Work

An Eight-Gene Prognostic Signature Capturing EGFR-Driven, Immune-Cold Biology in Cutaneous Melanoma

Year: 2026
Type: preprint
Abstract: Background: Cutaneous melanoma remains the deadliest skin cancer despite advances in targeted therapy and immunotherapy. About 14–27% of melanomas carry mutations in the NF1 tumor suppressor gene, and these patients have limited options once immunotherapy fails. Recent work has shown that NF1 loss drives EGFR upregulation and EGFR-dependent growth in patient-derived models, that NF1-mutant melanomas form immune-cold meta-niches with impaired antigen presentation and CD8 T cell exclusion, and that they harbor a proliferative, CDC20-dependent transcriptional program amenable to combination targeting. However, no bulk-tissue, multi-cohort gene expression signature has yet operationalized this EGFR-driven, immune-cold biology in association with clinical outcomes, and it remains unknown whether such a signature tracks with NF1 mutation status or captures a broader EGFR-driven phenotype. Methods: From 242 NF1/RAS pathway genes, we built a signature using TCGA-SKCM bulk RNA-seq. WGCNA identified genes co-expressed with NF1; LASSO Cox regression selected the final signature. We evaluated it in four independent GEO cohorts and characterized its biology with GSVA, immune deconvolution, and single-cell RNA-seq (GSE72056), and assessed associations with NF1 mutation status and clinical features. Results: The final 8-gene signature (PLA2G2D, FASLG, RASGRP1, FLT3, FGF1, PLA2G5, FGF2, EGFR) stratified patients into High and Low Risk groups with significantly different overall survival (log-rank p = 1.4×10⁻¹⁰). EGFR was the only risk-factor gene (coefficient = +0.118); the remaining genes were protective. The signature was independent of age and stage in multivariate Cox analysis (HR = 5.32, 95% CI 3.55–7.99, p < 0.001; C-index = 0.72), with time dependent AUCs of 0.72 (1-year), 0.73 (2-year), and 0.71 (3-year). GSVA showed that all top differentially active pathways were immune-related and downregulated in High-Risk patients (e.g., T cell receptor signaling, antigen presentation, NK cell cytotoxicity; FDR < 10⁻²⁸). Immune deconvolution confirmed depletion of CD8 T cells, CD4 T cells, NK cells, and cytolytic activity in High-Risk tumors. Single-cell analysis localized protective genes to immune cells (RASGRP1, 26.4%; FASLG, 9.8%), whereas EGFR was tumor-intrinsic and expressed in only 2.7% of cells. The signature was directionally consistent in three of four external cohorts (HR > 1), with a pooled meta-analysis HR of 2.25 (95% CI 1.03–4.89, p = 0.041; I² = 7.6%). It did not correlate with NF1 mutation status (p = 0.12; n = 72), consistent with the possibility that the signature captures a broader EGFR-driven phenotype rather than NF1 loss per se. This analysis was underpowered to detect a modest association. Conclusions: We present a bulk-tissue 8-gene signature that operationalizes the EGFR-driven, immune-cold biology recently defined mechanistically, spatially, and therapeutically in NF1 mutant melanoma. The signature captures a broader EGFR-driven phenotype that is not restricted to NF1-mutant tumors, as it did not correlate with NF1 mutation status in our cohort. It stratifies survival in TCGA-SKCM, is independent of stage, captures an immune-cold phenotype concordant with spatial findings, and is directionally consistent in external cohorts. These findings generate testable hypotheses for selecting patients for EGFR inhibitor and immunotherapy combinations and provide a bulk-computable biomarker for prospective testing. The signature is not yet a validated clinical assay, but it is a biologically anchored, hypothesis generating tool that operationalizes the EGFR-driven, immune-cold phenotype and is immediately testable in existing RNA-seq cohorts.
Author: Shafi Rehman
Language: English
Subfield: Oncology
Field: Medicine
Open Access status: green
OpenAlex ID: W7235200741
Cites: 21
Cited by:
Related to: 10