Epigenetic reprogramming shapes the cellular landscape of schwannoma

Abstract

Mechanisms specifying cancer cell states and response to therapy are incompletely understood. Here we show epigenetic reprogramming shapes the cellular landscape of schwannomas, the most common tumors of the peripheral nervous system. We find schwannomas are comprised of 2 molecular groups that are distinguished by activation of neural crest or nerve injury pathways that specify tumor cell states and the architecture of the tumor immune microenvironment. Moreover, we find radiotherapy is sufficient for interconversion of neural crest schwannomas to immune-enriched schwannomas through epigenetic and metabolic reprogramming. To define mechanisms underlying schwannoma groups, we develop a technique for simultaneous interrogation of chromatin accessibility and gene expression coupled with genetic and therapeutic perturbations in single-nuclei. Our results elucidate a framework for understanding epigenetic drivers of tumor evolution and establish a paradigm of epigenetic and metabolic reprograming of cancer cells that shapes the immune microenvironment in response to radiotherapy.

Fig. 1. Schwannomas are comprised of neural crest and immune-enriched molecular groups.

a Hierarchical clustering of the top 2000 most differentially methylated probes from 66 vestibular schwannomas. Significant gene ontology terms corresponding to hypomethylated probes in neural crest schwannomas (NCS) or immune-enriched schwannomas (IES), clinical metadata, and the molecular neuropathology (MNP) DNA methylation classification of central nervous system tumors are shown. b Integrated UMAP of 10,628 transcriptomes from harmonized schwannoma single-nuclei (n = 6) or single-cell (n = 3) RNA sequencing showing schwannoma cell (SC) types and tumor microenvironment cell types. c UMAPs from b with individual transcriptomes split according to a molecular group of origin. d Relative composition of SC types according to a molecular group of origin, colored as in (b). e Scaffold plot comprised of 375,355 Ki67+ immune cells from NCS (n = 3) or IES (n = 3) analyzed using mass cytometry time-of-flight (CyTOF). Manually gated landmark immune cell populations (black) are annotated. Schwannoma immune cell cluster is colored when the proportion of cells is statistically different between IES (positive) and NCS (negative). f CyTOF UMAPs of CD45+ immune cells with overlaid density plots for manually gated myeloid cells (top, 40,000 cells) or CD8 T cells (bottom, 6632 cells) in NCS (left) or IES (right). g UMAP feature plots of marker genes used to define myeloid or CD8 T cells in (f). h CyTOF proportion of schwannoma myeloid cells (left) or CD8 T cells (right) corresponding to M1 macrophages or PD1 + TEMRA CD8 T cells, respectively, in NCS (n = 3 independent tumors) or IES (n = 3 independent tumors). Lines represent means, and error bars represent the standard error of means (two-sided Student’s t-tests, *p = 0.0174, **p = 0.0065). i Representative preoperative magnetic resonance imaging of 66 vestibular schwannomas using T1 post-contrast, T2 diffusion-weighted, or apparent diffusion coefficient (ADC) sequences reveals NCS present as solid masses with reduced diffusion (dotted line) and IES present with cystic changes (arrows) and hydrocephalus. j Nomogram for schwannoma immune enrichment and molecular grouping based on non-invasive clinical and magnetic resonance imaging features. Source data are provided as a Source Data file.

Fig. 2. Radiotherapy is sufficient for epigenetic reprogramming of neural crest to immune-enriched schwannoma.

a Pairwise Pearson correlation coefficients grouped by hierarchical clustering of DNA methylation profiles from the patient (Pt) matched primary and recurrent schwannomas (n = 13). Arrows represent the reprogramming of primary NCS to IES at recurrence. b H&E-stained sections of a patient-matched primary NCS and recurrent IES, showing Verocay bodies (top arrows) and abundant lymphocytes (bottom arrows) with foamy and hemosiderin-filled macrophages (asterisks). Scale bar, 100 μm. Similar findings were observed in two additional matched pairs. c UMAP of 38,754 transcriptomes from single-cell RNA sequencing of HEI-193 cells after treatment with 0Gy, 1.8Gy × 5, or 12.5Gy × 1 (n = 3 independent replicates per condition) of radiotherapy revealing 15 distinct schwannoma cell states. d UMAPs from c with individual transcriptomes split according to triplicate treatment conditions. e Relative composition of cell states from c according to treatment conditions. f Perturb-seq gene-set expression heatmap of 3546 pseudobulked transcriptomes from HEI-193 cells following sgRNA perturbations (columns) across treatment conditions. Expression values are normalized to cells harboring non-targeting control sgRNAs (sgNTC) with 0 Gy. g Number of differentially expressed genes (DEG) from schwannoma cell Perturb-seq with radiotherapy (y-axis) versus 0Gy (x-axis). sgRNA perturbations with ≥40 DEG after radiotherapy compared to control are orange. sgRNA not meeting this threshold are dark gray. sgNTCs are light gray. h Differential gene expression analysis from PTPRG perturbation compared to sgNTC in 0 Gy (left) versus 1.8 Gy × 5 conditions (right). Significant positive (red) or negative (blue) gene expression changes are colored (p < 0.05, |log₂(fold change)| > 1), corresponding to gene ontology (two-sided Fisher’s exact test without adjustments for multiple comparisons). i Feature plot of integrated UMAP from harmonized schwannoma single transcriptomes (Fig. 1b) showing PTPRG expression in schwannoma cells. j TUNEL staining for apoptosis in HEI-193 cells following CRISPRi suppression of PTPRG versus sgNTC after treatment with 0 Gy (n = 3 independent cultures), 1.8 Gy × 5 (n = 3 independent cultures across 2 sgRNAs each), or 12.5 Gy × 1 (n = 3 independent cultures across 2 sgRNAs each) of radiotherapy. Fold changes are normalized to sgNTC in each treatment condition. Lines represent means, and error bars represent the standard error of means (two-sided Student’s t-tests, *p = 0.030, **p = 0.0044). Source data are provided as a Source Data file.

Fig. 3. Epigenetic regulators reprogram schwannoma cells and drive immune cell infiltration in response to radiotherapy.

a Experimental workflow for genome-wide CRISPRi screens using dual sgRNA libraries comprised of 20,528 targeted sgRNAs and 1025 non-targeted control sgRNAs (sgNTC). Libraries were transduced into HEI-193 cells that were subsequently treated with 0 Gy or 1.8 Gy × 5 radiotherapy (n = 3 per condition). sgRNA barcodes were sequenced and quantified as proxies for cell enrichment or depletion. b CRISPRi screen results showing the average rho log₂(sgRNA in radiotherapy conditions/sgRNA in control conditions) and two-sided Student’s t-test p values across three replicates. On-target hit genes (purple), epigenetic regulator hit genes (orange), and sgNTCs called hits (green) at a false discovery rate of 1% are shown. c Rho phenotypes of epigenetic regulator CRISPRi screen hits genes from (b). n = 3 replicate screens. d Crystal violet staining of HEI-193 following CRISPRi of KDM1A or KDM5C compared to sgNTC after treatment with 0 Gy, 1.8 Gy × 5, or 12.5 Gy × 1 of radiotherapy. Scale bar, 100 μm. e Quantification of cell density from (d) (n = 3 independent cultures, two-sided Student’s t-test). f Volcano plot of 425 peptides identified using proteomic mass spectrometry of conditioned media from triplicate HEI-193 cultures after radiotherapy or control treatment. Significant gene ontology terms of enriched peptides after radiotherapy conditions annotated (two-sided Fisher’s exact test). g Proteomic mass spectrometry parallel reaction monitoring targeted assay validating secreted peptide enrichment in conditioned media from HEI-193 after radiotherapy (n = 3 independent cultures, two-sided Student’s t-test) as in (f). h Transwell primary human peripheral blood lymphocyte migration assays using conditioned media from HEI-193 (n = 3 independent cultures, two-sided Student’s t-test) following CRISPRi suppression of APOA1, KDM1A, or KDM5C ± radiotherapy as a chemoattractant. i Targeted metabolite mass spectrometry of HEI-193 after treatment with 0 Gy, 1.8 Gy × 5, or 12.5 Gy × 1 of radiotherapy (n = 3 independent cultures per condition, two-sided Student’s t-test). Fold changes normalized to 0 Gy treatment for each metabolite. j Metabolic enzymes gene expression changes from bulk RNA sequencing of HEI-193 cells ± radiotherapy (n = 3 independent cultures) (Supplementary Data 6). Fold changes normalized to 0 Gy treatment for each gene. Lines represent means, and error bars represent standard error of means (two-sided Student’s t-tests, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.0001). Source data are provided as a Source Data file.

Fig. 4. Single-nuclei ATAC, RNA, and CRISPRi perturbation sequencing identify integrated genomic mechanisms driving schwannoma cell state evolution.

a Experimental workflow for single-nuclei ATAC, RNA, and CRISPRi perturbation sequencing (snARC-seq). Triplicate HEI-193 cultures were transduced with sgRNA libraries targeting 29 epigenetic regulators driving radiotherapy responses from genome-wide CRISPRi screens (Fig. 3c) and treated with 0 Gy or 1.8 Gy × 5 of radiotherapy prior to isolation of single-nuclei for sequencing. sgRNA identities were recovered from CROP-seq tags in single-nuclei RNA sequencing data. b UMAPs of ATAC (left) or RNA (right) sequencing of 855 single nuclei passing snARC-seq quality control from triplicate control or radiotherapy conditions (Supplementary Fig. 18). c Hierarchical clustering of differential gene activity scores between radiotherapy and control conditions for each snARC-seq perturbation (columns). Gene activity modules (rows) were derived from HEI-193 schwannoma cell states ± radiotherapy (Fig. 2c) or from human schwannoma cell types (Fig. 1b). Gene ontology of perturbed epigenetic regulators and CRISPRi screen growth (gamma) or radiation response (rho) phenotypes from genome-wide CRISPRi screens (Fig. 3c) are shown. d Hierarchical clustering of differential ChromVAR transcription factor motif deviations between radiotherapy and control conditions for each snARC-seq perturbation (columns). e Average profile plots of normalized ATAC signal at KLF13 or TCF3 motifs with ENCODE ChIP-seq peak annotations and differential accessibility following snARC-seq perturbation of KDM1A, KDM5C, or SETDB1. f Feature plot of integrated UMAP from harmonized schwannoma single-nuclei and single-cell RNA sequencing (Fig. 1b) showing genes near TCF3 motifs that are differentially accessible following SETDB1 snARC-seq perturbation. g Hierarchical clustering of human schwannoma RNA sequencing profiles using 56 differentially expressed SETDB1 targets with TCF3 motifs showing separation of NCS and IES molecular groups of schwannomas. Source data are provided as a Source Data file.

Fig. 5. An integrated model of schwannoma tumorigenesis and epigenetic reprogramming in response to treatment.

Two molecular groups of schwannoma, neural crest schwannoma and immune-enriched schwannoma, are driven by distinct mechanisms. Radiotherapy can induce immune-enriched schwannoma through metabolic and epigenetic reprogramming.