A coordinate deregulation of microRNAs expressed in mucosa adjacent to tumor predicts relapse after resection in localized colon cancer

Forty-eight patients with sporadic stage I-II colon cancer were considered and both tumor tissue and matched normal mucosa were collected at the time of curative surgery, see Additional file 1. The expression levels of the five selected miRNAs, already known to be modulated in advanced CRC (stage IV), were evaluated. Interestingly, our analysis confirmed that they were all significantly regulated in the early phases of the CRC tumor process. Specifically, miR-18a, miR-21, miR-182 and miR-183 were strongly up-regulated (p < < 0.001) in cancer tissue with respect to normal mucosa, whereas miR-139 was strongly down-regulated (p < < 0.001), see Fig. 1. This finding extends to stage I-II CRC the results previously obtained in stage IV CRC [5], showing that these five miRNAs accompany the CRC tumor process, from initial stages to advanced tumorigenesis.

To assess the role of tumor-associated inflammation in the modulation of these miRNAs, we also evaluated their expression levels in 10 patients affected by ulcerative colitis, a form of inflammatory bowel disease characterized by chronic and widespread inflammation of the colorectal mucosa. MiR-18a and miR-21 appeared weakly up-regulated in the inflammatory tissue vs. normal mucosa with p-value of 0.053 and 0.042, respectively. Conversely, miR-182, miR-183 and miR-139 were not significantly modulated (Additional file 1: Fig. S1). Although with a limited sample size, our data indicate that the modulations of miR-182, miR-183 and miR-139 are more specific of the tumor process. MiR-18a and miR-21, instead, appear weakly up-regulated also in inflamed bowel tissue, in line with previously published works [7, 8], thus suggesting that the strong up-regulation of these two miRNAs in the tumor tissue could be partially related to inflammation.

Based on recurrence-free survival (RFS), the 48 localized colon cancer patients in study were subdivided into a recurrent group (R, 23 patients with RFS > 55) and a non-recurrent group (NR, 25 patients with RFS < 55), see Additional file 1. The clinical characteristics of the patients are described in Table 1.

To investigate whether the selected miRNAs could be useful to predict tumor relapse, we applied the miRNA ratio approach [13, 14]. We calculated 10 ratios between the expression values of all possible miRNA pair combinations (see Additional file 1) in both the tumor tissue and the adjacent normal mucosa, and assessed their capability to predict relapse through univariate logistic regression analysis, Additional file 1: Table S1. None of the miRNA ratios resulted predictive when evaluated in the colon cancer tissue. Three miRNA ratios, evaluated in the tumor-adjacent mucosa, were found to be significant predictors of relapse within 55 months from resection: miR-21/miR-183 (p = 0.0011), miR-18a/miR-182 (p = 0.0053) and miR-18a/miR-183 (p = 0.0099), see Fig. 2a and Additional file 1: Table S2 for details about univariate logistic regression models. Corresponding areas under ROC curves were 0.83, 0.76 and 0.78, respectively, see Fig. 2b. It is noteworthy that the three significant miRNA ratios (miR-21/miR-183, miR-18a/miR-182 and miR-18a/miR-183) together with miR-21/miR-182 ratio, which appears weakly significant (p = 0.063, Additional file 1: Table S1), share a common characteristic. These miRNA ratios have at numerator a microRNA that in our setting appears associated also to inflammation (either miR-18a or miR-21) and at denominator a microRNA more specific of the tumor (either miR-183 or miR-182). We thus conclude that a coordinate deregulation of these two couples of miRNAs may be a useful predictor of relapse. To investigate how the individual miRNAs perform as predictive biomarkers, an additional univariate logistic regression analysis was executed and reported in Additional file 1: Table S3, showing that the miRNA ratio approach yields to more significant results. Moreover, a bivariate logistic regression analysis, taking as covariates all the possible combinations of miRNA pairs, was conducted (Additional file 1: Table S4). This analysis confirmed that, regardless of the chosen approach, it is the coordinated alteration of the two couples of miRNAs that we have identified to be predictive of relapse. The advantage of the miRNA ratio approach is to overcome the need for miR-200c as a normalizer.

More interestingly, the three significant miRNA ratios seem to be potential biomarkers when evaluated in the adjacent, morphologically normal, mucosa and not in the tumor tissue. This result, apparently counterintuitive, is in line with recent gene expression studies on the adjacent mucosa conducted both in CRC [6] and in head and neck cancer [15]. Indeed, it is emerging with increasing evidence that the crosstalk between tumor and microenvironment could also affect the adjacent mucosa and that also this tissue may be informative.