SIRT2 Regulates the SMARCB1 Loss-Driven Differentiation Block in ATRT Available to Purchase

Brain tumors are the most common cause of oncologic deaths in American children (1). An atypical teratoid rhabdoid tumor (ATRT) is a highly malignant pediatric brain tumor that is frequently metastatic (1). Despite aggressive chemotherapy, autologous stem-cell rescue, and radiation, patients have a survival rate of only ∼35% (1–5). ATRT tumor genomes demonstrate a very low mutational burden with a paucity of targetable lesions (6). Instead, ATRT is driven by the loss of SMARCB1 (and more rarely, SMARCA4; refs. 1, 7). SMARCB1 is an essential member of the switch/sucrose nonfermentable (SWI/SNF) complex and is involved in regulating gene expression by controlling chromatin remodeling and nucleosome incorporation (8, 9). Despite their quiescent genome, ATRTs exhibit substantial transcriptional heterogeneity. We originally showed that ATRT can be subclassified into four transcriptomic subgroups with increased BMP pathway gene expression, identifying a particularly poor outcome population (10). Two subsequent large studies clustered ATRT into three subgroups based on transcriptomic and DNA methylation analyses (11, 12). Further studies have demonstrated that SMARCB1 is required for SWI/SNF complex occupancy at enhancers to activate bivalent promoters and that genetic deficiencies in the SWI/SNF complex lead to the assembly of aberrant complexes and differential occupancy of this complex at super-enhancers (refs. 13, 14).

Given the singular genetic aberration and the transcriptional dysregulation, identifying novel therapeutic vulnerabilities remains a challenge (5). Our group and others have previously shown that EZH2 antagonizes SMARCB1 through epigenetic control of histone lysine methylation (15, 16). We reasoned that the alteration in the epigenetic state due to SMARCB1 deficiency renders ATRT cells dependent on unique epigenetic regulatory mechanisms. To test this hypothesis, we performed an unbiased epigenome-targeted functional short hairpin RNA (shRNA) screen and identified SIRT2, a NAD-dependent histone deacetylase, as a primary regulator of ATRT. Sirtuins have complex roles in both promoting and/or suppressing tumorigenesis (17). We showed that SIRT2 inhibition attenuates ATRT cell self-renewal, promotes differentiation, and suppresses in vivo growth of intracranial xenografts by inducing differentiation and activating apoptosis. Our data establish SIRT2 as a druggable pathway in ATRT.

The in vivo xenograft model was established as previously described (18). Female athymic nude (Foxn1 nu) mice (6 weeks old; weight, 18 20 g) from Harlan Laboratories were used. Mice were treated with either thiomyristoyl (TM; 50 mg/kg i.p. 3 times/week for 4 weeks) or tenovin-6 (50 mg/kg i.p. daily for 15 days). For bioluminescence analysis, mice were imaged using the Xenogen IVIS 200 in Vivo Imaging System (PerkinElmer). Tumor bioluminescence was analyzed using Living Image 2.60.1 software (Caliper Life Sciences, PerkinElmer). All animal procedures were approved by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee (protocol number: 00052).

All cell lines were cultured in a cell incubator (Thermo Fisher Scientific) at 37°C with 5% CO₂. The MAF737 (group 2A, TYR subtype) ATRT SMARCB1/INI1-deleted cell line was established from a surgical sample of a 12-month-old male obtained from Children’s Hospital Colorado. The sample was collected in accordance with local and federal human research protection guidelines and Institutional Review Board regulations (approval no. COMIRB 95-500). Consent was provided by a parent. The BT16 (group 2B, MYC subtype) ATRT cell line was received from Dr. Peter Houghton (Nationwide Children’s Hospital, Center for Childhood Cancer and Blood Diseases). MAF737 and BT16 cells were cultured in RPMI medium (Life Technologies, Gibco, #11875-093) supplemented with 10% FBS (Atlanta Biologicals, Inc., #S11550) and 1% penicillin/streptomycin (Life Technologies, Gibco, #15140-122). The CHB-ATRT1 (group 2B, MYC subtype) and SU-ATRT2 (group 2A, TYR subtype) cell lines were kindly provided by Dr. Mitra (University of Colorado Anschutz Medical Campus). CHB-ATRT1 (MYC) cells were derived from a posterior fossa ATRT at Boston Children’s Hospital. SU-ATRT2 (TYR) cells were derived from an intraventricular ATRT at Stanford University Medical Center (19). Both cell lines were cultured in Neurobasal medium (Life Technologies, Gibco, #10888-022), EGF (Shenandoah Technology, #100-26), FGF (Shenandoah Technologies, #100-146), and leukemia inhibitory factor (Millipore, #LIF1050).

The CHLA04 (SHH subtype) cell line was purchased from the ATCC and cultured in DMEM/F12 (Gibco; Thermo Fisher Scientific, #11330-032) with EGF and 2% B27 supplement (Gibco, #17504-44). Immortalized normal human astrocytes (NHA) were provided by Dr. Cynthia Hawkins (Sick Kids Hospital, Department of Pediatric Laboratory Medicine, Toronto, Canada) and cultured in DMEM (Gibco; Thermo Fisher Scientific, #11320-033) supplemented with 10% FBS (Atlanta Biologicals, #S11550), 1% penicillin/streptomycin, L-glutamine(Gibco, #25030-081), and sodium pyruvate (Gibco, #11360-070). All the cells were cultured according to provided protocols. Cells were passaged when flasks reached 80% confluency. Cells were detached from flasks with 0.25% Trypsin-EDTA (Gibco, # 23200072) followed by centrifugation at 1,200 rpm for 5 minutes. Cells were resuspended in growth media, and 20% were moved to a new flask. All the cell lines were authenticated by DNA fingerprinting through the University of Colorado Molecular Biology Service Center utilizing the short tandem repeat DNA Profiling PowerPlex-16 HS Kit (DC2101, Promega). Mycoplasma testing was performed routinely using Mycoplasma Detection Kit MP0035 optimized for use with JumpStart Taq DNA Polymerase (D9307, Sigma-Aldrich).

Tenovin-6 was purchased from MedChemExpress (HY 15510). TM was bought from MedChemExpress (HY-101278) and provided by Dr. Henning Lin (Department Chemistry and Chemical Biology, Cornell University, Ithaca, NY). DMSO (Sigma-Aldrich, #D2650) was used for control.

Lentiviral production and target cell transductions were performed as previously described (20). Lentiviral particles for transduction were made from a pooled shRNA library (Human pLVX-ZsGreen lentiviral shRNA-miR epigenetics–related genes, TransOMIC Technologies, TRH6110). For lentiviral production, HEK293FT cells were cotransfected with packaging plasmids psPAX and pMD2.G and shERWOOD-UltramiR-GFPshRNA library using Lipofectamine 3000 (Thermo Fisher Scientific, #100022050).

Functional genomic shRNA screening was performed as previously described by us (20). The MAF737 (TYR) cell line was transduced for 24 hours with a pooled lentivirus shRNA library consisting of 4,200 shRNAs targeting 408 epigenetic genes (backbone, Human pLVX-ZsGreen lentiviral shRNA-miR; TransOMIC Technologies, #TRH6110). Twenty-four hours after seeding, MAF737 (TYR) cells were infected with the pooled shRNA lentiviral library at a multiplicity of infection of 0.3 (n = 4). Seventy-two hours after transduction (day 0), one aliquot (control, before puromycin treatment) of cells was collected to isolate genomic DNA, and the remaining transduced cells were reseeded. Twenty-four hours after reseeding, cells were treated with puromycin (2 mg/mL) to select for the pure populations of transduced cells. These cells were maintained in culture medium for a further 21 days (day 21; passaged every 3 days). Cells were passaged to maintain a multiplicity of infection of 0.3 to ensure that each transduced cell had a single genomic integration in the presence of puromycin. The cells were collected 21 days (7–8 doubling times of the cells) after transduction. Genomic DNA was isolated and sequenced using an Illumina HiSeq 3000/HiSeq 4000 instrument. The sequencing results were analyzed using the R-based package DESeq2 (v1.34.0; Bioconductor.org) by comparing the shRNAs present on day 0 with day 21 (20) with a FDR of 0.5 and 0.1, respectively. Raw data for this study were generated at Genomics and Microarray Core Sequencing, University of Colorado Denver. Derived data supporting the findings of this study are available from the corresponding author upon request. Genes were considered as top hits when depleted in at least three of the four replicates (20).

The transcriptome sequencing [RNA sequencing (RNA-seq)] procedure was performed as previously published (15). RNA was isolated from treated cells using QIAGEN miRNAeasy kit, Cat. #217004 (Qiagen). Libraries were prepared and sequenced by Genomics and Microarray Core Facility (University of Colorado Denver). Derived data supporting the findings of this study are available from the corresponding author upon request. Universal Plus mRNA-Seq Library Kit with NuQuant (Tecan Life Sciences) was used. Libraries were sequenced on Illumina NovaSeq X. High-quality base calls at Q30R ≥ 80% were obtained with 40 million paired-end reads. Sequenced 150 bp paired-end reads were mapped to the human genome (GRCh38) by STAR 2.4.0.1, read counts were calculated by R Bioconductor package GenomicAlignments 1.18.1, and differential expression was analyzed with DESeq2 1.22.2 in R. Further analysis by gene set enrichment analysis (GSEA) was performed in GSEA v2.1.0 software with 1,000 data permutations.

BT16 (MYC) and MAF737 (TYR) ATRT cells were transfected with plasmids purchased from the Functional Genomics Shared Resource, University of Colorado (shSIRT2, pLKO1. #TRCN0000040218, shNull MISSION pLKO.1-puro nonmammalian shRNA control, Cat. #SHC002, SIRT2ORF Cat. # CCSBBroad304_02705.) A total of 2.5 μg plasmid DNA and 7.5 μL Lipofectamine 3000 (Thermo Fisher Scientific, Cat. #L3000-008) were used for transfection at 37°C. The cells were selected with puromycin (1 μg/mL) following transfection. Luciferase-expressing MAF737 (TYR) or BT16 (MYC) ATRT cell lines were obtained by transfection with the pLV[Exp]-Bsd-EFS>Luc2(ns):T2A:TurboGFP vector (VectorBuilder).

RNA was isolated from transfected BT16 (MYC) and MAF737 (TYR) cells, reverse transcribed, and quantified with qRT-PCR with TaqMan gene-specific probes (SIRT2, Thermo Fisher Scientific, #Hs01560289_m1, GAPDH Thermo Fisher Scientific, #Hs_02786624_g1; ref. 20). Relative gene expression was calculated using the ΔΔCT method with GAPDH as the endogenous control.

The BT16 (MYC) and MAF737 (TYR) ATRT cell lines were seeded in six-well plates in triplicate at a density of 500 cells/well, treated with the indicated doses of TM or tenovin-6, and cultured for 10 days. Cells were stained with 0.25% crystal violet in methanol for 15 minutes at room temperature. Crystal violet–positive colonies (>50 cells per colony) were counted using a precise electronic counter (Heathrow Scientific) and a light inverted microscope at ×2 magnification (Olympus S751; Olympus Corporation).

Five hundred CHB-ATRT1 (MYC) and SU-ATRT2 (TYR) cells were plated in a 1:1 mixture of 2.6% methylcellulose and complete growth medium. Cells were treated with the indicated drugs and grown for 10 days. Then colonies were stained with nitrotetrazolium blue chloride (Sigma-Aldrich, #6876) at 1.5 mg/mL in PBS for 24 hours at 37°C. Nitrotetrazolium blue chloride–positive colonies (>50 cells per colony) were counted using a precise electronic counter and a light inverted microscope at ×2 magnification as above.

Cell proliferation (cell index) was measured using the xCELLigence Real-Time Cell Analyzer instrument (Roche Diagnostics) as previously described (18). Cells were seeded in triplicate at 3,000 cells/well in a gold-plate E-Plate 96 (Agilent, #128499), treated, and monitored for growth for 10 days. Slopes were calculated as a measure of the rate of cell growth using the RTCA software.

Cells were seeded in triplicate at 50,000 cells/well in 24-well plates and treated with the indicated drug after 24 hours. After 72 hours, the cells were trypsinized, collected, suspended in ViaCount Reagent (Luminex, #4000-040), and counted on the Guava easyCyte 8HT Flow Cytometry System (EMD Millipore).

BT16 (MYC) and MAF737 (TYR) neurospheres were generated in a 24-well ultra-low cluster plate (500 cells per well in triplicate) in serum-free medium [neurobasal medium (Life Technologies, Gibco, #10888-022) supplemented with B-27 (Life Technologies, Gibco, Cat. #17504044), L-glutamine (Life Technologies, Gibco, #25030), penicillin/streptomycin (Life Technologies, Gibco, #15140-122), EGF (Shenandoah Technology, #100-26), and FGF (Shenandoah Technologies, #100-146)]. Spheres were treated with 5 μmol/L tenovin-6 or 10 μmol/L TM, and the medium was replaced every 3 days. Neurosphere proliferation assays were performed on an IncuCyte S3 Live Cell Analysis System (Essen BioScience) for 14 days. Measurements of neurosphere size over time were calculated using IncuCyte S3 software and normalized against day 1 values to determine growth.

For the aldehyde dehydrogenase (ALDH; ALDEFLUOR) assay (STEMCELL Technologies, #01700), BT16 (MYC) and CHB-ATRT1 (MYC) cells were treated with TM at their corresponding IC₅₀ values. After 48 hours of treatment, 1 × 10⁶ cells from each condition were collected, centrifuged, and resuspended in 500 μL ALDEFLUOR buffer. A measure of 5 μL of DEAB reagent was added to the negative control sample, and 5 μL of ALDEFLUOR reagent was added to each tube. The cells were incubated at 37°C for 45 minutes, centrifuged, and stained with 1:500 DRAQ5 (Thermo Fisher Scientific, #62254). ALDH accumulation was analyzed using an Amnis FlowSight flow cytometer (Millipore) with IDEAS software v6.1 (Luminex Corporation).

Limiting dilution assay was performed as previously described (21). BT16 (MYC) and CHB-ATRT1 (MYC) cells were treated with TM at their corresponding IC₅₀ values and seeded in 96-well ultra-low–attachment round-bottom tissue culture plates (Corning) in serum-free medium. Cells were seeded in 10 wells (10, 25, and 50 cells/well) or 30 wells (1 cell/well) per condition. The plates were monitored using an IncuCyte S3 Live Cell Analysis System (Essen BioScience) for a few weeks. Measurements of neurosphere size over time were calculated using IncuCyte S3 software. Published ELDA software (http://bioinf.wehi.edu.au/software/elda/) was used to calculate the comparative self-renewal potential of the cells.

Protein expression levels were determined by Western blotting as previously described (22). Proteins (30 μg in total) were separated using 4% to 20% SDS-PAGE (Bio-Rad Laboratories, #4561094), transferred to a membrane, and incubated with the primary antibody overnight at 4°C. The next day, the membrane was incubated with a secondary antibody α-mouse-HRP (Cell Signaling Technology, #7076, RRID: AB_330924) or α-rabbit-HRP (Cell Signaling Technology, #7074), developed with Immobilon Forte Western HRP Substrate (Millipore, #WBLUF0500), and imaged using Syngene G:Box Chemi-SL1.4 gel doc. Primary antibodies were from the following sources: anti-SIRT2 (EPR1667; Abcam, #ab134171, RRID: AB_2716787), anti–c-Myc (Y69; Abcam, #ab32072, RRID: AB_731658), α-tubulin (DM1A; Cell Signaling Technology, #3873, RRID: AB_1904178), and anti-acetylated α-tubulin (6-11B-1; Santa Cruz Biotechnology, #sc-23950, RRID: AB_628409).

The MRI images were obtained using Bruker 9.4 Tesla BioSpec MRI Scanner (Bruker Medical) as we previously described (20). T2-turboRare images of the sagittal and axial panels were acquired and analyzed using Bruker ParaVision NEO360 v2.0 software (Bruker Corporation). A region of interest was manually segmented on each anatomic slice, and the tumor volume (mm³) was calculated by a radiologist. The apparent diffusion coefficient (s/mm²) was calculated using diffusion-weighted imaging maps as a criterion for tumor cellularity.

Tissue from patient samples or experimental animals was fixed in 10% formalin for 3 days at room temperature and submitted to the University of Colorado Denver Tissue Histology Shared Resource for sectioning and staining. Paraffin-embedded sections (5 μm) were prepared for the immunodetection of anti-SIRT2 (1:100, Boster Biological Technology, #PB9160), anti-Ki67 (1:500; Thermo Fisher Scientific, #RM-9106), anti-FOXM1 (1:100; Cell Signaling Technology, #20459, RRID: AB_2798842), and anti-Caspase3 (1:1,000; Cell Signaling Technology, #9661, RRID: AB_2341188). Immunodetection was performed using a Benchmark XT auto stainer for 32 minutes at 37°C using a modified I-VIEW DAB detection system (Ventana Medical Systems, Roche Diagnositics). 3,3'-diaminobenzidine (DAB) was used to visualize antigen–antibody complexes using a standard protocol. All sections were counterstained with hematoxylin and mounted using a synthetic resin (Cancer Diagnostics) with a cover glass at room temperature. Negative controls were used to confirm the specificity of immunostaining, including omission of the primary antibody incubation step and substitution with the primary antibody diluent. The images were captured using a BZ-X710 all-in-one microscope (Keyence Corporation) and quantified using BZ-X viewer v.01.03.01.01.01 (Keyence Corporation).

A total of 3,000 BT16 (MYC) ATRT cells were seeded on poly-D-lysine–coated chamber slides (Corning, #35463). CHB-ATRT1 (MYC) cells were seeded onto poly-D-lysine/laminin–coated chamber slides (Corning, #354688). The next day, the cells were treated with TM at their respective IC₅₀ values, washed with PBS (Corning, #21-040-CV), fixed for 15 minutes at room temperature with 4% paraformaldehyde (Alfa Aesar; Thermo Fisher Scientific, #61899), blocked for 30 minutes at room temperature in 5% skimmed milk and 0.05% TritonX-100 (Sigma-Aldrich; Merck KGaA, #93443) in PBS, and incubated with the primary antibody for 12 hours at 4°C and the secondary antibodies for 1 hour at room temperature, as previously described by us (23). The following antibodies were used: Sall4 (Cell Signaling Technology, #8459, RRID: AB_10949321) and Alexa Fluor 488–conjugated secondary antibody 1:500 (BD Pharmingen; BD Biosciences, #560445). Then the cells were washed with 0.05% TritonX-100 in PBS 3 times and mounted with ProLong Gold Antifade reagent containing 4',6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich; Merck KGaA, #36935). Images were captured using a fluorescence microscope (BZ-X700; Keyence Corporation) at ×40 magnification.

A cleavage under target (CUT) and release using nuclease (RUN) assay was performed as described by Walker and colleagues (24) with some modifications. CUTANA ChIC/CUT & RUN kit version3, User Manual Version 3.3 from EpiCypher was used (Cat. #14-1048). Beads were prepared using 10 μL/sample of CUTANA Concavalin A Conjugated Paramagnetic Beads (EpiCypher, SKU:21-1401). A total of 500,000 BT16 cells per reaction was harvested using 0.05% trypsin, centrifuged, washed with DPBS without calcium and magnesium chloride (Gibco, #14190-144), and washed twice with 100 μL/sample of wash buffer [20 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L spermidine, and 1× Roche cOmplete EDTA-free Protease Inhibitor (1187358001)]. Then the cells were resuspended in 100 μL of wash buffer and incubated for 10 minutes on the magnetic strip with 10 μL of preactivated ConA beads (EpiCypher). Subsequently, 1 μL of cold antibody specific to the reaction was added to each sample in 49 μL of cold antibody buffer. Next, the following antibodies were used: IgG (rabbit IgG negative control, EpiCypher, 13-0042K), H3K4me3 (positive control, EpiCypher, 13-0041), H3K27Ac (Cell Signaling Technology, #8173), and H3K27me3 (Cell Signaling Technology, #9733). Samples were incubated overnight on a nutator at 4°C. The supernatant was removed, and cells were permeabilized with 50 μL of 5% digitonin buffer. An aliquot of 2.5 μL of CUTANA pAG-MNase (Epicypher) was added to each reaction, vortexed gently to evenly distribute the enzymes, and incubated for 10 minutes at room temperature. Then, samples were placed on a magnet, and the supernatant was removed. The step was repeated twice and followed by resuspension with 50 μL of cold digitonin buffer. An aliquot of 1 μL of 100 mmol/L CaCl₂ was added to each sample and then incubated on nutator for 2 hours at 4°C. A measure of 33 μL per sample of Stop Master Mix containing 0.5 ng/sample of Escherichia coli spike-in DNA was added to the samples, mixed by pipetting, and incubated in a thermocycler for 10 minutes at 37°C. Samples were quick-spun and placed on a magnet stand for the slurry separation, and the supernatant was transferred to a 1.5-mL tube (DNA/RNA LoBind tube, #4043-1021, Eppendorf Scientific). The DNA was purified using CUTANA DNA Purification Kit (EpiCypher, 14-0050) according to the manufacturer’s instructions, and 2 μL of DNA was used for quantification by High Sensitivity D1000Reagents (Agilent Technologies, Per# 5067-5585). Illumina sequencing libraries were prepared using CUTANA CUT & RUN Library Prep Kit (EpiCypher, 14-1001&14-1002) and sequenced using NovaSeq 6000.

The quality of the FASTQ files was accessed using FastQC (v.0.11.8)119 and MultiQC120. Illumina adapters and low-quality reads were filtered out using BBDuk (http://jgi.doe.gov/data-and-tools/bb-tools). Bowtie2 (v.2.3.4.3)121 was used to align the sequencing reads to the hg38 reference human genome. Samtools (v.1.11)122 was used to select the mapped reads (samtools view-b-q30) and sort the BAM files. PCR duplicates were removed using Picard MarkDuplicates tool (http://broadinstitute.github.io/picard/). The normalization ratio for each sample was calculated by dividing the number of uniquely mapped human reads of the sample with the lowest number of reads by the number of uniquely mapped human reads of each sample. These normalization ratios were used to randomly subsample reads to obtain the same number of reads for each sample using samtools view -s. Bedtools genomecov was used to create bedGraph files from the BAM files123. Bigwig files were created using deepTools bamCoverage124 and visualized using IGV125. Peaks were called using MACS2 (v2.1.2)126 using ENCODE recommendations. Irreproducible discovery rate (IDR) was used to identify the reproducible peaks between the replicates127. Further processing of the peak data was performed in R, using in particular the following tools: valR128 and DiffBind129. Average profiles and heat maps were generated using ngs.plot130. Differentially marked genes were calculated using DiffBind and DESeq2, based on the threshold of FDR < 0.05 and fold change ≥2. Raw data for this study were generated by Genomics and Microarray Core Sequencing, University of Colorado Denver. Derived data supporting the findings of this study are available from the corresponding author upon request.

Tumors were collected from the mice cerebellum, disaggregated with razor, filtered using a 100-μm cell strainer, centrifugated, and resuspended with 2% FBS in PBS. Cells were sorted for GFP color, nuclei were isolated (10× genomic kit, #cG000366), and cells were processed for single-cell RNA-seq. Single-cell RNA-seq was performed using Chromium Single Cell 3′ Library and Gel Bead Kit (v3.1 10× Genomics), and the Chromium X. Libraries were sequenced on Illumina NovaSeq 6000, with 150,000 sequencing reads per cell. Demultiplexing, alignment to the mm10 transcriptome, and unique molecular identifier–collapsing were performed using the Cell Ranger tool kit (v5.0, 10× Genomics). Raw sequencing reads were processed using Cell Ranger single-cell software suite (v. 6.1.1) with the default parameters (25). The reads were aligned to the human reference genome (GRCh38 V3.0.0). Quality control measures were applied to exclude low-quality cells and potential doublets, based on metrics such as total unique molecular identifiers, the number of detected genes, and the percentage of mitochondrial gene expression. Following quality control, normalization was performed using sctransform. Dimensionality reduction was then conducted using principal component analysis as a preliminary step, followed by t-distributed stochastic neighbor embedding to visualize the high-dimensional data in two dimensions. Clustering was performed using the shared nearest neighbor modularity optimization algorithm to identify discrete cell populations. Differentially expressed genes were identified using the “FindAllMarkers” function in Seurat (v 4.0.4; ref. 26). Genes with a fold change of more than 1 and an adjusted P value less than 0.05 were considered differentially expressed genes. The cell type identification was carried out by examining the expression of canonical marker genes within each cluster. Gene set coregulation analysis was performed using fqsea (v.1.33.2). Raw data for this study were generated by Genomics and Microarray Core Sequencing, University of Colorado Denver. Derived data supporting the findings of this study are available from the corresponding author upon request.

To identify enriched molecular pathways in each cell cluster detected by Seurat, Metascape (27) and DAVID bioinformatics resources (v 6.8; ref. 28) were performed on differentially expressed marker genes. Gene sets from Gene Ontology Biological Process (29) and Kyoto Encyclopedia of Genes and Genomes were used.

The cell-cycle phases of each cell line were determined using the CellCycleScoring function in Seurat (26).

Statistical analysis was performed using the GraphPad Prism v8 software (GraphPad Software). Either one- or two-way ANOVA or one- or two-tailed Student t test (unpaired) was used for comparisons between groups. Kaplan–Meier survival curve comparisons were performed using the log-rank (Mantel–Cox) test. P < 0.05 was considered to indicate statistical significance. Data were presented as the mean ± SEM. The experiments were independently repeated 3 times.

The data generated in this study are publicly available in Gene Expression Omnibus at GSE273455 (GSE273455; GSE273417, GSE273418, GSE274373, and GSE278558).

To systematically identify the epigenetic factors required for tumor cell survival in the absence of SMARCB1, we performed shRNA screening targeting 408 genes associated with transcription- and chromatin-related processes. A patient-derived SMARCB1-deleted ATRT cell line MAF737 (TYR) was transduced with a pooled lentiviral shRNA library consisting of 4 to 10 shRNAs per gene and 4,180 total shRNAs (30). Genomic DNA was isolated from transduced cells 4 and 21 days after transduction and sequenced using an Illumina HiSeq instrument. Differential enrichment of shRNA-targeting genes was assessed using the R-based package DESeq2 (v1.34.0; Bioconductor.org). Incorporated shRNA sequences were quantified in comparison with immediate post-transduction sampling, with depleted sequences representing candidate essential genes (Fig. 1A; Supplementary Table S1). shRNAs were ranked by the observed decrease in cell viability (Fig. 1A; Supplementary Table S1), identifying candidate genes with a fold change <0.8 and P < 0.05 (Fig. 1A; Supplementary Table S1). One of the top hits was SIRT2 (fold change = 0.309; P < 0.05). Gene hits were further confirmed using independent shRNA constructs in a colony focus assay (Supplementary Fig. S1A and S1B). Among the key hits was SIRT2, which when depleted, resulted in a significant loss of ATRT cell viability (Supplementary Fig. S1A and S1B). SIRT2 expression in patient samples from University of Colorado Colorado Children’s Hospital showed that SIRT2 is differently expressed across the subgroups MYC>SHH>TYR (Supplementary Fig. S1C). However, SIRT2 expression was not substantially increased compared with normal brain, suggesting that SMARCB1 loss generates a dependency without altering gene expression.

To evaluate the role of SIRT2 in ATRT, we used nonoverlapping shRNA lentiviral constructs to deplete SIRT2 in twoSMARCB1-deleted ATRT cell lines, BT16 (MYC) and MAF737 (TYR). We observed a marked decrease in colony formation following the shRNA-mediated reduction in SIRT2 expression (P < 0.05; Fig. 1B–D). We confirmed the genetic knockdown of shSIRT2 with qRT-PCR (P < 0.01; Supplementary Fig. S2A) and Western blot (Supplementary Fig. S2B and S2C). Moreover, SIRT2 depletion significantly decreased BT16 (MYC) and MAF737 (TYR) ATRT cell proliferation, as was demonstrated with a real-time cell growth assay (P < 0.05; Supplementary Fig. S2D). To rule out the possibility of off-target effects, rescue experiments were performed. By Western blot, we confirmed shSIRT2 knockdown in BT16 ATRT cells and restored SIRT2 level expression with open reading frame rescue vector for SIRT2 (Supplementary Fig. S3A and S3B). Then, using the colony formation assay and real-time cell growth, xCELLigence, we showed that cell growth is decreased in shSIRT2 cells and is rescued by reexpression of the SIRT2 open reading frame. Thus, we confirmed that SIRT2 is essential for ATRT cell growth (Supplementary Fig. S3C and S3D).

To evaluate SIRT2 depletion in vivo, we implanted SIRT2 sufficient or depleted ATRT cells into the cerebella of immunocompromised mice (n = 5). In both MAF737 (TYR) and BT16 (MYC) orthotopic xenograft models, SIRT2-depleted cells showed delayed time to tumor establishment and formed significantly smaller intracranial tumors (P < 0.01; Fig. 1E and F), further suggesting an essential role of SIRT2 in ATRT tumorigenesis. Analysis of xenograft tumor tissues revealed fewer Ki-67 and SIRT2 positive proliferating cells in SIRT2-depleted tumors compared with parental xenografts (Supplementary Fig. S4A and S4B).

Recent progress has identified several novel SIRT2 inhibitors (31). We found that tenovin-6, an inhibitor of both SIRT1 and SIRT2 (32), decreased the clonogenicity of BT16 (MYC) and MAF737 (TYR) cells at low micromolar concentrations (IC₅₀ = 5 and 0.3 μmol/L, respectively; Supplementary Fig. S5A and S5B). Moreover, we observed a dose-dependent decrease in the viability of ATRT cells from different subgroups (Supplementary Fig. S5C and S5D). Tenovin-6 attenuated ATRT cell viability at low concentration but only slightly affected the viability of NHA (Supplementary Fig. S5C–S5E). More recently, studies identified TM, a specific SIRT2 inhibitor (33). TM is more potent with higher activity against SIRT2 than tenovin-6 (31). TM strongly suppressed ATRT clonogenicity in BT16 (MYC) and MAF737 (TYR) cells (Fig. 2A and B). Moreover, TM potently inhibited the clonogenic growth of CHB-ATRT1 (MYC; group 2B, MYC) and SU-ATRT2 (group 2A, TYR), two short-term cultures of primary patient tumor cells (Fig. 2C and D) with relatively high c-MYC expression. Although TM treatment changed the viability of cells with high MYC expression, BT16 (MYC), CHB-ATRT1 (MYC), and ATRT cells with relatively high MYC expression, MAF737 (TYR), and SU-ATRT2 (TYR), TM treatment did not alter the growth of CHLA04 (SHH) and NHA (Supplementary Fig. S5F–S5H). Our data correlate with previous findings that the anticancer effect of TM is associated in part with its ability to decrease c-MYC levels and that TM has limited effects on nontransformed cells (33).

To examine whether SIRT2 inhibition alters the cell cycle, we performed a cell-cycle assay on BT16 (MYC) and MAF737 (TYR) cells treated with their tenovin-6 IC₅₀ values. After 72 hours of treatment, tenovin-6 decreased DNA synthesis (S-phase fraction) by 25% in BT16 (MYC) cells (P < 0.05; Supplementary Fig. S6A) and by 4% in MAF737 (TYR; P < 0.05; Supplementary Fig. S6B). To check whether the selective SIRT2 inhibitor, TM, also works through the cell cycle, we treated BT16 (MYC) and MAF737 (TYR) ATRT cells and CHB-ATRT1 (MYC) and SU-ATRT2 (TYR) primary ATRT cells with their TM IC₅₀ values for 72 hours. TM treatment significantly decreased S-phase in all cell lines evaluated with P < 0.05. In BT16 (MYC) cells, S-phase was decreased by 7% (Supplementary Fig. S6C), in MAF737 (TYR) cells by 13% (Supplementary Fig. S6D), in CHB-ATRT1 (MYC) primary cells by 10% (Supplementary Fig. S6E), and in SU-ATRT2 (group 2A, TYR) primary cells by 8% (Supplementary Fig. S6F).

More importantly, pharmacologic inhibition of SIRT2 resulted in the induction of apoptosis in both ATRT cell lines and primary patient cultures (Supplementary Fig. S7). Tenovin-6 treatment increased apoptosis in BT16 (MYC) cells by 188% (Supplementary Fig. S7A) and by 124% in MAF737 (TYR) cells (Supplementary Fig. S7B), P < 0.01. TM treatment also increased apoptosis in BT16 (MYC) and MAF737 (TYR) cells by 73% and 19%, respectively (Supplementary Fig. S7C and S7D), P < 0.05. Moreover, TM treatment significantly increased apoptosis in CHB-ATRT1 (MYC) primary cells by 68% (Supplementary Fig. S7E) and SU-ATRT2 (TYR) cells by 146% (Supplementary Fig. S7F), P < 0.01.

To examine the effect of SIRT2 depletion on the transcriptional landscape of ATRT, we performed RNA-seq of BT16 (MYC) cells treated with 1 μmol/L of tenovin-6 for 72 hours. We identified 5,634 differentially expressed genes in SIRT2-depleted cells (Fig. 2E). Gene Ontology of differentially expressed genes demonstrated a higher-ranked enrichment in transcriptional programs involved in neuronal differentiation (Fig. 2E). Conversely, we observed a decrease in the expression of genes important for regulating progenitor cell state and self-renewal (Fig. 2E). Additional ontology analysis revealed significant enrichment in gene sets involved in the cell differentiation and apoptotic pathways (Supplementary Fig. S8A), which was consistent with previous findings on cell-cycle arrest and apoptosis (Supplementary Figs. S6 and S7; ref. 34).

GSEA revealed a significant decrease of progenitor cell (radial glial cell)–associated networks and an increase in neuronal differentiation–associated genes (Fig. 2F). Consistent with previous reports, MYC-associated gene expression was also suppressed (Fig. 2F; ref. 33). Interestingly, genes regulated by the SWI/SNF complex were reinduced by depletion SIRT2, further suggesting that SIRT2 acts in an antagonistic manner to SWI/SNF activity (Fig. 2F).

To further validate the effects of SIRT2 depletion, we examined the expression of select differentiation–associated genes in BT16 (MYC) ATRT cells with genetic SIRT2 knockdown. SIRT2 depletion resulted in the increased expression of TLX2 and NRG1, all of which are associated with neuronal differentiation (Supplementary Fig. S8B). Conversely, genes associated with cell motility/metastasis (Ephb2), oncogenesis (FOXM1), and mitosis (PLK1) were suppressed by SIRT2 depletion (Supplementary Fig. S8C). There was an overlap, but not a complete similarity, between the gene expression changes due to shSIRT2 and tenovin-6 treatment, suggesting both common pathways and independent activity of tenovin-6 (Supplementary Fig. S8D and S8E). Tenovin-6 treatment (1 μmol/L for 72 hours) resulted in 637 differentially expressed genes (log₂ > 0.5 and P < 0.05) mapping to multiple oncogenic networks (Supplementary Fig. S8F), and TM treatment (30 μmol/L for 72 hours) resulted in 560 differentially expressed genes (log₂ > 0.5 and P < 0.05) mapping to multiple oncogenic networks (Supplementary Fig. S8G). These data suggest that SIRT2 regulates the balance between differentiation and self-renewal gene expression programs in the context of SMARCB1 loss.

Using immunofluorescence confocal imaging, we found that the expression of Sall4, a stem cell factor associated with ATRT, significantly decreased in TM-treated BT16 (MYC) and CHB-ATRT1 (MYC) primary ATRT cells (P < 0.05; Fig. 2G). Conversely, SIRT2 depletion resulted in increased expression of MEF2A (p<0.05) which is associated with neuronal differentiation (Fig. 2H ). Overall, these findings indicated that SIRT2 is required for the maintenance of pluripotency gene networks in ATRT.

To investigate whether SIRT2 downregulation changes the SMARCB1-driven chromatin reprogramming, we evaluated the occupancy of a series of histone marks by the CUT and RUN assay. BT16 (MYC) ATRT cells were treated with 20 μmol/L of TM for 24 hours followed by CUT and RUN assay with antibodies for H3K4me3, H3K27ac, and HeK27me3. On a genome-wide basis, TM treatment decreased H3K4me3 occupancy (Fig. 3A), slightly increased H3K27ac occupancy (Fig. 3A), and had little effect on H3K27me3 occupancy (Fig. 3A). Histograms depicting peak height versus density show that IgG immunoprecipitation has very small peak heights, whereas sample immunoprecipitations have distributions of larger peak heights (Supplementary Fig. S9A, left); H3K4me3 shows genome-wide decreases in peak height, whereas H3K27me3 and H3K27ac show minimal differences due to TM treatment (Supplementary Fig. S9A, right). GSEA shows depletion of H3K4me3 at gene sets for MYC targets, mTORC1 signaling, and radial glia (stem cell) gene expression enrichment in H3K27ac at genes involved in reactive oxygen species production (Supplementary Fig. S9B). H3K27me3 gene sets did not show any significant depletion or enrichment at the FDR <0.15 level (Supplementary Fig. S9B).

Further pathway analysis of genes in which H3K4me3 was depleted by TM revealed multiple proliferative and prosurvival pathways such as VEGF signaling, cell-cycle regulation, and Rho GTPase activity (Fig. 3B). Conversely genes in which H3K27ac was enriched were associated with differentiation pathways such as cilia development, axon genesis, and microtubule assembly (Fig. 3B). Further evolution of specific genes revealed significant depletion of H3K4me3 at both the MYC and MEF2A gene promoters (Fig. 3C) which was consistent with our RNA-seq data. Interestingly, there was significant enrichment of the activating H3K27ac mark at the SMARCA4 gene promoter (Fig. 3D), suggesting activation of the SWI/SNF complex by SIRT2 when SMARCB1 is lost. Similarly, there was enrichment of multiple prodifferentiation genes with NFIB shown as an example (Fig. 3D).

Because SIRT2 depletion inhibited the expression of pluripotency-associated radial glial–associated genes, we next tested whether SIRT2 inhibition altered ATRT cell tumor sphere growth as a measure of cell self-renewal. TM treatment potently inhibited the growth of new tumor spheres in stem cell conditions compared with control-treated cells (Fig. 4A and B, P < 0.05). In BT16 (MYC) and primary patient CHB-ATRT1 (MYC) cells, TM significantly diminished the activity of ALDH1, a marker of the brain tumor–initiating cell fraction with a given cell population (Fig. 4C; ref. 35). Furthermore, TM treatment decreased the capacity for self-renewal as measured by the extreme limiting dilution assay (Fig. 4D; Supplementary Fig. S10A and S10B). In both cell lines, stem cell frequency was significantly decreased by SIRT2 depletion (Fig. 4D), P < 0.001.

Next, we sought to determine whether SIRT2 dependency could be exploited for therapeutic effect in vivo. MAF737 (TYR) cells transduced with a luciferase-expressing construct were stereotactically injected into the mouse cerebellum, and tumor growth was monitored using bioluminescence. Mice were randomized and treated with vehicle or TM (50 mg/kg, 3 days a week) for 4 weeks. We performed patient–analogous response characterization of murine ATRT xenografts using a longitudinal MRI with volumetric analysis. We observed decreased tumor growth and less invasion of the surrounding anatomic structures in mice receiving TM than those in vehicle control (Fig. 5A). Volumetric analysis of the tumors on MRI demonstrated a significantly smaller tumor size (P < 0.05) in TM-treated animals than in vehicle control animals (Fig. 5B). Similarly, the apparent diffusion coefficient measured on MRI significantly increased in the TM treatment groups, indicating a decrease in tumor cellularity (Fig. 5C). Treatment with TM resulted in a significantly decreased bioluminescent signal, suggesting attenuation of tumor growth (Fig. 5D and E). There was no significant change in the weight of TM-treated animals, suggesting limited drug toxicity (Fig. 5F). Kaplan–Meier Survival analysis demonstrated a significant improvement in overall survival (OS) compared with those receiving vehicle control (Fig. 5G), P < 0.05. TM treatment significantly suppressed the proliferative index of Ki-67in vivo (Fig. 5H), P < 0.05. Concomitantly, TM induced the activation of Caspase3in vivo, suggesting the induction of apoptosis (Fig. 5I and J), P < 0.05.

To confirm that SIRT2 silencing with TM treatment had a therapeutic effect in vivo, we generated intracranial orthotopic ATRT tumors from another cell line, BT16 (MYC) ATRT. BT16 (MYC) cells transduced with a luciferase-expressing construct were stereotactically injected into the mouse cerebellum, and tumor growth was monitored using bioluminescence (Supplementary Fig. S11A). Mice were treated as described above for MAF737 (TYR) cells. TM-treated animals with BT16 (MYC) tumors showed a significant improvement in OS compared with the vehicle control (P < 0.05; Supplementary Fig. S11B). IHC staining revealed decreased FOXM1 oncogene and Ki67 proliferation marker staining accumulation in mouse tumors treated with TM (Supplementary Fig. S11C–S11F), P < 0.05. Hematoxylin and eosin staining was histologically consistent with ATRT (Supplementary Fig. S12). Similarly, treatment with tenovin-6 (50 mg/kg IP daily for 15 days) resulted in a significant decrease in the bioluminescent signal and tumor size at day 15, suggesting attenuation of tumor growth (Supplementary Fig. S13A and S13B), P < 0.05. The 15-day treatment resulted in significantly smaller tumors as measured by T2 MRI imaging (Supplementary Fig. S13A and S13B), P < 0.05.

To study the impact of TM on differentiation in vivo, we performed single-cell RNA-seq on GFP-positive BT16 (MYC) cells isolated from either vehicle (control) or TM-treated xenografts. After filtering cells with overall low expression and high mitochondrial gene expression, we used the Seurat software suite to perform dimensionality reduction and unsupervised cell clustering on 597 vehicle (control) cells and 1,203 cells treated with TM. We identified seven (C0–C6) distinct cell clusters in both treatment groups (Fig. 6A). A prominent population of cells with high MYC expression, characterized by elevated ribosomal gene expression (RPL27A), was observed in both conditions (Fig. 6B). As expected, the expression of MYC was decreased following TM treatment (Fig. 6C). Additionally, we found that TM treatment led to a marked reduction in the cycling cell population (MIK67 and TOP2A) as well as the stem cell population, characterized by the expression of SOX2 and NES (Fig. 6C). GSEA revealed that clusters 2 and 4 are enriched for stemness-related gene expression (Fig. 6D). TM treatment reduces the expression of markers in clusters 2 and 4 (Fig. 6E). A comparison of gene expression in clusters 2 and 4 between TM and DMSO treatments reveals downregulation in the LIM stem cell up and GOBERT differentiation down gene sets (Fig. 6E). In summary, these single-cell RNA-seq studies demonstrate that TM inhibits tumor growth and prolongs mouse survival by promoting differentiation. This was accompanied by the depletion of tumor cells with high expression of stem cell–affiliated genes.

ATRT is a particularly aggressive tumor in children driven by SMARCB1 loss (1, 36, 37). Owing to the genomic quiescence of these tumors, novel therapeutic targets have been difficult to identify (3, 5). SWI/SNF chromatin remodeling complexes are context-dependent mediators of chromatin compaction that regulate pluripotency and differentiation (38, 39). Although germline deletion of SMARCB1 results in early embryonic lethality, somatic depletion of SMARCB1 results in epigenetic dysregulation such as abnormal MYC chromatin occupancy in ATRT (13, 18, 40). SMARCB1 is required for SWI/SNF complex occupancy at typical enhancers to activate bivalent promoters. Genetic deficiencies in the SWI/SNF complex lead to the assembly of aberrant complexes and differential occupancy of super-enhancers (13, 14). The polycomb complex protein EZH2 is antagonistic to SMARCB1 deletion, and its inhibition is a therapeutic vulnerability in ATRT (15, 16). Interestingly, a large fraction of SMARCB1-binding sites is occupied by EZH2, but without the H3K27me3 mark, further emphasizing the global epigenetic dysregulation in ATRT (42). Given the urgent need to identify drug targets for ATRT, we sought to identify druggable epigenetic vulnerabilities in the context of SMARCB1 loss during ATRT. We identified SIRT2 as a key regulator of ATRT cell viability. SIRT2 is a member of the sirtuin family of NAD⁺-dependent protein-modifying lysine deacetylation enzymes (17), one of the seven mammalian homologs of yeast silent information regulator 2, and is associated with an extended lifespan (41). Although primarily cytoplasmic, SIRT2 can translocate to the nucleus, in which it deacetylates histone 4 at lysine 16 (H4K16) during mitosis (42). Nonhistone substrates of SIRT2 have also been identified, including α-tubulin, FOXO3, and TP53 (17). More recently, SIRT2 was shown to deacetylate MYC, thus preventing its degradation and stabilizing MYC protein (33, 43). We recently demonstrated that MYC activity drives ATRT tumor growth and that decreasing MYC levels by BRD4 inhibition results in the inhibition of tumor growth in vivo (18). Thus, targeting MYC via SIRT2 inhibition is an additional therapeutic option.

Selective depletion of SMARCB1 using shRNA in neural progenitor cells blocks a differentiation-associated gene signature and induces a gene signature similar to ATRT (44). Similarly, p53−/− and SMARCB1−/− iPSCs generate tumors in vivo and recapitulate a stem cell–like gene signature found in human patients with ATRT (45). Our data demonstrate that SIRT2 inhibition can restore gene expression programs lost due to SMARCB1 loss and reverse the transcriptional differentiation block in ATRT in vivo. SIRT2 inhibition reverses the chromatin reprograming driven by SMARCB1 deletion and reestablishes active histone marks at gene promoters associated with differentiation, further emphasizing the dynamic nature of chromatin remodeling in this tumor.

Treatment of ATRT is largely dependent on high-dose chemotherapy with autologous stem cell rescue (5). The results from the recent large Children’s Oncology Group study, ACNS 0333, demonstrate a 4-year EFS and OS of 37% and 43%, respectively (2). The results from the NCI-COG Pediatric MATCH trial targeting EZH2 with tazemetostat did not meet its primary efficacy endpoint with an objective response rate of 5% with a 6-month progression-free survival of 35% (46). These and other clinical studies clearly show that therapy for ATRT is inadequate. Our preclinical data suggest that targeting SIRT2 in ATRT with TM enhances survival of orthotopic tumor-bearing mice and thus could be an additional agent in the armamentarium against ATRT.

In summary, we used an unbiased RNAi screen to identify SIRT2 as the top dependency in SMARCB1-deleted ATRT. Validation using genetic and chemical methods confirmed ATRT cell dependency on SIRT2. SIRT2 inhibition effectively suppresses pluripotency-associated genomic programs, decreases tumor-sphere formation in ATRT cells, and attenuates tumor cell self-renewal while prolonging the survival of tumor-bearing mice. Thus, SIRT2 inhibition with TM is a promising therapeutic approach that warrants further investigation and clinical development.

H. Lin reports other support from Sedec Therapeutics outside the submitted work, as well as a patent for SIRT2 inhibitor issued. No disclosures were reported by the other authors.

I. Alimova: Conceptualization, data curation, formal analysis, methodology, writing–original draft. D. Wang: Formal analysis. J. DeSisto: Formal analysis, methodology. E. Danis: Software, formal analysis, validation. S. Lakshmanachetty: Formal analysis, validation, visualization. E. Prince: Formal analysis. G. Murdock: Visualization. A. Pierce: Visualization, methodology. A. Donson: Formal analysis. I. Balakrishnan: Formal analysis, visualization. N. Serkova: Formal analysis. H. Lin: Resources. N.K. Foreman: Conceptualization. N. Dahl: Formal analysis. S. Venkataraman: Visualization, methodology. R. Vibhakar: Data curation, supervision, writing–review and editing.

The authors appreciate the contribution to this research made by E. Erin Smith, Allison Quador, and Jessica Arnold from the University of Colorado Pathology Shared Resource and Research Histology Division (Colorado, USA). The authors also thank N. Serkova and Jenna Steiner from the University of Colorado Animal Imaging Shared Resource (Colorado, USA) for acquiring all mouse MRI scans and data analysis. The authors thank the University of Colorado Cancer Center Functional Genomics Core Facilities for plasmids and RNA-seq. The authors are also grateful to the University of Colorado Cancer Center Flow Cytometry Shared Resource for invaluable guidance and assistance. The University of Colorado Pathology Shared Resource, Research Histology Division, Cancer Center Flow Cytometry Shared Resource, and Animal Imaging Shared Resource were supported by the Cancer Center Support Grant (P30CA046934), NIH Grant S10OD023485, and NIH Grant S10OD027023. This work was supported by the Morgan Adams Foundation Pediatric Research Program (R. Vibhakar), American Cancer Society grant CSCC-RSG-23-991677-01 (R. Vibhakar), and Department of Defense grant CA170677 (R. Vibhakar).