In this preliminary investigation of miRNA expression in human melanoma tissue treated with targeted therapies, we report significant upregulation of 15 miRNAs in metastases following combination targeted therapy with the mTOR inhibitor Temsirolimus and the anti-VEGF antibody Bevacizumab for advanced melanoma. Remarkably, the observed change was upregulation with treatment, for all 15. Twelve of the 15 have tumor suppressor activities in melanoma or other cancers (Table 1). Because this regimen has clinical activity, it is possible that these altered miRNAs may have a role in that activity.
It is also interesting that no miRNAs were significantly altered 24 hours after treatment with Temsirolimus alone, despite the critical nature of the signaling pathway targeted by Temsirolimus. Rapamycin has been shown to modulate miR-1 expression; however, this relationship was identified in differentiating myoblasts and in mouse regenerating skeletal muscle, not in human melanoma cells. More importantly, mTOR’s influence over miR-1 expression was mediated through MyoD, a transcription factor specific for skeletal myogenesis . Since miRNA expression depends on intrinsic cellular factors, this relationship is unlikely to be found in human melanoma cells. Importantly, the lack of change observed with mTOR inhibition alone is consistent with the lack of clinical activity seen with Temsirolimus alone in metastatic melanoma  and may provide some insight into the lack of clinical impact with this agent alone. It is possible that treatment with Temsirolimus alone for greater than 24 hours would alter miRNA expression profiles more significantly. However, we would expect some changes within 24 hours, especially since we have observed consistent decreases in phospho-S6Kinase in these metastases 24 h after Temsirolimus therapy .
We did not test the effects of Bevacizumab alone in the trial; so, it is possible that the significant alteration of miRNA levels seen with combination treatment is due to Bevacizumab alone rather than the combination. However, the in vitro analysis revealed minimal effect of Bevacizumab alone on miRNA expression in most of the 4 tested melanoma cell lines. In addition, single agent therapy with Bevacizumab has had variable results in melanoma patients, with response rates of 0% (0/16) and 17% (6/35) in two studies [10, 11]. mTOR is important in cell survival during stress, and VEGF blockade can induce hypoxic stress. Thus, there is rationale for the combination effect to exceed the effect of either agent alone, and this is consistent with the synergistic anti-tumor activity we have observed in vitro . Future studies may clarify the mechanism of synergy of this combination therapy.
To obtain preliminary data on whether miRNA changes observed in the tumors may be explained by direct effects on melanoma cells themselves, we analyzed the effect of either one or both agents on miRNA expression in human melanoma cell lines. These data reveal the heterogeneity of individual melanomas. However, striking and global increases in almost all 15 miRNAs are induced by combination treatment in the VEGFR2+ melanoma VMM18, where VEGF can have a direct effect on the melanoma cells themselves , with more transient effects for DM13 (also VEGFR2+). In the VEGFR2neg lines, VMM39 and DM122, upregulation of miRNAs with combination treatment may be explained by blockade of direct effects of VEGF on VEGFR3, which is widely expressed on human melanomas and is phosphorylated in both of these cell lines . Thus, by combined effect of mTOR inhibition and VEGF blockade on VEGFR2 and VEGFR3 signaling, the effect of this combination therapy may be explained in part by direct effects of both agents on melanoma cells. However, some observed changes in miRNA expression in biopsies are likely due to other cells in the tumor microenvironment as well.
miRNA expression is mediated through strict regulation of both transcription and post-transcriptional maturation . The targeted therapies used in this study may target those processes directly or indirectly. Numerous drugs alter miRNA expression in cancer, including cisplatin and 5-fluorouracil in esophageal cancer  and 1α,25 dihydroxyvitamin D3 and testosterone in prostate cancer . MiR-320a and miR-29a/b were upregulated with treatment in those studies, respectively, which was also observed in the present study. It is possible that combination Temsirolimus and Bevacizumab similarly directly induces miRNA expression. Alternatively, upregulation may represent a broad molecular response to treatment, downstream of the anti-tumor activity of the drugs. Other potential regulators of miRNA expression include the miRNA targets themselves. For example, both MYC and LIN28B negatively regulate let-7 expression at the level of transcription and processing, respectively [49, 50]. Such auto-regulatory loops likely account for the lack of precise linear inverse correlations observed with analysis of miRNA and target mRNA expression (see Additional file 5: Table S2).
A potential mechanism by which upregulation of these miRNAs may exert an anti-tumor effect involves the influence of miR-125b and miR-100 over the Akt/mTOR pathway. miR-125b was the miRNA most upregulated with combination treatment in this study (Table 1). It is a putative tumor suppressor in melanoma [27, 29], and its expression is lower in metastasizing vs. non-metastasizing melanoma . Overexpression of miR-125b can produce senescence in melanoma cells . A potential target of miR-125b is Akt3, which is overactive in melanoma and whose expression increases during melanoma progression . Downregulation of miR-125b may contribute to progression of melanoma via Akt3 upregulation . Thus, upregulation of miR-125b may contribute to melanoma regression. miR-100 was also upregulated with combination treatment. It targets mTOR and the mTOR-associated protein raptor [26, 42]. Overexpression of miR-100 enhances in vitro sensitivity to rapamycin in ovarian cancer cell lines . The observed upregulation of miR-125b and miR-100 with combination treatment may reflect additive or synergistic inhibition of the Akt3/mTOR pathway with combination treatment, mediated by three mechanisms: direct inhibition of mTOR by Temsirolimus, translational inhibition of mTOR by upregulation of miR-100, and inhibition of the Akt3 pathway by upregulation of miR-125b.
Another putative tumor suppressor among the 15 miRNAs is the let-7 family. Let-7b is significantly downregulated in primary melanomas compared to benign nevi, inhibits cyclin D1 in melanoma cells, and inhibits cell cycle progression and anchorage-independent growth when overexpressed in melanoma cells . Furthermore, the let-7 family (all with a similar sequence necessary for target recognition) suppresses the oncogene HMGA2 . The effects we observed in melanoma cells are most striking for let 7b, and its strong inverse correlation with LIN28B expression supports further investigation of this miRNA-mRNA pair as a possible mediator of therapeutic effects of this combination therapy. Definitive association of let7b and LIN28B require luciferase reporter assays; such studies have performed for human hepatocellular cancer and confirm the role of let7b as a negative regulator of LIN28B .
Results of the clustering analyses (Figure 4) suggest other miRNAs, such as miR-193a-3p and -199a-5p (Figure 4B) that may also be worth investigating as possible molecular markers of treatment response. miR-193a was found to function as a tumor suppressor in several cancer types  and is under expressed in melanomas containing a BRAF mutation . miR-199a-5p and -199a-3p are both processed from pre-miR-199a, whose promoter region is important for expression of both miRNAs . miR-199a-3p targets mTOR and c-Met in its role as a tumor suppressor in hepatocellular carcinoma and enhances susceptibility to hypoxia when its levels are restored . Thus, it is interesting that miR-199a is upregulated in responders compared to non-responders, with a combination therapy that is presumed to act in part through hypoxia-induced cell death (Bevacizumab). A larger clinical study is needed to validate whether miRNAs within these signatures may predict treatment response; however, we provide a foundation for future development of a prognostic model.
A limitation of this study is the modest number of patients enrolled and studied, which was constrained by the sample size of the phase II clinical trial . The accrual goal was 20; the actual accrual was slightly lower, at 17. Three of 17 were taken off study drugs before day 23. Overall 8 patients had biopsies at all three time point (days 1, 2, 23) – this is a small number, limited by realities of the accessibility of tumor for biopsy and the requirements to manage patient safety in accord with the protocol. However, there is substantial statistical power in the analysis because these were sequential biopsies from the same patient in each of those cases. Studies with similar and smaller sample size have also been informative for miRNA studies of human tissues [56, 57]; however, it will be valuable to test these findings in a larger dataset when available. Nonetheless, the study represents, to our knowledge, the first study of miRNA expression in melanoma metastases before and after combination targeted therapy, and one of few that evaluates tumor on repeat biopsies.
Another limitation of this study is the fact that we did not analyze the expression of all potential targets of the 12 tumor suppressor miRNAs in the treatment samples. This was beyond the scope of the present study, and instead we focused on targets likely to have a functional effect relevant to melanoma and/or the treatments used in this study. Future studies expanding on these findings may reveal other targets with functional significance with regards to upregulation of these 15 miRNAs. Furthermore, the RNA extracts used in the target analysis were prepared using two different methods. Thus, the preliminary target validations are acknowledged to be just a pilot data set. Finally, we collected both frozen tissue and FFPE tissue from these patients, but we intentionally did the work on FFPE samples because they will have broad relevance for studies involving archival specimens, and because of published work that validates the accuracy of miRNA expression in FFPE samples [58–60].
We report preliminary results that establish the basis for further expansion. In the future, an independent and larger set of samples should be used to validate these preliminary results. After validation of these findings, further functional studies are needed to determine the mechanism of induction of these miRNAs and their role in the mechanism of action of combination Temsirolimus and Bevacizumab.