Fig. 5: Tumor immune composition by MethylCIBERSORT identifies clusters of divergent immunogenicity.

a DNA methylation data were deconvoluted according to immune cell populations (MethylCIBERSORT). This revealed two different types of BPDCN (Treg and CD14 driven; named IC1 and IC2, respectively) presenting with significantly differing immune cell subsets in regard to the markers CD14 (monocytes/macrophages), CD19 (B-cells), CD4 (T-helper cells) as well as fibroblasts, NK-cells and T-regulatory cells (T-regs). b–e A borderline BPDCN/AML pDC-like case analyzed by MethylCIBERSORT and a comparative immunohistochemical assessment of the tumor microenvironment is presented. b H&E staining reveals a cutaneous infiltrate covered by an intact epidermis. c Giemsa staining reveals small blastoid cells with partly roundish occasionally monocytoid nuclei, small nucleoli, and weakly basophilic cytoplasm with increased mitotic activity. d Staining for myeloid peroxidase reveals expected negativity in the malignant infiltrate alongside a few positive, tumor-infiltrating myeloid cells. e However, CD14-Expression highlights both a typical negative BPDCN population, as well as a relevant monocytoid population, including few tumor-infiltrating monocytes alongside a larger subgroup of malignant cells. f Mosaic plot visualizing the overlap between transcriptional phenotype C1/C2 and IC1/IC2. g Overall, complete and partial response rates according to C1/C2 and IC1/IC2. h Oncoplot displays mutational patterns of 9 genes that were found to be more differentially mutated between IC1 and IC2. Additionally, the heatmap illustrates TME cell proportions for each individual sample. i, j Progression-free (PFS) and Overall survival (OS) analysis for patients with available clinical follow-up according to IC1 vs IC2 identifies a significant inferior prognostic impact for the IC1 subtype regarding PFS accompanied by a trend towards inferior OS.