Genomic copy number burden distinguishes the diagnosis and prognosis of primary dermal melanoma.

Abstract

Primary dermal melanoma (PDM) is a subtype of melanoma with histological features indiscernible from nonepidermal metastatic cutaneous melanoma deposits, although patients with PDMs have more favourable outcomes than patients with dermal metastases. Despite the histological similarity, it is not known whether the mutational landscapes of PDMs and dermal metastases are also similar, or whether distinct mutational patterns might exist in PDMs. To identify genomic features that distinguish PDMs from nonepidermal cutaneous melanoma metastases, thereby enhancing the accuracy of clinical diagnosis and prognostication. Twenty PDMs and 23 nonepidermal cutaneous metastases were sequenced using the TSO500 panel. Genomic features distinguishing PDMs and metastases were identified using the regularized random forest model and feature engineering. Predictive biomarkers of patient event-free survival were identified using Lasso-Cox regression. PDMs and nonepidermal cutaneous metastases harboured point mutations characteristic of melanomas of epidermal origin. Compared with PDMs, metastases exhibited distinct copy number variations (CNVs) that led to gains of oncogenes and losses of tumour suppressor genes. Extreme focal amplifications were uniquely identified in cases of nonepidermal cutaneous metastases but were absent in PDMs. Large-scale CNVs and focal amplifications were identified as discriminative biomarkers between PDMs and metastases. A clinical diagnosis of PDM and an absence of BRAF CNVs collectively were associated with a reduced risk of death or disease progression after surgery vs. a clinical diagnosis of metastasis or of PDM with BRAF CNVs. We found that CNVs were the strongest discriminative features between PDMs and nonepidermal cutaneous metastases, with large-scale and focal CNVs abundant in metastases but not in PDMs. Notably, integrating BRAF copy number analysis with standard clinical criteria for diagnosing PDMs improved risk stratification after surgery. These data highlight the ability of panel-based genomic sequencing that enables the detection of CNVs to inform the clinical management of melanoma.