Lack of association between miR-146a rs2910164 C/G locus and colorectal cancer: from a case–control study to a meta-analysis

Abstract Previous studies suggested that miR-146a rs2910164 (C/G) locus was predicted to influence the risk of cancer. However, the relationship of miR-146a rs2910164 locus with colorectal cancer (CRC) susceptibility was controversial. We recruited 1003 CRC patients and 1303 controls, and performed a case–control study to clarify the correlation of miR-146a rs2910164 locus with CRC risk. Subsequently, a comprehensive meta-analysis was conducted to verify our findings. In the case–control study, we suggested that miR-146a rs2910164 variants did not alter CRC risk (CG vs. CC: adjusted P=0.465; GG vs. CC: adjusted P=0.436, CG/GG vs. CC: adjusted P=0.387 and GG vs. CC/CG: adjusted P=0.589), even in subgroup analysis. Next, we conducted a pooled-analysis to identify the correlation of miR-146a rs2910164 locus with CRC risk. In this pooled-analysis, 7947 CRC cases and 12,168 controls were included. We found that miR-146a rs2910164 polymorphism did not influence the risk of CRC (G vs. C: P=0.537; GG vs. CC: P=0.517, CG/GG vs. CC: P=0.520 and GG vs. CC/CG: P=0.167). Our findings suggest that miR-146a rs2910164 C/G polymorphism is not correlated with the susceptibility of CRC. In the future, more case–control studies are needed to confirm our results.


Introduction
In 2018, it was reported that over 1.8 million new colorectal cancer (CRC) cases were diagnosed and 881,000 CRC-related deaths occurred worldwide [1]. In China, CRC ranked both fifth in terms of malignancy incidence and mortality [2]. Trend of cancer incidence indicated that a significant upward trend was found for CRC in developing countries [3]. Previous studies attributed CRC to aging, unhealthy lifestyle and environmental factors [4][5][6][7]. It should be very important to understand susceptibility factors of CRC as they could be beneficial for the evaluation of prevention strategies and susceptibility. Accumulating evidences have shown that an individual's inherited factors also contribute to the development of CRC.
MicroRNA (miR-) involves ∼22 nucleotides. Recently, miRs have been widely investigated and accepted to be implicated in a number of human diseases, especially in malignancy. Adami et al. indicated that miR-146a was overexpressed in primary gastric tumor tissues without the status of chronic Helicobacter pylori infection and expressed a related lower level in progressed gastric cancer [8]. In addition, a recent study reported that miR-146a expression level decreased in progressed CRC patients with lymph node metastasis [9]. Thus, miR-146a was considered to be implicated in the development and progression of cancer [10,11].
In the past decade, several studies have focused on the correlation of miR-146a rs2910164 C/G with various malignancy susceptibility. MiR-146a rs2910164 C/G locus was suggested to be relevant to an increase susceptibility of CRC in several studies [12,13]. However, two publications suggested that miR-146a rs2910164 G allele had a lower susceptibility of CRC compared to those with C allele [14,15]. Notably, several pooled-analyses were performed regarding the relationship of CRC risk with this polymorphism [16][17][18][19][20][21][22][23]. However, the previous observations were diversity or even conflicting. Rong et al. and Liu et al. conducted meta-analyses and found that miR-146a rs2910164 polymorphism may confer risk to CRC [17,19]. Other meta-analyses reported that miR-146a rs2910164 was not related to CRC [16,18,[20][21][22][23]. The controversial findings might be due to the limited sample sizes. The aim of the present study was to explore the potential correlation of miR-146a rs2910164 polymorphism with CRC risk more extensively. We first conducted a case-control study focusing on the correlation of miR-146a rs2910164 with the risk to CRC. Subsequently, we carried out a meta-analysis to further determine the relationship between miR-146a rs2910164 locus and CRC risk.

Case-control study
We recruited 1003 sporadic CRC patients between October 2014 and August 2017. They were diagnosed via pathology in Fujian and Jiangsu provinces (China). The mean age of CRC patients was 61.10 + − 12.17 years. In addition, 1303 healthy subjects with a mean age of 61.40 + − 9.61 years were included as controls. The CRC cases and controls were matched by sex and age. The detailed information have been presented in our previous study [24][25][26][27]. After giving a written consent in the present study, all participants were questioned the information of risk factor and individual history. The Ethics Committee of Jiangsu University approved protocol of the present study.
Ethylenediamine tetraacetic acid-anticoagulated blood sample was stored in a −80 • C refrigerator. Genomic DNA was isolated from lymphocytes by using a DNA Kit (Promega, Madison, U.S.A.). MiR-146a rs2910164 variants were analyzed by a SNPscan Kit. To confirm the obtained results, 4% DNA samples were randomly selected and reciprocally tested by another person. The reproducibility of genotyping was 100%.
Genotype frequency of miR-146a rs2910164 variants and risk factors were compared by using χ 2 test. Crude and adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated to predict the association of the miR-146a rs2910164 variants with risk to CRC. A P<0.05 (two-tailed) was used to determine the significance. Hardy-Weinberg equilibrium (HWE) was used to confirm whether the distribution of miR-146a rs2910164 variants accorded with Mendelian inheritance patterns by a Chi-square software. All the comparisons were conducted using SAS 9.4 version software (SAS Institute, Cary, U.S.A.).

Meta-analysis
To verify the correlation between miR-146a rs2910164 locus and the risk of CRC, we subsequently carried out a systematic review and meta-analysis. All studies focusing on the association of miR-146a rs2910164 polymorphism with CRC were identified by a widely internet-based search of China Biology Medicine (CBM), PubMed and Embase databases (published up to March 31, 2019) with the terms of: (miR-146a or hsa-mir-146a or miRNA-146a or MicroRNA-146a or rs2910164) and (colorectal or colon or rectum) and (polymorphism or SNP or variant) and (cancer or carcinoma). Other potential papers were obtained by searching of the references of in eligible publications.
The following included criteria were used: (a) studies based on the association between miR-146a rs2910164 and CRC risk; (b) genotype data were presented; and (c) the genotype distribution in controls consistent with HWE. Two authors (S. Zhang and J. Jiang) extracted the following information independently: the surname of first author, year of publication, country, ethnicity of participant, the region of CRC, method of genotyping and frequencies of miR-146a rs2910164 variants in cases and controls.
Crude ORs with their 95% CIs were harnessed to evaluate the strength of correlation between miR-146a rs2910164 and CRC risk. Chi-square-based Cochran's Q-test and I 2 test were carried out to assess the heterogeneity among the eligible studies, which were defined as statistical significance if I 2 >50% or P<0.1. The pooled ORs were calculated by a random-effect model (DerSimonian and Laird method), if we found significant heterogeneity [28,29]. Otherwise, the fixed effects model (the Mantel-Haenszel method) was used [30]. Sensitivity analysis was conducted by one-way method, which removed an individual study in turn to address the stability of our findings under all genetic models. In addition, Begg's test and Egger's test were used to assess the bias of publication. A P<0.1 was defined as the level for significance of publication bias. The Newcastle-Ottawa Scale (NOS) was used to assess the quality of the included studies. Galbraith radial plot was harnessed to identify the source of heterogeneity. All statistical tests were conducted by using the STATA software (Version 12.0; Stata Corporation, College Station, Texas). A P values (two-sided) less than 0.05 was defined as statistically significant association between miR-146a rs2910164 and CRC risk. By using Power and Sample Size Calculator software [31], power value (α = 0.05) of this meta-analysis was calculated.

Case-control study
This case-control study consisted of participants enrolled in a study conducted in Jiangsu and Fujian provinces (China) from October 2014 to August 2017. The detailed information of the present study was described in our previous studies [24,25]. Briefly, patients with 431 colon cancer and 572 rectum cancer were recruited. CRC cases were unrelated and confirmed histologically. Non-cancer subjects were included as controls from the same populations with age and sex matched. And all subjects were unrelated Chinese Han populations. Table 1 lists demographic information and risk factors that were collected in the present study. Table 2 summarizes the corresponding information of miR-146a rs2910164. The distribution of miR-146a rs2910164 variants in control group is consistent with HWE (P=0.706). Data of genotypes and environmental factors are summarized in Supplementary Table S1. Table 3 shows the genotype distribution of miR-146a rs2910164 in two groups. MiR-146a rs2910164 variant frequencies were 37.55% (CC), 47.96% (CG) and 14.49% (GG) in CRC patients and 39.69% (CC), 46.23% (CG) and 14.08% (GG) in controls. In the studied populations, minor allele frequency (MAF) of miR-146a rs2910164 C/G polymorphism was 0.372, which was similar to SNP database for Chinese populations (MAF = 0.430, Table 2). We found that there were no significant association between miR-146a rs2910164 polymorphism and susceptibility of
We found that there was no significant association between miR-146a rs2910164 polymorphism and the risk of CRC (G vs. C: P=0.537; GG vs. CC: P=0.517, CG/GG vs. CC: P=0.520 and GG vs. CC/CG: P=0.167; Table 6 and Figure 2). Additionally, subgroup analyses were performed by the ethnicity and region of CRC. Null association between miR-146a rs2910164 and the susceptibility of CRC was also found (  Figure 3).
We used one-way method to evaluate the sensitivity of this analysis. The results suggested that any individual study could not materially alter the association between miR-146a rs2910164 and CRC risk (Figure 4). Since significant heterogeneity among the studies was found, we conducted subgroup analyses and Galbraith radial plot to identify the source of heterogeneity in a dominant model. Because ethnicity and type of CRC could affect the observations of meta-analysis, subgroup analyses were carried out (Table 6). We found that rectal cancer and Asians might lead to major source of heterogeneity. In Galbraith radial plot, we could find five outliers [12,13,[35][36][37], which contributed significant heterogeneity to this analysis ( Figure 5). NOS scores of the included studies were summarized in Table 5. According to the criterion mentioned in previous study [42,43], we found that only two investigations were poor quality studies [36,44]. When we omitted these poor quality investigations, the findings were not changed.
For this pooled-analysis, the power value (α = 0.05) was 0.258 in G vs. C, 0.235 in GG vs. CC, 0.275 in GG/CG vs. CC and 0.579 in GG vs. CC/CG genetic model.

Discussion
In CRC patients, miR-146a expression level decreased in progressed tumors [9]. Since miR-146a plays important roles in the development of CRC, we presumed that miR-146a rs2910164 variants could influence the susceptibility to CRC. Here, we first performed a case-control study and subsequently conducted a meta-analysis on the potential correlation between miR-146a rs2910164 polymorphism and the susceptibility of CRC. In the case-control study, we recruited 1003 sporadic CRC patients and 1303 controls. In meta-analysis, the present study included 21 independent studies with 20,115 subjects (7947 CRC patients and 12,168 controls). Nevertheless, null correlation has been identified between miR-146a rs2910164 locus and the risk of CRC, even in different subgroup.
To date, some studies have been carried out to assess this potential correlation between miR-146a rs2910164 polymorphism and CRC susceptibility in different populations. However, conflicting findings were found. Consequently, we designed a case-control study matching with age and sex. We got a null correlation between this SNP and the risk of CRC. These findings were similar to some previous publications, which the genotype frequency of the miR-146a rs2910164 G allele had no significant difference in CRC cases versus non-cancer controls [35,36,39,40]. It was worth mentioning that the paper was published by Ying et al. [40], where the investigation included more than 1000 CRC patients and 1000 controls. However, other publications, identified a significant associations or even suggested opposite findings. Dikaiakos et al. [15] and Chae et al. [14] reported a decreased risk of the miR-146a rs2910164 G allele to CRC. Recently, two publications found an association of the increased risk with miR-146a rs2910164 G allele [12,13]. It could be interpreted that these investigations were carried out in different ethnicity, region, age and with variable sample size. A recent meta-analysis has found an association for the European populations in the recessive model   [45]. Compared with the meta-analysis mentioned above, the current meta-analysis has included more case-control studies with larger sample sizes. Thus, our findings might be more credible.
In the large-scale pooled-analysis, we found that miR-146a rs2910164 might not confer the risk to CRC. However, the interaction between genes and environment factors might be implicated in the occurrence of CRC. Reviewing early publications, some reasons were identified for the conflicting findings. First, different ethnicity may lead to the different results, since the genotype frequencies of the miR-146a rs2910164 were diverse between Asians and Caucasians. For example, in the present study, MAF value of miR-146a rs2910164 (C/G) was 0.404 in Asians, while the value was 0.631 in Caucasians. On the other hand, most of the included case-control studies did not focus on some vital environmental risk factors (e.g., low intake of dietary fiber, alcohol consumption, tobacco use, obesity, overweight and being physically inactive). It is possible that miR-146a rs2910164 has influence the development of CRC, while those environmental factors mentioned above may cover up the role of C→G variation in rs2910164. Thus, in the future, more case-control studies should be designed to explore the relationship of miR-146a rs2910164 on the risk of CRC, especially focusing on the interaction of gene-environmental factors.
In the present study, heterogeneity might be noted, which could significantly affect the explaination of our findings. When we carried out subgroup analyses, significant heterogeneity was found in colon cancer and Asians subgroups. Combined the Galbraith radial plot ( Figure 5) and the forest plot ( Figure 2) in a dominant model, we identified five outliers [12,13,[35][36][37], which might contribute to prominent heterogeneity. The quality evaluation was used to improve the preent study. In the meta-analysis, we found that only two investigations were poor quality study [36,44]. In subgroup analysis, when we omitted them, the findings were not altered materially.
There are, whereas, some potential limitations in the present study. First, this case-control study only focused on Chinese Han populations. Second, for lack of parameter of risk factors, we only calculate the crude ORs and CIs to assess the correlation between miR-146a rs2910164 and CRC risk in subsequent meta-analysis. Finally, due to the limited participants in subgroups, the power of subgroup analysis might be insufficient.
To conclude, our findings have not supported a relationship between the miR-146a rs2910164 and risk of CRC. The importance of this locus as a genetic predictor for the development of CRC may be very inappreciable and the significance of hereditary screening in healthy individuals may be lack of evidence. In the future, larger studies with well-matched controls are needed to confirm our findings.

Data Availability
The genotypes and environmental factors are summarized in Supplementary Table S1.