SWI/SNF is an evolutionarily conserved multi-subunit chromatin remodeling complex that regulates epigenetic architecture and cellular identity. Although SWI/SNF genes are altered in approximately 25% of human malignancies, evidences showing their involvement in tumor cell–autonomous chromatin regulation and transcriptional plasticity are limiting. This study demonstrates that human primary acute myeloid leukemia (AML) cells exhibit near complete loss of SMARCB1 (BAF47 or SNF5/INI1) and SMARCD2 (BAF60B) associated with nucleation of SWI/SNFΔ. SMARCC1 (BAF155), an intact core component of SWI/SNFΔ, colocalized with H3K27Ac to target oncogenic loci in primary AML cells. Interestingly, gene ontology (GO) term and pathway analysis suggested that SMARCC1 occupancy was enriched on genes regulating Rac GTPase activation, cell trafficking, and AML-associated transcriptional dysregulation. Transcriptome profiling revealed that expression of these genes is upregulated in primary AML blasts, and loss-of-function studies confirmed transcriptional regulation of Rac GTPase guanine nucleotide exchange factors (GEF) by SMARCB1. Mechanistically, loss of SMARCB1 increased recruitment of SWI/SNFΔ and associated histone acetyltransferases (HAT) to target loci, thereby promoting H3K27Ac and gene expression. Together, SMARCB1 deficiency induced GEFs for Rac GTPase activation and augmented AML cell migration and survival. Collectively, these findings highlight tumor suppressor role of SMARCB1 and illustrate SWI/SNFΔ function in maintaining an oncogenic gene expression program in AML.
Implications: Loss of SMARCB1 in AML associates with SWI/SNFΔ nucleation, which in turn promotes Rac GTPase GEF expression, Rac activation, migration, and survival of AML cells, highlighting SWI/SNFΔ downstream signaling as important molecular regulator in AML. Mol Cancer Res; 16(5); 791–804. ©2018 AACR.
This article is featured in Highlights of This Issue, p. 743
SWI/SNF (BAF) chromatin remodelers are evolutionarily conserved, large (∼2 MDa) multi-protein complexes, which utilize energy derived from ATP hydrolysis to mobilize nucleosomes (1). SWI/SNF core components include SMARCB1 (BAF47, SNF5 or INI1), SMARCC1/SMARCC2 (BAF155 and BAF170), and one of the mutually exclusive ATPase subunits, SMARCA4 (BRG1) and SMARCA2 (BRM). SWI/SNF complexes often include cell type–specific, lineage-restricted subunits, and play important roles in pluripotency and cellular reprogramming (1, 2). Cancer genome sequencing studies have identified SWI/SNF complexes as one of the most commonly mutated (∼25%) chromatin modulators in human cancer (3, 4). However, mutational profiling alone may not always inform transcriptional dependencies embedded in tumorigenesis.
Emerging evidences indicate that SWI/SNF subunits critically regulate murine hematopoiesis. Recent studies have shown that SMARCD2 mediates granulopoiesis through CEBPϵ-dependent mechanism (5, 6). Actl6a (BAF53a) plays essential role in hematopoietic stem cell (HSC) function (7). Mutant allele of Arid1a (BAF250a) determines pool size of fetal liver HSC populations (8). In addition, SWI/SNF was also implicated in murine leukemia development. SMARCA4 was shown to regulate proliferation of murine leukemic cells (9, 10). SMARCB1 plays tumor suppressor role in several cancers, and frequent deletion of SMARCB1 is observed in chronic myeloid leukemia patients (11). Loss of Smarcb1 in vivo leads to fully penetrant malignant rhabdoid tumors (12, 13).
Rac GTPases belong to small Rho GTPase family and are involved in regulation of a diverse array of cellular functions including cell proliferation, survival, adhesion, migration, actin assembly, and transcriptional activation (14, 15). Similar to Ras superfamily proteins, Rac GTPases cycle between inactive GDP-bound and active GTP-bound conformations, regulated by specific guanine nucleotide exchange factors (GEF), to transduce signals to effector proteins (14). Recent studies have suggested that Rac GTPases play integral roles in myeloid leukemia cell homing, engraftment, survival, and trafficking within the bone marrow microenvironment (15–18). Attenuation of Rac GTPase signaling in synergy with Bcl-2 inhibition has been shown as a modality for combination targeted therapy in MLL-AF9 leukemia (19). Myeloid leukemia cells are characterized with elevated Rac GTP level; however, molecular regulation of Rac activation in leukemia pathophysiology remains incompletely understood.
Here we identify that in human primary acute myeloid leukemia (AML) cells, SMARCB1 deficiency associates with nucleation of SWI/SNFΔ. SMARCC1, an intact core component of SWI/SNFΔ, colocalized with H3K27Ac to target tumor oncogenic loci, including Rac GTPase GEFs, in AML cells. Loss of SMARCB1 induced recruitment of SWI/SNFΔ and associated histone acetyltransferases (HAT) to target GEFs for Rac GTPase activation and promoted AML cell migration. Collectively, these findings highlight tumor suppressor role of SMARCB1 and illustrate SWI/SNFΔ function in maintaining an oncogenic gene expression program in AML.
Materials and Methods
Human AML (n = 67) bone marrow aspirates (1–2 mL each) were obtained from Park Clinic, Kolkata from untreated, freshly diagnosed patients after written, informed consent according to Institutional Human Ethics Committee (HEC) approval and following Indian Institute of Chemical Biology (CSIR-IICB) Institutional Review Board (IRB) set guidelines. Sample collection was part of routine diagnosis and the inclusion criterion for this study was histopathologic confirmation of bone marrow aspirates or biopsies, karyotyping, and immunophenotypic analyses (20). Bone marrow aspirates were also collected from age-matched normal individuals (n = 6) after informed consent, who turned out to be pathologically negative for AML (20). Individual case information is presented in Supplementary Tables S1 and S2. Umbilical cord blood samples (40 mL each) were obtained from Deb Shishu Nursing Home (Howrah, West Bengal, India) from term pregnancies after written, informed consent according to CSIR-IICB HEC approval and following IRB set guidelines. Low density (1.077 gm/cc) nuclear cells from AML bone marrow, normal bone marrow, or cord blood samples were isolated by Ficoll (Sigma) separation and cryopreserved in liquid nitrogen. Normal blood specimens were obtained from age-matched healthy volunteers (n = 3) after written, informed consent and nucleated cells were isolated using RBC lysis (BD Pharmingen).
Reverse transcription and quantitative PCR
Total RNA was isolated by using TRIzol (Life Technologies) according to manufacturer's recommendation. RNase-free DNase treatment were carried out to remove any genomic DNA contamination using DNase I recombinant, RNase free kit (Roche). RNA amount was quantified and cDNA was prepared using TaqMan Reverse Transcription Reagents (Applied Biosystems). Gene expression levels were determined by quantitative PCR performed using cDNA with SYBR Select Master Mix (Applied Biosystems) on the 7500 Fast Real-Time PCR System (Applied Biosystems). GAPDH was used as a housekeeping gene. Relative expression levels were calculated using the 2−ΔΔCt method (21–24). qRT-PCR primer details are available in Supplementary Table S3.
Array Comparative Genomic Hybridization and analysis
Array Comparative Genomic Hybridization (CGH) and analysis were carried out by Genotypic Technology Private Limited. Genomic DNA quantity and purity was assessed by the NanoDrop ND-2000 UV-Vis Spectrophotometer (NanoDrop Technologies) and the integrity was assessed on a 0.8% Agarose Gel. Genomic DNA with OD260/OD280 >1.8 and OD260/OD230 ≥ 1.3 was used for microarray experiments. DNA was considered to be of good quality when a single clear band was seen when run against a reference ladder. A total of 0.5 μg of DNA in 10.1 μL was taken into a microfuge tube and digestion master mix containing restriction enzymes (Alu I, 5U and Rsa I, 5U) was added. The samples were incubated at 37°C for 2 hours followed by heat inactivation of enzymes at 65°C for 20 minutes. Samples were labeled using Agilent sure tag DNA Labeling Kit (catalog no: 5190-3399). Control samples were labeled with Cy3 and test sample with Cy5. The labeled samples were cleaned up using Amicon Ultra columns 30-kDa size exclusion filter. DNA yield and incorporation of labeled dye (specific activity) was measured using NanoDrop spectrophotometer. One micrograms each of Cy3- and Cy5-labeled sample was added with human cot-1 DNA (catalog no. 5190-3393), Agilent aCGH/CoC Blocking agent (part number: 5188-6416), and hybridization buffer (part number: 5188-6420). The labeled samples in above hybridization mix were denatured at 95°C for 3 minutes and were incubated at 37°C for 30 minutes. The samples were then hybridized at 65°C for 24 hours. After hybridization, the slides were washed using Agilent aCGH Wash Buffer1 (Agilent Technologies, part number 5188-5221) at room temperature for 5 minutes and Agilent aCGH Wash Buffer 2 (Agilent Technologies, part number 5188-5222) at 37°C for 1 minute. The slides were then washed with acetonitrile (part number: A2094) for 10 seconds. The microarray slides were scanned using Agilent Scanner (Agilent Technologies, part number G2600D). Image analysis was performed using Agilent Feature Extraction software, feature extracted raw data was analyzed using Agilent Genomic Workbench 7.0 software. The data were normalized using Lowess normalization. Significant regions having amplification and deletions were identified among each of the samples. Genomic view and chromosome view of the amplification and deletion region for each sample were generated. Graphical representation has been done using Human UCSC Genome Browser by loading the data in wiggle file format. Details are included as Supplementary array CGH Files.
Genomic DNA was isolated using the QIAamp DNA Blood Mini Kit (Qiagen, catalog no. 51104). Isolated genomic DNA was then bisulfite converted, that is, unmethylated cytosines were converted to uracil using EpiTect Bisulfite Kit (Qiagen, catalog no. 59104) according to the manufacturer's protocol. Methylation-specific primers for the gene of interest were made using MethPrimer. Bisulfite converted DNAs were amplified using methylated DNA (M pair)-specific primers. Methylation-specific PCR at SMARCB1 promoter loci in primary AML blasts was compared with normal BMNC. The fold change levels of the methylated DNA were calculated with respect to GAPDH (unrelated control). Relative methylation levels were plotted after normalizing it with GAPDH. qMSP primer details are available in Supplementary Table S3.
Coimmunoprecipitation, histone acid extraction, and immunoblotting
Nuclear extracts for immunoprecipitation experiments were prepared using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) and diluted in 1× RIPA (Cell Signaling Technology) containing protease and phosphatase inhibitor cocktails. About 300 μg extracts were incubated with 2.0 μg of respective antibodies against SMARCC1 (clone R-18, sc-9746, Santa Cruz Biotechnology), p300 (A300-358A, Bethyl Laboratories), CBP (clone D6C5, 7389S, Cell Signaling Technology), BRD4 (clone E2A7X, 13440S, Cell Signaling Technology), or rabbit IgG (P120-101; Bethyl Laboratories), and incubated overnight at 4°C with gentle rocking. Fifty microliters of protein A/G agarose beads (Cell Signaling Technology) were added and incubated for 3–4 hours at room temperature. The beads were then washed 6 times with 1× RIPA supplemented with 300 mmol/L NaCl and resuspended in 1× SDS gel loading buffer. The proteins were separated in SDS-PAGE and transferred to PVDF membrane (Millipore) and subsequently probed with respective antibodies. Detailed list of antibody is included in Supplementary Table S4. All antibodies were used at a dilution of 1:1,000 unless specified otherwise. Total cell lysate for immunoblotting was prepared by incubating cells in 1× RIPA for 15 minutes followed by brief sonication. Supernatants were collected following centrifugation at 16,000 × g for 15 minutes at 4°C. Protein concentration was determined using Pierce BCA Protein assay kit (Thermo Fisher Scientific). SDS-PAGE was used to separate proteins, transferred to PVDF membrane, and probed using respective antibodies. Densitometry analyses were performed using NIH Image J software.
Sucrose density gradient centrifugation
A total of 700 μg–1.0 mg nuclear extracts isolated from pooled (n = 5–7) primary AML BMNCs were prepared and diluted in 300 μL of 1× RIPA. The extracts were overlaid on a 10 mL 20%–50% sucrose gradient (in 1× RIPA) in 13 × 89 mm polyallomer tubes (Beckman Coulter). The tubes were then centrifuged in a SW-41 Ti swing out rotor at 30,000 rpm for 12 hours at 4°C. A total of 500-μL fractions were collected and separated in SDS-PAGE, transferred to a PVDF membrane, and subsequently probed with specific antibodies.
ChIP-seq, ChIP-qPCR, and analyses
ChIP-seq experiments were carried out at Core Technologies Research Initiative (CoTeRI), National Institute of Biomedical Genomics (NIBMG, University of Kalyani, Kolkata, West Bengal, India). A total of 1 × 107 primary bone marrow nuclear cells (BMNC), isolated from three independent (biological replicates) AML patients (Supplementary Table S1 case nos. 82, 83, and 80 as AML 01, AML 02, and AML 03, respectively), per chromatin immunoprecipitation (ChIP) set were crosslinked with formaldehyde (Sigma) in culture media. After crosslinking, chromatin was extracted and sonicated to fragment lengths between 150 and 900 bp in chromatin extraction buffer containing 10 mmol/L Tris pH = 8.0, 1 mmol/L EDTA pH = 8.0, 0.5 mmol/L EGTA pH = 8.0. Chromatin was incubated with ChIP-grade antibodies to SMARCC1 [sc-9746 (R-18), Santa Cruz Biotechnology), H3K27Ac (ab4729, Abcam), H3K27Me3 (07-449, Millipore), or rabbit IgG (clone P120-101; Bethyl Laboratories) or mouse IgG (clone G3A1; 5415S; Cell Signaling Technology) or goat IgG (sc-2028, Santa Cruz Biotechnology) during overnight at 4°C with rotation. Six micrograms of antibody was used per immunoprecipitation. All antibodies were used at 1:1,000 dilution. Detailed list of antibody is included in Supplementary Table S4. Protein A/G Agarose beads (Cell Signaling Technology) were then added and incubated for 2 hours at 4°C. The beads were washed with chromatin extraction buffer and by increasing salt concentration for four times. The chromatin was eluted from the beads in chromatin elution buffer at 65°C with gentle vortexing. The eluted chromatin was treated with RNAse for 30 minutes at 37°C. Reverse cross-linking was performed by treating the eluted chromatin with Proteinase K (Sigma) at 65°C for 2 hours. The DNA was finally precipitated by phenol–chloroform extraction; precipitated DNA was dissolved in TE buffer, and subjected to ChIP-seq analyses. For ChIP-qPCR experiments, 4–5 × 106 normal CD34+ cells or 293T cells and 2 μg of antibody, or ChIP DNA obtained from primary AML BMNCs were used. ChIP-qPCR primer details are available in Supplementary Table S5.
Size distribution of the ChIP-enriched DNA was checked using High Sensitivity chips in 2100 Bioanalyzer (Agilent) for each sample and quantitation was performed in Qubit Fluorometer (Invitrogen) by picogreen method. ChIP-seq library preparation was performed using TruSeqChIP Sample Prep Kit (Illumina) according to the manufacturer's instructions. Ten nanograms of input ChIP-enriched DNA was used for ChIP-seq library preparation. Final libraries were checked using High Sensitivity chips in 2100 Bioanalyzer (Agilent). Average fragment size of final libraries was found to be 280 ± 8 bp. Paired-end sequencing (2 × 100 bp) of these libraries were performed in HiSeq-2500 (Illumina). Quality control analysis of the raw data using NGS-QC ToolKit was done and HQ reads with filter criteria of bases having ≥20 Phred score and reads with ≥70% were filtered. Paired-end reads (.fastq format) were aligned with Bowtie software using –best and –m 2, that is, mismatches against reference genome Ensembl build GrCh37/hg19 (considering 2% input as the baseline) and saved in SAM format, which was then converted to sorted BAM file using SAMTOOLS. PCR duplicates were removed using SAMTOOLS rmdup. Peak calling was performed using MACS14 model building with P value cutoff of 0.05. Annotation of the identified peaks was performed with PeakAnalyzer.
Functional enrichment analysis (Gene Ontology and Pathway) was done using The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8. Gene list were uploaded and converted to its respective ENTREZ Gene ID. The converted gene list was submitted to DAVID and Functional Annotation Clustering was carried out which comprises of Gene Ontology and Pathway analysis. R bioconductor package ChipSeeker was used to generate heatmap, average profile distibutions, and pie charts. Unique gene names were used to plot the Venn diagram represented by peaks in the respective samples either upstream or downstream or overlap to the genetic region. Bigwig/bed files were imported into Integrated Genome Viewer (IGV) and snapshots of particular genomic loci were captured. For motif analysis, the MEME ChIP V4.10 was used to analyze the motif using the peak sequences, with default parameters. and for transcription encoding motif, Jaspar database (Jaspar) was used using MEME ChIP tool. Details are included as Supplementary ChIP-seq Files.
RNA-seq and analyses
RNA-seq experiments were performed by Bionivid Technology. Total RNA was isolated from BMNCs from the identical AML cohort (n = 3) and age-matched normal (n = 2) hematopoietic cells using TRIzol (Life Technologies) according to manufacturer's instruction. DNase treatment was carried out to remove any genomic DNA contamination using DNase I recombinant, RNase free kit (Roche). RNA amount was quantified and the sequencing library prepared using TruSeq RNA Sample Prep Kit v2 (Illumina). Paired-end sequencing was performed on HiSeq 4000 using TruSeq 3000 4000 SBS Kit v3 (Illumina).
Raw data resulted in an average of 35.39 × 106 reads in normal hematopoietic cells and 36.86 × 106 reads of 101-bp length in primary AML cells. Quality control using NGSQC toolkit yielded around 34.28 × 106 HQ reads in normal hematopoietic cells and 36.05 × 106 HQ reads in primary AML cells. Around 88.2% of the HQ reads from normal and 85.12% HQ reads from primary AMLs could be mapped to the Homo sapiens (hg 38) genome reference sequence using TOPHAT, suggesting a good quality of RNA sequencing. Transcripts were given a score for their expression by Cufflinks-based maximum likelihood method. A total of 24,784 transcripts were identified using Cuffdiff validation to be expressed in either normal or primary AML cells representing 13,847 genes. Transcript type analysis revealed 95.5% of the transcripts to be of “Full Length” or “Known Transcripts” and 4.5% as “Potentially Novel Isoforms” Transcripts as per Cufflinks Class Code distribution. This indicates a largely complete transcription machinery activity in both normal and AML cells. Significant Biology for Differentially Expressed Transcripts was performed with GO-Elite_v.1.2.5 Software. A cutoff of P value less than 0.05 was considered for filtering the significantly enriched GO pathways. In testing for differential expression, we consider log2FC > 2 (upregulation) and log2FC ← 2 (downregulation). For gene-set enrichment analysis (GSEA), differentially expressed genes from individual comparisons were preranked on the basis of fold change such that maximally upregulated genes fall topmost in the list. This was used as an input to perform GSEA (GeneSpring). GSEA was performed on “H: Hallmark gene set” representing well defined biological states or processes available on Molecular Signature Database. Details are included as Supplementary RNA-seq Files.
Gene enrichment and functional annotation analysis
Gene ontology (GO) analysis of the shared gene set (Supplementary ChIP-seq Shared GENELIST) was carried out using DAVID v6.8 (https://david.ncifcrf.gov/). The P value used in the analysis is a modified one, termed as EASE score threshold (maximum probability). The threshold of EASE Score is a modified Fisher exact P value used for gene enrichment analysis. It ranges from 0 to 1. Fisher exact P value = 0 represents perfect enrichment. Usually P value is equal or smaller than 0.05 to be considered strongly enriched in the annotation categories.
shRNA-expressing lentiviral constructs targeting against SMARCB1 (pLKO-shSMARCB1, 39587) and SMARCB1-overexpression vector HA_INI1/BAF47 was a gift from Dr. Olivier Delattre (Institut Curie, Paris, France). shRNA-expressing lentiviral construct targeting against SMARCB1 (pLKO.1-puro-CMV-TGFP, TRCN0000295966) was purchased from Sigma. Empty vector SHC003 was purchased from Sigma. Lentiviral packaging constructs PAX2 (Addgene; 12260) and pMD2.G (Addgene; 12259) were purchased from Addgene.
HSPC isolation, lentivirus preparation, and transduction
CD34+ HSPCs were isolated from freshly collected normal BMNCs and cord blood nuclear cells or from cryopreserved specimens using CD34 Microbead positive selection kit (Miltenyi Biotec) following manufacturer's protocol (21, 23). For lentivirus preparation, 293T cells were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mmol/L l-glutamine (all from Gibco) at 37°C with 5% CO2. Cells were seeded in T-225 flasks at 70% confluence and transfected with the target plasmid DNA, PAX2 (Addgene; 12260), and pMD2.G (Addgene; 12259) using calcium-phosphate transfection method (23, 25). After overnight incubation, butyrate induction was given for 8 hours. Supernatant containing lentiviral particles were collected after 36–40 hours of incubation at 37°C with 5% CO2, and ultracentrifuged at 25,000 rpm for 90 minutes at 4°C using (Sorvall, Thermo Scientific). Virus pellet was resuspended in X-VIVO (Lonza) and aliquoted in several tubes and stored at −80°C. HL60 and U937 cell lines were stably transduced with lentiviral particles containing mock (sh-Control) or sh-SMARCB1 plasmids in a U-bottom 96-well nontissue-culture–treated plate. A total of 1 × 105 cells were incubated overnight with virus particles at a MoI of 5 and polybrene was added at a final concentration of 8 μg/mL to initiate lentiviral infection.
Cells, drug treatments and survival, and proliferation assays
Human AML cell lines (obtained from Dr. Jose Cancelas, Cincinnati Children's Hospital, Cincinnati, OH) were maintained in Iscove's modified Dulbecco's medium (IMDM) supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mmol/L l-glutamine (all from Gibco) at 37°C with 5% CO2. 293T cells (obtained from Dr. Jose Cancelas, Cincinnati Children's Hospital, Cincinnati, OH; refs. 15, 23) were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mmol/L l-glutamine (all from Gibco) at 37°C with 5% CO2. Adherent cells were transfected at 70% confluency using the calcium phosphate transfection method (16, 22, 23, 25). Cell lines have been freshly authenticated using STR profiling (Supplementary Cell Lines_STR Profiles).
JQ1 (cat SML0974) was purchased from Sigma. For calculating IC50, parental or lentivirus transduced AML cell lines were treated with varying doses of JQ1 from 0.1 to 50 μmol/L. Viable cell counts were taken after 48 hours of drug treatment and the total number of GFP+ cells were analyzed by flow cytometry. Cell counts were normalized and plotted against logarithm of the inhibitor concentration using GraphPad Prism5 to measure the IC50. For proliferation assay, lentivirus transduced cell lines were grown in triplicate in regular media for 6 days. Trypan blue–negative cell numbers were determined at respective time points and the total number of GFP+ cells were analyzed by flow cytometry.
AML BMNCs were grown in regular media supplemented with cytokines in presence of 500 nmol/L 5-azacytidine (Sigma, catalog no. A1287) or DMSO (vehicle) for 72 hours. Media was calibrated with fresh 5-azacytidine and cytokines after every 24 hours. Posttreatment, total RNA was isolated and gene expression levels were determined by qRT-PCR.
PAK1 pulldown and Rac GTPase activation assay
Cells were lysed using 400 μL of MLB buffer (1×) with repeated pipetting, centrifuged at 14,000 × g for 5 minutes at 4°C and supernatants were used for PAK1 pull down assay (Upstate, Millipore) as described earlier (15, 16, 18, 26, 27). To the supernatant, 10 μL of Rac1-conjugated agarose beads were added and incubated for 45 minutes at 4°C with gentle rocking. The beads were centrifuged at 14,000 × g for 10 seconds at 4°C. After removing the supernatants the beads were washed three times with MLB buffer (1×) and resuspended in 40 μL of protein loading buffer, boiled for 5 minutes, separated in 12% polyacrylamide gel, and transferred to PVDF membrane and probed using respective antibodies. Presence of active Rac GTP was determined by using antibody against Rac GTP in the pulldown fraction and normalized against total Rac present in the lysate. Densitometry analyses were performed using NIH Image J.
HL60 cells were transduced with lentiviral particles expressing sh-SMARCB1 or sh-Control in a U bottom 96-well nontissue culture–treated plate. Cells were incubated overnight with virus particles at a MoI of 5 and polybrene was added at a final concentration of 8 μg/mL. Forty-eight hours posttransduction, 50,000 cells were seeded in duplicate in the top chamber of a 24-well Transwell (Corning Incorporated) in 100 μL of IMDM. After 4 hours of incubation at 37°C with 5% CO2, the total migrated cells were counted from bottom chamber containing 600 μL of IMDM supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 100 ng/mL CXCL12 (PeproTech). Total number of GFP+ cells was analyzed by flow cytometry.
Flow cytometric analysis
AML cell lines were transduced with lentiviral particles expressing sh-SMARCB1 or sh-Control and coexpressing GFP at a MoI of 5. After 48 hours, cells were harvested at 500 × g for 5 minutes and washed two times with cold PBS and resuspended in 500 μL of PBS supplemented with 2% human serum and 7-AAD at a final concentration of 1 μg/mL. 7AAD−/GFP+ cells were analyzed in LSRFortessa (Becton Dickinson) using FACSDiva software (Becton Dickinson).
Correlation and survival analysis of TCGA AML cohort
Cross-cancer analysis of SMARCB1 and DOCK5 expression, as well as SMARCB1 and DOCK5 expression heatmap cluster of TCGA AML cohort, was derived using cBioPortal for Cancer Genomics interface (28, 29). For correlation analysis, mRNA expression (RNA Seq V2 RSEM) data were obtained from TCGA database and plotted using GraphPad Prism5. For correlation as well as Kaplan–Meier survival analysis, mRNA expression (RNA Seq V2 RSEM) was used for clustering of samples; all samples showing expression level above mean value for the particular gene were considered “hi,” whereas all samples showing expression level below mean value were considered “lo.”
Statistical analyses were performed using GraphPad Prism 5. Statistics were calculated with Student's t test. Quantitative data are expressed as mean ± SEM. unless specified otherwise. For IC50 calculation, cell counts were normalized and plotted against logarithm of the respective inhibitor concentration using GraphPad Prism5. Gene expression correlation plots were derived using the TCGA datasets of AML cohort available through cBioPortal. Densitometry analyses were performed using Image J software (NIH). For all statistical analyses, the level of significance was set at 0.05.
All sequencing data have been submitted to the database with accession numbers as follows. ChIP-seq: GSE108976. RNA-seq: SRA accession: SRP127783; BioProject ID: PRJNA428149.
Human primary AML cells show loss of SMARCB1 expression and SWI/SNFΔ nucleation
We set out to identify SWI/SNF contribution to human AML pathogenesis. Gene expression analysis identified a significant loss of SMARCB1 (SNF5 or BAF47) in human primary AML bone marrow nuclear cells (BMNCs; P < 0.0058, n = 67) compared with age-matched normal bone marrow (NBM) CD34+ hematopoietic stem/progenitor cells (HSPC; Fig. 1A; Supplementary Tables S1 and S2). SMARCB1 expression was also downregulated in established AML lines (Fig. 1B). Array CGH analysis did not detect amplification, deletion, or copy number gain or loss, or any other genetic alteration at the SMARCB1 locus in our AML cohort (Supplementary array CGH Files). However, compared with normal hematopoietic cells, AML blasts showed a substantial increase in repressive DNA methylation at CpG islands of the SMARCB1 promoter (Fig. 1C), accounting for SMARCB1 downregulation observed in AML. In agreement with this result, inhibition of DNA methyl transferases in vitro restored SMARCB1 levels in SMARCB1lo AML blasts (Fig. 1D).
Apart from SMARCB1, expression of SMARCD2 (BAF60B), SMARCE1 (BAF57), and ARID2 (BAF200) were also significantly reduced in AML BMNCs compared with NBM CD34+ cells (Fig. 1E and F). Consistent with the mRNA downregulation, SMARCB1 and SMARCD2 protein levels were dramatically lost in primary AML cells (Fig. 1G). Expression of SMARCC1 (BAF155), core subunit, and SMARCA4 (BRG1), ATPase subunit of SWI/SNF remained intact in AML (Fig. 1G). Coimmunoprecipitation experiments indicated association of endogenous SMARCC1 with the remaining SWI/SNF complex in AML BMNCs as well as HL60 cells (Fig. 1H). Sucrose density gradient analysis further confirmed presence of an endogenous, residual, nuclear SWI/SNF complex (hereafter called SWI/SNFΔ) in primary AML cells (Fig. 1I). Collectively, these data identify loss of SMARCB1 and SWI/SNFΔ nucleation in human AML.
SMARCC1 (SWI/SNFΔ) is involved in maintenance of oncogenic gene expression program in primary AML cells
Mammalian SWI/SNF complex organizes nucleosome occupancy at target promoters and enhancers (30), thereby regulating gene expression. Recent studies have shown SMARCB1 deficient SWI/SNF complex to be essential for rhabdoid tumor survival (31, 32). To elucidate the function of SWI/SNFΔ in AML, we investigated genome-wide occupancy of endogenous SMARCC1, which would indicate SWI/SNFΔ binding, in primary AML cells using chromatin immunoprecipitation-sequencing (ChIP-seq). SMARCC1 ChIP-seq identified about 14,000 genes on average in three independent (biological replicates) AML BMNCs (Fig. 2A and B; Supplementary Fig. S1A–S1C). SMARCC1 localized approximately 10% at promoters, approximately 41% at gene body, and approximately 49% at transcription start site (TSS)-distal intergenic regions (Fig. 2C and D). In rhabdoid tumor, it has been shown that SWI/SNF binding at TSS-distal enhancer loci, marked by H3K27Ac, is essential for its oncogenic role (31). Also SMARCB1 deficiency has been shown to regulate H3K27Ac level at enhancers (32, 33). In general, H3K27Ac is enriched at sites of active transcription; therefore, SWI/SNF function is typically associated with transcription activation. H3K27Ac ChIP-seq analysis indicated SMARCC1 overlapped with H3K27Ac at 2,660 genes (Supplementary ChIP-seq Files) that are shared among the three biological replicates of AML. Detailed statistical analysis is included in Supplementary ChIP-seq Files as “Supplementary ChIP-seq_P values of Venn diagram genes.” Enrichment of SMARCC1 at these shared genes was approximately 10% at promoters, approximately 55% at gene body, and approximately 35% at TSS-distal intergenic regions (Fig. 2E and F).
To identify the role of SMARCB1 in directing SWI/SNFΔ function, we compared the AML ChIP-seq data with primary, normal hematopoietic nuclear cells expressing SMARCB1-containing intact SWI/SNF complex (Supplementary Fig. S2A). ChIP-seq analysis identified about 16,500 genes occupied by SMARCC1, out of which 13,987 also showed H3K27Ac (Supplementary Fig. S2B). The distribution pattern of SMARCC1 was also similar to that in AML, with 4%, 37%, and 58% being the occupancy at promoter, intron, and TSS-distal intergenic region respectively (Supplementary Fig. S2B). Hence unlike in rhabdoid tumor, SMARCB1 deficiency does not affect the overall occupancy of SWI/SNF complex in AML, indicating that the regulation may rather be gene or loci specific. Motif analysis of SMARCC1-binding sites in AML identified enrichment of several transcription factors, notably KLF4, HOXA13, and HOXD13 that are implicated in AML (Supplementary Fig. S2C and Supplementary ChIP-seq Files). Motif analysis in normal hematopoietic cells identified enrichment of MEF2A, MEF2B, and MEF2D (Supplementary Fig. S2C and Supplementary ChIP-seq Files). Therefore, SWI/SNF is associated with different sets of transcription factors in AML and normal hematopoietic cells. This suggests that although SMARCB1 loss does not alter overall chromatin affinity of SWI/SNF, it may differentially determine recruitment to and regulation of altered gene sets.
Next, to identify gene sets regulated by SWI/SNFΔ, we performed functional annotation clustering of the SMARCC1 and H3K27Ac cooccupied genes. Gene Ontology (GO) terms and pathway analysis showed an enrichment of transcripts associated with Rac GTPase-dependent cell migration, hematopoietic self-renewal, and transcriptional regulation (Fig. 3A). Interestingly, among these gene sets, we noted SMARCC1 occupancy at several Rac GTPase GEFs. To validate differential expression of gene sets identified as SWI/SNFΔ targets, we next performed transcriptome analysis. RNA-sequencing of paired samples identified, among the SMARCC1 and H3K27Ac cooccupied genes, 280 were significantly upregulated in AML, compared with normal (n = 2) hematopoietic cells (Supplementary RNA-seq Files). Among the upregulated genes were VAV3 and the DOCK family of Rac GEFs. Importantly, although these Rac GTPase GEFs show SMARCC1 binding also in controls, but the binding sites are significantly distinct from that observed in AML (Fig. 3B; Supplementary Fig. S2D). In addition, H3K27Ac did not cooccupy SMARCC1 binding sites in normal hematopoietic cells (Fig. 3B; Supplementary Fig. S2D). Together, genes showing SWI/SNFΔ overlap with H3K27Ac and resultant upregulation represent putative SMARCB1-dependent targets.
To further confirm differential SWI/SNFΔ binding due to SMARCB1 loss, ChIP-qPCR analysis was performed using SMARCC1 and H2K27Ac antibodies in normal CD34+ cells and SMARCB1-deficient primary AML blasts (Fig. 3C; Supplementary Fig. S3A). SMARCC1 and H3K27Ac cooccupancy were significantly higher at the GEFs in AML blasts compared with the identical loci in normal CD34+ cells (Fig. 3C; Supplementary Fig. S3A). Upregulation of the GEFs expression was additionally validated by qRT-PCR analysis, which shows that they are elevated in AML blasts compared to control (Supplementary Fig. S3B). Collectively, these results indicate differential locus-specific binding of SWI/SNF at target GEFs in AML.
SMARCB1 levels correlate with DOCK expression and AML pathophysiology
The Cancer Genome Atlas (TCGA) cross-cancer analysis reveals SMARCB1 median expression level to be minimum in AML patients, after mesothelioma (Fig. 4A). This as well as our AML cohort indicates that SMARCB1 downregulation is a general phenomenon observed in AML. Unlike SMARCB1, median expression of SMARCD2 was not downregulated, and was at par with other cancers (Supplementary Fig. S3C). Therefore, we focused our analysis on SMARCB1 for subsequent studies. We wanted to evaluate whether SMARCB1 levels have any prognostic significance in AML. To this end we studied the survival trends of AML patients corresponding to their expression of SMARCB1 from TCGA database. In many AML subtypes, well-characterized oncogenic translocations are enough to drive leukemogenesis. To eliminate the effect of complex translocations, only patients with normal karyotype were considered. Patients with lower SMARCB1 levels showed correspondingly poorer nonsignificant overall (P = 0.561) as well as disease-free (P = 0.230) survival.
DOCK family of Rac GTPase GEFs being one of the primary targets of SWI/SNFΔ as identified from ChIP-seq and transcriptome analysis, we next wanted to check whether there was any correlation between SMARCB1 and DOCK members in AML. TCGA cross-cancer analysis showed that DOCK5 has highest expression in AML among multiple cancers (Fig. 4B). This is in stark contrast to SMARCB1 expression, which led us to hypothesize that SMARCB1 and DOCK5 expression must be inversely correlated. Indeed, TCGA database analysis confirmed significant negative correlation between SMARCB1 and DOCK5 expression in AML (Fig. 4C). Patients with reduced SMARCB1 showed significantly upregulated DOCK5 levels (Fig. 4C). Like DOCK5, the other atypical Rac GTPase GEFs, DOCK2, DOCK8, and DOCK10 also showed reciprocal correlation with SMARCB1 (Fig. 4C). Survival analysis further strengthened the SMARCB1-DOCK interdependence, demonstrating that patients with low SMARCB1 and high DOCK expression have poorer nonsignificant survival than patients with high SMARCB1 and low DOCK levels.
SMARCB1 deficiency induces GEFs expression by promoting H3K27Ac at target loci
SWI/SNFΔ binding analysis indicated that SMARCB1 deficiency is accompanied with elevated H3K27Ac at oncogenic loci in AML cells. In agreement with this, lentivirus-mediated silencing of SMARCB1 in normal CD34+ cells and in established AML cell lines (Supplementary Fig. S3D–S3G), which still express residual SMARCB1, induced GEFs expression (Fig. 5A and B; Supplementary Fig. S4A and S4B). Acetylation at H3K27, which controls transcriptional state and facilitates gene expression are mediated by HATs, and recently it has been shown that SWI/SNF coimmunoprecipitates with HATs in rhabdoid tumors (33). Coimmunoprecipitation studies indicated interaction of SWI/SNFΔ and HATs in AML cells (Supplementary Fig. S4C). We investigated whether loss of SMARCB1 can affect recruitment of SWI/SNFΔ and HATs at target GEFs. Although SMARCB1 loss did not apparently affect SWI/SNF interaction with the HATs (Supplementary Fig. S4D and S4E), interestingly SMARCB1 deficiency resulted in an increased occupancy of SMARCC1, HATs, and H3K27Ac levels at target GEFs loci (Fig. 5C; Supplementary Fig. S5A and S5B). Collectively, these findings provide mechanistic evidence for SMARCB1 loss and SWI/SNFΔ-mediated transcriptional regulation of GEFs expression in AML.
Loss of SMARCB1 induces Rac GTPase activation
Rac GEFs control the activation of Rac GTPase signaling; therefore, we asked whether induction in GEFs expression in SMARCB1-deficient cells affect Rac activation. Consistent with the role of SMARCB1 in preferential recruitment of SWI/SNF, HAT and H3K27Ac, and expression of DOCK genes, loss of SMARCB1 resulted in approximately 2-fold activation of Rac GTPase (Fig. 6A and B). This was accompanied with increase in cell migration, suggesting its tumor suppressor role (Fig. 6C). In our AML discovery cohort, Rac GEFs emerged as important downstream candidates of SWI/SNFΔ. Mechanism of SWI/SNFΔ-mediated oncogene induction is through recruitment of HATs and increased H3K27 acetylation. HATs as well as several SWI/SNF subunits contain acetyl-lysine binding bromodomains, which is a target of BET inhibitors (34). We therefore sought to determine whether SMARCB1-deficiency would alter sensitivity of AML cells to BET inhibition. SMARCB1-deficient AML cells were significantly more sensitive to BET inhibition than controls (Supplementary Fig. S5C). Collectively, these results indicate that loss of SMARCB1 in AML cells results in increased occupancy of SWI/SNFΔ along with HATs to target GEFs (Fig. 6D), which induces GEFs expression, activation of Rac GTPase signaling and cell migration.
In this study, we present evidence that in human primary AML cells there is loss of SMARCB1, which is associated with nucleation of SWI/SNFΔ. Leukemic SWI/SNFΔ retains core components SMARCC1 and SMARCA4. Recent reports suggest that loss of Snf5, yeast homolog of mammalian SMARCB1, induces formation of aberrant SWI/SNF complex (35), and SMARCB1-deficient malignant rhabdoid tumors depend on SMARCA4 for transformation (36). SMARCB1 has previously been implicated in rhabdoid tumor, with biallelic inactivating mutations sufficient to drive malignant transformation (13). A separate study has shown that SMARCA4 regulates proliferation of murine leukemic cells (9). In addition, Smarcd2-deficient mice fail to generate functionally mature myeloid cells (5, 6). Therefore, individual subunits of SWI/SNF complex have been shown to coordinate diverse cellular functions regulating disease and development. Transcriptional plasticity is an emerging aspect in tumorigenesis (37, 38). Herein, our study provides evidence and reinforces the importance of leukemic SWI/SNFΔ with altered subunit stoichiometry configuration toward maintenance of oncogenic gene expression program precisely in human primary AML cells. It strengthens the importance of epigenetic perturbation of the SWI/SNF complex in tumorigenesis. Essentially, our data indicate that in AML, SMARCB1 deficiency associates with altered SWI/SNFΔ, and leukemic cells depend on SWI/SNF core components for transcriptional dysregulation and survival.
Importantly, in our study among the SMARCC1 and H3K27Ac cooccupied genomic targets, we noted SMARCC1 occupancy at several Rac GTPase GEFs, which play important roles in cell survival, trafficking, and small GTPase signaling (17, 39, 40). SMARCB1 deficiency upregulated expression of the GEFs, and was associated with Rac GTPase activation and hypermigration of AML cells. Earlier, we and others have shown that Rac GTPases critically regulate leukemia cell engraftment and survival (15–18). VAV3, Rho/Rac GTPase GEF, was implicated in leukemogenesis (41, 42). DOCK2 is a noncanonical GEF for Rac GTPases, and DOCK2 inhibition in vivo attenuates AML cell survival (43, 44). SNF5 was implicated in regulation of RhoA-dependent cytoskeleton organization and migration of malignant rhabdoid tumor cells (45). In our study, SMARCC1 occupancy was also enriched at KIT, ASXL1, SF3B1, TET2 target loci, and silencing of SMARCB1 induced their expression. Many of these genes are somatically mutated at relatively high frequency in myeloid malignancies with poor prognosis (46, 47), suggesting that SMARCC1/SWI/SNFΔ would help sustain expression of mutant oncoproteins in AML. Collectively, these findings account for SWI/SNFΔ involvement in maintenance of AML cell gene expression program.
We demonstrate that a fraction of SMARCC1 and H3K27Ac cooccupied genomic targets in AML cells were enriched in TSS-distal intergenic regions (∼30%). SWI/SNF was shown to play chromatin remodeling function at both promoters and enhancers (30, 48). Recent reports have demonstrated an interdependency of SWI/SNF and HAT function (31). SMARCB1 levels were shown to regulate not only expression of p300, BRD4, and mediator, but also control interaction of SWI/SNF with p300 in rhabdoid tumor (31). Intact SWI/SNF function, overlapping with H3K27Ac is needed for the maintenance of lineage-specific enhancers, regulating cell fate and differentiation (33). SMARCB1 deficiency, however, shifts SWI/SNF recruitment from enhancers to oncogenic super-enhancer regions (31). However, in contrast to some of these reports where experiments were performed in different cell types, our results indicate that loss of SMARCB1 induces Rac GEFs expression that is associated with elevated SMARCC1 and H3K27Ac occupancy at target loci. This is similar to an earlier study demonstrating that Snf5 localizes to Gli1-regulated promoters and that loss of Snf5 leads to activation of the Hedgehog–Gli pathway in malignant rhabdoid tumors (49). Essentially, these findings indicate that SWI/SNF function and epigenetic plasticity secondary to absence of specific SWI/SNF subunits are cell type and context-dependent phenomenon.
Genetic perturbations are usually attributed to genetic deletions and inactivating mutations. Apart from this DNA methyltransferases play an important role in silencing expression of key tumor suppressor genes (50). Genotyping of our AML cohort indeed corroborates this, showing that SMARCB1 loss seen in AML is not due to genetic deletion, but rather increased DNA methylation at CpG islands of proximal promoter region. In addition, we demonstrate that expression of SMARCB1 corresponds to prognosis in AML patients. SMARCB1 expression is the lowest in AML among multiple cancers; and also within AML cohort, patients with low SMARCB1 levels to have comparatively shorter overall as well as disease-free survival period compared with patients with higher SMARCB1 expression. That SWI/SNFΔ mediated Rac GEF regulation is indeed important in AML pathogenesis is demonstrated by reciprocal correlation between SMARCB1 and various members of the DOCK family in AML. Low SMARCB1 corresponds to elevated DOCK family expression. Moreover survival analysis considering SMARCB1 and DOCK levels reflect the importance of this correlation, with AML patients having low SMARCB1 and high DOCK expression displaying poorer disease-free survival compared with patients with high SMARCB1 and low DOCK expression.
To conclude this study, we elucidate that in human primary AML cells SMARCC1, an intact core component of SWI/SNFΔ, colocalized with H3K27Ac to target oncogenic loci. Loss of SMARCB1 induced Rac GTPase GEFs expression, Rac activation and promoted AML cell migration and survival. In summary, these findings inform epigenetic signal integration downstream of SWI/SNF toward oncogenic gene expression program maintenance in AML.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: A. Sengupta
Development of methodology: S.S. Chatterjee, M. Biswas
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.S. Chatterjee, M. Biswas, D. Banerjee
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.S. Chatterjee, M. Biswas, L. Debraj Boila, A. Sengupta
Writing, review, and/or revision of the manuscript: L. Debraj Boila, A. Sengupta
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Sengupta
Study supervision: A. Sengupta
This study is supported by funding from DST (SB/SO/HS-053/2013; to A. Sengupta), DBT (BT/PR13023/MED/31/311/2015; to A. Sengupta), DBT Ramalingaswami Fellowship (BT/RLF/RE-ENTRY/06/2010; to A. Sengupta), and CSIR, Govt. of India (NWP/BIODISCOVERY, BSC 0120; to A. Sengupta). S.S. Chatterjee, and M. Biswas acknowledge fellowships from CSIR and UGC, respectively.
The authors thank Dr. Prasanta Mukhopadhyay for providing umbilical cord blood samples. We acknowledge Dr. Olivier Delattre for sharing plasmids and Addgene for shipping DNA constructs. We also thank Drs. Arindam Maitra and Subrata Patra, CoTERI, National Institute of Biomedical Genomics (NIBMG) for conducting ChIP-seq experiments; Dr. Madavan Vasudevan, Madhura and Shemi Ramesh, Bionivid Technology for ChIP-seq analysis and RNA-sequencing; Genotypic Technology, Bangalore for array CGH experiments and IICB flow cytometry core for services.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.