies have shown that cholesterol accumulation in the arterial wall may be the starting cause of AS, which leads to an imbalance between the lipoprotein influx and the cholesterol efflux. At present, the cholesterol efflux pathways are mainly found in the following: ABCA1 pathway, scavenger receptor type B1 pathway, ATP-binding cassette transporter G1 pathway, and less efficient pathways such as passive efflux. Macrophages and ABCA1-mediated reverse cholesterol transport (RCT) account for approximately 90% of cholesterol excretions.[

ABCA1, a conserved transmembrane-spanning protein, plays crucial roles in the efflux of cellular cholesterol to HDL. Moreover, ABCA1 affects cellular inflammatory cytokine secretion by modulating cholesterol content in the plasma membrane and within intracellular compartments. In humans, ABCA1 mutations can cause a severe HDL-deficiency syndrome characterized by cholesterol deposition in tissue macrophages and prevalent AS. Disrupting Abca1 in mice promotes accumulation of excessive cholesterol in macrophages, and physiological manipulation of ABCA1 expression affects AS. Transplantation of bone marrow from Abca1−/− mice into Ldlr−/− or apoE−/− recipients caused an increase in AS.[ Based on the above clinical and experimental studies, it is confirmed that ABCA1 indeed play an important role during the pathogenesis of CAD.

The ABCA1 gene spanning 149 kb genomic DNA is located on human chromosome 9q22-q31 consisting of 50 exons and 49 introns and belongs to the superfamily of ATP-binding cassette.[ It is abundantly expressed on the plasma membrane and Golgi complex.[ To date approximately 100 gene mutation sites have been reported in ABCA1 including coding regions and noncoding regions such as rs2230806, rs2422493, rs146292819, rs4149339, −191G/C, and so on. Among them, rs2230806 (R219K, 107620867C>T) is the most widely studied common missense polymorphism. However, the relationship between rs2230806 and CAD was not consistent in the reported researches. Based on these findings, we carried out an update meta-analysis combining the latest data in different countries on 14,085 CAD patients and 20,263 controls to investigate rs2230806 and its effect on CAD risk and further classified the combined populations according to ethnicity, geographical location in China, and the source of control population. To further investigate the sample size impact on the results, large and small samples were divided in the Asian and Caucasian.

Methods

Study selection

A systematic search was conducted on online electronic databases using the PubMed, Web of Science, Embase, Cochrane Library, EMBASE, Google Scholar, China National Knowledge Infrastructure, and Wan Fang databases (the last search update was June 28, 2019). The following search terms were used: “genetic polymorphism or single nucleotide polymorphism or SNP (single nucleotide polymorphism) or mutation or variants,” “coronary artery disease or coronary heart diseases or acute coronary syndrome or myocardial infarction or ischemic heart disease,” “ATP binding cassette transporter A1 or ABCA1” and “rs2230806 or R219K or G596A, or 107620867C>T.” As this article is a meta-analysis of the previous published studies, hence patients’ consent and approval of the ethics committee were not required.

Eligible studies in this meta-analysis had to meet the following inclusion criteria: evaluation of the association between rs2230806 and CAD; case-control design; studies focusing on humans; sufficient data for estimating an odds ratio (OR) with 95% confidence interval (CI); the genotype in the control group should be agreed with the Hardy-Weinberg equilibrium (HWE); published in Chinese or English. Exclusion criteria were as follows: comment, review, and editorial articles; duplication of previous publications; family-based studies of pedigrees.

Data extraction

Two investigators (QF and YFZ) independently extracted data of the eligible studies and the result was reviewed by a third author (FZ). For each of the publication, the extracted information included: first author's name, published year, country of study population, ethnicity, geographical location in China, the source of control, total number of cases and controls, genotype numbers of cases and controls, genotyping methods.

Quality assessment

The study quality was assessed using the Newcastle–Ottawa scale (NOS) for case-control studies. The study quality based on 8 items ranged from 0 to 9 points. If the score was higher than 6, then the study was considered high quality.

Statistical methods

All statistical analyses were performed with the STATA 12.0 (StataCorp, College Station, TX). The relationship between rs2230806 and CAD risk by calculating pooled OR and 95% CI. The following 5 genetic models were employed to calculate the OR and 95% CI of rs2230806 under the allele, dominant, recessive, homogenetic, and heterogenetic model. Genotype distribution of controls with HWE was assessed using a χ2test. Q-statistical test and I2 test were used to evaluate the heterogeneity. All outcome measures were determined using random-effects models based on DerSimonian and Laird method to obtain the frank heterogeneity.[ Subgroup analyses were performed to detect the source of heterogeneity based on ethnicity, geographical location in China, source of control population, and the sample size in different races. In addition, stratification analyses were also performed under 5 genetic models. Sensitivity analysis was calculated by removing each individual study to evaluate effect of the result. The significance of publication bias was checked separately in all and different groups by Egger and Begg tests, and P value less than.05 was considered as statistically significant.

Results

Characteristics of eligible studies

A total of 476 potentially eligible papers were initially identified based on keyword-related and manual search. After screening titles and abstract review, 20 studies were discarded for duplicates and 413 studies were further excluded for comment, review, editorial and so on. Finally, 43 articles including a total of 34,348 subjects (14,085 CAD cases and 20,263 healthy controls) were involved in our meta-analysis.[ All the eligible studies were ranged from 2001 to 2019. The detailed flow diagram was shown in Figure 1. The characteristics and genotype frequencies of eligible studies were summarized in Tables 1 and 2. Twenty-eight studies were on East Asian populations and 12 on Caucasians. Eleven studies were from the South and 13 were North according to geographical location in China. Thirty-one studies were hospital-based and 12 were population-based in the controls. In addition, according to the sample size, the Asian populations divided into large sample (≥500) and small sample (<500), similarly, large sample (≥1000) and small sample (<1000) were grouped in the Caucasians. Polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) assay was used in 33 studies, 4 studies used probe method, and 2 studies utilized DNA sequencing to detect the genotypes. HWE of genotype distribution among the controls was tested in all the included studies.

Figure 1

Flow diagram of the study selection process.

Table 1

Characteristics of all eligible studies included in this meta-analysis.

Table 2

Genotype frequencies in the studies included in the meta-analysis.

Meta-analysis results

Overall analyses

As shown in Table 3 , the pooled results indicate a statistically significant association between rs2230806 polymorphism and the risk of CAD under different genetic models: an allelic genetic model (OR=0.745, 95% CI = 0.687–0.809, P < .001), a recessive genetic model (OR = 0.683, 95% CI = 0.603–0.774, P < .001), a dominant genetic model (OR = 0.703, 95% CI = 0.633–0.781, P < .001), a homozygote genetic model (OR = 0.573, 95% CI = 0.488–0.672, P < .001), and a heterozygote genetic model (OR = 0.761, 95% CI = 0.693–0.837, P < .001). Forest plot under allelic model is demonstrated in Figure 2.

Table 3

Overall and stratified meta-analysis under different genetic models.

Figure 2

Meta-analysis with a random-effects model for the overall association between rs2230806 and CAD under the allele model.

Subgroup analyses

In the subgroup analysis by ethnicity, for the Asian ethnicity, a significant association between rs2230806 and CAD risk was found in all 5 genetic model: an allelic genetic model (OR = 0.686, 95% CI = 0.633–0.744, P < .001), a recessive genetic model (OR = 0.621, 95% CI = 0.561–0.688, P < .001), a dominant genetic model (OR = 0.645, 95% CI = 0.595–0.700, P < .001), a homozygote genetic model (OR = 0.505, 95% CI = 0.450–0.566, P < .001), and a heterozygote genetic model (OR = 0.707, 95% CI = 0.649–0.770, P < .001). Stratified according to the sample size, we found that the association in the large and small sample size was also significant (Table 3 ). For Caucasian populations, the association did not return significant in all 5 genetic model: an allelic genetic model (OR = 0.887, 95% CI = 0.786–1.001, P = .051), a recessive genetic model (OR = 0.915, 95% CI = 0.804–1.041, P = .177), a dominant genetic model (OR = 0.857, 95% CI = 0.730–1.007, P = .060), a homozygote genetic model (OR = 0.903, 95% CI = 0.791–1.031, P = .132), and a heterozygote genetic model (OR = 0.870, 95% CI = 0.741–1.021, P = .088). Similarly, stratified by the sample size, a statistically significant association was observed in the small sample under 4 genetic models except the recessive model, but not in the subjects from the large sample under 5 genetic models (Table 3 ).

We also performed stratified analyses by geographical location in China and source of the control population. The results suggested the polymorphism was associated with decreased CAD risk in any group under 5 genetic models (Table 3 ).

Heterogeneity analysis

Strong evidence of heterogeneity among the whole population was demonstrated in all 5 models (R vs K: I2 = 76.3%, Pheterogeneity < .001; KK vs RK + RR: I2 = 55.8%, Pheterogeneity < .001; KK + RK vs RR: I2 = 71.9%, Pheterogeneity < .001; KK vs RR: I2 = 68.2%, Pheterogeneity < .001; RK vs RR: I2 = 60.2%, Pheterogeneity < .001). Next, we detected the cause of heterogeneity in any of genetic models by ethnicity, geographical location, source of control population, and the sample size in Asian and Caucasian. Statistical analysis showed the main contributors of heterogeneity resulted from Caucasians, after further subgroup analysis, the heterogeneity was significantly decreased in the small population (R vs K: I2 = 29.5%, Pheterogeneity = .203) whereas significant heterogeneity was observed in the large population (R vs K: I2 = 83.4%, Pheterogeneity < .001).

Sensitivity analysis

To better examine the influence on the pooled OR, we performed a sensitivity analysis omitting any individual study using STATA 12.0 software. The result suggested no significant difference was found, suggesting our results were relatively stable and credible (Table 4).

Table 3 (Continued)

Overall and stratified meta-analysis under different genetic models.

Publication bias

As outlined in Figure 3, the shape of funnel plots in overall population between rs2230806 and CAD under the allelic model indicated publication bias in the current meta-analysis. Considering ethnicity heterogeneity, we make a statistics analysis for different races. The result showed publication bias was not found in the Asian group, with similar in Caucasian populations. The Egger test (t = 0.590, P = .563 in Asian; t = −0.070, P = .944 in Caucasian) and Begg test (t = 0.560, P = .574 in Asian; z = 1.58, P = .115 in Caucasian) were also performed to investigate the symmetry of the funnel plot.

Figure 3

Forest plot for publication bias test between rs2230806 and CAD in the allelic model. A, Overall population. B, Asian population. C, Caucasian population.

Discussion

CAD is a serious threat to human health with the acceleration of social development and the prevalence of unhealthy lifestyles. A large number of epidemiological investigations and clinical trials have shown lipid accumulation and vascular wall inflammation may be 2 key factors in the pathogenesis. Although decreasing LDL-C levels is an important therapeutic intervention for reducing cardiovascular events, however, it has been observed in the clinic that the level of LDL-C in some patients is in a suitable range or reduced to less than 70 mg/dL by lipid-lowering drugs, atherosclerotic lesions remain in progress. Thus, there are other lipid indicators involved in AS such as HDL-C. The antiatherosclerosis effect of HDL-C can inhibit the development of AS and serum HDL-C level is significantly negatively correlated with the incidence of CAD because of the important role of HDL-C in cholesterol metabolism, especially RCT. More and more genes have been discovered in the HDL lipid metabolism, such as HMGA, TNNT1, and so on. Some of them may have similar regulatory mechanisms. For example, ABCA1 and HMGA are both targeted by miR-33.[

Animals and human studies document that functional effects in the ABCA1 pathway are important determinants of CAD.[ ABCA1 expression and ABCA1-dependent efflux of cell cholesterol are closely associated with increased surface binding of apoA-1.[ Abca1-deficient mice increased circulating levels of chemokines, cytokines, and growth factors, which are most evident after the injection of LPS.[ Genetic mutations in the ABCA1 cause Tangier disease and familial hypoalphalipoproteinemia.[

Plenty of evidence demonstrated that candidate gene case-control association studies were frequently used method for identifying the susceptibility genes for CAD.[ The polymorphism of ABCA1 gene becomes a research hotspot in the genetic mechanism of CAD. Currently, several meta-analyses evaluating the association between ABCA1 polymorphism and CAD have been published.[ Besides, a large number of epidemiological studies have also been conducted. For instance, the study of Clee et al[ reported that the frequent R219K variant was associated with a decreased severity of atherosclerosis, decreased risk of coronary events, decreased TG, and increased HDL-C, suggesting the variant is associated with a gain of normal ABCA1 function and increased RCT. In another study, Evanset et al[ also suggested that the R219K polymorphism either directly or as a result of linkage disequilibrium with additional functional variant, had a protective effect against CAD and was associated with lower plasma triglycerides in subgroups of patients with hyperlipidaemia. While in the study of Brousseau et al[ we found that the mutant allele might cause a decrease in the HDL level and promote the development of CAD. Therefore, the possible role of rs2230806 in CAD is still debatable or even the opposite.

Based on combined analyses of 34,348 participants, our study demonstrated that the K allele of rs2230806 was significantly associated with decreased risk of CAD, especially under the recessive and homogenetic models. In addition, statistical result showed rs2230806 is significantly associated with CAD in Asian population, marginally significant in Caucasian and not significant in other group. The reason for racial difference phenomenon may be attributed to allele frequency and other factors, such as lifestyle discrepancy and so on.

Due to significant heterogeneity under five genetic models in the whole population, subgroup analysis was needed to explore the source of heterogeneity. After subgroup analysis by ethnicity, geographical location, the source of control population, and the sample size in the Asian and Caucasian, we observed that the protective effect of allele from small sample subgroups in Asian population was greater than that of large sample subgroups under all 5 genetic models, the same in Caucasian population. One possible explanation is small sample effect.[ We also found the main contributors of heterogeneity resulted from the large population in Caucasians. In addition, the statistical result mild alters after removing the large sample population of Caucasians 1 by 1 based on sensitivity analysis. Moreover, the heterogeneity from North China is greater than South China possibly due to genetic regulation in different regions and diverse living habits. After the source of control population, significant heterogeneity was observed in the hospital-based subgroup because such control subjects from a special group might not represent the general population.

Our finding is slightly different from the previous meta-analysis by Ma et al, Jiang et al, Yin et al.[ Current meta-analysis studies showed the role of rs2230806 in Asian populations is consistent with previous studies, and there is controversy in the Caucasian population. Ma et al indicated rs2230806 is a protective role for CAD risk both in Asians (OR = 0.76, 95% CI = 0.68–0.85) and Caucasians (OR = 0.89, 95% CI = 0.81–0.99). Jiang et al showed the polymorphism was significantly associated with decreased risk of CAD in Asian population (OR = 0.70, 95% CI = 0.61–0.81) not Caucasians (OR = 0.91, 95% CI = 0.80–1.04). Similarly, Yin et al also found the same result in Asian population (OR = 0.66, 95% CI = 0.59–0.74) and Caucasians (OR = 0.90, 95% CI = 0.76–1.07). Our meta-analysis included the largest sample size, and the result was consistent with Jiang et al and Yin et al, but we further stratified the Caucasian population according to the sample size, the association analysis, and heterogeneity results were significantly different. Statistics analysis found that the above results in the Caucasian population P value swinged around.05, the size of the sample may play a key role. Then, considering that HWE may have an impact on heterogeneity and all of our included samples meet the HWE. In addition, due to the ethnic differences in publication bias, we examine the bias in Asian and Caucasian population separately.

However, several drawbacks existed in our meta-analysis. First, CAD is a complex disease affected by many factors such as multiple microeffect genes, linkages between gene loci and gene–environment interactions. The positive association between the polymorphism of a certain gene and the disease only plays a weak role in the development of CAD. Second, disease subtypes or clinical pathological features are not considered. Third, all eligible studies were published in English or Chinese, which might cause potential language bias. Last, heterogeneity may have an impact on our analysis.

Conclusion

In conclusion, the K allele of rs2230806 was significantly associated with decreased risk of CAD, especially in Asian populations and small sample Caucasians. However, in the light of the limitations in our meta-analysis, further large-scale multicenter study is needed to evaluate gene–gene and gene–environment interaction.

Author contributions

Data curation: Qian Fan, Yanfang Zhu.

Formal analysis: Qian Fan.

Methodology: Qian Fan, Yanfang Zhu, Fang Zhao.

Software: Qian Fan, Yanfang Zhu.

Validation: Fang Zhao.

Conceptualization: Qian Fan, Fang Zhao.

Funding acquisition: Qian Fan.

Project administration: Qian Fan, Fang Zhao.

Writing – original draft: Qian Fan.

Writing – review & editing: Qian Fan, Fang Zhao.

Table 4

Results of sensitivity analysis of rs2230806 in allele model.