Comprehensive Review of Genetic Association Studies and Meta-Analysis on polymorphisms in microRNAs and Urological Neoplasms Risk

Gene expression is negatively regulated by microRNAs (miRNAs), which commonly act as tumor oncogenes or suppressors. Previous results were inconsistent concerning the relationship between polymorphisms in miRNAs and risk of urological neoplasms. Here, we conducted a comprehensive literature research on diverse databases aiming at enrolling all eligible studies up to August 31, 2016. A total of 13 publications comprising 29 case-control studies were enrolled for three polymorphisms in three miRNAs. Overall analyses suggested significant associations between miR-146a rs2910164 polymorphism and urological neoplasms risk in allelic, homozygote and recessive models. In the stratified analysis by ethnicity, we uncovered a significant association between rs2910164 polymorphism and risk of urological neoplasms in Asian populations in allelic, homozygote and recessive models. Highlighted, when stratified analysis was conducted by cancer type, rs2910164 polymorphism was also significantly associated with an increased risk of bladder cancer in allelic, homozygote and recessive models. Although for rs11614913 and rs3746444 polymorphisms, overall analyses suggested negative results, for rs11614913 polymorphism, when subgroup analysis was conducted by cancer type, a significantly decreased risk of renal cell cancer was identified in recessive model. In brief, current work indicated that miR-146a rs2910164 polymorphism is a risk factor for urological neoplasms, particularly for bladder cancer.

mechanism transcriptionally regulating gene expression. To date, plenty of evidences have indicated that mature miRNAs participated in the degradation or translational suppression of mRNA by linking to the 3′ untranslated region of target genes, and eventually related to the regulation of various critical biological activities, including cell metabolism, proliferation, differentiation, proliferation, and apoptosis, even working as tumor suppressors or oncogenes for the sake of participating in tumorigenesis through posting expression regulation of homologous target genes [3][4][5] . Some key miRNAs, which have been regarded as biomarkers, could ameliorate diagnosis and prediction of prognosis and treatment response for cancer patients 6 .
Single nucleotide polymorphisms (SNPs) are defined as a kind of genetic polymorphisms related to disease susceptibility, population diversity, drug metabolism and genome evolution [7][8][9] . It has been proposed that SNPs found in miRNA genes affect miRNA transcription, processing, and interactions with target mRNAs. SNPs in miRNA genes are widely-accepted to affect function in one of three ways: firstly, through the transcription of the primary transcript; secondly, through pri-miRNA and pre-miRNA processing; and also through effects on miRNA-mRNA interactions 10 . For example, miR-196a gene encodes many mature products including miR-196a and miR-196a*, and it also comprises rs11614913 polymorphism. In a study conducted by Hoffman et al. 11 , they demonstrated that relative to empty vector control, expression of miR-196a can be increased when transfected with pre-miR-196a-C or miR-196a-T in breast cancer cells. In addition, miR-196a was doubly more expressed in cells transfected with pre-miR-196a-C when compared with miR-196a-T. Namely, rs11614913 polymorphism can influence pre-miRNA processing into mature and functional form, and subsequently activating tumorigenesis. In recent years, plenty of investigators have studied the relationship between genetic polymorphisms in precursor or mature miRNA sequence and diverse cancer risks, including gastric cancer (GC) 12,13 , colorectal cancer (CRC) 14 , liver cancer 15 , gallbladder cancer (GBC) 16 , esophagus cancer (EC) 17 , bladder cancer (BCa) 18 and prostate cancer (PCa) 19 . However, these results were controversial and inconsistent. In current work, we performed an update meta-analysis at the aim of precisely verifying the relationship between genetic polymorphisms in miRNAs and urological neoplasms risk.

Materials and Methods
Search strategy and study selection. We carried out a literature research for all eligible articles that investigated the relationship of three polymorphisms in miRNAs (miR-146a rs2910164; miR-499 rs3746444;   . All the fields of the retrieved articles were screened through titles and abstracts. We also checked reference lists of the review articles and enrolled articles. Eligible studies were included while they met the following inclusion criteria: (1) assessment of microRNA 146a/499/196a2 polymorphisms and risk of urological neoplasms; (2) case-control design independently for human, and; (3) providing useful data of genotype frequencies. At the same time, the exclusion criteria were presented as follows: 1) duplicate data, 2) reports of clinical cases, comments, series, reviews and editorial and 3) insufficient data. Studies published in some other languages instead of English were also excluded. Articles involved with two or more case-control tests were regarded as two or more different studies. Data extraction. We reviewed carefully for each publication, and extracted the following data which conformed to the selection criteria: first author, year of publication, original country, ethnicity of the population studied, genotyping method, source of controls, cancer type, numbers for cases and controls of all genotypes, whether verified Hardy-Weinberg equilibrium (HWE). If original data of genotype frequency was not provided in relevant studies, we could send a request to the corresponding author for additional information.
Statistical analysis. We used crude ORs and 95% CIs to evaluate the relationship of the miRNA polymorphisms and urological neoplasms under five genetic models, including allelic (B vs. A), recessive (BB vs. BA + AA), dominant (BA + BB vs. AA), homozygous (BB vs. AA), and heterozygous (BA vs. AA) models 20 (A: wild allele; B: mutated allele). In addition, subgroup analyses were conducted by source of controls, genotyping methods, ethnicity, cancer type and HWE status. Chi-square-based Q-tests were used to check heterogeneity across all the included studies, and P < 0.05 level was considered statistically significant 21 . If there existed heterogeneity in the included studies, the random-effect model (DerSimonian and Laird method) was performed to assess the pooled OR, otherwise the fixed-effect models (DerSimonian and Laird method) were adopted 22 . We determined publication bias using Begg's funnel plot and Egger's test. Sensitivity analyses were conducted in order to assess the data stability. We removed each study involved in this meta-analysis one at a time, and the residual studies were analyzed. STATA version 12.0 was applied in carrying out all data analyses. Statistical significance was set at the level of P < 0.05. Table 1 showed the characteristics of all the eligible studies and genotype frequency distributions of three miRNA polymorphisms (miRNA-146a rs2910164, miR-499 rs3746444 and miRNA-196a2 rs11614913) included in current meta-analysis 16,17,[23][24][25][26][27][28][29][30][31][32][33] . The study selection processes were presented in Figs 1-3.

Study Characteristics.
For miRNA-146a rs2910164 polymorphism, a total of eleven studies with 3,647 cases and 4,413 controls met inclusion criteria. Eight of them were performed in Asian ethnicities and three in Caucasian   (Table 1). For miR-499 rs3746444 polymorphism, nine eligible studies with a total of 2,797 cases and 2,941 controls were enrolled 16,17,23,27,29,30,32,33 . Seven of them were performed on Asian ethnicities, one in Caucasian and African ethnicity, respectively. In addition, controls of five studies were H-B, and four were P-B. In addition, control groups of two studies were not consistent with HWE 30,32 .
For miRNA-196a2 rs11614913, we analyzed nine studies comprising 4,963 cases and 5,066 controls published between 2008 and July 2016 17,23,27,29,30,[33][34][35] . Four studies were Asian ethnicities and rest were Caucasian ethnicities. In addition, three studies were conducted based on P-B, and the remaining six studies were H-B. As for HWE status, genotype distributions of control groups in five studies were not consistent with HWE 16,23,29,30,32 . Quantitative Data Synthesis. All the calculated results were summarized in Table 2.
miR-499 rs3746444. Overall, no significant association was uncovered for rs3746444 polymorphism and risk of urological neoplasms. Moreover, in the stratification analyses by cancer type, source of control and ethnicity, also null results were uncovered. Similarly, when subgroup analysis was conducted by HWE status, after excluding two studies that were not in accordance with HWE 30,32 , overall results were not changed ( Table 2).  Sensitivity Analysis. We repeat this meta-analysis and omit every study one by one, in order to examine effects of all the eligible studies involved in our study. The results showed that there was no material alteration in corresponding pooled ORs for miR-196a2 rs11614913 and miR-499 rs3746444 (Supplementary Table S2). For miR-146a rs2910164 polymorphism, the study conducted by Wang et al. 25 was mainly considered as the reason of heterogeneity (Supplementary Table S2). After we excluded this study, there no longer existed heterogeneity, but still presented a negative association.

Test of Heterogeneity. Both overall comparisons and subgroup analyses identified that heterogeneity was
across the studies. Thus, we estimated source of heterogeneity within each model by performing further analyses stratified by source of control, ethnicity, HWE status and cancer type. For rs11614913 polymorphism, ethnicity (P heterogeneity = 0.005), source of controls (P heterogeneity = 0.007), HWE status (P heterogeneity = 0.009) and cancer type (P heterogeneity = 0.009) showed potential sources of between-study heterogeneity. For rs2910164 and rs3746444, no source was observed contributed to the substantial heterogeneity.
Publication Bias. Both Begg's funnel plot and Egger's regression tests were conducted to evaluate potential publication bias. In the funnel diagram, if publication bias was not existed, data obtained from each study will be presented with an inverted funnel-like symmetric distribution on the graph, whereas asymmetric inverted funnel graph suggested publication bias. As for miR-146a rs2910164 and miR-499 rs3746444 polymorphisms, neither Begg's funnel plot nor Egger's regression showed evidence of publication bias (Shape of funnel plot was symmetrical, which was further confirmed by Egger's regression test; Supplementary Table S3 and Figs 4-6). For miR-196a2 rs11614913 polymorphism, publication bias was existed both in H-B and HWE (N) subgroups. Therefore, we further conducted sensitivity analyses using the trim and fill method 36 , and imputed studies provide a symmetrical funnel plot (data not shown), indicating that no publication bias was existed in both H-B and HWE (N) groups.

Discussion
In human, miRNA genes are approximately 50% situated in genomic areas that are frequently related to tumor. Polymorphisms are the most common type of genetic factors within the human genome, which may cause differences of phenotype 37 ; and such polymorphisms in miRNA may influence the formation of miRNAs, pri-miRNAs, pre-miRNAs and/or mature miRNAs, and/or the selection of targets and therefore obviously influence an individual's cancer susceptibility 38 . Here, we assessed the relationship between three common miRNA variations (miR-146a rs2910164, miR-196a2 rs11614913 and miR-499 rs3746444) and their risk of urological neoplasms. Current study was a kind of stratified investigations based on large populations and various cancer types. Significant relationship between miR-146a rs2910164 polymorphism and urological neoplasms risk was found, particularly in Asians instead of Caucasians. Especially, in the stratification analysis by cancer type, rs2910164 polymorphism was significantly associated with an increased risk of BCa. Although for rs11614913 and rs3746444 polymorphisms, overall analyses uncovered negative results, however, for rs11614913 polymorphism, when subgroup analysis was conducted by cancer type, a significantly decreased risk of RCC was identified in a recessive model. In addition, as for Begg's funnel plot and Egger's test, the adjusted results indicated that publication bias was not existed. Several investigators have paid attention to a single polymorphism or several polymorphisms in miRNA and cancer risk. Anyway, none of these meta-analyses have exhaustively enrolled all the available miRNA polymorphisms or commonly studied miRNA polymorphisms. In the current analysis, we covered the previously published eligible publications, and added some updated works. By comparing with previous meta-analysis works, several advantages in current study should be pointed. First of all, a more comprehensive analysis was conducted with large sample sizes, for the purpose of strengthening statistical power and reliability of conclusions. Secondly, we performed various subgroup analysis by ethnicity, source of controls and so on, in order to provide the sources of heterogeneity and the tumor and/or race markers. Thirdly, data in our work was adjusted referring to a commonly used formula in genome-wide association study (GWAS); our conclusions suggested that miR-146a rs2910164 polymorphism was associated with an increased risk of urological neoplasms, particularly for BCa in Asians, whereas in the study conducted by Xiao et al. 35 , they indicated that miR-146a rs2910164 B allele is associated with a decreased risk of BCa and PCa in Asians.
Nevertheless, there were still several limitations that should be noted in present work. Firstly, for miR-196a2 rs11614913 polymorphism, relatively heterogeneity existed between some studies, although we conducted this analysis with severe inclusion criteria and explicit extraction data. Therefore, after stratified analysis by ethnicity, source of control, HWE status and cancer type, we observed that the subgroup heterogeneity reduced significantly. It can be assumed that the heterogeneity possibly derived from difference of ethnicity, source of control, HWE status and cancer type. Secondly, the small sample sizes of studies may cause low power for data, particularly in subgroup analyses. Thirdly, we did not assess potential interactions due to lack of relevant data across the included studies; this is of extreme significance because interactions between gene-gene and gene-environment may modify kinds of disease risk. Last, our study was in lack of African ethnicities; therefore, more studies based on African populations were needed to be included in our analysis for drawing a more comprehensive conclusion.
In summary, our conclusions demonstrate that rs2910164 polymorphism is a risk factor for urological neoplasms in Asians, particularly for BCa. Further studies with larger sample sizes are warranted to clarify the possible roles of these polymorphisms in urological neoplasms.