Molecular prognostic markers in cutaneous malignant melanoma

Abstract

With recurrences occurring in 10% of localized cutaneous melanoma patients despite appropriate management, the identification of molecular markers that supplement established clinicopathologic prognostic variables to triage additional high-risk patients are needed. A systematic review of the published melanoma immunohistochemistry literature was conducted to enumerate already prognostically useful proteins. Only 37 of 515 identified manuscripts were eligible, revealing incomplete REMARK criteria adoption. Collectively, robust data for 62 unique proteins were available. Increased expression of proteins that facilitate tissue invasion and metastasis and effectors of DNA replication/repair as well as reduced p16^INK4A and AP-2α were most consistently associated with unfavorable outcomes. Automated Quantitative Analysis (AQUA ^®) of immunofluorescence-based immunohistochemistry on melanoma tissues was then executed. Subcellular localization of HMB45 reactivity distinguished benign nevi from melanomas. Nevi typically exhibited nuclear staining in contrast to the established cytoplasmic reactivity of melanomas, a potentially useful tool for discriminating histopathologically ambiguous cases. Next, the prognostic relevance for 3 protein sets was evaluated on a cohort of primary melanomas collected at Yale-New Haven Hospital during 1959-1994. Hierarchical clustering of data from the 3 classical cadherins and their catenin adaptors identified 4 unique clusters with differential overall survival (χ² _{1og-rank,3d.f.} 10.54; p=0.01). The best survival occurred among those tumors over-expressing both N- and E-cadherin, likely recapitulating appropriate neural-crest development, with the worst survival among those with high N-cadherin alone, indicating a successful epithelial-mesenchymal transition. The second set, seventeen proteins related to melanoma oncogenesis, was entered into a genetic algorithm for building a multi-marker prognostic assay. Multiple iterations yielded a consistent 5-marker solution where tumors that met at least 4/5 conditions were of "low risk" and those with ≤3/5 conditions "high risk", the latter displaying significantly reduced survival after adjustment for known clinicopathologic covariates (HR=2.84 (95% CI=1.46-5.49)). Validation on a second cohort of 243 Clark level III-V melanomas all evaluated by sentinel lymph node biopsy was also significant (multivariable HR=2.72 (95% CI=1.12-6.58)). Finally, 6 sex steroid hormone receptors were considered. Although each was expressed in a subset of lesions, supporting a potential role for related therapeutics, none mediated the established female survival benefit. Validation of each finding in additional cohorts is warranted.