Functional interactions between neurofibromatosis tumor suppressors drive Schwann cell tumor de-differentiation and treatment resistance

doi:10.21203/rs.3.rs-1933793/v1

Download PDF

Article

Functional interactions between neurofibromatosis tumor suppressors drive Schwann cell tumor de-differentiation and treatment resistance

https://doi.org/10.21203/rs.3.rs-1933793/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 12 Jan, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Schwann cell tumors are the most common cancers of the peripheral nervous system and can arise sporadically or in patients with neurofibromatosis type-1 (NF-1) or type-2 (NF-2). NF-1 is caused by loss of NF1, a negative regulator of Ras signaling. NF-2 is caused by loss of NF2, a pleiotropic tumor suppressor that inhibits PAK signaling. Functional interactions between the NF1 and NF2 tumor suppressors and broader mechanisms underlying malignant transformation of the Schwann lineage are unclear. Here, we integrate DNA methylation profiling, whole exome sequencing, bulk and single-cell RNA sequencing, biochemistry, and pharmacology across human samples, patient-derived cell lines, and mouse xenografts to identify cellular de-differentiation mechanisms driving malignant transformation and treatment resistance in Schwann cell tumors. Our data show molecular groups of Schwann cell tumors are distinguished by de-differentiation trajectories that drive resistance to MEK inhibition, the only approved molecular therapy for patients with NF-1. Functional genomic screening for mediators of MEK inhibitor responses in NF1-deficient tumor cells reveals NF2 loss and PAK activation underlie Schwann cell tumor de-differentiation and MEK inhibitor resistance. In support of these findings, we identify a group of de-differentiated Schwann cell tumors with concurrent loss of NF1 and NF2, and find combination molecular therapy inhibiting MEK and PAK is an effective treatment for de-differentiated Schwann cell tumor xenografts. In sum, we elucidate a paradigm of de-differentiation driving malignant transformation and treatment resistance, uncovering a functional link between the NF1 and NF2 tumor suppressors that sheds light on a novel therapeutic vulnerability.

The neural crest derived Schwann cell lineage gives rise to schwannomas, neurofibromas, and malignant peripheral nerve sheath tumors (MPNSTs), which comprise the most common cancers of the peripheral nervous system¹. Despite their shared embryonic origin, the clinical course and molecular drivers of Schwann cell tumors are distinct. Neurofibromas and schwannomas are benign tumors that can be cured with surgery or radiotherapy, while MPNSTs metastasize and are often incurable². Neurofibromas and MPNSTs are associated with loss of NF1, a tumor suppressor that inhibits Ras/Raf/MEK/ERK signaling. Schwannomas are associated with loss of NF2, a pleiotropic tumor suppressor that modulates numerous downstream effectors including PAK signaling, the Hippo pathway, apoptosis, contact inhibition, and the proteasome³. Germline loss of NF1 causes neurofibromatosis type-1 (NF-1)⁴, and germline loss of NF2 causes neurofibromatosis type-2 (NF-2)⁵, which are among the most common cancer predisposition syndromes.

MPNSTs are the most aggressive Schwann cell tumors and can arise from NF1-deficient plexiform neurofibromas in the setting of second hits disrupting epigenetic differentiation^6–8. Although rare, NF2-deficient schwannomas can also undergo malignant transformation⁹, but it remains unclear if NF1 and NF2 interact during Schwann cell tumorigenesis or treatment response. Moreover, despite the recent approval of the MEK inhibitor selumetinib to treat neurofibromas in patients with NF-1^10,11, there are currently no effective molecular therapies for patients with MPNSTs¹². To address these gaps in our understanding of Schwann cell biology and the unmet translational need for new therapies to treat Schwann cell tumors, we performed multiplatform bulk and single cell molecular profiling combined with biochemical, pharmacologic, and functional genomic interrogation of human Schwann cell tumors, patient-derived cell lines, and mouse xenografts. Our results demonstrate Schwann cell de-differentiation and functional interactions between NF1 and NF2 underlie malignant transformation and treatment resistance, elucidating a novel druggable dependency in Schwann cell tumors.

Multiplatform bulk and single-cell molecular profiling reveals de-differentiation underlies malignant transformation of Schwann cell tumors.

DNA methylation profiling provides robust classification of central nervous system tumors, but how this approach applies to peripheral nervous system tumors is incompletely understood¹³. To elucidate the epigenetic landscape of Schwann cell tumors, DNA methylation profiling was performed on schwannomas (n = 66), plexiform neurofibromas (n = 11), or MPNSTs (n = 42) from patients who were treated at a single institution from 1991 to 2021. Neuropathology review using the most recent World Health Organization criteria validated all histologic diagnoses¹⁴. Unsupervised hierarchical clustering revealed 3 DNA methylation groups (Fig. 1a, Extended Data Fig. 1a, and Supplementary Table 1). Group 1 tumors were enriched for MPNSTs and were distinguished by CNVs deleting SUZ12 or EED, obligate members of the PRC2 epigenetic complex that is recurrently lost in MPNSTs^6–8 (Fig. 1a and Extended Data Fig. 1b-e). Both Group 1 and 2 tumors harbored CNVs deleting CDKN2A/B, a tumor suppressor implicated in Ras-induced senescence that can be lost during neurofibroma transformation to MPNST including in the transitory premalignant stage defined as atypical neurofibromatous neoplasm of uncertain biologic potential (ANNUBP)^6,8,15−17 (Fig. 1a and Extended Data Fig. 1b-e). Group 3 tumors were enriched for schwannomas and distinguished by minimal CNVs, with the exception of recurrent CNVs deleting chromosome 22q, which harbors NF2 (Fig. 1a and Extended Data Fig. 1d-e). Chromosome 22q loss was also enriched Group 1 compared to 2 tumors (60% NF2 loss Group 1 versus 24% NF2 loss Group 2, p = 0.02, Chi-squared test), typically in combination with NF1 or PRC2 alterations (Fig. 1a and Extended Data Fig. 1e). Differential DNA methylation and principal component analysis across Schwann cell tumor groups showed hypermethylated sites in Group 2 tumors were enriched for PRC2 target genes while hypermethylated sites in Group 1 tumors were enriched for cell differentiation transcription factors (Extended Data Fig. 1f-h). In sum, these data suggest loss of PRC2 or loss of NF2 drive cellular de-differentiation and distinguish DNA methylation groups of differentiated (Group 2) or de-differentiated (Group 1) Schwann cell tumors.

To understand genetic and gene expression features distinguishing DNA methylation groups of Schwann cell tumors, whole exome sequencing (n = 34), RNA sequencing (n = 17), or immunohistochemistry (n = 36) were performed on Group 1 or Group 2 tumors. Whole exome sequencing identified recurrent short somatic variants in the core PRC2 components SUZ12 or EED in Group 1 but not Group 2 tumors (Fig. 1a, Extended Data Fig. 2, and Supplementary Table 2). RNA sequencing revealed transcriptomic signatures separated according to DNA methylation groups (Extended Data Fig. 3a and Supplementary Table 3), and differential expression analysis showed enrichment of Schwann cell differentiation genes (S100B, SOX10) and SUZ12 target genes in Group 2 compared to Group 1 tumors (Fig. 1a and Extended Data Fig. 3b-d). Immunohistochemistry for H3K27 trimethylation, an epigenetic marker of PRC2 activity, or immunohistochemistry for the Schwann cell differentiation markers S100B or SOX10, demonstrated loss of all three in Group 1 tumors compared to Group 2 tumors (Fig. 1a, b and Extended Data Fig. 3e).

Intratumor heterogeneity represents a challenge for histologic diagnosis of neurofibroma or MPNST¹⁴ (Extended Data Fig. 3f). To define the cellular architecture across groups of NF1-deficient Schwann cell tumors, single nuclear RNA sequencing was performed on 19,276 nuclei from Group 1 (n = 3) or Group 2 (n = 3) tumors (Supplementary Table 4). Uniform manifold approximation and projection (UMAP) revealed a total of 13 cell clusters that were defined using cell signature gene sets or differentially expressed cluster marker genes¹⁸. Five clusters were shared across Group 1 and Group 2 tumors, and the remaining 7 clusters were distinct to Group 1 or Group 2 tumors (Fig. 1c and Extended Data Fig. 4a). All shared clusters corresponded to non-tumor cell populations such as smooth muscle, pericytes, macrophages, lymphocytes, or endothelia (Fig. 1c-d and Extended Data Fig. 4b-c). Group 1 tumor single-cell clusters were distinguished by enrichment of de-differentiation markers (MEST, POSTN, DLK1, SOX4) or proliferation markers (KI67, TOP2A) (Fig. 1d and Extended Data Fig. 4c). Group 2 tumor single-cell clusters were distinguished by enrichment of differentiation markers (SRPX2, TNXB) or suppression of proliferation markers (Fig. 1d and Extended Data Fig. 4c). Assignment of cell cycle phases based on consensus gene signatures demonstrated enrichment of G2/M phase cells in Group 1 tumors compared to Group 2 tumors (Fig. 1e and Extended Data Fig. 4c). Among non-tumor single-cell clusters, the Schwann cell cluster was comprised almost entirely of cells from Group 2 tumors (Fig. 1d and Extended Data Fig. 4b-c), consistent with enrichment of Schwann cell differentiation markers in Group 2 compared to Group 1 tumors from bulk RNA sequencing or immunohistochemistry (Fig. 1a, b). Moreover, Schwann cell differentiation markers were suppressed and PRC2 target genes were enriched in single-nuclear RNA sequencing of Group 1 tumors compared to Group 2 tumors (Fig. 1f).

Schwann cell de-differentiation underlies MEK inhibitor resistance

To determine if PRC2 drives Schwann cell tumor differentiation, we analyzed a panel of patient-derived neurofibroma or MPNST cells with NF1 and PRC2 mutations (Extended Data Fig. 5a). RNA sequencing showed enrichment of PRC2 target genes and suppression of differentiation genes in MPNST cells compared to neurofibroma cells (Extended Data Fig. 5b and Supplementary Table 5). Integration of RNA sequencing data from CRISPR PRC2 knockout neurofibroma cells¹⁹ revealed both MPNST and PRC2-deficient neurofibroma cells exhibited suppression of Schwann cell differentiation markers (S100B, SOX10), enrichment of de-differentiated early neural crest markers (EN1, SOX9, FOXF1), and enrichment of Ras/Raf/MEK/ERK target genes (DUSP6, SPRY2, ETV4) (Extended Data Fig. 5c-e). Hierarchical clustering of all 16 patient-derived neurofibroma or MPNST cell lines based on a consensus PRC2 target gene set comprised of 24 differentiation, early neural crest, or Ras/Raf/MEK/ERK target genes segregated PRC2-intact from PRC2-deficient cells (Fig. 2a). Moreover, analysis of published H3K27 trimethylation ChIP-seq data¹⁹ revealed epigenetic de-repression of early neural crest markers (EN1, SOX9, FOXF1) but not Ras/Raf/MEK/ERK target genes (DUSP4, SPRY2, ETV4) in PRC2-deficient cells, suggesting direct epigenetic mechanisms underlie Schwann cell tumor de-differentiation (Extended Data Fig. 5f).

The MEK inhibitor selumetinib is effective for neurofibromas but shows mixed results for MPNSTs^10–12. Thus, we examined the relationship between cell differentiation and selumetinib response across patient-derived neurofibroma or MPNST cell lines. PRC2-deficient MPNST cells demonstrated resistance to selumetinib compared to PRC2-intact cells (Extended Data Fig. 5g, h). To determine if PRC2 inactivation was sufficient for selumetinib resistance, we used CRISPR interference (CRISPRi) to suppress SUZ12 or EED in NF1-deficient NF95.11b neurofibroma cells (Fig. 2b). Consistent with CRISPR PRC2 knockout data (Fig. 2a and Extended Data Fig. 5c-e), CRISPRi suppression of PRC2 inhibited Schwann cell differentiation markers and was sufficient for selumetinib resistance (Fig. 2b).

To interrogate cellular mechanisms underlying response to MEK inhibition, we performed biochemical or transcriptomic analyses of NF1-deficient NF95.11b neurofibroma cells after treatment with selumetinib (Supplementary Table 6). Bulk RNA sequencing confirmed repression of Ras/Raf/MEK/ERK target genes (DUSP4, SPRY2) and revealed loss of Schwann cell differentiation markers such as S100B after selumetinib treatment compared to vehicle treatment (Extended Data Fig. 6). Immunoblotting of neurofibroma cell lysates after selumetinib treatment showed an initial repression of ERK phosphorylation (pERK) and early induction of apoptosis (cleaved Caspase-3, cleaved Caspase-7), followed by recovery of pERK in cells that persisted despite continued selumetinib treatment (Fig. 2c). To determine if de-differentiation underlies MEK inhibitor resistance in vivo, we performed single-cell RNA sequencing on 65,314 cells from JW23.3 male MPNST xenografts²⁰ grown in female recipient mice that were treated with selumetinib or vehicle control (Supplementary Table 7). UMAP analysis revealed 13 clusters of cells (Fig. 2d and Extended Data Fig. 7a). Tumor microenvironment cells in female recipient mice were distinguished from male MPNST xenograft cells using Xist expression from the X chromosome (Extended Data Fig. 7b), leading to identification of 9 tumor microenvironment clusters that were defined using cell signature gene sets or differentially expressed cluster marker genes (Extended Data Fig. 7c-d). Analysis of the remaining 4 tumor cell clusters revealed suppression of Schwann cell differentiation markers in the one cluster that was enriched after selumetinib treatment compared to vehicle treatment (Fig. 2e, f). In sum, these data demonstrate Schwann cell de-differentiation underlies MEK inhibitor resistance in vitro and in vivo.

NF2 inactivation drives de-differentiation and MEK inhibitor resistance in NF1-deficient Schwann cell tumors

To identify druggable mechanisms underlying MEK inhibitor resistance, we performed triplicate genome-wide CRISPRi screens in NF1-deficient NF95.11b neurofibroma cells treated with selumetinib or vehicle control (Extended Data Fig. 8a and Supplementary Table 8). As expected, sgRNAs targeting tumor suppressor genes such as TP53 or negative regulators of the Ras pathway (RASA2, SPRY2) were enriched following selumetinib treatment, consistent with selumetinib resistance (Fig. 3a and Extended Data Fig. 8b-c). In contrast, sgRNAs targeting the cell cycle (CCNE1, E2F4) or differentiation (CDH2, SETD9, KDM1B) were depleted following selumetinib treatment, consistent with selumetinib sensitization (Fig. 3a and Extended Data Fig. 8b-c). sgRNAs suppressing NF2 were among the strongest drivers of selumetinib resistance (Fig. 3a), and Nf2 expression was suppressed in de-differentiated, selumetinib-resistant single-cells following selumetinib treatment of MPNST xenografts (Fig. 3b).

To validate NF2 as a driver of MEK inhibitor resistance and elucidate functional consequences of combined NF1 and NF2 loss in the Schwann cell lineage, we used CRISPRi to suppress NF2 in NF1-deficient NF95.11b neurofibroma cells (Fig. 3c and Supplementary Table 9). RNA sequencing validated NF2 suppression and revealed loss of Schwann cell differentiation markers, SUZ12, and SUZ12 target genes upon NF2 suppression (Fig. 3c and Extended Data Fig. 9a-b). Moreover, NF1/NF2-deficient neurofibroma cells were resistant to selumetinib compared to NF1-deficient, NF2-intact neurofibroma cells (Fig. 3c).

Epigenetic mechanisms regulating cell differentiation remain challenging pharmacologic targets²¹. In contrast, PAK activation following NF2 loss²² represents a druggable dependency²³ that can be inhibited by small molecules under clinical development²⁴. In support of this hypothesis, CRISPRi suppression of NF2 in NF1-deficient NF95.11b neurofibroma cells induced PAK phosphorylation (pPAK) (Fig. 3d). High pPAK was observed in NF1/SUZ12-deficient ST88-14 MPNST cells independent of CRISPRi suppression of NF2 (Extended Data Fig. 9c-d and Supplementary Table 10), suggesting PAK activation occurs upon de-differentiation of Schwann cell tumors. NF1/NF2-deficient NF95.11b neurofibroma cells maintained pERK activity (Fig. 3e) and cell viability (Fig. 3c) in response to selumetinib compared to NF2-intact neurofibroma cells, suggesting loss of NF2 is sufficient to drive selumetinib resistance. In contrast, the PAK1 inhibitor NVS-PAK1-1 repressed pPAK1 and pERK in both NF2-intact and NF2-deficient NF95.11b neurofibroma cells (Fig. 3e). To determine if PAK inhibition was effective in vivo, we treated MPNST xenografts with vehicle, selumetinib, NVS-PAK1-1, or selumetinib plus NVS-PAK1-1 and found combination molecular therapy additively inhibited xenograft growth compared to vehicle or either selumetinib or NVS-PAK1-1 monotherapy (Fig. 3f).

Neurofibromatosis was first reported over 140 years ago²⁵, and the NF1 gene product’s function as a Ras GAP was determined over 30 years ago^26,27. Molecular therapies have only recently become standard of care for patients with neurofibromatosis^10,11,28, but with recent reports of the NF1 protein structure^29–31 additional therapeutic strategies are on the horizon. Given the fact that components of growth factor signaling and the Ras pathway, such as NF1, are mutated in nearly half of human cancers³², such translational insights may be transformative for clinical oncology. Here, we identify 3 DNA methylation groups of neurofibromatosis-associated peripheral nervous system tumors distinguished by cellular de-differentiation mechanisms underlying malignant transformation and treatment resistance. Loss of the epigenetic regulator PRC2 is sufficient to drive de-differentiation, linking Schwann cell development and treatment resistance. NF2 inactivation and pPAK induction drive de-differentiation and selumetinib resistance in NF1-deficient tumors, shedding light on a novel druggable dependency underlying the most common tumors of the peripheral nervous system. In sum, our data provide a functional link between the NF1 and NF2 tumor suppressor genes.

Despite the inarguable clinical success of selumetinib, additional therapies for patients with MPNSTs or recurrent neurofibromas or schwannomas are needed. Thus, our identification of PAK as a target for combination molecular therapy to treat Schwann cell tumors has potential clinical implications, particularly given the challenges of directly targeting epigenetic or de-differentiation mechanisms in patients. Beyond PAK inhibition, other signaling mechanisms downstream of NF2 loss such as the Hippo Pathway³³, the Rho/Rac/Cdc42 family of small GTPases³⁴, or IRF-mediated apoptosis³⁵ may also be candidates for future combination molecular therapies to treat neurofibromatosis-associated tumors³⁴.

Several key questions remain regarding de-differentiation and treatment resistance in neurofibromatosis-associated tumors. First, does cellular de-differentiation underlie MEK inhibitor resistance in human patients with NF-1, and how does the single cell architecture of tumor cell states or non-tumor cell types change in response to MEK inhibition in humans? As MEK inhibition becomes more common for patients with NF-1, serial molecular analyses will be critical to unravel mechanisms of treatment response or resistance to optimize combination or sequential molecular therapies. Second, is de-differentiation mediated malignant transformation or treatment resistance of neurofibromatosis-associated peripheral nervous system tumors a relevant mechanism for neurofibromatosis-associated central nervous system tumors, such as gliomas? Ongoing clinical trials are evaluating the efficacy of MEK inhibition for gliomas arising in patients with NF-1 (ClinicalTrials.gov NCT03871257), and multiplatform molecular analysis of these precious samples will no doubt help determine if the relationships between activation of the Ras family of small GTPases following NF1 loss, and induction of the Rho/Rac/Cdc42 family of small GTPases following NF2 loss, are conserved across oncologic contexts. If so, the strategy of dual Ras/Raf/MEK/ERK and Rho/Rac/Cdc42/PAK inhibition may be broadly useful across other Ras-mutated tumors, which comprise the most common molecular alterations in human cancers³².

Acknowledgments

We thank Angie Hirbe from the Washington University in St. Louis for providing JW23.3 mouse xenograft cells, Tomoko Ozawa and the staff of the University of California San Francisco (UCSF) Brain Tumor Center Preclinical Therapeutics Core, Anny Shai and the staff of the UCSF Brain Tumor Center Biospecimen and Pathology Core, and Eric Chow and the staff of the UCSF Center for Advanced Technology. We appreciate the thoughtful comments and critiques from members of the McCormick or Raleigh laboratories. This work was supported by a Children’s Tumor Foundation Young Investigator Award and a Francis Collins Scholar Award to H.N.V., NIH grant R35 CA197709 and DOD/CMRP award WH2010129 to F.M., and DOD/CDMRP award NF200021 to D.R.R.

Author contributions statement

All authors made substantial contributions to the conception or design of the study; the acquisition, analysis, or interpretation of data; or drafting or revising the manuscript. All authors approved the manuscript. All authors agree to be personally accountable for individual contributions and to ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated, resolved, and the resolution documented in the literature. H.N.V. designed, performed, and analyzed all experiments and bioinformatic analyses. E.P. performed biochemistry, qPCR, cell viability assays, and mouse xenograft experiments. C.D performed single-cell RNA sequencing and assisted with bioinformatic analysis. M.J.S. performed biochemistry, pharmacology, and helped with experimental design. S.J.L. assisted with bioinformatic analysis, CRISPRi cell line generation, and genome-wide screening. S.L. performed qPCR analysis and provided critical mouse support. C.H.L. performed pathology review and assisted with processing human tumor samples for genomic analyses. C.D.E performed biochemistry and assisted with mouse xenograft experiments. T.C.C. performed qPCR, immunofluorescence, and assisted with mouse xenograft experiments. S.T.M. extracted nucleic acids from tumor specimens and guided experimental design from patient samples. W.C.C. helped assemble human tumor resection specimens and assisted with bioinformatic analysis. A.T.R., S.E.B., and L.J. provided key insight into study design and provided clinical data. A.P. and M.P. assembled tumor resection specimens, provided clinical data, supervised C.H.L. for pathologic review, and assisted with study design. A.R.A. supervised H.N.V and C.D. for single cell sequencing experiments and aided with genomic analysis. F.M. and D.R.R. conceived, designed, and supervised the study.

Competing interests statement

The authors declare no competing interests.

Tables

Not applicable

This study complied with all relevant ethical regulations and was approved by the UCSF Institutional Review Board (13-12587, 17-22324, 17-23196, 18-24633). As part of routine clinical practice at UCSF, all patients included in this study signed an informed consent waiver to contribute de-identified data to scientific research projects.

Nucleic acid extraction for DNA methylation profiling, whole exome sequencing, or bulk RNA sequencing

DNA and RNA were isolated from cell lines, human samples, or mouse xenografts using the All-Prep Universal Kit (#80224, QIAGEN). For fresh frozen human samples or mouse xenografts, specimens were thawed in RLT Plus Buffer with beta-mercaptoethanol. Formalin fixed paraffin embedded tissue was de-paraffinized. All tumor or tissue samples were mechanically lysed using a TissueLyser II (QIAGEN) with stainless steel beads at 30 Hz for 90 seconds. QiaCubes were used for standardized automated nucleic acid extraction per the manufacturer’s protocol. For cell line samples, pellets were directly lysed in RLT Plus buffer with beta-mercaptoethanol. RNA quality was assessed by chip-based electrophoresis on a BioAnalyzer 2100 using the RNA 6000 Nano Kit (#5067-1511, Agilent Technologies), and clean-up was performed as needed using the RNeasy kit (QIAGEN). DNA quality was assessed by spectrophotometry, and clean-up was performed as needed using DNA precipitation. Only samples with high-quality DNA (A260/280>1.8, A260/230>1.6) and/or RNA (RIN>8) were used for DNA methylation profiling, whole exome sequencing, or bulk RNA sequencing.

DNA methylation profiling and analysis

Genomic DNA from human tumors were processed for methylation analysis using the Illumina Methylation EPIC Beadchip (#WG-317-1003, Illumina) according to the manufacturer’s instructions. Preprocessing and normalization were performed in R using the minfi Bioconductor package^36,37. Only probes with detection p<0.05 in all samples were included for further analysis. Data were normalized using functional normalization³⁷. Probes were filtered based on the following criteria: (i) removal of probes mapping to the X or Y chromosomes, (ii) removal of probes containing a common single nucleotide polymorphism (SNP) within the targeted CpG site or on an adjacent basepair, and (iii) removal of probes not mapping uniquely to the hg19 human reference genome. DNA methylation-based molecular neuropathology brain tumor classification¹³ or CNV estimation¹³ were performed as previously described.

Whole exome sequencing and analysis

Library preparation, exome capture, and sequencing were performed at the Institute for Human Genetics at UCSF. Sequencing libraries were prepared using the Kapa Hyper Prep Kit (#07962312001, Roche) and exome capture was performed using the Nimblegen SeqCap EZ Human Exome Kit v3.0 (Roche). Paired end sequencing with read length 100 base pairs was performed on an Illumina HiSeq4000. Whole exome data were analyzed following the Genome Analysis Toolkit (GATK) best practices^38,39. Raw FASTQ files were aligned to the reference genome with Bowtie2⁴⁰. Only uniquely aligned reads were included for further processing using the Genome Analysis Toolkit to carry out de-duplication, local realignment, and base quality score recalibration. Alignment quality metrics and header information were determined using the Picard suite (http://broadinstitute.github.io/picard/). Somatic variants (point mutations, small indels) were identified from matched tumor-normal samples (n=15) and using a panel of normal (PoN) samples with Mutect2, per GATK best practices, when a matched normal sample was not available (n=19). Variants were annotated using Snpeff⁴¹ and were further filtered to include only those marked as high/moderate/low priority, only those occurring in protein coding or splice site locations, and only those meeting the following hard filters: (i) >5 reads in tumor compared to normal samples, (ii) >10% variant reads in tumor, and (iii) >90% reference reads in normal. The full list of parameters and filters can be found in the headers of the VCF files that will be deposited in GEO.

RNA sequencing

Library preparation was performed using the TruSeq RNA Library Prep Kit v2 (#RS-122- 2001, Illumina) and 50 bp single end reads were sequenced on an Illumina HiSeq 2500 or NovaSeq to a minimum depth of 25M reads per sample at Medgenome, Inc. Quality control of FASTQ files was performed with FASTQC, and after trimming of adapter sequences, reads were filtered to remove bases that did not have an average quality score of 20 within a sliding window across 4 bases (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were mapped to the appropriate reference genome (hg19 or mm10) using HISAT2 with default parameters⁴². Transcript abundance estimation in transcripts per million (TPM) and differential expression analysis were performed using DESeq2⁴³. Differentially expressed transcripts with an adjusted p-value<0.1 were identified and filtered based on an expression cutoff (TPM>1) and a fold change threshold (log2FC>1) to prioritize biologically relevant gene sets. Clustering dendrograms and heatmaps were generated in R using TPM values and plotted as normalized row expression values with the heatmaps.2 function.

Immunohistochemistry (IHC)

IHC was performed as previously described⁴⁴ using formalin-fixed, paraffin-embedded tissue sections from tumor resection specimens on a combination of whole slide sections or tissue microarrays using the following primary antibodies: H3K27me3 (Cell Signaling Technology, #9733, clone C36B11, 1:50 dilution), SOX10 (Cell Marque, #383R-1, clone EP268, 1:50 dilution), S100B (Ventana, #760-2523, 1:2 dilution), or p53 (Dako, #GA61661-2, clone DO-7, 1:50 dilution). All IHC was performed on a Ventana Benchmark XT automated stainer (Roche) using standard techniques. IHC studies that were previously performed as part of clinical diagnostic workup, or stains obtained as part of prior research studies were reviewed for protein expression concordance⁴⁴. For quantitative analysis, percent staining for H3K27me3, SOX10, or S100B was estimated as the percentage of positive tumor cells on available stained tissue.

Single-nuclear or single-cell RNA sequencing and analysis

Frozen human neurofibroma or MPNST resection specimens were thawed on ice, minced with sterile razor blades, and mechanically dounced on ice in cell lysis nuclei extraction buffer until all macroscopically visible tissue dissolved into suspension. Cell suspensions were filtered through a 50 mm filter, centrifuged at 500g for 5 minutes at 4 degrees Celsius, and resuspended in 0.1% BSA in PBS. Nuclei were stained using DAPI (#D3571, Thermo Fisher Scientific) and counted. A total of 10,000 nuclei were loaded per single-nuclei RNA sequencing sample. For MPNST mouse xenograft single-cell RNA sequencing, tumors were minced with sterile razor blades and enzymatically dissociated with papain (#LS003, Worthington) at 37 degrees Celsius for 45 minutes. Samples were centrifuged at 500g for 5 minutes, resuspended in RBC lysis buffer (#00-4300-54, eBioscience), incubated for 10 minutes at room temperature, and resuspended in 5% FBS in PBS. Cell suspensions were serially filtered through 70 mm and 40 mm filters before being resuspended again in 5% FBS in PBS for manual cell counting using a hemacytometer. For cell culture single-cell RNA sequencing, cells were trypsinized and resuspended in 0.04% BSA in PBS before manual cell counting using a hemacytometer. A total of 10,000 cells were loaded per single-cell RNA sequencing sample.

Single-nuclei or single-cell RNA sequencing was performed using the Chromium Single Cell 3’ Library & Gel Bead Kit v3.1 on a 10x Chromium controller (10X Genomics) using the manufacturer recommended default protocol and settings. Samples were sequenced on an Illumina NovaSeq at the UCSF Center for Advanced Technology, and the resulting FASTQ files were processed using the CellRanger analysis suite (https://github.com/10XGenomics/cellranger) for alignment to the hg38 reference genome, identification of empty droplets, and determination of a count threshold. All downstream analyses were performed in Seurat (https://satijalab.org/seurat/) using the default pipeline. In brief, data were empirically filtered on a per sample basis to remove outliers with regard to gene count, UMI count, or mitochondrial genes followed by cluster identification, UMAP generation, and marker gene list generation using computed highly variable features and the top ten principal component dimensions as previously described^35,45. The Schwann cell differentiation gene module (GO 0014037) and PRC2 target gene module (M8448) were downloaded from www.gsea-msigdb.org.

Cell culture, cell viability assays, and in vitro pharmacology

Patient-derived neurofibroma (NF95.11b, NF95.6) or MPNST (SNF02.2, SNF94.3, SNF96.2, ST88-14) cell lines⁴⁶ were obtained from the Neurofibromatosis Therapeutic Acceleration Program or American Type Culture Collection. Cell lines were grown in Dulbecco’s Modified Eagle Medium (#11960069, Life Technologies) with 10% FBS and 1X Pen-Strep (#15140122, Life Technologies). Cell lines were regularly tested and verified to be mycoplasma negative (#LT07-218, Lonza). Viability assays were carried out with the CellTiter 96 Non-Radioactive Cell Proliferation Assay (#G410, Promega) and a Glomax Discovery Multimode Microplate Reader (Promega). For pharmacologic assays, cells were seeded at a density of 5,000 cells per well in a 96 well plate the night prior to treatment, after which cells were treated with drugs at the indicated concentrations for the indicated periods (or 48 hours if not indicated) prior to experimentation.

Immunoblotting

Whole cell lysates were harvested using RIPA buffer (50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 0.5% Deoxycholate, 0.1% SDS, 1% IGEPAL CA-630) with fresh protease (#P8340, Sigma) and phosphatase inhibitor (#P2850, Sigma) cocktails. A total of 10-20 mg of protein was loaded into pre-cast NuPAGE electrophoresis gels (Life Technologies). Samples were separated by SDS-PAGE, transferred to nitrocellulose or PVDF membranes, and blocked in either 5% bovine serum albumin or 5% skim milk in TBS buffer for 1 hour at room temperature. Primary antibodies were incubated overnight at the indicated dilution at 4 degrees Celsius and HRP conjugated secondary antibodies were incubated for 1 hour at room temperature followed by ECL based detection on film. A minimum of 2 biologic replicates were performed for all experiments. The following antibodies were used: pERK (Cell Signaling Technologies, #4370, 1:1,000 dilution), beta tubulin (Developmental Hybridoma Studies Bank, #E7, 1:10,000 dilution), pAkt (Cell Signaling Technologies, #4060, 1:1,000 dilution), pMEK (Cell Signaling Technologies, #9121, 1:1,000 dilution), pPAK (Cell Signaling Technologies, #2601, 1:1,000 dilution), Caspase-3 (Cell Signaling Technologies, #9662, 1:1,000 dilution), Caspase-7 (Cell Signaling Technologies, #9492, 1:1,000 dilution), or NF2 (Abcam, #ab88957, clone AF1G4, 1:2,000 dilution).

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

RNA was extracted from cell lines using the RNeasy Mini Kit (#74106, QIAGEN) according to manufacturer’s instructions, and cDNA was synthesized from RNA using iScript cDNA Synthesis kit (#1708891, Bio-Rad). Real-time qPCR was performed using PowerUp SYBR Green Master Mix (#A25918, Thermo Fisher Scientific) on a QuantStudio 6 Flex Real Time PCR system (Life Technologies). The following qRT-PCR primers were used: GAPDH-F (5’-GTCTCCTCTGACTTCAACAGCG-3’), GAPDH-R (5’-ACCACCCTGTTGCTGTAGCCAA-3’), SUZ12-F (5’-AGGCTGACCACGAGCTTTTC-3’), SUZ12-R (5’-GGTGCTATGAGATTCCGAGTTC-3’), EED-F (5’-GTGACGAGAACAGCAATCCAG-3’), EED-R (5’-TATCAGGGCGTTCAGTGTTTG-3’), NF2-F (5’-TTGCGAGATGAAGTGGAAAGG-3’), NF2-R (5’-CAAGAAGTGAAAGGTGACTGGTT-3’), S100B-F (5’-TGGCCCTCATCGACGTTTTC-3’), or S100B-R (5’-ATGTTCAAAGAACTCGTGGCA-3’).

CRISPRi cell line generation and genome-wide screening

Lentivirus containing pMH0001 (UCOE-SFFV-dCas9-BFP-KRAB, #85969, Addgene) was produced from transfected HEK293T cells with packaging vectors (pMD2.G #12259, Addgene, and pCMV-dR8.91, Trono Lab) following the manufacturers protocol (#MIR6605, Mirus). Neurofibroma NF95.11b cells were stably transduced to generate parental NF95.11b^{dCas9-KRAB-BFP}cells and selected by flow cytometry using a SH800 sorter (Sony). Subsequent gene specific knockdowns were achieved by individually cloning single-guide RNA (sgRNA) protospacer sequences into the pCRISPRia-v2 vector (#84832, Addgene) between BstXI and BlpI restriction sites. All constructs were validated by Sanger sequencing of the protospacer region. The following protospacers were used: sgNTC (GTGCACCCGGCTAGGACCGG), sgSUZ12-1 (GCTGAAACGTCTTTGGAAGG), sgSUZ12-2 (GGCAGCGGGTCGGAGATCGA), sgEED-1 (GAGTCTAGAGCCACCGTCCA), sgEED-2 (GCAGGGAGCAGGTAGCTGCT), sgNF2-1 (GTCGGGACGGGACCCCTAGA), or sgNF2-2 (GGACTCCGCGCGCCTCTCAG). Lentivirus was generated as described above and cells were selected to purity using 1 mg/mL puromycin for at least 5 days.

For genome-wide CRISPRi screening, a genome-wide dual sgRNA library was constructed by cloning the top 2 on-target sgRNAs for 23,483 genes as described previously⁴⁷ into the library expression vector pU6-sgRNA Ef1alpha Puro-T2A-GFP derived from pJR85 (#140095, Addgene) and modified to express a second sgRNA using the human U6 promoter. 1137 non-targeting sgRNA pairs were also included as negative controls in the screen. To generate lentiviral pools, HEK293T cells were transfected with the sgRNA library along with packaging plasmids as described, and viral supernatant was collected 72 hours following transfection. Lentiviral libraries were infected into NF95.11b^{dCas9-KRAB-BFP}cells, cultured for 2 days following infection, selected in 1 mg/mL puromycin for 2 days, and then allowed to recover in 10% FBS in DMEM for 1 day. Infection efficiency was evaluated by measuring BFP positivity on flow cytometry, and cell pellets were subsequently frozen down at this “T0” timepoint. The screen was subsequently carried out in biologic triplicate, with cells cultured in either 1 mM selumetinib or vehicle (DMSO) control for 10 days. Cell pellets were frozen down at this “T10” timepoint and processed for sgRNA abundance library preparation using Q5 High-Fidelity DNA Polymerase (NEB) and sequenced on an Illumina NextSeq-500 as previously described⁴⁸. Enrichment or depletion of sgRNA abundances were determined by down sampling trimmed sequencing reads to equivalent amounts across all samples, and then calculating the log2 ratio of sgRNA abundance in experimental conditions to sgRNA abundance in control conditions at T10, or between sequencing reads from T10 and T0 timepoints within experimental or control conditions. Statistical significance was calculated using Student’s t-test comparing replicates across conditions. sgRNAs with fewer than 100 reads at T0 were removed from subsequent analysis. Hits were selected by normalizing log2 ratios by the standard deviations of the non-targeting control sgRNAs, and genes were considered significant at a ‘threshold’ >5 calculated as the |normalized log2 ratios| * -log10(p-value) to prioritize biologically meaningful hits for mechanistic or functional investigation.

Mouse tumor xenografts and in vivo pharmacology

The study was approved by the UCSF Institutional Animal Care and Use Committee (AN174769) and all experiments were conducted in compliance with institutional and governmental regulations. Subcutaneous xenografts were performed by implanting 5 million JW23.3 MPNST xenograft cells into the flanks of 5-6 week old female NU/NU mice (Harlan Sprague Dawley). For pharmacologic experiments, mice were treated with 25 mg/kg selumetinib twice-daily by oral gavage in 0.5% methylcellulose solution with 0.2% v/v Tween-80, 100 mg/kg 1-ABT followed by 10mg/kg NVS-PAK1-1 2 hours later in 60% PEG400/40% water, or vehicle control gavaged once daily. Tumors were measured using calipers 3 times per week.

Statistical Analysis

All experiments were performed as repeated, independent biologic replicates, and statistics were derived from biologic replicates. The number of biologic replicates is indicated in each panel or figure legend. No statistical methods were used to predetermine sample sizes. Considering the rarity of MPNSTs and accounting for the number of genomic approaches used in this study, our total cohort size is similar to prior publications^6–8. The clinical samples used were retrospective and non-randomized, and all samples were equally interrogated within the constraints of sufficient tissue for each analytical method. Cells and animals were randomized to experimental conditions, and no clinical, molecular, cellular, or animal data points were excluded from analysis. Unless otherwise specified, data are plotted as mean with error bars representing the standard error of the mean. The statistical tests of choice were selected based on the input data and are noted in the methods and figure legends. All statistical tests were one-sided. Where appropriate, multiple hypothesis testing corrections were performed. Statistical significance thresholds are indicated in each figure legend and exact p-values are provided when possible.

Data availability

Human tumor DNA methylation (n=119), RNA sequencing (n=41), whole exome sequencing (n=34), or single-cell RNA sequencing data (n=6) reported in this manuscript will be deposited in the NCBI Gene Expression Omnibus. Cell line RNA-sequencing (n=6), selumetinib-treated cell line RNA-sequencing (n=10), CRISPRi NF2-deficient cell line RNA-sequencing (n=4), or single-cell RNA sequencing of mouse xenograft data (n=5) reported in this manuscript will similarly be deposited in the NCBI Gene Expression Omnibus. Additional RNA-sequencing and H3K27 trimethylation ChIP sequencing data from previously reported PRC2-intact or PRC2-deficient neurofibroma cell lines is available under GSE 118185 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118185) or GSE118183 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118183), respectively¹⁹. The publicly available GRCh37 (hg19, https://www.ncbi.nlm.nih.gov/assembly/GCF_000001405.13/) and GRCm38 datasets (mm10, https://www.ncbi.nlm.nih.gov/assembly/GCF_000001635.20/) were used in this study.

Code availability

The open-source software, tools, and packages used for data analysis in this study, as well as the version of each program, were ImageJ (v2.1.0), R (v3.5.3 and v3.6.1), FASTQC (v0.11.9), HISAT2 (v2.1.0), featureCounts (v2.0.1), Bowtie2 (v2.3), snpEff (v5.1), Mutect2 (v4.0), picard (v2.2), cellranger (v6.1.2), Seurat R package (v3.0.1), DESeq2 (Bioconductor v3.10), minfi (Bioconductor v3.10), ConsensusClusterPlus (Bioconductor v3.10), Heatmap.2 R package (gplots v3.13), and ggplot2 (v3.3.6). No custom software, tools, or packages were used.

Ostrom, Q. T., Cioffi, G., Waite, K., Kruchko, C. & Barnholtz-Sloan, J. S. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2014–2018. Neuro. Oncol. 23, III1–III105 (2021).
Farid, M; Demicco, EG; Garcia, R; Ahn, L; Merola, PR; Ciofi, A; Maki, R. Malignant Peripheral Nerve Sheath Tumors (MPNST). Oncologist 19, 193–201 (2014).
Vasudevan, H. N., Lucas, C. H. G., Villanueva-Meyer, J. E., Theodosopoulos, P. V. & Raleigh, D. R. Genetic Events and Signaling Mechanisms Underlying Schwann Cell Fate in Development and Cancer. Neurosurgery 88, 234–245 (2021).
Gutmann, D. H. et al. Neurofibromatosis type 1. Nat. Rev. Dis. Prim. 3, 1–17 (2017).
Asthagiri, A. R. et al. Neurofibromatosis type 2. The Lancet vol. 373 1974–1986 (2009).
Lee, W. et al. PRC2 is recurrently inactivated through EED or SUZ12 loss in malignant peripheral nerve sheath tumors. Nat. Genet. 46, 1227–1232 (2014).
De Raedt, T. et al. PRC2 loss amplifies Ras-driven transcription and confers sensitivity to BRD4-based therapies. Nature (2014) doi:10.1038/nature13561.
Zhang, M. et al. Somatic mutations of SUZ12 in malignant peripheral nerve sheath tumors. Nat. Genet. 46, 1170–1172 (2014).
Seferis, C., Torrens, M., Paraskevopoulou, C. & Psichidis, G. Malignant transformation in vestibular schwannoma: report of a single case, literature search, and debate. J. Neurosurg. 121, 160–166 (2014).
Dombi, E. et al. Activity of Selumetinib in Neurofibromatosis Type 1–Related Plexiform Neurofibromas. N. Engl. J. Med. 375, 2550–2560 (2016).
Gross, A. M. et al. Selumetinib in Children with Inoperable Plexiform Neurofibromas. N. Engl. J. Med. (2020) doi:10.1056/NEJMoa1912735.
Wu, L. M. N. & Lu, Q. R. Therapeutic targets for malignant peripheral nerve sheath tumors. Future Neurol. 14, (2019).
Capper, Jones, S. DNA methylation-based classification of central nervous system tumours. Nature 555, 469–474 (2018).
Louis, D. N. et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro. Oncol. 23, 1231–1251 (2021).
Rhodes, S. D. et al. Cdkn2a (Arf) loss drives NF1-associated atypical neurofibroma and malignant transformation. Hum. Mol. Genet. 28, 2752–2762 (2019).
Miettinen, M. M. et al. Histopathologic evaluation of atypical neurofibromatous tumors and their transformation into malignant peripheral nerve sheath tumor in patients with neurofibromatosis 1—a consensus overview. Hum. Pathol. 67, 1–10 (2017).
Pemov, A. et al. Low mutation burden and frequent loss of CDKN2A/B and SMARCA2, but not PRC2, define premalignant neurofibromatosis type 1-associated atypical neurofibromas. Neuro. Oncol. 21, 981–992 (2019).
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. U. S. A. 102, 15545–15550 (2005).
Wassef, M. et al. EZH1/2 function mostly within canonical PRC2 and exhibit proliferation-dependent redundancy that shapes mutational signatures in cancer. (2019) doi:10.1073/pnas.1814634116.
Qin, W. et al. TYK2 promotes malignant peripheral nerve sheath tumor progression through inhibition of cell death. Cancer Med. 8, 5232–5241 (2019).
Mohammad, H. P., Barbash, O. & Creasy, C. L. Targeting epigenetic modifications in cancer therapy: erasing the roadmap to cancer. Nat. Med. 2019 253 25, 403–418 (2019).
Kissil, J. L. et al. Merlin, the Product of the Nf2 Tumor Suppressor Gene, Is an Inhibitor of the p21-Activated Kinase, Pak1. Mol. Cell 12, 841–849 (2003).
Radu, M., Semenova, G., Kosoff, R. & Chernoff, J. PAK signalling during the development and progression of cancer. Nat. Rev. Cancer 2014 141 14, 13–25 (2013).
Hawley, E. et al. PAK1 inhibition reduces tumor size and extends the lifespan of mice in a genetically engineered mouse model of Neurofibromatosis Type 2 (NF2). Hum. Mol. Genet. 30, 1607–1617 (2021).
Von Recklinghausen, F. D. Uber ide multiplen Fibrome der Haut und ihre beziehung zu den multiplen Neuromen. Berlin: Hirschwald. 3–18 (1882).
Ballester, R. et al. The NF1 locus encodes a protein functionally related to mammalian GAP and yeast IRA proteins. Cell 63, 851–859 (1990).
Martin, G. A. et al. The GAP-related domain of the neurofibromatosis type 1 gene product interacts with ras p21. Cell 63, 843–849 (1990).
Jessen, W. J. et al. MEK inhibition exhibits efficacy in human and mouse neurofibromatosis tumors. J. Clin. Invest. 123, 340–347 (2013).
Chaker-Margot, M. et al. Structural basis of activation of the tumor suppressor protein neurofibromin. Mol. Cell 82, 1288–1296.e5 (2022).
Lupton, C. J. et al. The cryo-EM structure of the human neurofibromin dimer reveals the molecular basis for neurofibromatosis type 1. Nat. Struct. Mol. Biol. 28, 982–988 (2021).
Naschberger, A., Baradaran, R., Rupp, B. & Carroni, M. The structure of neurofibromin isoform 2 reveals different functional states. Nat. 2021 5997884 599, 315–319 (2021).
Sanchez-Vega, F. et al. Oncogenic Signaling Pathways in The Cancer Genome Atlas. Cell 173, 321–337.e10 (2018).
Wu, L. M. N. et al. Programming of Schwann Cells by Lats1/2-TAZ/YAP Signaling Drives Malignant Peripheral Nerve Sheath Tumorigenesis. Cancer Cell 33, 292–308.e7 (2018).
JA, M. et al. Genetic disruption of the small GTPase RAC1 prevents plexiform neurofibroma formation in mice with neurofibromatosis type 1. J. Biol. Chem. 295, 9948–9959 (2020).
Choudhury, A. et al. Meningioma DNA methylation groups identify biological drivers and therapeutic vulnerabilities. Nat. Genet. 54, (2022).
Aryee, M. J. et al. Minfi: A flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 30, 1363–1369 (2014).
Fortin, J. P., Triche, T. J. & Hansen, K. D. Preprocessing, normalization and integration of the Illumina HumanMethylationEPIC array with minfi. Bioinformatics 33, 558–560 (2017).
DePristo, M. A. et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat. Genet. 43, 491–8 (2011).
Van der Auwera, G. A. et al. From fastQ data to high-confidence variant calls: The genome analysis toolkit best practices pipeline. Current Protocols in Bioinformatics (2013). doi:10.1002/0471250953.bi1110s43.
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff. Fly (Austin). 6, 80–92 (2012).
Kim, D., Langmead, B. & Salzberg, S. L. HISAT: a fast spliced aligner with low memory requirements. Nat. Methods 12, 357–60 (2015).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Lucas, C.-H. G. et al. Histopathologic findings in malignant peripheral nerve sheath tumor predict response to radiotherapy and overall survival. Neuro-Oncology Adv. 2, 1–12 (2020).
Magill, S. T. et al. Multiplatform genomic profiling and magnetic resonance imaging identify mechanisms underlying intratumor heterogeneity in meningioma. Nat. Commun. (2020) doi:10.1038/s41467-020-18582-7.
Li, H., Chang, L. J., Neubauer, D. R., Muir, D. F. & Wallace, M. R. Immortalization of human normal and NF1 neurofibroma Schwann cells. Lab. Investig. 96, 1105–1115 (2016).
Nuñez, J. K. et al. Genome-wide programmable transcriptional memory by CRISPR-based epigenome editing. Cell 184, 2503–2519.e17 (2021).
Horlbeck, M. A. et al. Mapping the Genetic Landscape of Human Cells. Cell 174, 953–967.e22 (2018).

There is NO Competing Interest.

VasudevanNatGenetEditorialChecklistv18.pdf
VasudevanNatGenetReportingSummaryv18.pdf
VasudevanNatGenetSupplementaryTablesv18.xlsx
NCOMMS2231725Trs.pdf
Reporting Summary
floatimage4.png
Extended Data Fig. 1. DNA methylation profiling of Schwann cell tumors. a, DNA methylation profiling and hierarchical clustering of the 2000 most variable probes across schwannomas (n=66), neurofibromas (n=12), or malignant peripheral nerve sheath tumors (MPNSTs) (n=41) reveals 3 epigenetic groups of Schwann cell tumors. See also Supplementary Table 1. b, Histologic distribution across Schwann cell tumor DNA methylation groups. c, Molecular neuropathology classification (MNP) based on Schwann cell tumor DNA methylation profiles accurately classifies schwannomas but not neurofibromas or MPNSTs¹³. d, Copy number variant (CNV) number in Group 1 (n= 25) compared to Group 2 (n=25) Schwann cell tumors. CNVs were derived from Schwann cell tumor DNA methylation profiles and were defined as amplifications or deletions comprising one-third or more of a chromosome arm. Lines represent means. Error bars represent standard error of the means. ***p£0.0001, Student’s t tests. e, CNV analysis shows loss of chromosome segments containing the PRC2 components SUZ12 (p<0.0001) or EED (p=0.0003), or CDKN2A/B (p=0.0222), but not TP53 (p=0.1962), were enriched in Group 1 compared to Group 2 tumors. Fisher’s exact tests. f, Differential DNA methylation and gene ontology analysis of Schwann cell tumors demonstrate hypermethylation of SUZ12 and EZH2 targets, both core components of PRC2, in Group 2 tumors (n=25) consistent with maintenance of PRC2 activity in this group, and hypermethylation of cell differentiation pathways in Group 1 tumors (n=25) consistent with epigenetic silencing of differentiation programs in this group. CHEA, chromatin immunoprecipitation enrichment analysis. g, Principal component analysis of Schwann cell tumor DNA methylation profiles reveals an epigenetic gradient across schwannomas (n=66), neurofibromas (n=12), and MPNSTs (n=41). h, Gene ontology analysis across Schwann cell tumor DNA methylation profiling principal component analysis reveals alterations in cell differentiation pathways associated with PRC2 target genes upon malignant transformation. In particular, negative PC1 hypermethylated genes are enriched for cell differentiation transcription factors, consistent with epigenetic silencing of these genes upon malignant transformation. In contrast, positive PC1 hypermethylated genes are enriched for SUZ12 and EZH2 targets, consistent with active PRC2 mediated repression of these genes in neurofibromas and MPNSTs.
floatimage5.png
Extended Data Fig. 2. Whole exome sequencing of Group 1 or Group 2 Schwann cell tumors. Summary matrix of short somatic variants identified in Group 1 or Group 2 Schwann cell tumors analyzed by whole exome sequencing. See also Supplementary Table 2.
floatimage6.png
Extended Data Fig. 3. RNA sequencing and immunohistochemical analysis of Schwann cell tumors. a, RNA sequencing and principal component analysis of Group 1 (n=9), Group 2 (n=9), or Group 3 Schwann cell tumors (n=23) segregates samples based on DNA methylation groups. b, Volcano plot of differentially expressed genes from RNA sequencing of Group 1 (n=9) compared to Group 2 (n=9) Schwann cell tumors. c, Gene ontology analysis of differentially expressed genes shows enrichment of cell differentiation pathways and suppression of PRC2 target genes in Group 2 (n=9) compared to Group 1 Schwann cell tumors (n=9). CHEA, chromatin immunoprecipitation enrichment analysis. d, Differential expression analysis shows suppression of Schwann cell differentiation marker genes (S100B, SOX10) in Group 1 (n=9) compared to Group 2 (n=9) Schwann cell tumors. TPM, transcripts per million. ***p£0.0001, Student’s t tests. e, Quantitative immunohistochemical analysis of Group 1 (n=23) or Group 2 (n=13) Schwann cell tumors shows loss of H3K27me³ (p<0.0001) or S100B (p=0.0049), but not p53 (p=0.7356), in Group 1 compared to Group 2 tumor. Fisher’s exact tests. f, Tumor purity analysis from DNA methylation profiling of Group 1 or Group 2 Schwann cell tumors. ***p£0.0001, Student’s t tests. Lines represent means. Error bars represent standard error of the means. See also Supplementary Table 3.
floatimage7.png
Extended Data Fig. 4. Single-nuclear RNA sequencing of Group 1 or Group 2 Schwann cell tumors identified 13 cell clusters defined by consensus MSigDB gene expression programs or cluster marker genes. a, Single cell counts across UMAP clusters from each human Schwann cell tumor analyzed using single-nuclear RNA sequencing. Tumor histology and DNA methylation groups are annotated. b, Feature plots showing MSigDB gene expression programs for tumor microenvironment single-cell clusters corresponding to the endothelial differentiation gene set or the lymphocyte activation gene set. c, Feature plots for specific marker genes across UMAP cell clusters. Broadly, non-tumor cell types were distinguished by marker genes for pericytes (NOTCH3, PDGFRB), smooth muscle cells (SFRP4), lymphocytes (CD2, IL7R), endothelial cells (VWF, CD163), or Schwann cells (S100B, MPZ). Tumor cell clusters were defined based on top marker gene expression. For single-cell clusters from Group 1 tumors, such an approach led to the delineation of proliferating cells (MKI67, TOP2A), POSTN Group 1 tumor cells, MEST Group 1 tumor cells, DLK1 Group 1 tumor cells, or SOX4 Group 1 tumor cells. For single-cell clusters from Group 2 tumors, no proliferating compartment was found but SPRX2 or TNXB Group 2 tumor cell clusters were identified. See also Supplementary Table 4.
floatimage8.png
Extended Data Fig. 5. Transcriptomic, epigenomic, biochemical, and pharmacologic analysis of patient derived neurofibroma or MPNST cells. a, Patient derived neurofibroma or MPNST cells harbor NF1, CDNK2A, or SUZ12 mutations affecting one or both alleles⁴⁶. b, RNA sequencing volcano plot from differential expression analysis, principal component analysis, and gene ontology analyses reveal neurofibroma and MPNST cells are distinguished by enrichment of cell differentiation pathways or de-repression of PRC2 target genes, respectively. CHEA, chromatin immunoprecipitation enrichment analysis. c, RNA sequencing differential expression and gene ontology analyses show simultaneous CRISPR deletion of both PRC2 epigenetic complex members EZH1 and EZH2¹⁹, which can functionally compensate for one another, induces PRC2 target genes and regulators of cell differentiation in neurofibroma cells. d, RNA sequencing differential expression and gene ontology analyses show deletion of the obligate PRC2 epigenetic complex member SUZ12¹⁹ induces PRC2 target genes and regulators of cell differentiation in neurofibroma cells. e, RNA sequencing transcripts per million (TPM) expression shows suppression of Schwann cell differentiation markers (S100B, SOX10), enrichment of canonical Ras target genes (DUSP4, SPRY2, ETV4), or enrichment of neural crest development genes (EN1, SOX9, FOXF1) in PRC2-deficient compared to PRC2-intact Schwann tumor cells. PRC2-deficient (blue): ST88-14, SNF02.2, SNF94.3, SNF96.2, or NF05.5 cells with SUZ12 or simultaneous EZH1 and EZH2 knockout. PRC2-intact (red): NF9511b, NF95.6, and NF05.5 cells ± EZH1 or EZH2 knockout. Lines represent means. Error bars represent standard error of the means. *p<0.05, ***p£0.0001, Student’s t tests. f, Integration of H3K27me³ ChIP sequencing data from PRC2-intact versus PRC2-deficient neurofibroma cells¹⁹ with RNA sequencing data across neurofibroma and MPNST cells (Extended Data Fig. 5e) suggests direct epigenetic de-repression of development genes (EN1, SOX9, FOXF1), and indirect induction of Ras target genes (DUSP6, SPRY2, ETV4) in PRC2-deficient compared to PRC2-intact cells. g, Selumetinib dose-response viability curves for neurofibroma or MPNST cells following 48 hours of treatment reveals ST88-14 MPNST cells with bi-allelic SUZ12 mutations are resistant to MEK inhibition. h, Immunoblots following selumetinib treatment reveals greater doses are required for pERK inhibition in PRC2-deficient ST88-14 MPNST cells compared to PRC2-intact neurofibroma or MPNST cells. See also Supplementary Table 5.
floatimage9.png
Extended Data Fig. 6. Transcriptomic analysis of selumetinib responses over time in neurofibroma cells. a, RNA sequencing of NF95.11b neurofibroma cells after treatment with 1 mM selumetinib or vehicle control demonstrates selumetinib inhibits expression of the Schwann cell differentiation marker S100B and Ras target genes DUSP4 and SPRY2. TPM, transcripts per million. b, RNA sequencing and gene ontology analyses of NF95.11b neurofibroma cells after treatment with 1 mM selumetinib shows selumetinib treatment inhibits cell differentiation programs of neural crest derived lineages such as the nervous or skeletal systems. CHEA, chromatin immunoprecipitation enrichment analysis. See also Supplementary Table 6.

Download PDF

Journal Publication

published 12 Jan, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Functional interactions between neurofibromatosis tumor suppressors drive Schwann cell tumor de-differentiation and treatment resistance

Status:

Journal Publication

Version 1

Abstract

Figures

Main Text

Schwann cell de-differentiation underlies MEK inhibitor resistance

NF2 inactivation drives de-differentiation and MEK inhibitor resistance in NF1-deficient Schwann cell tumors

Discussion

Declarations

Methods

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1