We present the first study demonstrating the potential prognostic value of serum miRNAs in melanoma. Specifically, we provide evidence that serum miRNAs show promise as biomarkers for improved identification of high-risk melanoma patients at the time of primary diagnosis. By doing so, we introduce biological subclassifiers that could add to the utility of clinical prognostic indicators and help account for latent heterogeneity in melanoma. This is increasingly important as better adjuvant therapies are developed, as it would enable clinicians to target those likely to recur for aggressive treatment.
We identify miRNAs that are potential markers of disease progression in melanoma patients. In other malignancies, such as colon, liver, testicular, prostate, and ovarian cancers, serum tumor markers have become part of the standard of care in surveillance for recurrence, at times used in lieu of periodic imaging in asymptomatic patients . In melanoma, however, there are no widely available circulating markers that can facilitate early detection of relapse. Here, we show support for an association between serum miRNAs and tumor burden, a feature that if further developed, could be exploited by incorporation of serum miRNAs assays in routine surveillance for melanoma recurrence.
While there are several tissue-based and in vitro studies of miRNA dysregulation in melanoma [28, 32, 35–39], little is known about the “source” of serum miRNAs [19, 21]. There is speculation that the source is not limited to the tumor and can include miRNAs shed from the tumor microenvironment or from circulating immune cells that also reflect the myriad of events involved in the initiation and development of malignancy. This suggests that, though tissue-based analyses are useful for understanding cancer biology, there is a complimentary role for studying circulating miRNAs as an independent source of biologically and clinically relevant information.
Growing interest in circulating miRNAs as biomarkers has resulted in several studies exploring their potential in solid tumors. However, the emphasis has been placed on elucidating the diagnostic capabilities of blood-based miRNAs [16, 17]. In melanoma, one group reported on serum miR-221 levels and demonstrated a correlation with tumor thickness . The second used array screening to identify unique miRNA expression profiles in peripheral blood cells of melanoma patients compared to healthy controls , which requires analysis of fresh blood specimens, and limits the practicality of the approach.
While we are the first to identify the prognostic relevance in melanoma of three of the miRNAs included in the recurrence risk signature (miR-199a-5p, -33a, -424), the biological function of all the identified miRNAs have begun to be elucidated in tissue and in vitro studies. miR-150 directly targets MUC4 in pancreatic cancer cells, an aberrantly overexpressed transmembrane mucin promoting growth, invasion and metastasis. miR-150 overexpression inhibits growth, clonogenicity, migration and invasion, and enhances intracellular adhesion in pancreatic cancer cells . Clinically, miR-150 was elevated in melanoma tissues of patients with longer post-recurrence survival , also supporting a tumor suppressor role. Conflicting data exist regarding a pro-proliferative effect of ectopic expression of miR-150 in gastric cancer both in vitro and in vivo, at least in part due to repression of the tumor suppressor EGR2, resulting in translational arrest . Our findings are in line with an oncogenic role for miR-150, with higher circulating expression of miR-150 in patients with a high recurrence risk. However, our findings can also be due to the possibility that serum miRNAs reflect the systemic immune response as miR-150 has a role in the modulation of the T-cell development. Specifically, via modulation of NOTCH3, overexpression of miR-150 has adverse effects on T-cell proliferation and survival, resulting in decreased antitumor immunity and subsequent progression .
In our model, patients with high recurrence risk had lower levels of miR-15b, similar to findings in a recent study showing an association between reduced miR-15b expression, chemotherapeutic resistance and poor prognosis in patients with tongue squamous cell carcinoma . The same study identified BIM1 as a functional target of miR-15b, through which BIM1 overexpression due to reduction of miR-15b regulates epithelial to mesenchymal transition and chemoresistence. Also in line for a tumor suppressor role for miR-15b is its direct regulation of the critical anti-apoptotic Bcl-2 protein . On the other hand, a previous study in melanoma demonstrated that miR-15b downregulation in miR-15b high melanoma cell lines resulted in decreased proliferation and increased apoptosis and clinically, found increased tissue-based expression of miR-15b in melanoma FFPE samples from patients with shorter recurrence-free and overall survival . However, these differences may be attributable to host-specific effects reflected by measurement of circulating miRNAs as opposed to solely tumor-specific effects.
While we are the first to demonstrate the prognostic potential of miR-199a-5p in melanoma, with higher expression in high recurrence risk patients, its diagnostic potential was supported by a study that found miR-199a-5p overexpression in blood cells of melanoma patients compared to healthy controls . Our data, taken with a recent study that found miR-199a-5p was one of several miRNAs overexpressed in a small sample of brain-metastatic compared to primary colorectal carcinomas , invites speculation that miR-199a-5p has a role in the process of dissemination beyond its role in carcinogenesis, which could in part be due to modulation of Brm-type SWI/SNF activity . Future studies will be needed to further define the role of this miRNA in cancer metastasis, and specifically in melanoma.
Strong evidence exists supporting a tumor suppressor function for miR-33a via repression of proto-oncogene Pim-1  and inhibition of expression of cyclin-dependent kinase 6 (CDK6) and cyclin D1 (CCDN1) , both of which results in reduced cellular proliferation and cell-cycle progression. Though not previously studied in melanoma, our data is also suggestive of a tumor suppressive function for miR-33a, with lower levels of miR-33a associated with high recurrence risk scores. As with miR-33a, the prognostic or functional relevance of miR-424 has not been studied in melanoma, but roles have been previously defined for miR-424 in HIF-1α/HIF-2α mediated angiogenesis  and regulation of monocyte/macrophage differentiation . Given these roles of miR-424 in tumor formation and dissemination (i.e. angiogenesis) and in systemic immune responses, it is not surprising that, in measuring circulating miRNAs, we found patients with high recurrence risk to have elevated levels of miR-424 in the recurrence risk model. Though the biological role and mechanistic relationship of each of these components are not completely understood, particularly as they relate to melanoma, these miRNA prognostic classifiers can still be clinically useful .
The significant post-operative reduction and subsequent elevation at recurrence of the miRNAs identified as having potential in detection of disease relapse (miR-103, -221, -222, -423-5p) is consistent with existing data delineating their roles in cancer progression. miR-103, via downregulation of the enzyme Dicer, promotes cell migration and invasion in breast cancer cells in vitro . Clinically, high levels of miR-103 are associated with metastasis and poor outcome in breast cancer patients . In vitro and in vivo studies support oncogenic roles for miR-221/222, which function to increase invasion and migration capabilities as well as proliferative growth rate in melanoma by targeting of c-kit, p27, and p57 . Though the functional relevance and target genes of miR-423-5p have yet to be uncovered, serum levels were significantly higher in gastric cancer patients compared to healthy controls and it was identified as part of 5-miRNA signature for gastric cancer diagnosis . Together with our results that demonstrate higher levels at the time of recurrence, we can begin to speculate about a possible oncogenic role for miR-432-5p.
We acknowledge that our study has several limitations. First, we performed unbiased serum miRNA expression profiling using a system demonstrated to have the highest sensitivity for analyzing serum miRNAs . With 1,500+ transcribed miRNAs identified in the human genome , the use of a panel containing 355 miRNA assays could potentially be limiting. However, the miRNAs included were identified as being a comprehensive set of miRNAs expressed in serum from extensive genome wide qPCR screening of normal and diseased samples, including melanoma patients . Next, given the expected imbalance in recurrence status among stage I and III patients, we cannot speculate about a predictive model that would work across all stages. However, we are able to prioritize a manageable panel of miRNAs for future testing that show potential as circulating biomarkers with clinical relevance in identifying primary melanoma patients at high risk for recurrence.