Background

In early 1990s, microRNAs (miRNAs) were first discovered through analysis of developmental timing mutants in C. elegans by several groups, simultaneously [1],[2]. In the last decade, the study of miRNA biology has attracted remarkable attention, resulting in rapid advances, and revealing miRNAs as key gene regulators in diverse biological pathways [3]. miRNAs participate in regulation of stem cell functions [4], development [5], drug resistance [6], metabolism and metabolic disorders [7], cardiovascular [8] and malignant diseases [9],[10].

miRNAs are hairpin-derived RNAs ~20-24 nucleotides long, which post-transcriptionally repress the expression of target genes usually by binding to the 3′-untranslated region (3′ UTR) of messenger RNA (mRNA) in a broad range of organisms in both normal physiological and disease contexts [11],[12]. The majority of human miRNA loci are located within intronic regions and are transcribed by RNA polymerase II as part of their hosting transcription units [13]. Typically, following RNA polymerase II-mediated transcription, the long primary precursors (pri-miRNAs) are cleaved by the nuclear RNase III (Drosha) to release ~70 nt pre-miRNAs [14]. The resulting transcripts adopt a stem-loop structure and are then exported to the cytoplasm where they are subsequently processed by another RNase III (Dicer) to generate mature double-stranded ~22 nt miRNAs [15]. Subsequently, one strand of this duplex is incorporated into an Argonaute-containing RNA-induced silencing complex (RISC), resulting in the translational repression and/or degradation of their target mRNAs [16].

Recent evidences support that single nucleotide polymorphisms (SNPs), occurring in DNA sequences of miRNA-coding genes or in miRNA-binding site in mRNAs, contribute to the biogenesis and functions of miRNAs [17]. Gain/loss-of-function of miRNA polymorphisms may result in enhancing of the combination of the miRNA to the targets or losing control of the mRNAs, which may be associated with diseases [18].

Reports on the associations between SNPs in pre-miRNA or SNPs in binding sites in 3′ UTR of miRNA targeting mRNAs and human diseases have provided new insights into the molecular mechanisms of pathophysiological processes in humans. rs4846049 (G > T) of MTHFR gene is associated with increased risk for coronary heart diseases, and the potentially pathogenetic mechanism may be SNP-modified posttranscriptional gene regulation by miRNA-149 to MTHFR [19]. MiR-196a binding-site SNP regulates RAP1A expression contributing to esophageal squamous cell carcinoma risk and metastasis [20]. A miRNA-binding SNP (1010A/G) located within 3′-UTR of HOXB5 is associated with gene expression and may be a promising prognostic factor for bladder cancer [21]. The A to G base change of rs999885 may provide a protective effect against chronic HBV infection but an increased risk for HCC in HBV persistent carriers by altering the expression of the miR-106b-25 cluster [22]. miR-200b/200c/429-binding site polymorphism in the 3′-UTR of AP-2α gene is associated with cisplatin resistance, suggesting that SNP (rs1045385) A > C variation may be a potential prognostic marker for cisplatin treatment [23].

To date, many studies explored the association between rs3746444 T > C SNP in hsa-mir-499 and susceptibility to diseases, such as breast cancer [24]–[26], lung cancer [27], congenital heart disease [28], dilated cardiomyopathy [29], gallbladder cancer [30], squamous cell carcinoma of the head and neck [31], chronic obstructive pulmonary disease [32], liver cancer [33], rheumatoid arthritis [34], coronary artery disease [35], and colorectal cancer [36]. Especially, one study suggested that miR-499 rs3746444*T alleles might be protective for breast cancer [37], but another found that the miR-499 rs3746444 C allele increased cancer risk in the allelic contrast model and in the dominant model, especially in breast cancer [38], while another study showed that no significant associations were observed between rs3746444 in miR-499 and breast cancer susceptibility [39]. Recently, a meta-analysis suggested that polymorphism of hsa-mir-499 rs3746444 T > C was not associated with increased susceptibility to cancers [40],[41], while other systematic analysis supported that hsa-mir-499 rs3746444 polymorphism contributed to the susceptibility to cancers [42]–[45].

The results of these observations remain controversial and inconclusive. In the present study, we conducted a meta-analysis in order to derive more precise and more comprehensive estimation of the associations between the SNP hsa-mir-499 rs3746444 T > C and susceptibility to cancers to quantify the potential between-study heterogeneity.

Methods

Study selection

We performed a publication search in PubMed, EMBASE, ISI Web of Science, The Cochrane Library, ScienceDirect, EBSCO, Ovid, Wiley Online Library, and HighWire databases with the following search terms: (miR-499 OR rs3746444) AND (cancer), by two independent investigators (Chen Chen and Shenglan Yang, last search update: June 29, 2014). Hand searches were also performed to identify additional articles in the reference lists of included articles not retrieved by initial electronic search. Publication date and publication language were not restricted in our search. All studies matching the inclusion criteria were retrieved for further examination and data extraction. All of the investigators have received training in literature search, statistics and evidence-based medicine.

Inclusion and exclusion criteria

Studies included in current meta-analysis had to meet all the following criteria: (1) evaluated the associations between the hsa-mir-499 rs3746444 polymorphism and cancer risk; (2) studied on human beings; (3) diseases were confirmed by histology, imaging or pathology; (4) a case–control design; (5) detailed genotype data were provided for the calculation of odds ratio (OR) and 95% confidence interval (CIs); (6) if serial studies of the same population from the same group were reported, the latest study was included. Studies were excluded when they represented duplicates of previous publications, or were meta -analyses, meeting abstracts, letters, reviews, or editorial articles.

Data extraction

Two investigators (Chen Chen and Shenglan Yang) independently extracted data from the included studies using a standard protocol and data-collection form according to the inclusion criteria listed above, and reached consensus on all items. Data extracted from eligible studies included the first author’s name, year of publication, country of origin, ethnicity, cancer type, genotyping method, total numbers of cases and controls, and genotype frequencies of cases and controls. For study including subjects of different countries of origin group from same ethnicity, we combined them together. The ethnic descents were categorized as Caucasian or Asian. If different results were generated, the two authors would check the data and have had a discussion to come to an agreement. Two senior investigators (Yan Wang and Dao Wen Wang) were invited to the discussion if disagreement still existed.

Statistical analysis

For each study, the departure of frequencies of hsa-mir-499 polymorphism from expectation under Hardy-Weinberg equilibrium (HWE) was assessed by the goodness-of-fit chi-square test in controls. P < 0.05 was considered representative of a departure from HWE.

OR corresponding to 95% CI was used to assess the strength of association between hsa-miR-499 rs3746444 T > C polymorphism and susceptibility to cancer risk. The significance of the pooled OR was determined by the Z-test, and P < 0.05 was considered as statistically significant. Pooled ORs were calculated for allele frequency comparison (C versus T, TC versus TT, CC versus TT, TC/CC versus TT (dominant) and CC versus TC/TT (recessive)), respectively. Subgroup analyses were done by racial descent and cancer type. Statistical heterogeneity among the studies was estimated using chi-square-based Q-test, a P value greater than 0.1 indicates no significant heterogeneity and the pooled OR was estimated by the fixed-effects model (the Mantel-Haenszel method); otherwise, the random-effects model (the DerSimonian and Laird method) was employed [46]. Publication bias of literatures was evaluated by funnel plots and the linear regression asymmetry test by Egger et al. [47]. An asymmetric plot suggests a possible publication bias and the P value of Egger’s test less than 0.05 was considered representative of statistically significant publication bias.Sensitivity analysis was conducted by deleting one study at a time to examine the influence of individual data set on the pooled ORs.

All of the statistical tests were performed with Review Manage (v.4.2; Oxford, England) and STATA software version 11.0 (STATA Corporation, College Station, TX, USA). All the P values are two-sided.

Results

Study characteristics

After first search in PubMed, EMBASE, ISI Web of Science, The Cochrane Library, ScienceDirect, EBSCO, Ovid, Wiley Online Library, and HighWire databases, 70, 76, 116, 0, 2080, 86, 23, 104, and 202 articles, respectively, were retrieved. Of these, 55 articles were relevant to the search words (Figure 1). After further manual search of reference lists, one more study was included. Then, 25 studies were excluded (7 meeting abstracts, 13 meta-analyses, 1 without detailed allele frequency data, and 4 non-case–control studies) and finally 31 studies involving a total of 12799 disease cases and 14507 controls met the inclusion criteria and were subjected to further examination. Characteristics of included studies are summarized in Table 1, including ethnicity, genotype detection method and cancer type. All studies were case–control studies, including 8 liver cancer studies [33],[48]–[54], 4 breast cancer studies [24]–[26],[55], 4 gastric cancer studies [56]–[59], 3 colorectal cancer studies [36],[60],[61], 2 lung cancer studies [27],[62], 2 esophageal cancer studies [63],[64], and others (1 cervical cancer study [65], 1 bladder cancer study [66], 1 prostate cancer study [67], 1 head and neck cancer study [31], 1 gallbladder cancer study [30], 1 childhood acute lymphoblastic leukemia [68], 1 renal cell carcinoma [69] and 1 oral squamous cell carcinoma [70]). Cancers were histologically or pathologically diagnosed in all studies. There were 28 studies of Asian descents (16 Chinese, 4 Indian, 3 Korean, 2 Iranians and 3 others) and 3 studies of Caucasian descents. Several genotyping methods were employed in the studies including Taqman, polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP), DNA sequencing, high resolution melting analysis (HRMA), T-ARMS-PCR, MALDI-TOF MS and Sequenom MassARRAY. PRISMA checklist was generated to provide detailed description of this meta-analysis (Additional file 1: Table S1).

Figure 1
figure 1

Flow diagram of study identification.

Full size image
Table 1 Characteristics of studies included in the meta-analysis
Full size table

Meta-analysis results

The association between hsa-mir-499 rs3746444 polymorphism and susceptibility to cancer was analyzed in 31 independent studies. The results are shown in Table 2 and Additional file 2: Figures S1, S2 and S3. Significant association between rs3746444 polymorphism and susceptibility to cancer was identified in TC versus TT and TC/CC versus TT (dominant) models, when all the eligible studies were pooled (C versus T: OR = 1.08, 95% CI 0.99-1.17, P = 0.08; TC versus TT: OR = 1.11, 95% CI 1.00-1.23, P = 0.04; CC versus TT: OR = 1.02, 95% CI 0.85-1.22, P = 0.85; TC/CC versus TT (dominant): OR = 1.11, 95% CI 1.01-1.23, P = 0.03; CC versus TC/TT (recessive): OR = 0.98, 95% CI 0.81-1.17, P = 0.79). Next, subgroup analyses were performed. Twenty-eight out of the thirty-one included studies were conducted in Asian population. In ethnicity subgroup analysis, significantly increased risks were found in Asians (TC versus TT: OR = 1.14; 95% CI = 1.02-1.27, P = 0.02; TC/CC versus TT (dominant): OR = 1.14; 95% CI = 1.02-1.27, P = 0.02), consistently. This increased risk of cancer was especially significant in Iranian (C versus T: OR = 1.57; 95% CI = 1.23-2.01, P = 0.0003; CC versus TT: OR = 2.23; 95% CI = 1.42-3.51, P = 0.0005; CC versus TC/TT (recessive): OR = 2.23; 95% CI = 1.44-3.45, P = 0.0003) and Chinese population (TC versus TT: OR = 1.18; 95% CI = 1.02-1.35, P = 0.02). However, no significant association between hsa-mir-499 rs3746444 T > C polymorphism and cancer risk was found in Caucasians in any of the genetic models tested. Further subgroup analysis based on cancer type revealed a decreased risk for esophageal cancer (C versus T: OR = 0.80; 95% CI = 0.66-0.98, P = 0.03).

Table 2 Meta-analysis for the hsa-miR-499 rs3746444 T > C polymorphism and cancer risk
Full size table

Publication bias

Begg’s funnel plot and Egger’s test were performed to assess the publication bias of included studies. Symmetrical funnel plots were obtained in all comparison models (Figure 2). Further, Egger’s test confirmed the absence of publication bias in all genetic models (P > 0.05).

Figure 2
figure 2

Begg’s funnel plot for publication bias test. (A) C versus T; (B) TC versus TT; (C) CC versus TT; (D) TC/CC versus TT (dominant) and (E) CC versus TC/TT (recessive). Each point represents a separate study for the indicated association. Log OR, natural logarithm of OR.

Full size image

Sensitivity analysis

Exclusion of one study at a time was performed to examine the influence of the individual data set to the pooled ORs. Analyses were performed using STATA by means of the metainf program. Results showed that the pooled OR estimates were similar with those of the excluded studies (Figure 3).

Figure 3
figure 3

Sensitivity analysis of the influence of a single study on the overall meta-analysis estimate. (A) C versus T; (B) TC versus TT; (C) CC versus TT; (D) TC/CC versus TT (dominant) and (E) CC versus TC/TT (recessive). The solid lines correspond to the pooled OR and 95% CI. Circles and dashed lines correspond to the specific OR and 95% CI without omitted study.

Full size image

Discussion

In the present study, hsa-mir-499 rs3746444 polymorphism and susceptibility to cancer was evaluated in different genetic models in 12799 cancer cases and 14507 controls. Significant association between rs3746444 polymorphism and susceptibility to cancer was identified in TC versus TT and TC/CC versus TT (dominant) models. Subgroup analysis absed on ethnicity revealed significantly increased risk in Asians (TC versus TT and TC/CC versus TT (dominant), but there was no significant association in Caucasians in all genetic models. This association was especially marked in Iranians (C versus T; CC versus TT and CC versus TC/TT (recessive) and a significant association was also identified in Chinese (TC versus TT) However, subgroup analysis based on cancer type revealed a decreased risk for esophageal cancer (C versus T).

Many studies investigated the role of SNPs presented in precursor and mature miRNAs, and their influences on susceptibility and progression of various diseases. miRNA-associated SNPs can have direct or indirect effects: 1) direct effects are produced by the SNPs in the pri-miRNA, pre-miRNA or mature miRNA that impair or enhance miRNA processing or function; 2) indirect effects are derived from the SNPs in miRNA promoters that affect transcription and SNPs in an mRNA that create or destroy a target site [71].

Genome-wide patterns of human polymorphisms in miRNAs and miRNA target sites in 3′ UTRs of mRNAs revealed that only ~10% of human pre-miRNAs have documented SNPs, and <1% of mature miRNAs have SNPs in the functional seed region [72]. hsa-mir-499 rs3746444 polymorphism is located in the seed region of the mature miR-499 sequence. This T > C polymorphism-resulted mismatch may affect target mRNA expression. Several case–control studies have investigated the association between rs3746444 polymorphism and risks of various cancers. However, the results were controversial.

Different from past studies, we analyzed the specific association among different ethnic groups throughout Asia. In this meta-analysis, we first found that rs3746444 polymorphism was associated with increased cancer risk mainly in Asian population, especially in Chinese and Iranians. It is important to notice that, although rs3746444 polymorphism contributes to cancer risk in Chinese and Iranians, the genetic models are different. The genetic variations between Chinese Han population and Arabian population may contribute to the differences. Interestingly, we also found that rs3746444 polymorphism showed a decreased risk in esophageal cancer. This suggests that potentially different mechanisms may underlie tumorgenesis in different pathological backgrounds and the environment variations.

It has been suggested that miRNAs may influence gene expression of ~30% of protein-coding genes by binding to the mRNA incompletely, and a miRNA may affect multiple target mRNAs involved in same pathophysiological progress [3],[73]. This means that rs3746444 polymorphism may contribute to single gene function in certain disease model at molecular level. When taken together in epidemiological studies of populations, various effects of rs3746444 polymorphism on different genes may result in different associations with diseases at phenotype level.

Our meta-analysis pooled the largest numbers of cases and controls from included studies, which significantly increases the statistical power but some limitations should be considered. Firstly, lack of the consideration of combined genetic factors together with environmental exposures, while a more precise analysis needs to be conducted if more detailed data are available. Secondly, although sensitivity analysis showed that each study can not affect the overall results, some heterogeneity was evident in some of the comparisons. This may be due to different ethnicities and different pathology. And expanding the sample size may yield a better result. Thirdly, some unpublished data (such as negative report, not written in English and lack of studies in African population) may potentially influence the results of our meta-analysis. At last, different genotyping strategies may contribute to the bias in the analysis.

Recently several meta-analyses systematically reviewed the potential association of rs3746444 polymorphism with susceptibility to diseases [40]–[45],[74]–[77]. However, the results were controversial across these meta-analyses. Such discrepancies in the results of these systematic reviews may due to different search periods, databases searched, inclusion and exclusion criteria. Our efforts to search the maximum range of databases yielded the most comprehensive studies in the similar time period especially the most recent publications [53],[59]. Further, diagnosis of cancer is largely relied on confirmation by histology, imaging or pathology. Some of the other studies omitted this fundamental condition. Most importantly, deviations from Hardy-Weinberg equilibrium can inflate the chance of a false-positive association, and sensitivity analysis should be performed, pooling with and without studies not in HWE, to test the robustness of the results [78],[79]. Only few of other studies performed sensitivity analysis, which may limit the reliability of their conclusions. Although some studies claimed that their studies were in HWE, a careful examination of the studies revealed that the data did not reach HWE. In conclusion, in our study, relevant literatures selected from broad databases with stringent standards would be expected to provide a more reliable conclusion.

Conclusions

In summary, though with limitations, our meta-analysis suggested that hsa-mir-499 rs3746444 T > C polymorphism is associated with the risk of cancer, especially in Asians, mainly in Chinese and Iranians. However, rs3746444 T > C polymorphism is negatively associated with the risk of esophageal cancer. Larger studies from different ethnic groups and studying different types of cancers with detailed information are needed to further clarify the association between rs3746444 polymorphism and cancer risk.

Additional files