Related Experiment Video
Updated: Aug 5, 2026

Analysis of Lymph Node Volume by Ultra-High-Frequency Ultrasound Imaging in the Braf/Pten Genetically Engineered Mouse Model of Melanoma
Published on: September 8, 2021
Locus-specific HERV expression identifies an aggressive, NK-depleted, checkpoint-refractory acral melanoma phenotype
Jez L Marston1, Tongyi Fei1, Helena Reyes-Gopar2
1Division of Infectious Diseases, Weill Cornell Medicine, New York, NY, United States.
Background:
Acral melanoma (AM) is a non-ultraviolet (non-UV)-derived subtype of cutaneous melanoma that is associated with worse survival outcomes compared to UV-driven melanomas. AM tumors exhibit distinct molecular characteristics and poor responses to immune checkpoint blockade. Hitherto, the role of transposable elements (TEs), particularly human endogenous retroviruses (HERVs), in AM progression remains incompletely characterized.
Methods:
We quantified locus-specific TE expression from RNA sequencing (RNA-seq) of 36 AM samples from 33 patients and integrated the results with clinical outcomes, a pathologist-reviewed purity covariate, and immune cell deconvolution.
Results:
A three-locus HERV signature, composed of ERV316A3_6p25.1d, HERVH_6q21a, and ERVLE_9q21.32c, stratified AM tumors by overall survival with a leave-one-patient-out C-index of 0.778 (bootstrap 95% confidence interval [CI]: 0.614-0.914, permutation p = 0.015) and a Kaplan-Meier log-rank of p = 0.003 across risk tertiles. In a multivariable Cox model adjusted for primary site, tumor cellularity pathology, sex, ulceration, and Breslow depth, with patient-clustered robust standard errors, the signature remained strongly prognostic (HR: 2.19 per 1-standard deviation [s.d.] increase in risk score, 95% CI: 1.38-3.47, p = 0.001). High-risk tumors were natural killer (NK)-cell-depleted (Spearman ρ = -0.58, false discovery rate [FDR] p = 0.004) and showed coordinated elevation of LIN28A and HMGA2 (ρ = +0.55 and +0.45), consistent with reactivation of the LIN28A/let-7/HMGA2 oncofetal axis rather than a full pluripotency program. High-risk tumors were also associated with reduced odds ratio of response to immune checkpoint inhibitor (ICI) therapy (OR: 0.20 per 1-s.d. increase, p = 0.075), whereas reported mutation burden was not (OR: 1.02, p = 0.80), consistent with the unexplained low-TMB checkpoint responders previously reported in this cohort. A single locus, HML6_20p11.21, retained a full 9.14 kb proviral structure and a 333-aa open reading frame (ORF) with intact integrase, protease, and gag motifs. Peptide-human leukocyte antigen (HLA) binding predictions identified strong HML6-derived binders for every HLA class I allele in the typed cohort, and at least one strong binder in all 25 typed patients.
Conclusion:
Our findings reframe AM as a transcriptomically stratifiable disease, define a locus-specific HERV signature of aggressive, checkpoint-refractory AM, and nominate a structurally intact, pan-HLA-presentable HERV-K antigen candidate for further evaluation.

