Prognostic significance of microRNA-135 in patients with digestive system cancers: a systematic review and meta-analysis

Abstract Background: MicroRNA-135 (miR-135) is a well-known non-coding RNA that has been demonstrated to participate in tumorigenesis and cancer development; however, the clinical prognostic value of miR-135 in digestive system cancers remains controversial. This meta-analysis aims to explore the potential value of miR-135 as a prognostic marker for digestive system cancers. Methods: The PubMed, Embase, Cochrane Library, and Web of Science databases were searched for eligible articles published before 31 August 2019. Stata 12.0 software was used to analyze the overall survival (OS), disease-free survival (DFS), and recurrence-free survival (RFS) rates to access the prognostic value of miR-135 in digestive system cancers. We then used The Cancer Genome Atlas (TCGA) datasets to validate the meta-analysis results. Results A total of 1470 patients from 17 studies were included in this meta-analysis. The pooled results showed that enhanced miR-135 expression was significantly associated with poor OR (hazard ratio (HR): 1.790; 95% confidence interval (95% CI): 1.577–2.031; P=0.000), DFS (HR: 1.482; 95% CI: 0.914–2.403; P=0.110), and RFS (HR: 3.994; 95% CI: 1.363–11.697; P=0.012) in digestive system cancers. A sensitivity analysis confirmed the reliability of our findings, and no significant publication bias was observed. Conclusion: MiR-135 can be used as a novel biomarker for patients with digestive system cancers. We look forward to future large-scale clinical studies that will investigate the prognostic value of miR-135.


Introduction
Digestive system cancers are one of the most common malignancies throughout the world. Such digestive system cancers include esophageal, gastric, colorectal, liver, gallbladder, and pancreatic cancers (PCs). According to the latest reports, there are approximately 4.9 million new diagnosed cases and 3.5 million deaths for digestive system cancers that occur worldwide each year [1]. Although surgical techniques and immunotherapy have been developed, the prognosis of patients with digestive system cancers remains poor. With increased in-depth research into long non-coding RNA, miRNA, and epigenetics, we have gained a better understanding of cancer at the genetic level [2][3][4]. Therefore, we expect to identify reliable biomarkers that can be used as prognostic factors from these in digestive system cancers. Recently, miRNAs have also attracted increased attention and are considered to be potential biomarkers for cancer prognosis [5,6].
The iRNAs are short non-coding RNAs (ncRNAs) comprising approximately 18−23 nucleotides that mediate gene silencing by guiding Argonaute (AGO) proteins to target sites in the 3 untranslated region (UTR) of mRNAs [7]. Since their serendipitous discovery in nematodes, miRNAs have emerged as key regulators of various biological processes in animals [8]. To date, over 30000 miRNAs have been identified. Moreover, miRNAs have been applied to both basal experimental studies, as well as Phase I clinical trials for treating cancer [9]. MicroRNA-135 (MiR-135) is a member of the miRNA family that plays an important role in a variety of cancers, and has two isoforms, miR-135a and miR-135b. Although these isoforms are located on different chromosomes, mature miR-135a and miR-135b only differ in one nucleotide, which is not within the miRNA binding region. Thus, it has been reported that both miRNAs target the same genes [10,11].
Studies have reported that miR-135 is overexpressed in several human cancers and cancer cell lines, including PC [12], breast cancer [13], colorectal cancer (CRC) [14], gastric cancer [15], among others. However, the relationship between miR-135 expression and digestive system cancers remains controversial. Despite this discrepancy, an increasing number of recent studies have emphasized the role of miR-135 in predicting poor tumor prognosis. In addition, based on a bioinformatics analysis, miR-135 was considered worthy of further investigation as a potential novel biomarker for the diagnosis and prognosis of esophageal squamous cell carcinoma (ESCC) and PC [16,17]. In contrast, the prognostic value of miR-135 in gastric cancer was associated with the opposite result [18]. Therefore, the aim of the present study was to evaluate the association of miR-135 expression with the overall survival (OS), disease-free survival (DFS), and recurrence-free survival (RFS) of patients with digestive system cancers.

Quality assessment
Two investigators (Ce Chao andChen Sang) independently scored the included studies according to the Newcastle-Ottawa Quality Assessment Scale [20]. The final scores were presented from 0 to 10, with higher scores indicating better methodological quality.

Data extraction
Two investigators (Min Wang and Yanfei Li) independently extracted all the data from eligible studies. The following data were extracted: the first author, year of publication, type of cancer, size of sample, country of origin, follow-up time, detection methods, treatment measure, patient survival, and statistical methods. If the survival data were not reported, it was extracted from Kaplan-Meier curves for further calculation. Finally, the third investigator was consulted to review all the data .

Extraction and analysis of The Cancer Genome Atlas datasets
The level of miR-135 expression and clinical data for The Cancer Genome Atlas (TCGA) digestive system cancer (including cholangiocarcinoma, colon adenocarcinoma, esophageal carcinoma, liver hepatocellular carcinoma (HCC), pancreatic adenocarcinoma, rectum adenocarcinoma, and stomach adenocarcinoma) were extracted from the UCSC Xena website (https://xenabrowser.net/hub/) [21]. The level of miRNA expression was measured using the Illumi-naHiseq platform. There were 1488 patients with digestive system cancers who had both survival and miR-203 expression data in the TCGA Pan-cancer datasets for this subject [22]. The level of miR-135 expression was distinguished based on receiver operating characteristic (ROC) curves with a Youden's index correction [23]. The survival curves were plotted and tested using the Kaplan-Meier method and a log-rank test using Stata software version 12.0 (Stata Corporation, College Station, TX).

Statistical analysis
In this meta-analysis, all pooled results were analyzed with Stata software version 12.0 (Stata Corporation, College Station, TX). Pooled HRs with the corresponding 95% CIs were used to estimate the strength of the association between miR-135 and the clinical prognosis of cancer patients. The optimal method was to obtain survival data from the articles or calculate the HRs from O-E statistics and variance. Otherwise, Kaplan-Meier curves were analyzed using the Engauge Digitizer version 4.1 according to the method introduced by Parmar et al. [24]. The level of heterogeneity was assessed by the inconsistency index I 2 and χ 2 -based Q test. I 2 statistics with values >50% and a χ 2 test with P<0.05 indicated a strong heterogeneity across the studies. If the heterogeneity was not significant, the fixed-effect model was appropriate; otherwise, the random-effect model was adopted. Subgroup analysis was conducted for tumor type, digestive function, biomarker subtype, publication year, sample size, the method of HRs extraction, and quality score. The meta-analysis results were displayed as forest plots. Sensitivity analysis was performed to test the impact of an individual study on the pooled data by omitting each single study. Begg's funnel plots and an Egger's linear regression test were performed to estimate any potential publication bias [25]. False-positive report probability (FPRP) analysis and trial sequential analysis (TSA) were performed to assess the significant associations [26,27].

Prognostic value of miR-135 for the OS of digestive system cancers
In this meta-analysis, a total of 14 studies, including a total of 1225 patients, reported HRs and associated 95% CIs for OS. As shown in Figure 2, the pooled results indicate that elevated miR-135 expression was associated with a poor OS in patients with digestive system cancer (HR: 1.790; 95% CI: 1.577-2.031; P=0.000). Moreover, a heterogeneity analysis revealed that there was no between-study heterogeneity in the eight studies (χ 2 = 12.31; d.f. = 13; P=0.503; Although no inter-study heterogeneity was observed in this meta-analysis, a subsequent analysis was performed based on the cancer type, digestive function, biomarker subtype, publication year, sample size, the method of HR extraction, and quality score. As shown in Table 2 P=0.000). The association between miR-135 and the OS of patients was significant in studies with a sample size less than 100 (HR: 1.879; 95% CI: 1.571−2.247; P=0.000) and more than 100 (HR: 1.706; 95% CI: 1.427−2.038; P=0.000). We also detected a significant relationship between the level of miR-135 expression and the OS of patients in both studies with a quality score less than 6 (HR: 1.717; 95% A sensitivity analysis was performed to determine the reliability of the pooled results. The results proved to be solid by sequentially omitting each study (Figure 3). A Begg's funnel plot and Egger's test were conducted to assess the publication bias for the pooled OS. No publication bias was observed for the OS (Begg's test: P=0.622; Egger's test: P=0.932) (Figure 4).

Prognostic value of miR-135 for the DFS and RFS of digestive system cancers
Seven studies that included a total of 765 patients reported the level of miR-135 expression and DFS of digestive system cancers. Two studies consisting of 88 patients reported HRs with the 95% CI for RFS. As shown in Figure 5, miR-135 overexpression was associated with a poor RFS (HR: 3.994; 95% CI: 1.363-11.697; P=0.012), but DFS (HR: 1.482; 95% CI: 0.914-2.403; P=0.110) in patients with digestive system cancers.

FPRP analysis
We performed an FPRP assay for OS, DFS, and RFS to uncover potential false-positive results among significant associations, with a prior probability of 0.01 and an FPRP cut-off value of 0.2. Because the P-value of FPRP for OS was equal to 0, the corresponding FPRP value was infinitely close to 0 [26]. Thus, there was a significant association between OS and digestive system cancer. Moreover, there was a significant association between OS and digestive system cancer observed in the subgroups of cancer type (PC and ESCC), cancer classification (digestive gland and tract), source of control (curve and reported), sample size (large and small), biomarker subtype (miR-135a and miR-135b), publication year (2012−2016 and 2017−2019), and quality score (more than 6 and less than 6). The FPRP values for DFS and RFS were all >0.2, revealing that these associations were not truly significant (Table 2).

TSA
The cumulative Z-curve (blue line) was found to cross both the traditional boundary line and the trial sequential monitoring boundary (red line), and the cumulative information reached the required information size (RIS) for OS and RFS ( Figure 6A,C). Thus, additional studies for OS were not required. Additional studies for RFS were necessary due to the insufficient number of samples; however, the cumulative Z-curve (blue line) crosses the traditional boundary line rather than crossing the trial sequential monitoring boundary (red line), while the cumulative information reaches the RIS for DFS ( Figure 6B).

Validation by independent TCGA digestive system cancer datasets
To verify the meta-analysis results, we performed an analysis of TCGA database regarding the level of miR-135 expression and corresponding survival data. This analysis included seven types of cancer: cholangiocarcinoma, colon adenocarcinoma, esophageal carcinoma, pancreatic adenocarcinoma, liver HCC, rectum adenocarcinoma, and stomach adenocarcinoma. Moreover, the results showed that high miR-135 expression in the tissues was associated with a poorer OS for digestive system cancers (HR: 1.217; 95% CI: 1.083-1.368; P=0.002). And, the relationship between miR-135 subtype and prognosis of digestive system cancers is shown in Figure 7. In the analysis of single tumor types, we found that high miR-135 expression was associated with a poor OS for colon adenocarcinoma  Table 3). In the other tumor types, miR-135 expression was not associated with the prognosis of patients with cholangiocarcinoma, esophageal carcinoma, and rectum adenocarcinoma (Table 3). Therefore, TCGA results were consistent with the findings of our meta-analysis.

Discussion
This meta-analysis represents the first systematic evaluation of the relationship between the level of miR-135 expression and the prognosis of patients with digestive system cancers. A total of 17 studies consisting of 1470 patients were included in this meta-analysis. Our results revealed that a greater level of miR-135 expression was associated with a poorer OS and RFS in patients with digestive system cancers. Stratified analyses were performed, and significant results pertaining to the relationship between miR-135 and a poor OS were some of tumor type, digestive function, biomarker subtype, publication year, sample size, the method of HR extraction, and quality score. There was no significant heterogeneity and publication bias in this pooled result, which suggested that our results were reliable. In addition, the TSA indicated that the sample size was sufficient and the OS results were reliable due to the incalculable probability of false-positive reports in the FPRP analysis. In addition, the results were verified using TCGA database analysis. These findings suggested that miR-135 overexpression is indicative of a poor prognosis and higher recurrence in patients with digestive system cancers. However, the results required further validation for DFS and RFS due to the high probability of false-positive reports in the FPRP analysis. Stratified analyses showed that article factors, including publication time, sample size, studies quality, and methods of HRs, had no effect on the association between miR-135 expression and the OS of patients with digestive system cancers. Although mature miR-135a and miR-135b only differ regarding one nucleotide, this did not impact their prognostic effect because both of the miRNAs target the same genes. In the subgroup analysis, there was significant heterogeneity associated with the gastric cancer group due to the two contrary results of the incorporated studies. While Cheng et al. [31] found that decreased miR-135 expression was indicative of a poor OS, a study by Yan et al. [15] reported the opposite findings. However, TCGA database analysis revealed that miR-135 expression had prognostic significance for gastric cancer. Thus, future large-scale studies should be performed to verify this speculation and consolidate the pooled results. In addition, we used the data from TCGA datasets to validate the meta-analysis results. Similar results were found for pancreatic, colon, and hepatocellular cancers.
An increasing number of studies have shown that the deregulation of miRNA function is associated with an increasing number of human diseases, particularly cancer. Moreover, miRNA occupies an important position in regulating cancer proliferation, growth, invasion, and migration. Conventional non-invasive serological biomarkers, including carcinoembryonic antigen (CEA) and cancer antigen 19-9 (CA19-9), can be screened using blood tests that are currently being applied in the clinic. However, these biomarkers lack specificity for the early diagnosis of digestive system cancer [42,43]. Although some studies have found that CEA and CA19-9 could predict the prognosis of patients with digestive system cancer, including CRC and gastric cancer [44,45], the positive rate is too low. In contrast, miRNAs in digestive cancer patients may represent promising biomarkers that are stable and reproducible in both the peripheral blood and tissues. In the peripheral blood, circulating miRNAs primarily bind to proteins or are capsulated in exosomes [46]. Although measuring plasma or serum miRNA is non-invasive, the level of circulating miRNAs is susceptible to perturbations due to blood cells or hemolysis compared with tissue miRNA [47]. In addition, it is not difficult to measure tissue miRNA in patients after surgery. Moreover, the detection of tissue miRNA was more accurate, with fewer influencing factors. It was convincing that changes in miRNA expression and patient survival was consistent in both the peripheral blood and tissues. However, the results may differ on occasion; for example, when Zhang et al. [48] elucidated the prognostic value of miR-20a in human cancers, high expression of both circulating and tumorous miR-20a were associated with an unfavorable outcome only in gastrointestinal cancers. In this meta-analysis, all the included studies were related to tissue miRNAs with reliable results.
Kawaguchi et al. [49] proposed the use of miRNA as a next-generation biomarker for digestive system cancers. In addition, the authors thought that several miRNAs should be used in combination to evaluate the prognosis of one or more cancers. Recently, a database analysis demonstrated that a panel of miRNAs, including miR-135, was associated with a poor prognosis for both pancreatic and esophageal cancers [16,17,50]. Therefore, although meta-analyses have shown that several miRNAs, including miR-17-5p [51] and miR-133a [52], could predict the prognosis of digestive system tumors, a meta-analysis was required to assess the relationship between miR-135 and digestive system cancers. Moreover, miR-135 has been found to be involved in mediating the metabolic stress response in cancer. For example, miR-135 can modulate critical signaling pathways (e.g., JAK/STAT [14], P38MAPK/NF-κB [53], PTEN/PI3k/AKT [54], and RERG [34]) that indirectly regulate metabolism. Altered metabolism has long been recognized as a central hallmark of cancer. Yang et al. [12] established a key role for miR-135 in regulating glycolysis in vivo and found that miR-135 suppressed glycolysis and promoted PC cell adaptation to metabolic stress by targeting phosphofructokinase-1. In addition, a study by Wang et al. [14] found that miR-135 increased the levels of phosphorylated key kinases in the JAK/STAT pathway and functioned as a tumor promoter by targeting TRIM16. Furthermore, the deregulation of miR-135 function has been associated with both breast cancer [13] and Hodgkin lymphoma [55]. These findings provide evidence for the function of miR-135 as a potential prognostic factor and treatment for tumors.
There were some limitations associated with the present research. First, due to insufficient data, only 17 studies were recruited, comprising a total of 1470 patients in this meta-analysis. Moreover, some studies that yielded negative results were generally not published. Due to these factors, some subgroups only included one study in the subgroup analysis. Therefore, this result may be an outlier because the data were insufficiently comprehensive. Second, most of the included patients were Asian, which might reduce the universal applicability of the results. Thus, additional large-scale studies should be performed among different ethnicities. Third, the cut-off value of miR-135 was variable in different studies, which did not reach a consensus value in this meta-analysis. Fourth, HRs with associated 95% CIs and P-value extracted from the curve differed from the actual value because some of the studies did not provide HRs with 95% CIs.
In conclusion, the results of this meta-analysis indicate that miR-135 expression might be a novel prognostic factor for the survival of patients with digestive system malignancies. Further studies with a large-scale sample size should be performed to identify the association between miR-135 expression and digestive system cancer prognosis.