miRNAs have been shown to play a critical role in many biological processes, including cell proliferation, cell cycle, differentiation and apoptosis [1–5]. Their primary function was initially thought to be the direct inhibition of translation, but they are now recognized to target mRNAs for degradation . It has been suggested that the effect of a miRNA on its target mRNA depends on the strength of their binding and the degree of sequence complementarity. Under this paradigm, perfect pairing leads to mRNA degradation, while imperfect pairing results in translation inhibition . Until recently, most miRNA studies have focused on expression levels, but clinical data on miRNA activity are lacking and it is unclear if miRNA expression levels are a good surrogate for activity.
Sarcomas – a uniquely complex group of mesenchymal tumors – are perfect candidates for exploring the regulatory role of miRNAs with the aims of better understanding their biology, and developing clinical biomarkers and therapeutic targets. To our knowledge, there is limited information on the role of miRNAs in sarcoma. Subramanian et al. used miRNA expression levels to characterize various sarcoma subtypes with distinct miRNA profiles, thereby supporting the possible importance of miRNAs in the biology of these tumors .
Our goal was to determine miRNA activity in some of the most common sarcoma subtypes with a recently developed algorithm which uses sarcoma gene expression data as a surrogate. We identified several miRNAs that appear specifically deregulated in each sarcoma subtype, using normal tissue as a comparator. Despite the technical challenges associated with confirming in silico findings, we validated the deregulated activity of many of these candidate miRNAs using a paraffin-based cohort. The majority of these miRNAs were shared in all histological subtypes, suggesting that they are perhaps related to a general neoplastic transformation. Another subset, however, appeared to be unique to each sarcoma subtype. In order to further corroborate the miRNA specificity for each sarcoma subtype, we performed miRNA activity analysis using ovarian and head and neck cancer datasets and the same normal tissue cohort as a comparator. This analysis demonstrated that the majority of the sarcoma subtype-specific miRNAs were also truly unique to sarcoma subtypes. At the same time, our findings support the notion that certain common miRNA activity changes in sarcomas may be related to a general cancer phenotype as nearly all of these miRNAs were also activated in both the ovarian and head and neck tumors.
We were then interested in uncovering potential etiologies for differential activity, one example being mature miRNA sequence alterations. Using RNA-sequencing on two of the FFPE specimens from our validation cohort we found that several miRNAs which we identified as differentially activated in all sarcomas relative to normal tissue harbor possible sequence alterations. Whether this is indicative of mutation or post-transcriptional processing is unclear because we did not perform genomic DNA sequencing, but nevertheless, an impact on miRNA activity could be explained by either phenomenon. We reason that a miRNA base deletion could conceivably lead to either increased or decreased activity because target complementarity may be either increased or decreased as a result. Another explanation for differential activity could be the presence of a chromosomal translocation. We identified the chromosomal locations of miRNAs identified as sarcoma subtype-specific in our study, and we found that miR-221, which was uniquely silenced in synovial sarcoma in our analysis, is located at Xp11.3, very near the common synovial chromosomal translocation t(X;18)(p11.2, q11.2) . Rigorously investigating all possible reasons for differential activity is beyond the scope of this study, but our findings regarding potential miRNA sequence alterations suggest that mutation, post-transcriptional modification, and/or chromosomal aberrations may play a prominent role.
To explore how differential miRNA activity may manifest characteristic phenotypic states in cancer, we evaluated the relationship between miRNA activation and RAS signaling. We categorized sarcoma samples as RAS-active versus RAS-inactive using previously validated expression “read outs” of RAS activity . The data demonstrated that, in aggregate, sarcomas with active RAS were characterized by different miRNA activity profiles compared to sarcomas without active RAS and, interestingly, a subset of miRNAs appeared to “switch” activity between the two pathway “classes.” We also examined the distributions of RAS status with respect to histological subtype and found considerable variability in the rates of RAS activation. This suggests that RAS pathway activity may be sarcoma-subtype-specific per se. Performing the activity analysis separately on each of the histological subtypes revealed that significant “RAS-switching” miRNAs were present in only the non-myxoid liposarcomas and the synovial sarcomas. Interestingly, one of these miRNAs, miR-7, has been shown to promote tumorigenesis via regulation by a mechanism in which RAS signaling increases miR-7 transcription . We propose that the increased expression of miR-7 in some RAS-active sarcomas also leads to increased miRNA activity as determined by our computational approach. A very interesting finding is that many of the identified RAS-switching miRNAs have predicted mRNA targets which encode proteins in the endogenous miRNA processing machinery. In all, nearly half of these miRNAs target one or more of the processing protein transcripts, and we confirmed significantly decreased expression of these mRNAs in RAS-active tumors. We therefore hypothesize that miRNA repression of processing proteins contributes to the observed down-regulation of some miRNAs in human tumors. This seems plausible as a similar phenomenon of Dicer regulation has been described . These observations require further work to examine whether the miRNA activity changes are contributory or causal in the RAS activation process, and to examine the link between miRNA processing machinery, RAS, and miRNA activity.
In addition to exploring miRNA activity, our study addresses the question of whether miRNA expression levels are reasonable surrogates for activity in sarcomas. Our data suggest that there is an imperfect relationship between activity and expression levels, and that it may be stronger for highly variant (in terms of expression level) miRNAs. It has been suggested that dramatic changes in miRNA levels may predictably result in activity changes, but activity can change even with small changes in expression level for various other reasons [46, 47]. For instance, functional alterations of proteins that have a role in the RNA-induced silencing complex (RISC), such as Argonaute, can cause activity changes without affecting miRNA levels . miRNA mutations can also cause altered miRNA activity while leaving the miRNA expression levels measured by microarray intact. Finally, it has been shown that certain transcripts may act as miRNA “sponges,” whereby miRNA regulatory effects may be modulated without changing their expression levels . Supporting these notions is a comparison of our findings and those of Subramanian et al. based on expression levels . We found only two synovial sarcoma-specific miRNAs, miR-126 and miR-129, that have both lower expression levels and decreased activity in both studies. While this question merits further study, these observations support the notion that expression levels and target mRNA levels capture different aspects of miRNA regulatory activity in sarcomas.