PC has caused a serious health risk to people all around the world in recent decades, with a fatality rate approximately equal to the incidence rate (15). To some degree, new early detection methods appear to be more effective in improving overall prognosis for PC patients than new therapy drugs (16). The Food and Drug Administration has approved CA19.9 as the sole biomarker for clinical PC management (17, 18). However, because CA19.9 levels rise in other malignancies and even benign conditions, it has little practical utility as a PC diagnostic biomarker (19). A growing number of novel biomarkers or molecular targets for PC diagnosis, such as circulating miRNAs, have developed in recent years. Many miRNAs can serve as promising biomarkers for numerous diseases, including cancer, because of their stable existence in peripheral circulation (8, 20).
In our previous study, as the discovery phase, we combined several serum expression profiles of miRNAs to discover the most significant miRNA signatures helpful in PC diagnosis and then created innovative miRNA diagnosis models for PC using multiple bioinformatics methods. In the training phase, based on the area under the curve (AUC) scores, 27 differentially expressed miRNAs were selected. Afterwards, five diagnostic models comprising of distinct combinations of miRNAs, based on their significant expression algorithms and functional features, were introduced using multivariate cox regression analysis. In this study, we focused on the first previously proposed panel to assess the expression levels and diagnostic performance of miR-125a-3p, miR4530 and miR-92a-2-5p as a panel in plasma samples from pancreatic cancer patients and non-cancerous controls. For this aim a two-phase analysis was carried out to discover differential expression of considered miRNAs in plasma samples from PC and non-cancerous patients. In total, we analyzed 77 PC plasma samples vs 65 NCs for the testing and validation phases. RT-qPCR was used to assess miRNA signatures. As a result, we were able to confirm that PC had significantly higher levels of miR-125a-3p, miR-4530, and miR-92a-2-5p versus NC plasma samples. Regarding decrease or increase of expression patterns, this discovery does not appear to be consistent with the results of our previous bioinformatics analyses, but we should take into account that the literature has published controversial statements about the expression of these miRNAs in PC or other types of malignancies. On the other hand, preliminary research suggests that expression profiles of miRNAs derived from clinical samples such as blood or tissue vary depending on the assay method (microarray, PCR, sequencer) (21). Aside from that, we evaluated the expression of the considered miRNAs in an Iranian population as a separate genetics pool using a different way of analysis such as RT-qPCR, which is a more reliable method than Microarray. In a study by Kojima et al. (21), as well as our prior bioinformatics investigations, miR-125a-3p was found to be down-regulated in plasma samples of PC patients. Several other researches have also reported the down-regulation of this miRNA in plasma and tumor tissues of PC patients, according to the literature (22–24). However, our search for has-miR-125a-3p and Pancreatic cancer in the database of Differentially Expressed MiRNAs in Human Cancers (dbDEMC) (25) revealed that the expression of this miRNA was also up-regulated in two other studies (GSE71533 and GSE74562). Little is known about the expression levels of miR-92a-2-5p and miR-4530 in pancreatic cancer, despite the fact that some studies have reported contradictory results regarding the expression of this genes in pancreatic cancer and also other types of cancers (21, 25–28). Our Roc curve analyses also revealed that when used as a single marker, miR-125a-3p, miR-92a-2-5p, and miR-4530 had 85.3 %, 74.3 %, and 76.4 % accuracies in distinguishing PC from non-cancerous controls, respectively. We had hoped to improve the diagnostic performance by combining three miRNAs to test and validate one of the panels that we had previously shown (13). We finally achieved a more robust discriminant performance with a sensitivity of 80.4 %, a specificity of 87.2 %, and an accuracy of 86.2 % in detecting pancreatic cancer vs healthy controls by combining three miRNAs. Even though we showed that a combo of these three serum miRNAs can be used to discriminate PC, the biological functions of those miRNAs are still being debated. To that end, we conducted another round of bioinformatics analyses to assess the functional roles of the miRNAs under consideration in interactions with their target genes. The prediction of miRNA target genes revealed that these three miRNAs share some target genes. Figure 5A depicts the entire network of miRNA target genes and the patterns of their interactions with one another. KEGG pathway analysis of common target genes revealed that these genes are mostly involved in TGF-beta/signaling, Colorectal cancer, and MAPK signaling pathways. TGF signaling plays a dual role in pancreatic cancer, promoting and inhibiting it depending on the stage of the disease. Some studies have shown that TGF-β promotes apoptosis and inhibits cell cycle progression through G1 arrest, which has a potent antiproliferative activity in early stage pancreatic cancer (29, 30). However, some other studies have found that TGF- β stimulates pancreatic cancer invasion and metastasis during the advanced stages of carcinogenesis (31). TGF-signaling therapeutic approaches have already shown usefulness in pancreatic cancer for both preclinical and clinical trials (30). We identified three target genes (RHOA, MYC and THBS1) implicated in this pathway which may be targeted by miR-125a-3p and miR-92a-2-5p (Table 3). Regarding pancreatic cancer tumorigenesis, studies have demonstrated that overexpression of RHOA motivates pancreatic cancer cells to invade and migrate (32). MiRNAs has been reported to prevent the activation of human pancreatic stellate cells (hPSCs) by targeting thrombospondin 1 (THBS1) and the downstream TGF-b pathway, according to some studies (33). Among the genes mentioned, MYC is likely the most important in the carcinogenesis of pancreatic cancer. Several studies have reported the overexpression of this gene in PC and some resources proposed MYC-targeted therapy as an effective method for PC treatment (34, 35). The results of GO analysis also showed that Negative regulation of signal transduction, response to carbohydrate stimulus and response to drug are the most significant terms amongst the target genes of considered miRNAs, Most of them correlated to tumorigenesis of pancreatic cancer directly or non-directly (12, 36, 37). Regarding the direct correlation of miR-4530 expression with serum alkaline phosphatase levels that we discovered, we should note that some studies have introduced elevated serum alkaline phosphatase levels as a marker of poor prognosis of pancreatic cancer (38)(39).
Theoretically, cancer cells are thought to leak exosomes into the circulation, which contain numerous molecular markers such as proteins, DNA, and miRNAs, and transport them to distant parts of the body in order to form a new environment for future invasion (40). The differential expression of certain MiRNAs in tumor tissue samples from cancer patients should be in concordance with sera from the same patients, according to this concept; nevertheless, some research did not find this to be the case (41). As a result, other biological pathways for the secretion of circulating miRNAs should be addressed. Our findings showed that plasma samples from PC patients had up-regulation of miR-125a-3p, miR-92a-2-5p, and miR-4530, despite the fact that some researches had found down-regulation of these genes (particularly for miR-125-3p) in plasma or tissues from PC patients and introduced these miRNAs with tumor suppressive role. However, the change of this miRNAs in some other studies has produced mixed results. Even if the down-regulated levels of these miRNAs in plasma are linked to particular specific oncogenes involved in PC carcinogenesis, their role in PC has been described and has to be investigated further. In this paper, we present the first evidence of a link between increased plasma levels of miR-125a-3p, miR-92a-2-5p, and miR-4530 in PC patients. Much more importantly, our findings reveal that serum miRNAs, specifically a diagnostic score based on the combination of three predictive miRNAs, may identify patients with pancreatic tumors when compared to non-cancerous healthy controls. We believe that diagnosing pancreatic cancer with circulating markers in blood, which are relatively easy to obtain from most patients, is of higher quality, especially as an initial screening test. Overall, our study found new diagnostic circulating miRNAs in PC, while more researches with larger sample size is needed to confirm the diagnostic performance of the identified miRNAs in PC.