Background
Gastric cancer (GC), a major public health challenge, is one of the leading causes of cancer death worldwide [
1]. There are limited detection methods for early diagnosis and few effective screening procedures in some countries. The most reliable program for diagnosis is mainly based on endoscopy and biopsy [
2]. However, this is invasive and inconvenient for patients to undergo. Consequently, most patients can only be diagnosed precisely in advanced stages when the clinical outcomes are poor [
3]. Therefore, there is a great need to explore new accurate and efficient, preferentially non-invasive, markers for early detection of GC.
In recent time, accumulating evidences have suggested that microRNAs may serve as novel biomarkers for cancer detection. MicroRNAs are a class of small non-coding RNAs with intermediate posttranscriptional regulation of the target genes [
4]. A large number of studies have demonstrated that microRNAs play vital roles in a wide variety of physiological processes including cancer cell growth, differentiation, invasion, and metastasis [
5]. Moreover, a number of studies have indicated that circulating microRNAs have high degree of stability and tolerance even under unfavorable physiochemical conditions including extreme variations in pH, temperature and freeze–thaw cycles [
6]. It is also promising that microRNAs have outstanding stability in multiple clinical samples including plasma, serum, feces and tissue, which enables them to be detectable effortlessly [
7]. Given their critical involvement in the vital biological processes and perfect biomarker features mentioned above, microRNAs could be considered as good candidates for using as non-invasive markers, and the application of them as biomarkers for early detecting GC is viable [
8].
As one of the most representative microRNA biomarkers, microRNA-106 (miR-106) has been extensively studied by a great number of researches in several cancers. MiR-106 belongs to the miR-17 family, one of the most common studied onco-microRNA groups, which includes miRs-17, -20a, -20b, -93, -106a and -106b. MiR-106a is a member of the miR-106a-92 cluster located on chromosome Xq26.2 while miR-106b is located at 7q21 [
9,
10]. There have been several studies indicating that both of investigated miR-106 could be expressed in the same individuals of gastric tumour tissues. Several studies have previously reported that circulating miR-106 could specifically serve as a pivotal and promising biomarker for GC [
11]. Nevertheless, the suitability of circulating miR-106 in early detection and diagnosis of GC remains inconsistent due to different sample sizes, disease statuses, sample sources, detection methods, and other uncontrolled factors. Moreover, the potential molecular mechanism of miR-106 is still poorly understood for the present insufficient knowledge.
In the present study, a comprehensive meta-analysis was performed to obtain a better understanding of the clinical feasibility of miR-106 as promising biomarker for early detection and diagnosis of GC. By focusing not only on a single miR-106 marker, we explored whether combination biomarkers based on miR-106 are more effective than individual miR-106. Furthermore, an integrative bioinformatics analysis was carried out to evaluate the functions of miR-106 at the systems biology level.
Discussion
Over the decades, microRNAs have gained great attention in scientific researches for cancer detection, diagnosis, and treatment as they possess perfect biomarker characteristics and are highly involved in the occurrence and development of a variety of cancers [
35]. As one of the most researched microRNAs, miR-106 has been suggested by emerging evidences that it could be a novel potential biomarker for GC detection. However, the detection accuracy was inconsistent among a series of quantitative analyses. These conflicting conclusions prompted us to employ this comprehensive and up-to-date research so that a conclusion on the diagnostic power of miR-106 for monitoring GC can be drawn. Meanwhile, we conducted an integrative functional analysis of miR-106 to understand the question why it could help distinguish GC patients from normal controls.
In the present study, we found that miR-106 achieved the overall pooled sensitivity of 0.71, specificity of 0.82, and AUC of 0.80, indicating a moderate overall accuracy. Circulating miR-106 as a more researched diagnostic marker in GC detection compared with other sample sources yielded a pooled sensitivity of 0.71, specificity of 0.82, and AUC of 0.81. Subgroup analysis indicated that sample type and sample size may influence the diagnostic accuracy. Specifically, it was revealed that plasma-based assays and miR-106 assays with a large sample size had significantly better overall diagnostic accuracy than serum-based ones and miR-106 assays with a small sample size, respectively. Meanwhile, although located in different chromosomes, miR-106a yielded a similar diagnostic accuracy compared with miR-106b.
Up to now, most attention on biomarker prediction in GC has been absorbed in single biomarkers. Actually, single biomarker is hard to reveal GC evolutionary process at the systems biology level as GC is a highly heterogeneous disease. On the contrary, combination markers may be more reliable with greater power for explaining the internal mechanisms of GC [
36]. Therefore, we performed an analysis for miR-106-related combination markers in GC to investigate whether they were more powerful than miR-106 alone in detecting GC. It is worth noting that serum miR-106-related combination markers had a higher level of predictive power than serum miR-106 alone. However, it is difficult for us to conduct further investigations due to the limited number of studies enrolled in the analysis for miR-106-related combination markers.
We also performed integrative and comprehensive bioinformatics analysis to explore the function of miR-106 at the systems biology level. Most GO terms enriched by miR-106a and miR-106b target genes were both significantly associated with the regulation of transcription, apoptotic process at the BP level, basic cell structures at CC level along with the binding functions such as protein binding, DNA binding, transcription factor binding at MF level. Furthermore, the enriched KEGG pathways of miR-106a and miR-106b target genes were approximately the same, including pathways in cancer, p53 signaling pathway, cell cycle, TGF-beta signaling pathway and proteoglycans in cancer, which were highly associated with the occurrence and development of GC. Interestingly, although the different locations on chromosome, the targets of miR-106a and miR-106b fall into the similar functional modules, pathways or networks and then become more consistent when enriched to systems biology levels. In general, functionally concerned genes often emerge a coordinated expression to exert their roles in the same functional modules, indicating that miR-106a and miR-106b may have a synergistic effect in the initiation and progression of GC. The above results not only demonstrated the robustness of our study but explained why miR-106 could serve as a promising biomarker for detecting GC to some extent.
To further reveal the correlations among the target genes of miR-106a and miR-106b, we performed the PPI network analysis. Through PPI network construction, a series of hub genes were screened by three different network analysis methods. In our study, it was revealed that these key target nodes regulated by miR-106a and miR-106b both participated in FoxO signaling pathway, pathways in cancer, PI3K-Akt signaling pathway, cell cycle, and p53 signaling pathway. What’s more, module analysis of the PPI network revealed that the most significant modules of miR-106a and miR-106b targets network were both associated with p53 signaling pathway, pathways in cancer, FoxO signaling pathway, microRNAs in cancer, PI3K-Akt signaling pathway and cell cycle. PPI network analysis including hub genes identification and module analysis revealed the function of miR-106 again.
Based on above results, we found that several pathways were repeatedly mentioned in KEGG pathway analysis enriched by all the miR-106a and miR-106b targets, key hub targets and network modules, including p53 signaling pathway, pathways in cancer, FoxO signaling pathway, PI3 K-Akt signaling pathway and cell cycle. Pathways in cancer consist of several well-known signaling pathways including TGF-β, MAPK, Wnt and p53, which play important roles in cell apoptosis, proliferation, differentiation, invasion and metastasis. The well-studied p53 pathway, perhaps the most vital determinant of carcinogenesis, has been inextricably linked to establishment and progression of almost all types of cancer including GC [
37,
38]. Cell cycle, another very important signaling pathway, contributes to the malignant progression of various human cancers including GC due to the involvement in cell growth, differentiation and apoptosis, as well as cancer development and metastasis [
39]. Recent studies have proposed that the activation of the PI3K/Akt pathway may be responsible for the tumorigenesis by playing a pivotal role in control of cell cycle and survival of cell [
40]. The information gathered so far indicates that FoxO signaling pathway could play vital roles in mediating apoptosis and thus determines cell death and survival [
41]. In short, all the above pathways have been verified by the published literatures involved in the tumorigenesis and progression of GC, which may provide new ideas for the molecular mechanisms of miR-106 in GC.
Although mounting evidence from diagnostic tests indicated miR-106 as a promising GC marker, difficulty still remains for its application to clinical practice. There are several points we can do to optimize the miR-106 assay. Firstly, an appropriate standard cut-off value, consistent detection and normalization methods for miR-106 expression are required. Secondly, it was revealed from our results that plasma miR-106 may be a more powerful marker for detecting GC compared with serum miR-106. So plasma could be selected as the suitable sample source for further detection. Thirdly, as indicated in our study, sample size influenced the sensitivity and specificity. Larger sample size exhibited higher diagnostic accuracy. Thus, further large-scale prospective studies are warranted to develop integrative diagnostic models with more appropriate and better prediction capacity. Fourthly, although miR-106a and miR-106b are members of different paralogous clusters and located on different chromosomes, there has been some evidence in the literature, that both these two microRNAs can co-expressed in gastric tumor tissues. Based on our results, both miR-106a and miR-106b could be evaluated in diagnostic samples for diagnostic purposes of gastric cancer. In addition, combination biomarkers, which are combinations of several markers, have been shown to improve the prediction accuracy compared with single biomarker. According to our findings, single miR-106 was significant but not strong enough to undertake early diagnosis, while miR-106-related combination markers improved the diagnostic accuracy. The combination of miR-106 and other microRNAs may be the right way to solve the limited accuracy. Moreover, it has been reported that combination of protein-biomarkers and microRNAs may be an effective way to improve the diagnostic accuracy [
42]. So more attempts are required for evaluating the combination biomarkers in the further study.
Our study had several important strengths. First, we carried out a relatively thorough systematic search and applied a comprehensive analytic approach to investigate the diagnostic power of miR-106 in patients with GC. Next, we evaluated the diagnostic value of miR-106-related combination markers in GC for the first time. It was suggested that the combination of miR-106 with other microRNAs improved the diagnostic accuracy, which may provide a novel potential tool for progress in a clinical context. Moreover, we performed integrative and comprehensive bioinformatics analysis to explore the function of miR-106 at the systems biology level, explaining the reason why miR-106 could be used in the diagnosis of GC. However, the power of our study was limited by a few factors. Firstly, most studies in the diagnostic tests enrolled healthy participants as controls and were not blind in design, which limits the diagnostic performance. Secondly, some key information including stage of cancer, sex proportions and age distributions was not known, so further analysis could not be carried out. Thirdly, there were no studies investigating the non-Asian population, which may cause potential heterogeneity from ethnicity. Fourthly, the sources of sample were inconsistent including plasma (n = 5), serum (n = 5), gastric juice (n = 1) and tissue (n = 1). Accordingly, subgroup analysis by specimen could not be performed for the limited individual sample size. In addition, the number of studies and sample sizes enrolled in the analysis for miR-106-related combination markers are limited, which make it difficult for us to conduct further investigations. As the miR-106 combination markers are all different in all six studies, it still remains an open question which should be combined with miR-106 for improving the diagnostic power.