Melanoma to rhabdomyosarcoma plasticity in the setting of immunotherapy

Abstract

Acquired resistance to immune checkpoint inhibitors (ICIs) remains a significant challenge in the treatment of metastatic melanoma. Phenotypic plasticity, such as dedifferentiation and transdifferentiation, is an increasingly recognized mechanism for treatment resistance. We present a case of a man in his 70s with metastatic melanoma who experienced progression through sequential treatments including pembrolizumab in combination with the HDAC inhibitor entinostat, and ipilimumab. During treatment a histologically distinct pleomorphic rhabdomyosarcoma (RMS) emerged at metastatic sites. Longitudinally acquired tumor samples representing both phenotypes were analyzed using whole-exome sequencing (WES), RNA sequencing (RNA-seq) and high-plex tissue imaging (spatial proteomics). WES revealed driver mutations (e.g. NRAS, NF1) and loss-of-heterozygosity (LOH) shared between phenotypes indicating a common ancestral clone. Phylogenetic analysis demonstrated an early divergence of the phenotypes, with each later acquiring unique mutations. RNA-seq showed mutually exclusive expression of lineage-specific markers as well as epithelial-mesenchymal transition and myogenic gene set enrichment in the RMS samples. High-plex imaging identified distinct tumor microenvironments, with RMS lesions enriched in CD163⁺ macrophages.

Figure 1:. The rhabdomyosarcoma and melanoma tumors share many genomic features and arose from the same common ancestor within the patient

A) Clinical course for the patient, including relative times at which the tumors were detected and sampled for WES and H&E showing the cellular morphology (H&E) and staining for a common melanoma marker (Sox10), and for a common rhabdomyosarcoma marker (Desmin). B) Comut plot showing clinical metadata (treatment, biopsy site) and tumor purity for the samples alongside some of the driver mutations and copy number alterations in these tumors and how these are largely shared across samples. C) Plot of the copy number alterations across the samples, including those shared across all samples (gray arrows), uniquely shared across melanoma samples (dashed brown arrow), and those uniquely shared across rhabdomyosarcoma samples (dashed blue arrow). The LOH in chr16 is subclonal in RHAB_M1 and RHAB_M2a so it is only colored here in RHAB_M2b where it is clonal. D) Plot of lineage tree summarizing large scale CNA changes and mutational differences (from pyclone) shared by all tumors, melanoma tumors (brown), and rhabdomyosarcoma tumors (blue) with key driver mutations labeled. ”M” refers to the melanoma lineage and “R” refers to the rhabdomyosarcoma phenotype lineage.

Figure 2:. The rhabdomyosarcoma and melanoma tumors are characterized by different patterns of gene expression

A) Top 50 genes upregulated in melanoma samples versus rhabdomyosarcoma samples and vice versa. Selected genes were chosen after the data was ranked by log2FC and filtered by p<0.05 based on a two-sided t test (without multiple hypothesis correction) B-C) Results showing selected results from GSEA of genes ranked by log2FC between rhabdomyosarcoma phenotype tumors and melanoma phenotype tumors (all pathways shown significant with an FDR<0.001). D) Most significantly enriched Cancer Cell Line Encyclopedia (CCLE) GSEA results (top, enriched in melanoma on left and rhabdomyosarcoma phenotype on right). E) GO Molecular Function and GO Biological process pathways in rhabdomyosarcoma phenotype tumors as compared to melanoma phenotype tumors (results from Enricher with top 3000 genes ranked by log2FC for each phenotype)

Figure 3:. Rhabdomyosarcoma and melanoma tumors are not mixed in phenotype, and are characterized by different patterns of NGFR and EGFR expression and distinct immune microenvironments

A. Example images from Rhabdomyosarcoma (T4_A1) and Melanoma (T4_B3). Tumor (SOX10, S100 and Desmin), immune (CD45) and stromal (SMA) markers are selected B. Distribution of different cell types in each samples. Single-cell clustering was performed with Gaussian Mixture Model (GMM), and 6 main cell types (Rhab, Mel, Stroma, Lymph, Macro, Other) were identified. CasePR2 is the control Rhabdomyosarcoma sample from another patient. The melanoma samples are colored with red and Rhabdomyosarcoma are labeled black. C. The example spatial map of cell types. D. M2 Macrophage (CD163+) fractions in different samples. Left panel: individual sample counts, middle panel: counts by timepoint, right panel: counts by tumor types. E. M1 Macrophage (CD68+) fractions in different samples. Left panel: individual sample counts, middle panel: counts by timepoint, right panel: counts by tumor types. F. NGFR+ cell fractions in different samples. Left panel: individual sample counts, middle panel: counts by timepoint, right panel: counts by tumor types. G. EGFR+ cell fractions in different samples. Left panel: individual sample counts, middle panel: counts by timepoint, right panel: counts by tumor types. H. Induction of NGFR after treatment