Melanoma is the deadliest form of skin cancer and has increasing rates of incidence and mortality in many countries (cancer.org). Recently, we identified a plasma-based 38-microRNA signature of cutaneous melanoma using samples from patients with melanoma with stage 1a–IV disease and nonmelanoma controls (MEL38) 1. The algorithm was validated with independently generated data sets representing melanoma or nonmelanoma control blood, melanoma cell lines and formalin-fixed, paraffin-embedded skin biopsy tissue.
The MEL38 score ranges from 0 to 10, positively correlating with the probability of a melanoma diagnosis. The AUC of the 38-gene score was reported as 0.79–0.94, depending on the patient population, specimen type and genomic platform used. The 38 genes in the signature have documented roles in angiogenesis, invasion, drug resistance and tumour suppression. Additional clinical and technical validation studies are under way to more accurately define diagnostic performance and clinical utility of MEL38.
One potential clinical application of MEL38 is in the post-treatment setting, where it may be used to verify that the tumour has been completely excised and/or there are no other malignant lesions. Patients with melanoma report high degrees of uncertainty, anxiety and depression about the success of their initial treatment and long-term prognosis 2. A genomic biomarker that could be used after melanoma treatment may assist in managing these valid concerns as well as providing a new tool for doctors to monitor disease recurrence.
Changes in circulating microRNA levels following melanoma resection have only recently been described. In 2016, Latchana et al. 3 used an 800-microRNA panel to screen plasma from six patients with stage III or IV disease, before and 3 weeks after surgery. Analysis of variance revealed 25 microRNAs with differential expression between time points; however, no classification algorithm was developed or applied. Friedman et al. 4 evaluated a panel of seven microRNAs, (chosen for previously supported roles in cancer progression and/or diagnostic utility) in 10 matched serum samples collected before and 1–2 weeks after resection of primary melanoma. The mean expression level of all seven genes decreased in the postresection sample; however, only two were statistically significant.
The goal of this study was to perform a preliminary evaluation of the potential of the MEL38 algorithm as an objective, quantifiable marker of melanoma treatment response.
Patients and methods
Plasma from four patients with melanoma was collected into EDTA tubes before surgical excision (T1) and again 12–14 days later during a follow-up visit (T2) by Cureline Inc. (Brisbane, California, USA). Two nonmelanoma control samples were also obtained. Specimen and clinical data are shown in Table 1. All biopsies were verified as having clear margins by the referring pathologist. Written informed consent was obtained from each patient.
MicroRNA isolation and Nanostring profiling was performed by Canopy Biosciences (St Louis, Missouri, USA) as previously described 1. Spike-in oligonucleotides corresponding to miR-254 and miR-414 were added to each sample to control for intersample variation.
Data analysis was performed also as previously described, including negative and positive control adjustment and scaling to housekeeping gene expression. Data processing was performed using nSolver 4.0 (Nanostring Inc., Seattle, Washington, USA) and R-studio 5. The MEL38 gene support vector machine classification algorithm was applied to the corresponding genes from each sample and the scores scaled to between 0 and 10. Individual samples with fewer than 50% of the MEL38 genes detected above background were processed with the Bioconductor ‘impute.knn’ function with default settings 6.
Nanostring microRNA profiles from four patients with melanoma before and after lesion excision and two nonmelanoma controls were generated, resulting a total of 10 samples. Before spike-in gene normalisation, the raw data were visually examined (Fig. 1). Spike-in mir-414 exhibited substantially lower and less variable expression than mir-254, suggesting an error in the design of this oligonucleotide or in the laboratory processes. The raw data for mir-254 closely resembled the intensity and intersample variation observed in our earlier work; therefore, mir-254 alone was used for normalisation in this study.
The MEL38 scores generated from plasma collected at the time of melanoma diagnosis (T1) were significantly higher than the two normal nonmelanoma controls, as shown in Fig. 2. The mean MEL38 score for the pre-excision plasma was 8.84, compared with 3.95 for the control samples (t-test P<0.001).
The mean MEL38 score of the postexcision (T2) plasma samples was 7.47. A t-test of T2 versus nonmelanoma control samples indicated the difference was also statistically significant (P=0.011), although to a lesser extent than the T1 pre-excision samples, reflecting the more ‘normal-like’ status of the T2 samples.
Next, a one-sided paired t-test was performed on the T1 and T2 MEL38 scores from all four patients. The mean MEL38 change between patient-matched time points was -15%, which was statistically significant based on a one-sided paired t-test (P=0.030). A general linear model was applied to the T1/T2 data points to test the association of patient age, tumour stage and days between plasma collections with the MEL38 score. None of the variables achieved statistical significance (data not shown) with the T1 versus T2 score being closest at P=0.064. Overall, these data suggest a ‘normalisation’ of the circulating microRNA profiles of these four patients after surgical excision, as quantified by the MEL38 signature.
At the individual microRNA level, a number of biologically relevant expression changes were observed between the two collection time points further supporting the hypothesis of MEL38 normalisation in response to successful treatment, in this case surgical excision with clear margins. Some of the individual genes present in MEL38 with notable changes in expression include the following:
- hsa-miR-548l: which increased 2.1-fold between T1 and T2 samples and has been demonstrated to have tumour suppression functionality in lung and liver cancer 7.
- hsa-miR-152-3p: expression increased 1.8× between T1 and T2, has been shown to regulate the PI3K (phosphatidylinositol 3-kinase)-AKT pathway and to be more highly expressed in melanoma cell lines with a less invasive characteristics 8.
- hsa-miR-520d-3p: has been shown to be upregulated during the early stages of melanoma progression and, in this study, decreased by 1.6-fold between T1 and T2 8.
- hsa-miR-4532: which is upregulated in two melanoma cell lines resistant to BRAF inhibitor treatment versus an BRAF inhibitor sensitive cell line, decreased by 2.7-fold between time points 9.
This pilot study represents for the first time that a validated, multi-gene biomarker of melanoma has been applied to patient samples obtained before and after melanoma treatment (excision). Although caution should be taken when interpreting results from studies with small sample sizes, the results suggest that a patient’s circulating microRNA profile does revert back to a more ‘normal-like’ MEL38 score after the successful removal of the tumour.
In the four patients analysed in this study, a mean reduction of the MEL38 score of 15% was observed between diagnosis and a 12–14 day follow-up visit. The differences in MEL38 between T1 and T2 time points were significantly different, and each time point was significantly different to normal nonmelanoma control samples. Although we do not yet have data concerning the biological and technical reproducibility of the MEL38 signature, other multi-index genomic assays report a technical error of 1–5% 10–12. Combined, these observations suggest that at 12–14 days after excision, the circulating microRNA profile of a patient with melanoma with a completely excised tumour has been altered, but not yet returned to a completely-normal-like state. Further studies using additional time points with larger numbers of clinically and pathologically diverse patients are warranted, possibly including treatments other than excision (e.g. adjuvant therapy).
There was no apparent association with melanoma stage and the degree of MEL38 reduction between T1 and T2, although larger studies are needed to determine the reliability of this observation. The MEL38 algorithm was trained as a binary classifier (i.e. negative or positive for the presence of melanoma) and our previously published validation work showed that clinical stage at diagnosis was not significantly associated with the value of the score. It is possible that in future studies, an association between stage and length of time required to achieve MEL38 normality may be detected. Other factors such as patient age, comorbidities, immune status or previous history of melanoma may also be involved in modulating the post-treatment MEL38 signature.
The individual microRNAs that comprise MEL38 have documented roles in melanoma development, progression or treatment response. When the T1 versus T2 changes were evaluated in the context of gene function, a number of interesting trends were apparent. For example, a number of microRNAs with tumour suppression abilities had increased expression after surgical excision. Conversely, microRNAs associated with the early stages of invasion and treatment resistance were decreased at the second time point as compared with the first.
In addition to the challenge of a limited sample size, another caveat to these data is the reliance of only one spike-in normalisation gene, owing to the detection failure of mir-414. As Nanostring microRNA profiling benefits from per-sample scaling to the geometric mean of multiple spike-in genes, there is a possibility of inadequate or excessive normalisation being applied from the use of only one control gene. Future MEL38 validation studies should include a larger number of spike-in genes and possible use of control samples to avoid having insufficient benchmark data for optimal normalisation.
The data suggest that the circulating microRNA profile of a patient with melanoma, as measured by the MEL38 signature, becomes more ‘normal like’ after melanoma excision. The ability of a biomarker to measure a patient’s response to melanoma treatment may allow doctors to further personalise treatment options and follow-up schedules as well as reduce patient anxiety.
Conflicts of interest
The authors of this study are the founders of Geneseq Biosciences, a Melbourne, Australia-based molecular diagnostic company. Provisional patent applications have been filed on the methods used to develop and apply the techniques contained in this study.
1. Van Laar R, Lincoln M, van Laar B. Development and validation of a plasma-based melanoma
biomarker suitable for clinical use. Br J Cancer 2018; 118:857–866.
2. Beesley VL, Smithers BM, Khosrotehrani K, Khatun M, O’Rourke P, Hughes MC, et al. Supportive care needs, anxiety, depression and quality of life amongst newly diagnosed patients with localised invasive cutaneous melanoma
in Queensland, Australia. Psychooncology 2015; 24:763–770.
3. Latchana N, Abrams ZB, Howard JH, Regan K, Jacob N, Fadda P, et al. Plasma microRNA
Levels following resection of metastatic melanoma
. Bioinform Biol Insights 2017; 11:1177932217694837.
4. Friedman EB, Shang S, de Miera EV, Fog JU, Teilum MW, Ma MW, et al. Serum microRNAs as biomarkers for recurrence in melanoma
. J Transl Med 2012; 10:155.
5. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2017. Available at: https://www.R-project.org/
6. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004; 5:R80.
7. Liu CH, Zhou R, Yang LF. Hsa-mir-548l inhibits the invasion and metastasis of non-small-cell lung cancer by targeting ERK1 and PKN2. Am J Respir Crit Care Med 2013; 187:A3437.
8. Mueller DW, Rehli M, Bosserhoff AK. miRNA expression profiling in melanocytes and melanoma
cell lines reveals miRNAs associated with formation and progression of malignant melanoma
. J Invest Dermatol 2009; 129:1740–1751.
9. Kim JH, Ahn JH, Lee M. Upregulation of microRNA
-1246 is associated with BRAF inhibitor resistance in melanoma
cells with mutant BRAF. Cancer Res Treat 2017; 49:947–959.
10. Van Laar R, Flinchum R, Brown N, Ramsey J, Riccitelli S, Heuck C, et al. Translating a gene expression signature for multiple myeloma prognosis into a robust high-throughput assay for clinical use. BMC Med Genomics 2014; 7:25.
11. Glas AM, Floore A, Delahaye LJ, Witteveen AT, Pover RC, Bakx N, et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006; 7:278.
12. Baehner FL. The analytical validation of the Oncotype DX Recurrence Score assay. Ecancermedicalscience 2016; 10:675.