Different Effects of Three Polymorphisms in MicroRNAs on Cancer Risk in Asian Population: Evidence from Published Literatures

MicroRNAs (miRNAs) are a class of small non-protein-coding RNAs, which have emerged as integrated and important post-transcriptional regulators of gene expression. It has been demonstrated that single nucleotide polymorphisms (SNPs) exist in protein-coding genes. Accumulated studies have evaluated the association of miRNA SNPs with cancer risk, especially in Asian population, which included a series of related studies. However, the results remain controversial for the different genetic backgrounds, living habits and environment exposed. To evaluate the relationship between SNPs in miRNAs and cancer risk, 21 studies focused on Asian population were enrolled for the pooled analysis for three polymorphisms rs2910164, rs11614913, rs3746444 in three miRNAs miR-146aG>C, miR-196a2C>T, miR-499A>G using odds ratios (ORs) with 95% confidence intervals (CIs). For rs2910164 polymorphism, C allele was observed association with decreased overall cancer risk. In addition, subgroup analysis revealed of rs2910164 C allele decreased hepatocellular carcinoma (HCC), cervical cancer and prostate cancer risk among Chinese population. For rs11614913 polymorphism, TT genotype was observed to be associated with decreased cancer risk, especially for cancer type of colorectal cancer (CRC), lung cancer and country of Korea, North India. Whereas, rs3746444 G allele was an increased cancer risk factor in Chinese population, especially for breast cancer. In conclusion, this meta-analysis indicated that rs2910164 C allele was associated with decreased cancer risk in Chinese population. However, the association varied from different cancer types. Furthermore, TT genotype of rs11614913 was associated with decreased cancer risk. While different cancer types and countries contributed to different effects. Whereas, rs3746444 G allele was a risk factor in Chinese population, and the association varied from different cancer types.


Introduction
MicroRNAs (MiRNAs) are a family of naturally occurring, small noncoding RNAs of 21-24 nucleotides in length that regulate gene expression by base pairing with target mRNAs at the 39UTR, leading to mRNA cleavage or translational repression [1,2]. MiRNAs encoded in the genome are transcribed by RNA polymerase II or RNA polymerase III in the nucleus, where they cleaved by Drosha and Dicer sequentially [3,4]. It has been suggested that miRNAs are important post-transcriptional regulators of gene expression that control diverse physiological and pathological processes. Accumulating evidence indicates that aberrant expressions of miRNAs were indicated to involve in tumorigenesis, development and prognosis of many cancers [5,6,7,8,9].
Single nucleotide polymorphisms (SNPs) were found in most genes, and recently SNPs of miRNAs have been paid much more attention. Studies have reported that miRNA SNPs could alter expressions or functions of miRNAs, and related to cancer risk. Meanwhile, studies have reported polymorphisms in miRNA genes, biogenesis pathway of miRNAs and their target binding sites. Moreover, polymorphisms in miRNA genes could directly influence the expressions and functions of miRNAs. Recently, miR-146aG.C (rs2910164), miR-196a2C.T (rs11614913) and miR-499A.G (rs3746444) were drawed close attention and were expected to demonstrate the association with many cancers [10,11,12,13,14,15,16,17]. However, the results were generally inconsistent and inconclusive. Therefore, this meta-analysis focused on these three polymorphisms to deliberate their associations with cancer risk, which have surveyed in many populations. According to the recent studies, consistent conclusions were observed in Caucasian population while conflicting results were found in Asian population [13,18,19,20,21,22,23] due to the different countries, the numbers of study population, cancer types. To draw a conclusion of three polymorphisms and cancer risk in Asian population, an analysis of pooled published studies was required. This meta-analysis explored the associations between polymorphisms of three miRNAs and cancer in Asian population.

Literature and Inclusion Criteria
Using the combined words ''miR-146a/miR-196a2/miR-499'', ''cancer'' or ''carcinoma'', ''genetic variation'' or ''polymorphism'', a comprehensive systematic bibliographic searching was applied through the medical databases PubMed, EMBASE and Web of Science for all medical published up to October 15th, 2012. In addition, studies were identified by manual search of the reference listed in the retrieved studies. Data from studies were accepted in our meta-analysis only if the study met all of the following criteria: (1) published in English; (2) available cancer risk and miR-146a/miR-499/miR-196a2 polymorphism data related to Asian population; (3) case-control studies; (4) sources of cases and sufficient available data to estimate an odds ratio (OR) with 95% confidence interval (CI); (5) available genotype frequency. Moreover, the studies were eliminated if there are no raw data in the studies, or they are case-only studies, case reports, editorials, and review articles (including meta-analyses).

Data Extraction
Information was reviewed carefully extracted from all the eligible articles independently by two of the authors (Yeqiong Xu and Ling Gu) according to the inclusion criteria listed above. The characteristics information of enrolled studies was extracted from the study: the first author's last name, year of publications, country of subjects, cancer type, the source of controls, genotyping method, matching numbers of genotyped cases and controls, polymorphism site and P for HWE (Table 1). If discrepancies and differences were existed after data collection, discussion was carried out to get consensus.

Statistical Analysis
The strength of association between the three SNPs and cancer risk was assessed by odds ratios (ORs) with 95% confidence intervals (CIs). The pooled ORs were estimated for dominant model, recessive model, homozygote comparison, heterozygote comparison and allelic comparison, respectively. Stratified analyses were also performed by cancer type (HCC, CRC, cervical cancer, prostate cancer, breast cancer, gastric cancer, lung cancer, and other cancers group which combined the cancer types containing less than two individual studies), country (China, Korea, Japan, North India and other countries group which combined the countries containing less than two individual studies) source of control and genotyping method. Heterogeneity across the studies was evaluated by using the Chi-square test based Qstatistic test, and it was considered significant when P heterogeneity (P h ) ,0.05. The data were combined using both fixed-effects (the Mantel-Haenszel method) and random effects (the DerSimonian and Laird method) models. A random-effect model was employed when heterogeneity existed [24][25], Otherwise, the fixed-effect model was employed to pool the results [26]. Moreover, to assess the stability of the results, a sensitivity analysis was performed. Publication bias was checked graphically by using funnel plots and statistically using the Egger's linear regression test. For the controls of each study, the genotype frequencies of the three polymorphisms of miRNA were assessed for Hardy-Weinberg equilibrium using a web-based program (http://ihg.gsf.de/cgibin/hw/hwa1.pl). All statistical tests were performed with STATA 10.0 and all the P values were two-sided.

Characteristics of Studies
This study enrolled 31 eligible papers ( Figure 1) according to the inclusion criteria. For rs2910164 polymorphism, 21 studies with available data were enrolled in the pooled analysis. These studies consisted of China (13 studies), Korea (2 studies), North India (3 studies) and Japan (2 studies) related to HCC (4 studies), gastric cancer (4 studies), cervical cancer (2 studies), prostate cancer (2 studies), breast cancer (2 studies) and other cancers (7 studies). In addition, the controls of most studies were population-based, and the main genotyping method was PCR-RFLP (Table 1).
For rs11614913 polymorphism, 21 studies provided available data, which were classified into CRC (4 studies), HCC (3 studies), breast cancer (3 studies), lung cancer (3 studies), gastric cancer (2 studies) and other cancers (6 studies). Meanwhile, these studies with data of 12 studies of Chinese population, 4 studies of Korean population, 3 studies of North Indian population and 2 studies of other countries. To analyze polymorphisms, genotyping by polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) were performed by the most studies, in which 16 were population-based and 5 were hospital-based (Table 1).
For rs3746444 polymorphism, 13 studies covered China (6 studies), Korea (2 studies), North India (3 studies) and other countries (2 studies) related to HCC (3 studies), breast cancer (2 studies) and other cancers (8 studies) were included in the pooled analysis. What's more, these studies contained 11 of populationbased controls and 2 of hospital-based controls.
There was no evidence that rs11614913 T allele associated with the risk of cancer. Meanwhile we conducted subgroup analysis of cancer type, country, source of controls and genotyping method. In the subgroup analysis of cancer type, a decreased risk was observed in CRC (OR = 0.84, 95% CI: 0.76-0.92, P h = 0.281) and lung cancer (OR = 0.87, 95% CI: 0.80-0.95, P h = 0.854). Country subgroup analysis revealed T allele was associated with decreased cancer risk in Korean population (OR = 0.87, 95% CI: 0.79-0.96, P h = 0.608). In the subgroup analysis of genotyping method indicated a decreased cancer risk with T allele determined by Taqman (OR = 0.87, 95% CI: 0.77-0.98, P h = 0.191).
For rs3746444 polymorphism, a significant increased cancer risk was found in the population with G allele (OR = 1.18, 95%  Table 2. Stratified analyses of the miR-146aG.C (rs2910164) polymorphism and cancer risk.

Sensitivity Analysis
Sensitivity analysis was performed to assess the stability of the results and assess the source of the heterogeneity by sequential removal of individual eligible study. For rs2910164 polymorphism, studies by Okubo [14] and Tian [10] were the main origin of heterogeneity. The heterogeneity was decreased when these two studies removed (CC+GC vs GG: P h = 0.067, I 2 = 34.9%). For rs11614913 polymorphism, sensitivity analysis indicated that studies by Chu [15], Dou [16], George [17] and Srivastava [24] were the main origin of heterogeneity. The heterogeneity was decreased when these four studies removed (TT+CT vs CC: P h = 0.054, I 2 = 38.4%). For rs3746444 polymorphism, sensitivity analysis indicated that studies by Tian [10], Chu [15], Kim [25] and Okubo [14] were the main origin of heterogeneity. The heterogeneity was decreased when these four studies removed  Table 3. Stratified analyses of the miR-196a2C.T (rs11614913) polymorphism and cancer risk. (GG+GA vs AA: P h = 0.066, I 2 = 45.4%). In addition, no other single study was observed to impact the pooled OR by sensitivity analysis.

Publication Bias
Begg's funnel plot and Egger's test were performed to assess the publication bias of enrolled literature. The shape of the funnel plot indicated obvious asymmetry in rs11614913 dominant model comparison ( Figure 5A). Thus, Egger's test was used to provide statistical evidence of funnel plot asymmetry (t = 2.15, P = 0.045) (shown in Table 5), which suggested the existence of publication bias in this meta-analysis. To adjust this bias, a trim-and-fill method illustrated by Duval and Tweedie [26] was utilized ( Figure 5B). As a result, the conclusion with or without the trimand-fill method did not change, which indicated that our results were statistically robust. While all models of rs2910164 and rs3746444 didn't show any publication bias (P.0.05) (shown in Table 5).

Discussion
As we all know, the association between the SNPs in proteincoding genes and the risk of cancer has been explained thoroughly, little cancer association studies concerning miRNA SNPs have been reported. In the present case-control study, associations of three miRNA polymorphisms (miR-146aG.C, rs2910164;miR-196a2C.T, rs11614913; miR-499A.G, rs3746444;) and cancer susceptibility were estimated. The polymorphisms of these three miRNAs may influence the effect of their targets, which contributed to the tumorigenesis, development and prognosis of many cancers. It is believed that tumor necrosis factor receptorassociated factor 6 and interleukin-1 receptor-associated kinase 1 are two potential targets of miR-146a [27], which could decrease the levels of these two proteins and reduce the activity of the NF-kB signaling pathway involving in tumorigenesis [28]. The main targets of miR-196a2 are homeobox (HOX) gene cluster and Annexin A1 (ANXA1). HOX genes include HOXB8, HOXC8, HOXD8, and HOXA7, which are known regulators of oncogenesis [29]. Meanwhile, ANXA1 is known as a mediator of apoptosis and an inhibitor of cell proliferation [30]. MiR-499 mainly targets to transcriptional repressor SOX6 [31], which reduce the level of  fibroblast growth factor (FGF)-3 affecting cell proliferation and differentiation [32,33]. In this meta-analysis, 31 eligible studies were enrolled to assess the association between three miRNA polymorphisms and cancer risk. We demonstrated that rs2910164 C allele, rs3746444 A allele and rs11614913 TT genotype were associated with significantly decreased risk of cancer.   Former studies regarding G to C variation in the miR-146a precursor showed that G-allelic miR-146a precursor displayed increased production of mature miR-146a compared with Callelic and G allele in rs2910164 was associated with the predisposition of several cancers [18,34,35]. In overall pooled results from 21 studies, we concluded rs2910164 C allele was associated with decreased cancer risk. Furthermore, stratified analyses by cancer type revealed that rs2910164 CC genotype reduced risk of HCC, cervical cancer and prostate cancer, however no significant associations were observed in breast cancer and gastric cancer, which indicated that rs2910164 polymorphism might have different effects in distinct cancers. The results were consistent with the previous studies [13,18,36,37]. While disaccords appeared in gastric cancer [14,38,39]. Inconsistent results might be caused by limited studies enrolled in this meta-analysis. Different study design and approach to select participants should also be taken into account. Followed stratified analyses by country indicated decreased cancer risk were discovered only in Chinese population, which reflected differences in genetic background and the environment exposured might produce different effects on cancer risk. At last, source of controls and genotyping method stratified analyses were also conducted in this meta-analysis. The results showed that different sources and methods could play different roles in cancer risk. As age and gender are risk factors for many cancers, which must be considered in this meta-analysis. The study conducted by Zeng et al [38] concluded that rs2910164 GG+GC genotype among males and subjects aged#58 years was associtated with increased gastric cancer risk. The similar phenomenon was observed in the study by Zhou et al [39], which showed elevated gastric cancer risk was more evident among younger subjects (,65 years) with rs2910164 GG genotype. However, they observed no significant difference in the stratification of sex. As age increases, accumulated exposure to environmental carcinogens and genomic alterations would facilitate carcinogenesis [39]. Therefore, the age is believed to be an important risk factor for cancers, which was inconsistent with our results. More genomic alterations and environmental carcinogens may contribute to late-onset gastric cancer and stratification analysis by age should be more cautious. As for the pre-miR-146a sex-specific effect, the exact mechanism remains unclear. Therefore, well-designed, unbiased, large case-control studies were urgently needed to achieve a more accurately result.
Recently, rs11614913 polymorphism in pre-miRNAs has been reported to contribute to susceptibility of breast cancer [13], lung cancer [10], glioma [16], and influence survival of non-small cell lung cancer [40] by altering the expression of mature miR-196a and its binding to target mRNA. Our results showed that TT genotype was associated with decreased risk of CRC, breast cancer and lung cancer, which was consistent with previous findings [20,23,41,42]. The controversy between our study were also apparent, no correlations were achieved between rs11614913 polymorphism and susceptibility of HCC and gastric cancer, while Li et al [12] and Okubo et al [14] presented contrary opinions respectively. In addition, contrast to Chinese and Korean population, rs11614913 CT genotype trend to increase cancer risk in North Indian population. Based on the above points, we deduced that cancer type and country differences made rs11614913 polymorphism have distinct effects. Cancer was a complex disease, and numerous factors would lead to tumorigenesis. Meanwhile, different cancers had different pathogenesis. Therefore, rs11614913 polymorphism might have distinct effects according to cancer types. Inconsistent results about different countries might be caused by differences in living habit, genetic background and the environment. In addition, this meta-analysis enrolled only 21 studies for rs11614913 polymorphism, inadequate study would be an influence factor. At the same time, the results also showed that different sources and methods contributed to different cancer risk. To further reveal the association between rs11614913 polymorphism and cancer risk, more well-designed studies based on homogeneous cancer patients and unbiased larger sample sizes were wanted.
As for rs3746444 G allele, an increased cancer risk was discovered in the pooled analysis. And subgroup analyses of cancer type and country showed that rs3746444 G allele association with increased risks were observed in breast cancer and Chinese population respectively. Meanwhile, a significant association was also observed for comparison of GG+GA vs AA in North Indian population. The results suggested different cancer types and countries could lead to distinct effects of rs3746444 polymorphism. Hu et al [13] showed that significantly increased breast cancer risk was associated with variant genotypes of hsa-mir-499 rs3746444 in Chinese women, which consistent with our results. Finally,source of controls stratified analysis was also conducted. The results indicated distinct source of controls was also an important influence factor to affect the association between rs3746444 polymorphism and cancer risk. However, the results were based on 13 studies enrolled in the analysis, which could affect the results owing to small amount of studies. To draw a more precise conclusion, more related studies needed.
According to the test of heterogeneity conducted above, it's not difficult to perceive different countries contributed to heterogeneity of rs11614913 and rs2910164 polymorphisms, which indicated miRNAs might play different roles according to countries. Meanwhile, different cancer types were also a main factor contributed to heterogeneity. Significantly decreased associations were found mainly in HCC, lung cancer, cervical cancer, which suggested miRNAs might have different affections in different cancer types.
After all, this meta-analysis still existed some limitations. Firstly, publication bias which we have detected in rs11614913 polymorphism, and in other polymorphisms publication bias might also exist while we didn't detect owing to quantitative restrictions of studies. Secondly, there was no uniform definition of controls, although most of the controls were mainly selected from healthy populations, a few of them were patients. Thirdly, the detailed information (such as age, sex, menstrual history, life-style and environmental factors) was not considered so that our unadjusted estimates should be confirmed by further studies.
In conclusion, this meta-analysis measured the association of three miRNA polymorphisms and cancer risk. We observed TT genotype of rs11614913 polymorphism was associated with decreased cancer risk, especially for CRC and lung cancer in Korean and North Indian population. Moreover, rs2910164 C allele was associated with decreased overall cancer risk especially for HCC, cervical cancer and prostate cancer risk in Chinese population. Whereas, rs3746444 G allele was a risk factor in Chinese population, especially for breast cancer. However, further studies based on larger, stratified population to facilitate evaluation the association between miRNAs and cancer risk.