SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

The last few years have witnessed the dawn of the molecular era in melanoma treatment. With the advent of successful therapy targeting mutant BRAF, melanoma is leading the field of cancer research in the molecular approach to therapy of advanced disease. Attempting to keep pace with advances in therapy are advances in the molecular assessment of melanoma progression, facilitated by the availability of genome-wide approaches to interrogate the malignant phenotype. At the DNA level, this has included approaches such as comparative genomic hybridization. At the RNA level, this has consisted of gene expression profiling using various assay methodologies. In certain instances, markers identified using these platforms have been further examined and developed using fluorescence in situ hybridization and immunohistochemical analysis. In this article, we will review recent progress in the development of novel molecular markers for melanoma that are nearing clinical application. We will review developments in the molecular classification of melanoma, in the molecular diagnosis of melanoma, and in the molecular assessment of melanoma prognosis.


Molecular classification of melanoma

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

The clinical heterogeneity of melanoma has resulted in the development of classification schemes that use different factors to capture the various subtypes of melanoma. The classical subtypes of melanoma were defined to include superficial spreading, nodular, acral lentiginous and lentigo maligna melanoma.[1] However, the lack of prognostic significance associated with them limited their clinical utility.

Studies using comparative genomic hybridization (CGH) to assess melanoma at the genomic level identified recurrent chromosomal aberrations that appeared to be associated with distinct subtypes of melanoma.[2, 3] Integration of CGH data with mutational analysis of specific oncogenes and with extent of sun exposure resulted in the development of a molecular classification scheme for melanoma.[4] While melanomas associated with intermittent sun exposure were characterized by BRAF or NRAS mutations, melanomas defined by either chronic or little sun exposure were characterized by KIT mutations, and the absence of BRAF/NRAS mutations. One attractive feature of this model lay in its potential ability to predict the likely mutational source based on location of the primary tumour and extent of sun exposure. However, this model has been undermined by the low prevalence of KIT mutations even in enriched subtypes, and because BRAF or NRAS mutations can be present regardless of the location of the primary tumour or the extent of sun exposure.[5]

Currently, in the clinical setting, mutational analysis for BRAF and NRAS is routinely used to classify melanomas at the molecular level and to determine eligibility for targeted therapy regimens. In addition, PTEN levels can further classify a subset of BRAF-mutant melanomas, in which PTEN loss results in activation of the PI3K/AKT pathway. Overall, BRAF mutations are present in up to 50% of melanomas (approximately half of which have PTEN loss), with NRAS mutations occurring in up to 20% of melanomas, almost exclusively with wild-type BRAF.[6-8]

Finally, the molecular basis of melanomas underlying the progression of melanomas lacking these mutations (termed ‘triple-negative melanomas’) has been poorly understood. Recent studies identified a role for the pleckstrin-homology domain interacting protein (PHIP) gene in melanoma classification. PHIP was identified as the top gene overexpressed in a cDNA microarray analysis comparing global transcriptional profiles of melanoma metastases vs. primary tumours.[9] Short-hairpin RNA-mediated targeting of PHIP suppressed the metastatic potential of human melanoma, in part through its activation of AKT.[10] AKT is activated in melanoma primarily due to either NRAS mutation or BRAF mutation coupled with PTEN loss.[8, 11] Additional analyses indicated that PHIP-overexpressing melanomas were predominantly characterized by wild-type NRAS, BRAF and PTEN, suggesting that PHIP levels may be used to classify this poorly understood subset of melanomas.

Molecular diagnostic markers for melanoma

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

The pathological diagnosis of melanoma can be difficult, as no single histological attribute can reliably be utilized to distinguish between naevus and melanoma. In addition, the relative significance assigned to the many histological variables used to assess melanocytic neoplasms is uncertain and differs between different pathologists. As a result, discordance in the diagnosis of melanocytic neoplasms can be high,[12, 13] even when ‘classical’ cases have been reviewed by a panel of expert pathologists.[14-18] These observations highlight the need for molecular adjuncts to the pathological diagnosis of melanoma.

Several immunohistochemical (IHC) markers are used in the clinical setting in the evaluation of pigmented lesions (e.g. S100, HMB45, Melan-A/MART1 and MITF). While these markers are useful in distinguishing melanocytic from nonmelanocytic neoplasms, they are not capable of reliably distinguishing benign from malignant melanocytic neoplasms.[19-22] In addition, given that BRAF mutations occur commonly in both naevi and melanomas,[23] the tumour's mutational status cannot be used to determine its malignant potential. As a result, additional molecular markers are required in the assessment of pigmented lesions.

Recently, a multimarker diagnostic assay was developed, based on an analysis of 693 melanocytic neoplasms, to assist with this differential diagnosis.[24] Five markers were incorporated into an IHC assay, derived from an earlier gene expression profiling analysis that demonstrated differential gene signatures associated with naevus vs. melanoma when a small cohort of freshly acquired specimens was analysed.[9] Each of the five markers (ARPC2, FN1, RGS1, SPP1 and WNT2) was shown to be significantly overexpressed in melanomas when compared with naevi using absolute marker expression scores. In addition, the pattern of marker staining (in the junctional zone vs. the base of the neoplasm) was significantly different between naevi and melanomas. Whereas in melanomas marker expression was uniformly present, naevi frequently exhibited loss of marker expression at the base of the lesion (termed ‘top–bottom difference’), a pattern that was observed for each of the five markers examined. Development of a diagnostic algorithm that incorporated both intensity scores and ‘top–bottom differences’ for the five markers in a training set of 534 melanocytic neoplasms yielded a sensitivity of 91% and specificity of 95% in the diagnosis of melanoma.

The multimarker IHC assay, using the same diagnostic algorithm developed in the training set, was then analysed in several additional cohorts with greater relevance to the histopathological diagnosis of melanoma, and found to have a high degree of accuracy in the setting of dysplastic naevi, Spitz naevi and melanoma arising in association with a naevus. Finally, the multimarker assay was found to diagnose correctly 75% of initially misdiagnosed melanocytic neoplasms, suggesting its potential utility to correct or prevent a significant proportion of errors arising from the routine pathological evaluation of melanocytic neoplasms. This assay is currently undergoing additional validation in distinct cohorts of melanocytic neoplasms.

Separately, CGH analyses of melanocytic neoplasms also indicated distinct findings in the assessment of naevi vs. melanomas,[25] suggesting the potential utility of copy number changes in the molecular diagnosis of melanoma. While melanomas consistently exhibited copy number changes in several chromosomal loci, naevi did not routinely harbour these aberrations. In addition, a proportion of Spitz naevi were characterized by presence of isochromosome 5p, which was not commonly observed in melanomas.[25] Subsequently, a molecular diagnostic assay was developed using fluorescence in situ hybridization (FISH), focused on several of the loci identified by CGH analysis.[26] This multicolour assay included four probes (6p25, 6q23, centromere 6 and 11q13) that were initially examined in a training cohort of 301 cases. Examination of a diagnostic algorithm developed in the training set in a distinct cohort of 169 melanocytic neoplasms resulted in a sensitivity of 87% and specificity of 95% in the diagnosis of melanoma. In one study performed by a different group of investigators,[27] the multicolour assay achieved a lower degree of diagnostic accuracy when applied to a cohort of ambiguous melanocytic neoplasms. Recent studies have evaluated the incorporation of additional markers to the assay,[28] and have suggested its utility in the assessment of sentinel lymph nodes (SLNs) in patients with melanoma.[29]

Taken together, these results indicate that substantial progress has been made in the molecular diagnosis of melanoma, resulting in the development of molecular markers that can be helpful in assisting pathologists, clinicians and patients in this differential diagnosis. However, significant challenges still remain to optimize the molecular diagnosis of melanoma. While an attractive feature of both these assays is the in situ assessment of marker expression or copy number, in which the molecular changes can be assessed along with the morphology of the cells being assayed, analysis of the melanocytic lesion in this manner can also result in a high degree of interobserver variability. As a result, whether different observers would reach the same molecular diagnostic conclusion when using the same clinical assay remains to be seen.

In addition, given the varying methodological approaches that are introduced in the adoption of laboratory-based assays, it is unlikely that the same diagnostic algorithm can be reliably used by different laboratories and still achieve a high degree of diagnostic accuracy. For example, in the case of the FISH-based assay, some academic centres merely report presence or absence of copy number gains or losses in one or more of the loci examined, without utilizing a rigorous diagnostic algorithm. The diagnostic accuracy of this clinical practice is unclear and is likely lower than that achieved using validated algorithms. Finally, the in situ assays have less capacity to incorporate additional diagnostic markers that are identified, in part due to tissue availability, when compared with assays using quantitative polymerase chain reaction that can multiplex up to 20 genes.

Molecular prognostic markers for melanoma

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

Molecular markers for metastatic melanoma

Predicting the behaviour of melanoma in individual patients can be difficult. While the staging classification for melanoma incorporates several powerful histological prognostic factors, additional factors are needed in order to predict accurately prognosis at the individual patient level. Despite significant attention from the melanoma research community, to date, no molecular markers are routinely used in the prognostic assessment of patients with melanoma. Genome-wide evaluation of melanoma has provided insights into melanoma progression and has identified a number of candidate molecular prognostic markers for melanoma. Due to the availability of fresh tissue, greater emphasis was initially placed on the gene expression profiles associated with metastatic melanoma. The original gene expression profiling analysis of melanoma identified two molecular subtypes of metastatic melanoma with possible prognostic significance.[30] This analysis suggested that the more aggressive subtype of melanoma was characterized by reduced expression of MART-1 and WNT5A, and elevated fibronectin expression. Subsequently, transcriptomic profiling of distinct stages of melanoma progression also identified two molecular subtypes of melanoma with potentially divergent prognosis.[9] This analysis demonstrated that the more aggressive subtype of melanoma had a gene expression signature overlapping with that found in the radial growth phase of primary melanoma, suggesting that the metastatic programme is already present in the earliest stages of melanoma progression.

Subsequent studies have identified other good- vs. poor-prognosis gene signatures. In one study,[31] analysis of the gene expression profiles of 57 patients with stage IV melanoma identified four subtypes of metastatic melanoma: (i) immune response group (characterized by CXCL12, IL1R1, IFNGR1, LCK and HLA class I antigen expression), associated with a better outcome; (ii) pigmentation differentiation group (with high TYR, MITF, PMEL (previously SILV), DCT and low WNT5A levels); (iii) proliferative group (characterized by high CCNA2, E2F1 and BUB1 expression), associated with poor survival, and frequent BRAF/NRAS mutations or CDKN2A deletions; and (iv) normal-like group (with high KIT, EGFR, FGFR3, KRT10 and KRT17 levels).

Finally, therapy of metastatic melanoma with various immunotherapy regimens has shown durable responses in a small subset of patients. Accordingly, microarray analysis of metastatic melanomas has been intensively examined for the identification of markers of immune responsiveness. In an original analysis of 63 subcutaneous melanoma metastases,[32] while subsets of metastases that were predictive of immune responsiveness could not be identified, 30 genes were associated with response to interleukin-2-based therapy, including several in the interferon signalling pathway. More recently, microarray analysis of patients with metastatic melanoma undergoing peptide vaccine therapy identified a major cluster of samples based on the presence or absence of T cell-associated transcripts,[33] and identified a chemokine gene signature enriched in tumours with prominent T-cell infiltrates. Finally, using microarray analysis, a gene expression signature of immune responsiveness was developed by investigators at GlaxoSmithKline that appears to be associated with clinical responses induced by the MAGE-A3 protein.[34] This gene signature is being further examined in a phase III adjuvant trial of this immunotherapeutic agent in the setting of node-positive melanoma.

Molecular markers for primary melanoma

Until recently, gene signatures for primary melanoma have been difficult to assess in a large cohort of specimens due to the requirement for fresh tissue for profiling analysis. Nevertheless, intriguing observations were made using microarray analysis of primary melanomas. Winnepenninckx and colleagues performed gene expression profiling using an oligonucleotide microarray on 58 primary melanoma samples,[35] and identified 254 genes associated with distant metastasis-free survival of patients with melanoma, including several minichromosome maintenance genes and geminin, whose prognostic correlation with overall survival (OS) was demonstrated at the protein level. In another study, cDNA microarray analysis of 34 vertical growth-phase melanomas, and supervised hierarchical analysis of metastasizing vs. nonmetastasizing melanomas identified a link between a gene signature of epithelial-to-mesenchymal transition and melanoma metastasis.[36] IHC analysis of several of the identified markers demonstrated a prognostic role for SPP1, CDH2 and SPARC on relapse-free survival (RFS) by univariate analysis.

More recently, the availability of profiling platforms that can analyse partially degraded RNA has enabled the assessment of larger numbers of paraffin-embedded samples. Conway et al.[37] examined gene expression profiles of 254 archived primary melanomas using the DASL chip (Illumina Inc., San Diego, CA, U.S.A.), which incorporates 502 genes, and identified SPP1 (osteopontin) as the most strongly correlated with RFS in this cohort. SPP1 expression levels were then investigated in a separate cohort of 218 primary tumours, and found to be independently predictive of RFS. In addition, Harbst et al.[38] analysed 223 archived primary melanoma specimens using the whole genome DASL assay, and were able to recapitulate the four previously mentioned subtypes identified in melanoma metastases. Subsequently, a binary classifier was developed, with a ‘high-grade’ class (incorporating the proliferative and pigmentation groups) and a ‘low-grade’ class (incorporating the high-immune and normal-like groups). This high-grade gene signature correlated with increased tumour thickness and mitotic rate, and was associated with significantly reduced OS and RFS. By multivariate analysis, this binary classifier was independent of tumour stage, and was predictive of survival in two other publicly available gene expression databases. Mutational analysis indicated a higher frequency of BRAF mutations in low-grade melanomas, and a trend for higher frequency of NRAS mutations in high-grade tumours.

In addition, markers identified by various means have been analysed at the protein level for their prognostic significance in primary melanoma. In one study, expression of 38 candidate markers was assessed using the automated quantitative analysis method for immunofluorescence-based IHC in a cohort of 192 formalin-fixed primary melanomas, and an algorithm developed using five markers (ATF2, p21/WAF1, p16/CDKN2A, beta-catenin and fibronectin) to define low-risk and high-risk groups.[39] Multivariate analysis, including the genetic algorithm, tumour thickness, age, sex and SLN status in a validation cohort of 246 tumours, demonstrated an independent prognostic impact for the genetic algorithm.

Separately, three markers (SPP1, NCOA3 and RGS1) that emanated from gene expression profiling analyses were incorporated into a multimarker IHC assay and examined in a training set of 395 primary melanomas.[40] Increasing multimarker scores were significantly and independently predictive of SLN status and disease-specific survival (DSS) in this cohort. By multivariate analysis, the multimarker index was the most powerful factor predicting DSS, even after the inclusion of SLN status and mitotic rate into the model. This multimarker assay was then examined in a German cohort of 141 primary melanomas, and found to have independent prognostic significance in the prediction of DSS. More recent work using the original tissue cohort examined identified an independent prognostic role for PHIP in the prediction of melanoma survival.[10] Overall, these studies have indicated that markers of tissue-specific metastasis can be developed for melanoma, as some of the markers identified have their greatest prognostic impact on distant metastasis (e.g. RGS1 and PHIP), while others impact survival through their prediction of lymph node metastasis (e.g. NCOA3 and SPP1). A list of promising molecular markers for melanoma is provided in Table 1.

Table 1. Molecular markers of primary melanoma
MarkerBiomarker typePlatformReference
  1. IHC, immunohistochemistry.

WNT2 DiagnosticIHC [24]
ARPC2 DiagnosticIHC [24]
CDKN2A PrognosticIHC [39]
MCM4 PrognosticRNA, IHC [35]
MITF PrognosticDNA, RNA [31, 38]
ATF2 PrognosticIHC [39]
NCOA3 PrognosticIHC [40, 41]
WNT5A PrognosticRNA [30, 31]
FN1 Diagnostic and prognosticIHC [24, 30, 39]
SPP1 Diagnostic and prognosticIHC, RNA [24, 36, 37, 40, 42]
RGS1 Diagnostic and prognosticIHC [24, 40, 43]

Conclusions and future directions

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

This is a highly exciting time for molecular investigations of melanoma biology. Genome-wide analyses of distinct stages of melanoma have offered significant insights into melanoma progression, and have resulted in the development of gene signatures that have been developed in various cohorts. In the realm of diagnostic markers, these assays have neared clinical application. In the realm of prognostic markers, further analysis is indicated, especially in prospectively defined, multicentre tissue sets, in order to validate the promising gene signatures identified in order to validate their clinical utility. Future progress in the development of molecular markers for melanoma will likely emanate from newer analytical platforms. These include microRNA (miRNA) expression profiling and whole genome sequencing. While early studies using these platforms have identified a plethora of differentially expressed miRNAs or mutated genes in melanoma, significant additional work will be required in order to understand the utility of these genes as clinically useful biomarkers for melanoma. In addition, it is anticipated that further developments in the proteomics field will both facilitate the identification of novel molecular markers for melanoma and provide a new potential diagnostic platform.

References

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography

Biography

  1. Top of page
  2. Summary
  3. Molecular classification of melanoma
  4. Molecular diagnostic markers for melanoma
  5. Molecular prognostic markers for melanoma
  6. Conclusions and future directions
  7. References
  8. Biography
  • Image of creator

    M. Kashani-Sabet is director of the Center for Melanoma Research and Treatment at the California Pacific Medical Center and senior scientist at the center's Research Institute. Dr Kashani-Sabet earned his medical degree from the State University of New York at Stony Brook, where he also completed an internship in internal medicine. He then completed a residency in dermatology at the University of California, San Francisco, as well as a postdoctoral fellowship in cutaneous oncology, and received training in both dermatology and medical oncology. Dr Kashani-Sabet maintains clinical interests in melanoma and cutaneous lymphoma, and research interests in targeted therapy, ribozymes, siRNAs, tumour metastasis, prognostic factors and tumour biomarkers. He has served as principal or co-investigator in several clinical trials of novel agents for melanoma and cutaneous lymphoma. He is a recipient of grant funding from the National Institutes of Health and the Department of Defense, and has served on several study sections for the National Cancer Institute. He serves on the editorial board of Journal of Translational Medicine, Journal of Skin Cancer, Cancer Gene Therapy and Oncology Letters, and has served as an ad hoc reviewer for numerous journals, including The Lancet, Annals of Internal Medicine and Journal of the National Cancer Institute. The author of over 100 peer-reviewed publications and 40 review articles/book chapters, as well as editor of two textbooks, Dr Kashani-Sabet has been an invited speaker at numerous regional, national and international meetings and symposiums.