Accumulating reports have indicated that miRNAs are detectable in blood and that circulating miRNAs have the potential to be new biomarkers in patients with different diseases including cancer. Circulating miRNAs must demonstrate different hallmark characteristics to considered reliable biomarkers [15, 39]: (i) stable and readily quantifiable in clinical samples; (ii) expressed by cancer cells at moderate or high levels; (iii) present at undetectable or very low levels in specimens from individuals without cancer; (iv) provide a predictive or prognostic clinical information; and (v) exhibit biological functions mechanistically linked to tumour progression.
Several studies have explored the use of miRNA expression levels in gastric tissues, sera and plasma samples to improve the diagnosis or prediction of GC [40–46]. Most reports focused on the diagnostic potential of quantifying miRNAs in blood; however, data regarding their possible prognostic role in solid tumours including GC are limited, as shown in Additional file 1, Table S3.
There are two main observations in the present study. First, miR-200c levels in the blood were significantly increased in GC patients compared with controls. These increased values were highly specific for a GC diagnosis and were associated with disease stage. Second, the blood expression levels of a single microRNA, miR-200c, provided prognostic information for patients with GC independent of a comprehensive panel of other established clinical predictors.
The miRNA-200 cluster has been shown to regulate the epithelial-mesenchymal plasticity that may be crucial at different stages of metastasis through direct targeting of the ZEB–cadherin 1 axis [21–25]. However, in vitro and functional studies have yielded conflicting results regarding the net effect of miR-200 deregulation in the metastatic process [47–50]. Recent reports have indicated that tumour colonisation at metastatic sites might be enhanced by the expression of miR-200c. The xenograft model data have suggested that although miR-200 expression can hinder the intravasation of tumour cells, those that reach the circulation may be more proficient at colonising distant organs [47, 48]. Our findings are consistent with these experimental data and with the clinical correlations observed between the up-regulation of miR-200c in tumours and poor prognosis in individuals with colorectal adenocarcinoma , oesophageal squamous cell carcinoma  and breast cancer [48, 53].
In spite of the growing evidence highlighting its relevance in various cancers, very few studies have systematically explore the role of the miR-200 family in GC. MiR-141 was significantly down regulated in gastric cancer tissues compared with pair-matched adjacent non-tumour tissues [54, 55]. Nevertheless, a recent report  found that miR-200a and miR-141 were significantly overexpressed in gastric cancer compared with those in normal gastric tissue. In addition, high miR-200a tumour expression was associated with a poor OS. Kurashige et al. have recently shown  that the downregulation of miR-200b in GC was associated with diffuse histologic type, depth of tumor, tumor size, lymph node metastasis, and lymphatic invasion. The upregulation of miR-200b was correlated with increased E-cadherin and low ZEB2. However, there were no differences in the tumour expression of miR-200c among histological types or other clinicopathological parameters.
To ascertain whether the miR-200 family expression profile can differ between GC and non-tumour mucosa and to analyse the association among miR-200a, 200b, 200c and miR-141 and histological characteristics, we used a large, public microarray database. The results of our in silico analyses demonstrated that the expression of miR-200a, -b, -c and miR-141 were similar in non-tumour gastric mucosae and gastric tumour tissue. Furthermore, miR-200a, -b, -c and miR-141 were not differentially expressed between intestinal and diffuse types of gastric carcinoma. In that sense, the miR-200 signature in GC was validated on an external data set. In our study, as shown in Table 2, there were no significant differences in the blood levels of miR-200c among histological types or other clinicopathological parameters. Similar data have been recently reported . These findings suggest that elevated blood miR-200c levels can be detected throughout the wide spectrum of gastric adenocarcinomas and therefore underscore its potential role as a clinical biomarker.
However, tumour or cellular miRNA-expression patterns can differ from miRNA patterns released into the blood [58, 59]. In addition, potential differences in the microRNAs expression profile between primary tumours and corresponding CTC or matching clinical metastases have not been systematically investigated. In that sense, the miR-200-a, -b and -c and miR-429 levels were increased in lung metastases compared to primary breast tumours . Also, the expression of miR-200c/miR-141 cluster was significantly upregulated in liver metastasis from colorectal cancer, as compared with that in primary tumours . Thus, circulating miRNAs may not always be directly associated with the changes occurring in primary tumor tissues.
When we considered the different reports regarding the potential diagnostic and clinical relevance of the blood-borne miRNA expression in cancer, a considerable degree of inter-study heterogeneity was noticed. Differences in the detection and quantification methods (microarrays, qRT-PCR and high-throughput sequencing technology), the types and numbers of miRNAs evaluated (pre-miRNA or mature form, expression profile or a single marker) and sample sources and timing (serum, plasma or blood cells obtained pre- or post-operatively), as well as in the clinical and pathological data of the included patients ought to be considered as potential causes of heterogeneity.
At present, there is no agreement on the most advantageous source from which to isolate circulating miRNA and the use of serum or plasma over whole blood for systemic miRNA analysis is debatable. One of the crucial problems is the efficient and reproducible extraction of small amounts of miRNA from plasma or serum. Therefore, higher yields of miRNAs have been consistently obtained from whole blood samples compared with matched serum or plasma samples and lower quantification cycles were performed in whole blood compared with matched serum and plasma samples in qRT-PCR experiments .
Recent reports have indicated that blood cells are major contributors of circulating miRNA . Hence, one can hypothesise that increased levels of expression of epithelial-specific miRNAs in blood, including miR-200c, might indicate the circulation of tumour cells. However, the origins of circulating miRNAs are not yet clearly understood. In theory, analysis of miRNAs obtained from whole blood may be advantageous, detecting not only those miRNA derived from circulating blood cells comprising tumour cells but also those secreted in subcellular GivenNames such as exosomes or those associated with RNA binding proteins and diverse tissues [62, 63].
Any PCR-based technique still has the disadvantage of potentially detecting minimal amounts of miRNA expression in a non-disease-specific manner. Some of the proposed miRNA cancer biomarkers have been found to be highly expressed in one or more blood cell types and plasma levels of these miRNA have been correlated to blood cell counts . Pritchard et al. reported that miR-200c was found in the blood and blood cells of controls, with the highest expression in neutrophils. However, patients with diverse metastatic cancer and severely ill conditions that could be considered as confounding factors were included in this study as “controls”. Conversely, we did not find any correlation between miR-200c levels and neutrophil counts in our series. In addition, miR-200c levels did not differ in subgroups defined according to neutrophil counts.
From a clinical perspective, assessment of miRNAs in the PB obtained after definitive loco-regional treatment reflects the “minimal residual disease” status that might better predict the clinical behaviour and/or therapeutic response. Postoperative sampling time combines, in theory, the baseline level of CNA, the potential release of CTC due to the surgical manipulation and the rapid death of in transit cells within the blood stream but with reduced survival ability. Our study shows that increased miR-200c levels are detected even in patients with very low tumour burdens (i.e., early-stage disease and after potentially curative R0 surgical resections).
Remarkably, we found that levels of miR-200c measured in the PB of GC patients independently correlate with OS and PFS. A clear clinical association of the expression levels of a single circulating miRNA (miR-200c) with poor survival outcomes indicated by multivariate analysis has been demonstrated. However, large prospective and follow up studies will be necessary in the near future to confirm the clinical relevance of circulating miRNAs, including miR-200c, as independent prognostic indicators for cancer.