Micro-RNAs as diagnostic or prognostic markers in human epithelial malignancies

The ability to examine FFPEspecimens, a universally standard histologic processing procedure, allows the expeditious discovery and evaluation of potential biomarkers, given their possible link to clinical databases with mature follow-up. Transcript (mRNA) profiling is technically challenging with FFPE samples due to significant RNA degradation during formalin fixation [41, 42], and continued deterioration with storage over time [43]. In contrast, miRs are not significantly affected by fixation, and can be readily extracted from FFPE samples due to their small sizes (approximately 22 nt in length) and remarkable stability [44, 45]. Hence, this greatly enhances the ability to evaluate miRs as cancer biomarkers, leading to a multitude of reports describing miR expressions in many epithelial malignancies, summarized as per anatomical site in Table 1.

As already mentioned, miR-21 up-regulation is the most commonly observed aberrant miR in human cancers, with oncogenic consequences [46] (Table 1). It was first reported in glioblastoma [16], but also described for epithelial cancers such as head and neck, breast, colon, lung, prostate, and others [12, 44, 47]; often associated with worse outcome [40]. Over-expression of miR-21 has been shown to increase cell proliferation, migration, invasion and survival [48]; in contrast, suppression of miR-21 induced apoptosis and decreased cell proliferation and invasion [49].

Mir-155 is another commonly dysregulated miR, wherein the majority of studies report its over-expression associated with tumorigenesis in lymphomas, breast, lung, colon, pancreatic cancers, and others [50]. Aside from these two miRs, there is usually minimal overlap of dysregulated miRs described amongt different studies, even when examining the same cancer type; the same variation as previously observed for mRNA profiling. Perhaps this might relate to multiple redundant "wirings" in human cancers [51], wherein just as four distinct mRNA profiles can all predict for breast cancer relapse [52], a similar phenomenon might also apply to miR profiles, although this remains to be definitively proven.

There is emerging interest in the investigation of miRs as non-invasive biomarkers in circulating blood. This was first described in B-cell lymphoma, reporting elevated levels of miR-155, -210 and -21 in patients' sera, with miR-21 associating with relapse-free survival [53]. In epithelial cancers, Mitchell et al. first identified tumor-derived miRs in plasma samples, and suggested that variations in miR abundance reflected tumor burden [54]. MiRs have been detected as free miRs in either plasma or serum, or contained within microvesicles such as exosomes; the latter being minute, natural membrane vesicles secreted by a variety of different cell types [55]. In addition to miRs, exosomes also carry intact and functional mRNA [56], with the probable purpose of transferring information and signals throughout the body [55]. Association of epithelial cancer and exosome miRs was first illustrated in ovarian cancer, wherein tumor-derived miR profiles strongly correlated with levels of peripheral blood-derived exosomal miRs [57]. Similar observations have also been reported for lung cancer [58, 59].

As shown in Table 2, the list of potential blood miR biomarkers is even more diverse than those from tissue studies (Table 1). The greatest degree of overlap was reported for miR-21, miR-196a and miR-210 from four different pancreatic cancer studies [60‐63]. As observed for the tissue studies, miR-21 and miR-155 are also the two most common aberrant miRs in circulation with putative diagnostic and prognostic value (Table 2). However, down-regulation of miR-155 was reported in one serum study of ovarian cancer [64]. There is some controversy surrounding miR-155; the majority of reports suggest an oncogenic role; however, in a lung cancer study, its up-regulation predicted for worse outcome for adenocarcinomas, but improved outcome for squamous cell carcinoma patients [65]. One possible tumor suppression function for miR-155 was demonstrated in miR-155 deficient mice, which appeared to reduce oncogenic translocations generated by activation-induced cytidine deaminase (AICD) [66]. Micro-RNA expression levels in circulation can also relate to hormone receptor status in that estrogen negative breast cancer sera samples had higher levels of miR-21 and miR-10b [67]; in contrast. miR-155 was detected for progesterone receptor positive patients [68].

In summary, there are multitudes of reports describing the potential value of miRs as both diagnostic and prognostic bio-markers for human malignancies. None to date, however, have been translated into clinical practice, likely a reflection of its complex biology, and lack of validation studies using appropriately-powered sample sizes.

Despite the promising data supporting the potential value of miRs as biomarkers, many challenges remain. First, robust platforms, as well as appropriate statistical and bio-computational analyses must be utilized in order to identify potential candidate miR signatures for predicting outcome. Furthermore, such candidate signatures must be validated using independent cohorts statistically powered to confirm the existence of a predictive signature. Second, the selection of the appropriate reference controls is extremely important for normalization of biological variation. Recent reports have observed that some of the commonly-utilized reference miRs, such as RNU43, RNU44 or RNU48, in fact fluctuate with the biological entity of interest [69]; hence it is critical to determine the most stable miRs for each condition under examination. Third, it is conceivable that given the "upstream" effects of miRs, and their biological complexities which we are just starting to unravel, their pattern of expression might be too subtle and variable to serve as robust predictive signatures. Nonetheless, pursuit of investigations such as prognostic signatures, or their measurements in sera/plasma are definitely warranted, particularly when using appropriately-sized population cohorts.