ε2 allele and ε2-involved genotypes (ε2/ε2, ε2/ε3, and ε2/ε4) may confer the association of APOE genetic polymorphism with risks of nephropathy in type 2 diabetes: a meta-analysis

Background Diabetic nephropathy (DN) contributes to end-stage renal failure. Microvascular injury resulted from reactive oxygen species is implicated in the pathogenesis of DN. Genetic polymorphism of Apolipoprotein E (APOE) influences the antioxidative properties of the protein. The relationship of APOE polymorphism with the risks of nephropathy in type 2 diabetes (T2DN) remains elusive. Methods An up-to-date meta-analysis was conducted on the basis of studies selected from PubMed, WanFang database, Embase, Vip database, Web of Science, Scopus, and CNKI database. Results A total of 33 studies conferring 3266 cases and 3259 controls were selected on the basis of criteria of inclusion and exclusion in this meta-analysis. For APOE alleles, the pooled odds ratio (OR) of ε2 vs. ε3 was 1.89 (95% confidence intervals [95% CI]: 1.49–2.38, P < 0.0001). With regard to APOE genotypes, ε2/ε2, ε2/ε3, and ε2/ε4 increased the risk of T2DN (ε2/ε2 vs. ε3/ε3: OR = 2.32, 95% CI: 1.52–3.56, P = 0.0001; ε2/ε3 vs. ε3/ε3: OR = 1.97, 95% CI: 1.50–2.59, P<0.0001; ε2/ε4 vs. ε3/ε3: OR = 1.69, 95% CI: 1.18–2.44, P = 0.0046). Conclusions This meta-analysis found that the APOE ε2 allele and the ε2-involved genotypes (ε2/ε2, ε2/ε3, and ε2/ε4) are the risk factors of T2DN.


Background
Diabetic nephropathy (DN) contributes to end-stage renal failure [1]. Microvascular injury resulted from reactive oxygen species is implicated in the pathogenesis of DN [2,3]. Elucidating risk factors of DN, such as genetic and environmental factors, is needed for controlling this disease.
Genetic factors complicated in DN etiology confer useful insights into the etiology of the disease [4]. Oxidative stress is also involved in the complex web of pathological events that confer susceptibility to DN [5,6]. Excessive generation of reactive oxygen species (ROS) gives rise to imbalanced redox signaling, resulting in renal injury on the long term; moreover, oxidative stress is also linked to changes in the structure and function of apolipoprotein E (APOE), as its coding gene is implicated in DN pathology [7,8]. Two single nucleotide polymorphisms (SNPs) (rs7412 and rs429358) existing on exon 4 of APOE gene combine to generate three major alleles: ε3 is characterized by cytosines in both positions, while substitution rs7412C > T defines ε2 and rs429358C > T determines ε4. The two SNPs confer APOE3 with arginine at residue 158 and cysteine on residue 112, APOE2 carrying cysteine on both positions, and APOE4 carrying arginine on both positions. Moreover, combinations of these alleles generate six APOE haplotypes (ε2/ε2, ε2/ε3, ε2/ε4, ε3/ε3, ε3/ε4, and ε4/ε4). Allele variation in ApoE locus accounts for 0-20% of ε2, 60-90% of ε3, and 10-20% of ε4, respectively [9]. Allele ε3 is accepted as "wild-type" as it is the most common, and ε2 and ε4 are variants. The association between the two SNPs and T2DN risk is conflicting. Lin et al. found that ε2 polymorphism increased the susceptibility to T2DN in Asian population [10]. ε2 carriers and ε3/ε4 genotype carriers had increasing risks of developing T2DN [11]. However, the differences in sample sizes, sample sources, disease status, genotyping method, and other uncontrolled factors generate the above disagreeing results.
Meta-analysis, featured in summarizing results quantitatively from a wide range of studies, is a powerful   [12]. Therefore, an up-to-date meta-analysis was performed to further investigate the association by including these new published articles.

Articles search
The meta-analysis was conducted by searching the relative articles published before July 31, 2019 from PubMed, WanFang database, Embase, Vip database, Web of Science, Scopus, and CNKI database. The combinations of keywords were used for searching PubMed, Embase, Web of Science, Scopus were (["APOE" OR "Apolipoprotein E"] AND ["Diabetic nephropathy"]). Furthermore, the equivalent Chinese keywords were utilized for searching the Chinese databases.

Inclusion/exclusion criteria
The articles selected in the meta-analysis were based on inclusion criteria (case-control design; type 2 DM with DN; and association of APOE with DN risks) and the exclusion criteria (case reports or reviews; duplicate reports; type 1 DM; and missing data of allele or genotype frequencies).

Data extraction and quality assessment
The information from the included articles was extracted, such as the last name of first author and data of APOE allele or genotype. According to the Newcastle-Ottawa scale (NOS), the quality of the included articles was evaluated. If an included article met a condition, a score of one point was allocated; otherwise, no point (0 score) was allocated. Each of the included articles was awarded the sum of all points (total Quality Score) [13]. Moreover, the quality of these articles was evaluated by the two investigators (Zhaorui Cheng and Jikang Shi) independently. If an agreement for an included article was not reached by the two investigators, the third investigator (Shuang Qiu) settled inconformity finally. Low-quality articles were also selected to avoid selection bias.

Statistical analysis
Chi-square test of goodness of fit was used for evaluating Hardy-Weinberg equilibrium (HWE) for each Fig. 2 Forest plot for association between nephropathy in type 2 diabetes risk and ApoE ε2 allele vs. ε3 allele based on a random-effects model included article among control groups, and HWE was rejected when P < 0.05. The strength of association between APOE polymorphisms and T2DN risks was assessed using Odds ratios (OR) and 95% confidence intervals (95% CI) owing to binary outcome variable. Both Chi-square test-based Q-statistic and quantified by I 2 -statistic were adopted to evaluate heterogeneity. Because genotype can represent the combined effect of alleles, the comparisons of APOE genotypes were performed. For heterogeneity between studies given by I squared > 50%, random-effect models were applied; otherwise, if I squared < 50%, fixed-effect models were used [14]. Subgroup analyses were conducted to find main heterogeneity sources. Metaregression was carried out to further reveal heterogeneity sources and the contribution to heterogeneity. Sensitivity analysis was conducted to evaluate the stability of overall results. Publication bias was examined by funnel plots, and quantified using the Begg's and Egger's tests: P < 0.05 was considered significant publication bias [15]. Bonferroni correction was carried out in multiple comparison; thus, P < 0.025 was considered as statistically significant. R Studio (Version 1.1.383) (RStudio, Inc., MA, USA) for Windows was used for all data management and analyses.

Trial sequential analysis (TSA)
Dispersed data and repeated significance testing give rise to an increased risk of random error in traditional meta-analysis. TSA adjusts threshold for statistical significance, reducing the risk of type I error by required information size (RIS). In addition, TSA is used to estimate statistical reliability. In the meta-analysis, TSA software (TSA, version 0.9.5.5; Copenhagen Trial Unit, Copenhagen, Denmark, 2016) was used. The overall type I error was set at 5%, the statistical power was 80%, and the relative risk was reduced by 20% [16]. When the Z-curve crossed trial sequential monitoring boundary or RIS was reached, additional studies were not required; otherwise, additional studies were required.

Characteristics of included articles
A total of 33 eligible articles were eventually chosen, after abstracts and full texts of 837 published articles originally collected were scrutinized according to the  Table 1).

Subgroup analysis
For APOE alleles, when ε2 was compared with ε3, the association of increased T2DN risk was significant in Chinese population (OR = 2.04, 95% CI: 1.58-2.62); however, when ε4 was compared with ε3, the protective association of T2DN risk was significant in other population (OR = 0.68, 95% CI: 0.51-0.91) (  Forest plot for association between nephropathy in type 2 diabetes risk and ApoE genotype ε3/ε4 vs. ε3/ε3 genotype based on a random-effects model Whereas, ε3/ε4 genotype decreased T2DN risk in other population (ε3/ε4 vs. ε3/ε3: OR = 0.61, 95% CI: 0.44-0.84), but ε4/ε4 genotype were not associated with T2DN risk in neither of the populations ( Table 2). The source of heterogeneity was not found using meta-regression analysis, although each factor decreased overall heterogeneity.

Sensitivity analysis and publication bias
Results of sensitivity analysis in this meta-analysis revealed that there was no individual article influencing the corresponding pooled ORs and 95% CIs (Table 3 and Table 4), indicating that results of this meta-analysis are robust.
Beggʼs funnel plot and Eggerʼs test identified that significant publication bias was not found between either allele and either genotype and T2DN risk (all P>0.05).

Trial sequential analysis
With regard to the relationship of ε2 with T2DN risks and for the relationship of the genotypes (ε2/ε2, ε2/ε3, and ε2/ε4) with T2DN risks, the sample size reached RIS, and the Z-curve crossed the trial sequential monitoring boundary (Supplementary Figure S2). For the relationship of the ε4/ε4 genotype with T2DN risks, the sample size reached RIS (Supplementary Figure S3). For the relationship of ε4 with T2DN risks and for the relationship of the ε3/ε4 genotype with T2DN risks, the sample size and Z curve were not up to the requirements (Supplementary Figure S3).

Discussion
This meta-analysis further investigated the association between the APOE polymorphism and T2DN risks using up-to-date data, indicating that ε2 allele may increase T2DN risks; moreover, ε2/ε2, ε2/ε3, and ε2/ε4 genotypes increase T2DN risks. The ε2 allele and the ε2-involved genotypes may confer the association of APOE polymorphism with T2DN risk.
Meta-analyses between ε2/ε3/ε4 of APOE and DN risks have been performed to recognize the function of variants in APOE. In 2011, Li et al. found that ε2 increases T2DN risk in patients with diabetes [50]. In  [10]. In 2015, Li et al. validated that ε2 may act as promotion factors of nephropathy in type 2 diabetes, but ε4 is not associated with T2DN risk [12]. This metaanalysis further corroborated that the ε2 allele and the ε2-involved genotypes may confer the association of APOE genetic polymorphism with T2DN risk. Additionally, the association of ε2 with increased T2DN risks was further identified in Chinese population, and ε4 and ε3/ ε4 genotype were associated with decreased T2DN risks in other population.
Heterogeneity affects interpretations of results [51]. Although the source was not pinpointed, each separate factor did decrease the overall heterogeneity. Sensitivity analyses and TSA were further performed to assess the robustness of the deductions, reflecting a reliable conclusion.
Oxidative stress affects APOE via amino acid residues 112 and 158, suggesting that oxidative stress may be a source of heterogeneity [52]. Reduced glutathione provides major antioxidative activity; however, glutathione levels were remarkably reduced in patients with DN compared with those in patients with diabetes and healthy controls [53]. The meta-analysis documented the relationship of ε2 allele and the genotypes (ε2/ε2, ε2/ε3, and ε2/ε4) with T2DN risk, suggesting that APOE2 in patients with T2DN cannot balance oxidative stress involved in T2DN progress, and oxidative stress may generate heterogeneity in patients with T2DN. APOE is interfered by oxidative stress in structure and function. APOE contains two domains (the lowdensity-lipoprotein receptor [LDLR] binding region [residues 136-150] and the principal lipoproteinbinding region [residues 244-272]), highlighting the implication of the LDLR-binding region of APOE in DN progress. The affinity of APOE3 to LDLR is similar to that of APOE4; however, the binding ability of APOE2 is significantly lower [54]. Moreover, the cysteine-to-arginine substitution in APOE2 at position 158 affects LDLR-binding activity by forming of a Fig. 6 Forest plot for association between nephropathy in type 2 diabetes risk and ApoE genotype ε2/ε2 vs. ε3/ε3 genotype based on a random-effects model new salt bridge between Arg150 and Asp154, further affecting the interaction between APOE2 and LDLR [55]. Thus, oxidative stress interferes the structure and function of APOE by dysregulating the affinity of APOE to LDLR possibly, and the dysregulation of LDLR correlates with DN risk directly [56]. Furthermore, renal lipid accumulation is observed in human DN [57], and knockout of ApoE increases foam cellrich soft plaques and aggressive renal dysfunction in mice substantially [58].

Study strengths and limitations
There are some strengths in this study. First, the up-todate articles were collected extensively, rendering this study more statistical power to draw valid conclusion on this issue. Second, TSA was the first utilized to evaluate the association of APOE genetic polymorphism with T2DN risk, conferring reliable evidence to reach the conclusion.
Some limitations exist in this study. First, the main source of heterogeneity was not identified, although subgroup analysis and regression analysis were conducted, and further studies based on larger sample size and multiple ethnicity and region are required. Moreover, the other factors, which could contribute to heterogeneity, are not retrieved. Second, data of oxidative stress status, which possibly reflects renal injury more directly than APOE genetic polymorphism, are not available in literatures. Third, the casecontrol design could prove an association, rather than a causal relationship, thereby needing prospective cohort studies in future.