Genetic and signaling changes that drive melanoma development and progression can be identified through molecular profiling. These changes can be detected in the circulation in the form of circulating free DNA (cfDNA), including circulating tumor DNA (ctDNA), microRNA (miRNA) and long non-coding RNA (lncRNA) that are shed into the blood stream by cancer cells. Detection sensitivity of cfDNA is a persistent problem, however, and a number of studies have sought to optimize the yield and stability of cfDNA by comparing a range of tubes during blood collection [
48,
49], and a range of commercial cfDNA purification kits [
50,
51]. The hope from such studies is to standardize practices in the field with the aim to enhance both sensitivity and consistency.
Commercial cfDNA purification kits typically employ a spin column-based or magnetic bead-based approach. Spin columns are more time consuming and costly, but appear to be the more consistent with higher yields than the magnetic-based systems [
48,
49]. Both approaches have the capacity to process large volumes of plasma, an important consideration for maximizing sensitivity, and can be partially or fully automated, which is attractive for high throughput, especially in a diagnostic setting. However, there is currently no standard best practice for cfDNA extraction.
Circulating tumor DNA (ctDNA)
ctDNA is highly fragmented single or double stranded DNA shed by tumor cells into the circulation [
52]. ctDNA has a size distribution of 130–170 bp, which is equivalent to the size of nuclease-cleaved nucleosomes, and suggestive that cell apoptosis is the principal source of ctDNA. Nevertheless, the precise mechanism of ctDNA release remains to be determined and may potentially include tumor cell necrosis, secretion from metabolically active tumor cells, or phagocytosis of necrotic tumor cells by macrophages [
53]. ctDNA has a short half-life, ranging from 16 min to 13 h [
54,
55], due to its rapid clearance from circulation via the kidneys, liver and spleen [
56].
The utility of ctDNA in identifying heterogeneous resistance mechanisms to EGFR targeted therapy has been well outlined in non-small cell lung cancer [
57], with the National Comprehensive Cancer Network guidelines recommending the use of liquid biopsy as an alternative to tissue biopsy for initial T790M mutation testing [
58]. However, the role of identifying mutations through ctDNA guiding treatment decisions has not been established, with only case reports available [
59].
Levels of ctDNA in cancer patients are associated with disease volume and can be influenced by tumor location, vascularity and cellular turnover [
60,
61], and ctDNA is often undetectable in the majority of early stage melanoma patients [
62]. However, in late stage melanomas, longitudinal assessment of ctDNA levels, including BRAF and NRAS mutations in ctDNA, in melanoma patients receiving immunotherapy was predictive of response [
63]. A favorable ctDNA profile (i.e. undetectable ctDNA either at baseline or during treatment) was associated with a better objective response, progression free and overall survival compared to patients with an unfavorable ctDNA profile (i.e. detectable ctDNA at baseline which remained detectable during therapy) [
63]. Similarly, baseline levels of ctDNA were lower in melanoma patients with better outcomes on targeted therapy (reviewed in [
64,
65]). In metastatic uveal melanoma, ctDNA was associated with tumor burden and overall survival [
66].
In addition to ctDNA quantitation, epigenetic changes in ctDNA such as methylation can also be detected and analyzed. Epigenetic modifications of ctDNA, especially the evaluation of methylation signatures [
67], is a promising avenue for biomarker discovery [
68,
69]. The stability of CpG island methylation, and the high rate of occurrence early in cancer make methylation analysis of ctDNA a reliable and sensitive biomarker target [
67]. Analysis of methylated ctDNA requires bisulfite conversion, which involves the deamination of unmethylated cytosines to uracil to allow discrimination of unmethylated from methylated cytosines. This involves PCR-based amplification using discriminating methylation specific primers for individual methylation sites, or non-discriminating primers coupled with sequencing for a more global gene analysis [
70]. Other common methylation analysis techniques are based on the use of methylation sensitive restriction enzymes [
70]. The analysis of methylated ctDNA using methylation-specific PCR in metastatic melanoma has yielded promising associations, namely in hypermethylation of the promoter region of Ras association domain family protein 1 (RASSF1A) significantly correlating with overall survival [
71] and hypermethylation of estrogen receptor α predicting progression-free and overall survival [
72].
Many different platforms have been used to detect ctDNA including quantitative PCR (qPCR), digital droplet PCR (ddPCR) and next generation sequencing [
73] (reviewed in [
61,
64], Table
3). Several technologies have also been developed to improve detection rate, including crosslinking ctDNA to magnetic beads (BEAMing), enrichment for mutant alleles (i.e. SCODA, synchronous coefficient of drag alteration; COLD-PCR, co-amplification at lower denaturation temperature PCR) and targeted hybrid selection and capture (i.e. CAPP-Seq, cancer personalized profiling by deep sequencing) (reviewed in [
74]).
There are a number of challenges in ctDNA detection and analysis. The proportion of ctDNA is low compared to total background cfDNA, and it is imperative that white blood cell lysis, which increases the cfDNA fraction, is avoided during pre-analytical steps such as blood collection, processing and storage. Plasma is the preferred source for ctDNA compared to serum due to greater cell lysis that occurs during the clotting process [
75]. However, there is also a lack of consistency in blood processing and plasma preparation, which may affect ctDNA quantitation, especially since ctDNA has a short half-life and there is a time-dependent increase in cfDNA in blood collection tubes [
76]. Factors such as time from blood collection to plasma separation, and the temperature for storage and transportation of collected blood are crucial in minimizing cell lysis and maintaining a stable cfDNA pool. Several blood collection tubes have been manufactured from companies such as Streck, Roche, Qiagen and CellSearch which minimize cell lysis and stabilize the total cfDNA pool by the inclusion of various additives/preservatives.
Currently, detection of BRAF and NRAS mutations in ctDNA has shown significant value in predicting treatment response and outcome in melanoma [
63,
65,
77] and the recent inclusion of ctDNA analysis in clinical trials [
64] further highlights its imminent implementation in clinical practice.
MicroRNAs (miRNAs) and long noncoding RNAs (lncRNAs)
MicroRNAs are short (20–200 nucleotides) noncoding RNA molecules that regulate gene transcription processes to affect cell proliferation, apoptosis, differentiation and survival. Long noncoding RNAs (lncRNAs), spanning more than 200 nucleotides, also have direct roles in transcriptional, post-transcriptional and epigenetic gene expression modulation [
78]. Both miRNAs and lncRNAs are secreted by cells into the circulation, and unlike ctDNA, they are relatively stable as they are predominantly secreted in vesicles, or in complex with other proteins such as high density lipoprotein and RNA-binding proteins [
79‐
82]. miRNAs and lncRNAs have been implicated in regulation of tumor development, progression and metastasis, and as such, have been proposed as potential cancer biomarkers (reviewed in [
78,
83]).
Detection of miRNAs requires selective and sensitive amplification methods including isothermal exponential amplification and rolling cycle amplification, capillary electrophoresis-based assays, and use of quantum dots, Raman spectroscopy, gold nanoparticle probes and duplex specific nucleases [
84]. Currently, levels of miRNAs in serum or plasma are normalized against housekeeping control miRNAs, such as U6, miR-451 and miR-16, or with spiked-in controls. However, levels of these controls may be deregulated in cancer and spiked in controls may not be practical when dealing with large numbers of biological samples [
85]. Unbiased RNA detection methods are also required for lncRNA detection and analysis, and these typically include tiling arrays, where cDNA is hybridized to microarray slides containing overlapping oligonucleotides that cover the complete genome, serial analysis of gene expression (SAGE) and cap analysis of gene expression (CAGE), which involve sequencing of short cDNA sequences [
86].
Expression of miRNAs and lncRNAs have shown diagnostic, prognostic and predictive value in melanoma [
87]. However, it is important to emphasize that miRNAs and lncRNAs are not tumor specific and it is difficult to attribute whether changes in abundance are due to the cancer or to secondary conditions such as inflammation [
88]. Elevated levels of miRNA-221 have been observed in early melanomas compared to healthy controls and melanoma in situ and increasing miRNA-221 levels further correlated with increased stage [
89]. Additionally, a panel of five miRNAs (miRNA-150, miRNA-15b, miRNA199a-5p, miRNA-33a and miRNA-424) classified primary melanoma patients into high-risk compared to low-risk of recurrence, and dynamic changes in longitudinal samples reflected tumor burden [
90]. Several lncRNAs are also upregulated in melanoma including SPRY4-IT1, BANCR, HOTAIR, UCA1 and MALAT-1 [
91]. Levels of UCA1 and MALAT-1 were significantly upregulated in melanomas compared to normal controls, and were significantly higher at later stage (stage III and IV) compared to early stage melanomas (stage I and II) [
85]. Overall, these studies implicate miRNAs and lncRNAs as promising prognostic and predictive biomarkers for melanoma but because there have been few studies in this area, and no additional studies have been performed to validate these findings, use of miRNAs and lncRNAs as biomarkers have not been translated into clinical use.
Exosomes
Exosomes are small (30–150 nm; equivalent to viruses) membrane bound vesicles, produced by all cells and capable of transporting DNA, RNA and proteins between cells. Tumor cells actively secrete exosomes, and these can deliver tumor-specific cargo (DNA, RNA and protein) to other body sites to modify tumor survival, proliferation and treatment response [
92,
93]. Secreted exosomes can be isolated based on their physical properties using ultracentrifugation, size-based methods, precipitation-based assays, immune-affinity capture and microfluidics (reviewed in [
94]). Characterization of quality and integrity of isolated exosomes typically include transmission electron microscopy, which captures vesicle morphology and size, coupled with a complementary analysis which measures size distribution and concentration of exosomes (reviewed in [
95]).
Exosome levels can be monitored directly in the circulation via cell surface markers such as the tetraspanin proteins, CD63, CD81 and CD9. For example, ExoScreen can detect and quantitate exosome surface proteins with streptavidin-coated donor beads that capture analyte-specific biotinylated antibodies, and acceptor beads conjugated to secondary antibodies that recognize an epitope of the analyte [
96]. ExoScreen is superior to immunoblotting detection of exosomes, as it does not require exosome purification or concentration (Table
3). Exosome cargo may also serve as cancer biomarkers, and elevated levels of exosome-derived miRNA-17, miRNA-19a, miRNA-21, miRNA-126 and miRNA-149 were identified in patients with sporadic metastatic melanoma compared to healthy individuals [
97]. Further promise in this area is illustrated by the recent report that miRNA-211-5p was induced within exosomes in response to vemurafenib treatment of BRAF-mutant melanoma cells, and as such, may represent a potential biomarker or therapeutic target [
98].
Circulating tumor cells (CTCs)
CTCs represent a small proportion of cells in the circulation, and are detected at a rate of approximately one CTC per million leukocytes [
100]. The detection of CTCs is further complicated by their short half-life of 1 to 2.4 h in circulation [
101]. CTC counts reflect tumor burden and the presence of CTCs strongly correlated with poor outcome in several cancers [
102,
103]. Furthermore, changes in CTC counts during therapy have been associated with treatment response [
104,
105].
Currently, there are more than 400 clinical trials incorporating CTCs as diagnostic biomarkers for patients with advanced late stage cancers. However, CTC detection in early stage disease remains challenging [
16] despite the newly-developed technologies in cell isolation and enrichment, and analysis methods (reviewed in [
106]). CTC isolation and enrichment techniques include microfluidics-based approaches, surface marker selection of tumor cells, size-based filtration methods such as ISET (isolation by size of epithelial tumor cells) [
107], and cell exclusion through negative depletion of immune blood cells using anti-CD45 antibodies [
106,
108]. Microfluidics-based platforms utilize magnetic and/or electrophoretic separation systems to capture antibody-tagged CTC complexes [
109] but recent developments have enabled isolation based on biophysical characteristics of tumor cells. One example of method development in this area is the slanted spiral microfluidics technique, which has shown high recovery rate (>80%) of CTCs whilst depleting 99.9% of white blood cells from blood [
110,
111]. CellSearch (Veridex) is an FDA-approved technique using EpCAM coated beads to isolate CTCs from blood but this surface marker selection is limited to carcinomas that expresses EpCAM [
104], and currently, this technique is only approved for prognostic evaluation of metastatic breast, prostate, lung and colon cancer.
Circulating melanoma cells (CMCs) have been detected in blood by qPCR of melanocyte specific genes or by enrichment using melanocyte surface markers; CMCs have shown prognostic value in identifying disease stage [
112], progression [
113,
114] and overall survival [
115] but, sensitivity of CMC detection is low (reviewed in [
115,
116]). It remains to be determined whether improvements in enrichment protocols and detection sensitivity could improve CMC detection, especially in early stage melanoma, and currently there has been limited implementation of CMC analysis in the clinic.
Circulating immune cells
The presence of tumor-infiltrating immune cells predicts melanoma response to immunotherapy [
117‐
119] and the possibility of using immune cell profiling as a surrogate for tumor tissue analysis is appealing. The cellular complexity of blood requires the use of advanced technologies to detect and discriminate multiple cell populations simultaneously.
The capacity to profile different immune populations has vastly improved due to refinements in multiparameter flow cytometry and mass cytometry techniques. These technologies allow phenotypic and functional characterization of individual cells using multiple parallel tags. Flow cytometry uses antibodies conjugated to fluorochromes that bind to cell surface or intracellular markers to allow cellular characterization; these fluorochromes have now expanded to allow for routine analysis of up to 15 different parameters [
120,
121]. Mass cytometry (CYTOF, cytometry by time of flight) also profiles single cells with high throughput and multiple parameters. Instead of using antibodies conjugated to fluorochromes, these antibodies are attached to heavy metal ions that can then be identified using mass spectrometry. Using these metal-conjugated antibodies, mass cytometry has surpassed the multiplexing capacity of flow cytometry, offering up to 40 different parameters [
122,
123]. Standardized panels have now been developed for PBMCs and whole blood immunophenotyping [
124] and these panels were recently used in the diagnosis of blood-based cancers such as leukemia and lymphoma [
125]. There is also potential in combining these platforms to incorporate identification of peptide-MHC multimers in order to characterize reactivity of specific T cell subsets [
126,
127].
Several recent studies have demonstrated the utility of immune profiling to discover predictive melanoma biomarkers. In a study with 209 melanoma patients treated with ipilimumab, low absolute monocyte counts, and high absolute eosinophil count, T regulatory cells and relative lymphocyte counts were associated with a favorable outcome [
128]. Similarly, in 616 patients treated with pembrolizumab, high relative eosinophil and lymphocyte counts were associated with favorable overall survival [
129]. Immune cell phenotyping of PBMCs from patients with stage IV melanoma before and after treatment with anti-PD-1 therapy identified a reinvigorated exhausted CD8+ T cell subset (Eomes
hi and Tbet
lo) expressing the proliferative marker Ki67. Numbers of circulating Ki67+ CD8+ T cells correlated with tumor burden before and after therapy, and more importantly, a ratio of T cell reinvigoration to tumor burden greater than 1.94 significantly associated with better objective response, progression free survival and overall survival [
130]. Whether immune profiling of liquid biopsies will be implemented in standard clinical practice will depend on additional studies to validate the predictive value of these immune cell biomarkers. However, it is apparent that an increasing number of clinical trials are including analysis of absolute lymphocyte count as part of their analytical pipeline (Table
1).