Although miR-31 was expressed at relatively low levels compared with some of the other candidates, high expression was associated with advanced tumor stage at diagnosis, and particularly with pT-stage, in accordance with previous results [9, 10]. There are multiple predicted targets for miR-31, but few have been fully validated at present. One proposed target of miR-31 that has been experimentally investigated is Special AT-rich Binding protein 2 (SATB2), which is involved in transcriptional regulation and chromatin remodeling . In an immunohistochemical study performed in 146 colorectal tumors, low expression of SATB2 was associated with metastasis development and poor prognosis . Another target that has been shown to be regulated by miR-31 is the T lymphoma Invasion And Metastasis gene 1 (TIAM1), which is a guanidine exchange factor for Rac GTPase and when over-expressed, it prevents TGF-β and TNF-α dependent motility and invasion in CRC cell lines . The postulated effects of miR-31 on SATB2 and TIAM1 are consistent with the associations between miR-31 expression and advanced tumor stage, observed by us and others, but clearly, the regulatory activity of miR-31 is still incompletely understood in CRC.
MiR-92a was included in the analyses because it has been proposed as an early-detection biomarker in plasma and stool [13, 21]. In general, one would expect an early-detection biomarker to be ubiquitously expressed in the tissue of interest, and although several tumors in our study had relatively high levels of miR-92a, low levels were found in a substantial proportion of the samples. Also, over-expression of miR-92a has been found in other cancer types, such as hepatocellular carcinoma and leukemia [22, 23], which suggest that further evaluation is necessary to determine its specificity and sensitivity as an early-detection biomarker. Although miR-92a was not primarily included in this study for its prognostic relevance, it was recently proposed as a key oncogenic component of the miR-17-92 cluster through targeting and down-regulating the proapoptotic protein Bim in CRC, suggesting that the functional role of miR-92a in CRC should be further elucidated .
MiR-21 is one of the more extensively studied miRNAs in CRC and was included in our study because of its proposed association with advanced tumor stage and outcome in CRC [12, 14]. In the present work, miR-21 exhibited the highest relative expression and the widest expression range of the examined candidates, but no significant associations with clinicopathological data or outcome were found. Although some investigators have identified this miRNA as clinically relevant, other exploratory studies of miRNA expression in CRC have not been able to verify these findings [25–27]. It has been speculated that discrepancies might be explained by the composition of patient cohorts, particularly regarding tumor localization, as the association between miR-21 and survival has primarily been documented in colon cancer . However, in our cohort no differences were found when comparing the clinical relevance of miR-21 expression in colon and rectum cancer. In most of the previous studies, miR-21 expression was reported relative to paired normal tissues, whereas only tumor tissue was available from our patients, which might influence interpretation of results . However, among the reports that did not identify miR-21 as relevant for outcome in CRC, both analysis of tumor tissue alone and paired tumor and normal samples were used, suggesting that this may not be the only explanation for the discrepancies.
When the primary objective is to identify cancer specific molecules, the inclusion of normal tissues is necessary, whereas, in the current project the aim was to evaluate previously identified potential biomarkers, which is a different setting. Importantly, normal tissue is often not obtainable for analysis, and expression of molecular targets in normal tissues might vary considerably between patients, and not necessarily in concert with the corresponding tumor sample. Thus, it is probably both practicable and necessary to develop assays that are independent of normal tissue. Another related challenge concerns the definition of biologically relevant cut-off levels, which have not been determined for specific miRNA in different tissues. We explored multiple cut-off levels, but associations with clinicopathological parameters and outcome for all the candidates remained relatively similar.
MiR-101, miR-145 and miR-106a have previously been associated with cancer-relevant biological processes, such as growth, proliferation and inhibition of apoptosis, or with clinical outcome in CRC [7–9], but few associations with clinicopathological parameters or outcome were found in our cohort. MiR-101 was hardly detectable in tumor samples, which is in accordance with its proposed function as a tumor suppressor that is lost during tumorigenesis. Interesting recent findings in pancreatic cancer suggest miR-101 as a key regulator of stem cell protein markers; its loss favoring the stem cell phenotype and its re-expression constituting a possible therapeutic strategy . Down-regulation of miR-145 was also identified as an early event in CRC carcinogenesis, which might explain why associations with clinical variables in invasive tumors were absent in our tumor panel. The biological relevance of miR-145 in CRC has, however, been repeatedly confirmed, and this miRNA is also being explored as a therapeutic target [30, 31]. MiR-106a was in a recent review identified as consistently up-regulated in CRC (relative to normal colon) which would be in agreement with our findings . It has also been identified in stool samples in CRC patients, and has been suggested as an early detection biomarker , but even if extensively studied in several cancer forms, its function and clinical relevance remain unclear.