110 Computational methods to monitor treatment response and toxicity in immune-checkpoint-inhibitor treated metastatic melanoma using methylated cell-free DNA

Abstract

Objectives/Goals: Immune checkpoint inhibitors (IO) have dramatically improved survival outcomes in patients with metastatic melanoma. Still, many patients do not respond these treatments, and others may experience harmful adverse events (irAEs). Thus, there an unmet need for biomarkers for real-time monitoring and management of patients exposed to IO therapies. Methods/Study Population: Serial serum samples were collected from patients with BRAFV600-mutant metastatic melanoma treated with ipilimumab/nivolumab (IO, n = 14) or dabrafenib/trametinib (TT, n = 10). Methylated cell-free DNA (cfDNA) was isolated and sequenced using enzymatic methyl-seq. We develop a robust computational pipeline to identify the top 250 cell-type specific regions of differential methylation (DMRs) across 24 cell-types. Using these differentially methylated regions, a deconvolution tool was developed to determine the abundance of cell type-specific cfDNA in patient serum, and changes in abundance were tracked over treatment time-course to assess response treatment and identify signals of adverse events. Results/Anticipated Results: We demonstrated improved precision in DMR detection evidenced by a higher area under the receiver operator characteristic curve (AUROC) of 0.85 on average. Pathway and functional annotation analysis revealed melanocyte-specific methylation marker regions regulated genes related to melanocyte development and differentiation, including MITF, SOX9/10, and FOXD3. We show these regions are conserved through the transformation to malignant melanoma, indicating melanocyte cfDNA abundance can be used as a marker for tumor burden. We characterize the dynamics of melanocyte-derived cfDNA over the course of treatment in responders and nonresponders to both IO and TT. We observe that changes in concentrations of cfDNA from other cell types correlate with clinically observed irAE-mediated damage to normal tissue. Discussion/Significance of Impact: We demonstrated the utility of decoding the origins of cfDNA fragments obtained from serial liquid biopsy samples. Using cell-specific methylation marks, we identified a signature from the primary melanoma to assess response to treatment, while also obtaining a signal from other tissues throughout the body to monitor immune related adverse events.