SETD2 mutations do not contribute to clonal fitness in response to chemotherapy in childhood B cell acute lymphoblastic leukemia

Abstract

Mutations in genes encoding epigenetic regulators are commonly observed at relapse in B cell acute lymphoblastic leukemia (B-ALL). Loss-of-function mutations in SETD2, an H3K36 methyltransferase, have been observed in B-ALL and other cancers. Previous studies on mutated SETD2 in solid tumors and acute myelogenous leukemia support a role in promoting resistance to DNA damaging agents. We did not observe chemoresistance, an impaired DNA damage response, nor increased mutation frequency in response to thiopurines using CRISPR-mediated knockout in wild-type B-ALL cell lines. Likewise, restoration of SETD2 in cell lines with hemizygous mutations did not increase sensitivity. SETD2 mutations affected the chromatin landscape and transcriptional output that was unique to each cell line. Collectively our data does not support a role for SETD2 mutations in driving clonal evolution and relapse in B-ALL, which is consistent with the lack of enrichment of SETD2 mutations at relapse in most studies.

Keywords:

• Relapsed acute lymphoblastic leukemia
• SETD2
• clonal evolution

Introduction

Despite overall 5-year survival rates approaching 90% for pediatric patients with newly-diagnosed B cell acute lymphoblastic leukemia (B-ALL), outcomes for children with relapsed disease remain poor [1]. We and others have shown that the selective pressures of therapy lead to outgrowth of resistant cells carrying genetic and epigenetic alterations associated with resistance to one or more agents used in treatment [2–9]. Recurrent mutations in epigenetic regulators have been found de novo as well as enriched at relapse in ALL, implicating them as culprits in the development of chemotherapy resistance [10]. One such recurrent alteration appears in the gene encoding the histone 3 lysine 36 (H3K36) tri-methyltransferase SETD2 [11]. SETD2 inactivating mutations have been observed in a wide variety of human cancers and the majority are heterozygous in nature. Many studies have shown that SETD2 mutations play a role in disease progression [12–15]. SETD2 mutations have been found in more than 10% of de novo ALL cases, and in some cases are enriched at relapse again implicating them as candidate drivers of clonal evolution [10]. These mutations are most frequently due to frameshifts and are enriched (22%) in patients with KMT2A rearrangements (KMT2Ar) [11]. Decreased levels of H3K36 tri-methylation associated with reduced SETD2 have direct effects on multiple cellular processes including DNA damage repair (DDR), transcriptional regulation, alternative splicing, and others [16].

The mechanisms by which dysregulation of any of these processes contribute to chemoresistance in leukemia remain incompletely understood. In acute myeloid leukemia (AML), diminished levels of SETD2 due to loss-of-function mutations correlate with outcome and changes in the DDR pathway, which leads to resistance to DNA damaging chemotherapy agents [17,18]. Mechanistic studies of the role of SETD2 mutations in B-ALL have not yet been undertaken, therefore, we sought to examine the role of SETD2 - H3K36me3 axis in B-ALL.

Materials and methods

Cells and reagents

The B-lineage leukemia cell lines Reh (ATCC, Manassas, VA), KOPN-8, 697, RCH-ACV (DSMZ, Braunschweig, Germany), and UOCB1 (a gift from Dr. Terzah Horton at Texas Children’s Cancer Center/Baylor College of Medicine), and the AML wildtype (WT) and SETD2 deficient MOLM-13 cell lines (a gift by Dr. Benjamin Ebert, Brigham and Women’s Hospital [17]) were grown in RPMI1640 medium. HEK293T (ATCC) cells were grown in DMEM medium. All media were supplemented with 10% FBS, 1% penicillin/streptomycin under 5% CO2 at 37 °C. No cell lines were used beyond passage 20. Each leukemia line was validated by short tandem repeat analysis through ATCC. Cell lines were routinely monitored for mycoplasma contamination by PCR using ATCC Universal Mycoplasma Detection Kit (20-1012K).

Drug preparation

Stock solutions of doxorubicin (Dox) (Sigma-Aldrich, St. Louis, MO), prednisolone (Pred) (Pharmacia, St. Paul, MN), 6-thioguanine (6-TG), 6-mercaptopurine (6-MP) cytarabine (AraC), etoposide and vincristine were serially diluted in RPMI before use at the indicated concentration.

CRISPR knockout

The CRISPR knockout (KO) lines were generated using the two vector pLenti system: lentiCas9-Blast (Addgene, #52962) and lentiGuide-Puro (#52963) (gifts from the Feng Zhang). Guide RNAs were designed to target exon 3 of SETD2 (guide B: 5′-GTAGATCAGAAAGAGAGCGA; guide M: 5′-AATGAACTGGGATTCCGACG; guide E: 5′-GAAGTCATCCATGACACAGG) and were cloned into the lentiGuide-puro following the protocol by Zhang et al. [19]. B-ALL cell lines were first infected with the plentiCas9-Blast and following selection with Blasticidin, were then infected with the plentiGuide-Puro constructs (including an empty control). Following selection with puromycin, cells were single cell cloned into 96 well plates to select individual KO clones. KO for each clone was verified using TOPO TA cloning followed by Sanger sequencing (Supplementary Figure S1).

Viral preparation

HEK293T cells were transfected with pLenti constructs and packaging plasmids coding for VSV-G, Gag-pol, and Rev using lipofectamine 2000 (Invitrogen, Carlsbad, CA). Viral supernatant was used to spin infect cell lines with 8 µg/ml Polybrene (Millipore, Darmstadt, Germany). After 72 h, infected cell lines were selected with 0.75–3 µg/ml of Puromycin or 3ug/ml Blasticidin. For overexpression studies, the SETD2 ORF (not full length) was PCR amplified from the SETD2-GFP plasmid, which was a gift from Sérgio De Almeida (Addgene plasmid #80653; http://n2t.net/addgene:80653) to include a FLAG-tag at the N-terminus [20]. Following amplification, the SETD2 PCR product was cloned into pInducer20-Blast (Addgene, #109334) using Invitrogen’s Gateway® technology protocol. All plasmids were verified by Sanger sequencing.

Immunoblotting

Immunoblotting was performed using standard techniques as described previously [21]. Primary antibodies included: SETD2 (Sigma-Aldrich, HPA042451), H3K36me3 (Abcam #ab9050), phospo-Chk1 (317) (2344s; Cell Signaling Technology, Danvers, MA), pan-Chk1 (2360s; Cell Signaling Technology), phospho-Histone H2AX (Ser139) (ɣH2AX) (2577S; Cell Signaling Technology), total H2AX (2595S, Cell Signaling Technology), p53 (OP43; Millipore), and Phospho-p53 (Ser15) (9284S, Cell Signaling Technology). Secondary antibodies used were IRDye® anti-rabbit (680RD 926-68071; 800CW 926-32211) and anti-mouse (680RD 926-68070; 800CW 926-32211) (LI-COR, Lincoln NE). Signals were visualized using the Odyssey Imaging System (LI-COR).

Proliferation and cytotoxicity assays

To measure proliferation, cells were plated at an initial concentration of 500,000 cells/ml. On indicated days, cells were collected, resuspended 1:1 with trypan blue and counted using the TC20™ Automated Cell Counter (Bio-Rad, Hercules, CA). Cytotoxicity was assessed by plating cells at a density of 60,000–80,000 cells/well and exposing them to a panel of agents used in ALL therapy for 72 hours. Cell viability was then measured using CellTiter-Glo® Luminescent Cell viability Assay (Promega, Madison, WI). Luminescent readings were normalized to the untreated control for each cell line to account for any differences in baseline growth kinetics or survival. Percentages were plotted and used to calculate the IC50s using nonlinear regression with a four parametric variable slope on GraphPad Prism 8 (GraphPad Prism Software Inc., La Jolla, CA). Each experiment was plated in triplicate and repeated at least three times. For the experiments with the SETD2 overexpressing cell lines, cells were plated in media with 500 ng/ml of doxycycline before and during the experiment.

Cell cycle assays

Cells were treated with vehicle or 6-TG for 120 h. Cells were then fixed with 70% ethanol at each time point. After all time points were collected and fixed, cells were RNAse treated (QIAGEN, Valencia, CA) and stained with Propidium Iodide (PI) (Invitrogen) at a final concentration of 100 µM, and analyzed by flow cytometry using the FACSCalibur (Becton Dickinson, Franklin Lakes, NJ, USA). Live cells were captured on forward/side scatter and doublets excluded by pulse area vs. width. DNA content was analyzed using FlowJo V10 software (v10, Tree Star Inc., Ashland, OR, USA).

Apoptosis assays

For co-culture apoptosis experiments, HS-5 cells were seeded and allowed to adhere overnight. The B-ALL cells, pre-loaded with 5uM CellTrace™ CFSE (ThermoFisher Scientific, C34554) were plated on the HS-5 cells at 4x10⁵ cells/mL in RMPI media and treated with indicated drugs for 72 h. Following drug treatment, cells were lifted from plate using trypsin and then stained with Annexin V-PE and 7AAD (BD Pharmingen, San Diego, CA, USA). Apoptosis was then measured by flow cytometry using the FACSCalibur. The CSFE positive cells were gated and analyzed for the percentage of Annexin V positive and negative cells by FlowJo software.

Whole exome sequencing (WES)

SETD2 mutant and WT KOPN-8 cells were single cell cloned and 3 sub clones from each group were subsequently treated with 6-TG for 9 days. Following treatment, DNA was extracted from the treated cells as well as the pretreated clones using the QIAGEN DNeasy Blood and Tissue Kit. WES libraries were generated by the NYU Genome Technology Center using Nextera Flex for Enrichment Prep. WES was performed using the Illumina HiSeq2500. WES data was processed with triplicates. Adapters and low-quality bases were trimmed with Trimmomatic (v0.36) [22,23]. Reads were aligned to the reference genome (hg19) using BWA-MEM [24]. Duplicate reads were removed using Sambamba (v0.6.8) [25]. Reads were realigned and recalibrated using GATK (v4.2.1.0) [26,27]. Capture efficiency and depth of coverage were also determined using GATK. SNVs and small insertions/deletions between matched tumor-normal samples were called using Strelka [28].

RNA-sequencing (RNAseq)

RNA was extracted using the QIAGEN RNeasy Micro Kit and quality was verified by an Agilent Bioanalyzer 2100 (PICO chip). RNAseq libraries were generated by the NYU Genome Technology Center using the low input Clontech SMART-Seq kit and sequenced on the Illumina NovaSeq6000. Cell line RNAseq data was processed with triplicates. Paired-end reads were aligned to the human reference genome (hg19) using the STAR aligner with default parameters [29]. Counts were obtained using featureCounts [30]. Bigwig tracks were obtained for visualization on individual samples using deeptools (v3.1.0) [31]. Downstream analysis including normalization was performed using DESeq2[32]. Genes were categorized as differentially expressed if abs(L2FC >0.32) and padj <0.05. Significantly differential genes were selected for pathway analysis with Enrichr and KEGG 2021 to determine pathways significantly altered (p<.05) as indicated by combined scores (p-value and z-score) [33,34].

Chromatin accessibility profiling (ATACseq)

ATAC libraries were generated based on the protocol by Buenrostro et al. [35]. Briefly, cells are resuspended in cold lysis buffer (10 mM Tris Cl, 10 mM NaCl, 3 mM MgCl₂, 0.1% (v/v) Igepal CA-630, pH 7.4) and centrifuged for 1 min at 500xg. Nuclei were tagmented using Nextera (Illumina) Tagmentation DNA buffer and enzyme. PCR amplification was performed as described in protocol. Libraries were sequenced on the Illumina NovaSeq6000. ATAC-seq data was processed with two replicates. Paired-end reads were aligned to the human reference genome (hg19) with Bowtie2 (v2.3.4.1) [36]. Reads with a mapping quality <30 were removed. Duplicated reads were removed using Sambamba (v0.6.8) [25]. Remaining reads were analyzed by applying the peak-calling algorithm MACS2 (v2.1.1) [37]. Bigwig tracks were obtained for visualization on individual samples using deeptools (v3.1.0) [31]. Differential ATACseq analysis was performed using DiffBind [38]. Peaks were categorized as significantly differential if abs(L2FC >0.32) and FDR <0.01. Nearest genes were annotated using ChIPseeker [39]. RNAseq data was integrated with ATACseq data to investigate whether gene expression changes associated with changes in chromatin accessibility. Differentially expressed genes (abs(L2FC) >0.32, padj < 0.05) within differentially accessible peaks restricted to promoter and gene body (abs(L2FC) >0.32, false-discovery rate (FDR) <0.01) were identified and box plots were generated to display changes in ATACseq relative to gene expression.

Data availability

All raw and processed data files generated through high-throughput sequencing for this publication (WES, ATAC-seq, and RNA-seq) have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) and are accessible through GEO Series accession number GSE231724.

Results

Knockout of SETD2 in B-ALL does not confer a proliferative advantage or increase clonal potential despite demonstrating reduced H3K36me3

To determine whether SETD2 alterations confer a clonal advantage in pediatric B-ALL, we generated a panel of isogenic SETD2 mutant cell lines using CRISPR/Cas9 to target exon 3. Cell lines with different genetic subtypes were used for the CRISPR knockout generation. Individual clones were grown and sequenced to confirm mutations (Supplementary Table S1). Loss-of-function (LOF) heterozygous frame shift mutations were successfully generated in two cell lines, 697 (E2A::PBX1) and KOPN-8 (KMT2Ar), and confirmed with reduced, but not complete loss of, H3K36me3 expression (Figure 1(a), Supplementary Fig. S1). This is consistent with previous work on cell line models and patient samples where reduction of H3K36me3 is observed [11,14,40]. Although loss of both alleles of SETD2 is atypical we identified a clone with compound heterozygous frameshift mutations of SETD2 among the 697 engineered KO clones (697 SETD2-/-) (Figure 1(a), Supplementary Fig. S1b). Interestingly, only in-frame mutations continued to proliferate in the other two cell lines, Reh (ETV6::RUNX1) and UOCB1 (TCF3::HLF), which both harbor copy number loss of one allele (Supplementary Table S1). We also included the previously developed AML lines with SETD2 KO (MOLM-13) reported by Mar et al. [17] as a control. SETD2 mutant cells showed no differences in proliferation or clonal potential compared to control (Figure 1(b,c)).

Figure 1. Knockout of SETD2 does not lead to increased proliferation or colony formation. (A) Western blot analysis of chromatin-bound nuclear compartment lysate from AML cell line MOLM-13 and B-ALL cell lines 697 and KOPN-8. Cell line models include an empty gRNA control clone (control), and/or Mono-allelic (-/+) and bi-allelic (-/-) KO clones. Numbers below represent H3K36me3/H3 ratios each normalized to control as measured by ImageJ [61]. (B) Proliferation curves of 697 and KOPN-8 control and KO cell lines counted by trypan blue. Each cell line was plated in triplicate with bars representing the mean ± standard deviation. (C) Representative images (top) and quantification (bottom) of MethoCult^TM colony forming assay of 697 and KOPN-8 cell lines. Each cell line was plated in duplicate. Colonies were stained with MTT and counted after 14 days in culture. Statistical significance determined by wilcoxon test (unpaired). All experiments were repeated at least 3 times.

SETD2 loss in B-ALL lines does not induce intrinsic chemoresistance

We hypothesized that B-ALL clones with SETD2 KO would be more resistant to DNA-damaging agents than their isogenic controls, similar to the previously reported findings in AML [17]. To address this, we exposed SETD2 KO cells to various DNA damaging chemotherapeutic agents, including cytarabine, 6-thioguanine, doxorubicin, and etoposide. However, we found no significant differences in drug response among any of the SETD2-KO clones in either KOPN-8 or 697 cell lines compared to their isogenic controls (Figure 2(a,b)). We confirmed that the MOLM-13 AML heterozygous and compound heterozygous clones displayed resistance to cytarabine compared to the control clone, but not to 6-TG, doxorubicin, or etoposide (Figure 2(c)). Finally, it is well known that bone marrow stroma has a protective effect on leukemia cells when exposed to chemotherapy [41], however we saw no significant difference in apoptosis between 697 SETD2 KO and WT clones when co-cultured with the HS-5 stromal line (Supplementary Fig. S2).

Figure 2. Knockout of SETD2 confers chemoresistance in AML, but not in B-ALL. (A-C) Representative cytotoxicity curves assessed by CellTiter-Glo^® of (A) KOPN-8, (B) 697, and (C) MOLM-13 lines exposed to cytarabine, 6-thioguanine, doxorubicin, and etoposide. X-axis is in Molar (M) (1x10^e). Each cell line plated in triplicate and ± bars represent standard deviation. All experiments were repeated at least 3 times.

Although we confirmed KO in one or both alleles by sequencing our CRISPR single-cell clones, the decrease in H3K36me3 was modest in the KOPN-8 cells compared to MOLM-13 possibly accounting for lack of phenotypic changes. Furthermore, although there are no known genetic mutations that are co-observed with LOF SETD2 mutations, it is possible the SETD2 mutations collaborate with other oncogenic pathways as we have demonstrated for NSD2 gain-of-function mutations [21]. To explore these possibilities, we created a rescue model where we over-expressed SETD2 under a doxycycline inducible promoter in B-ALL cell lines that naturally harbor SETD2 LOF mutations and therefore were more likely to contain oncogenic co-drivers or possible cooperative genetic mutations. B-ALL cell lines Reh (ETV6::RUNX1), RCH-ACV (E2A::PBX1), and UOCB1 (TCF3::HLF) were engineered to overexpress SETD2 (SETD2+) upon treatment with doxycycline. As a control, all three cell lines were also infected with the empty vector (EV). SETD2+ lines all displayed an increase in SETD2 expression relative to their respective EV control following treatment with doxycycline (Figure 3(a), Supplementary Fig. S3). Increased H3K36me3 was also observed upon induction of SETD2 (Supplementary Fig. S3b). However, overexpression of SETD2 did not result in changes in chemosensitivity to cytarabine, 6-mercaptopurine, doxorubicin, or prednisolone (Figure 3(b)).

Figure 3. Overexpression of SETD2 does not restore chemosensitivity in B-ALL. (A) Western blot of whole cell lysate of REH, RCH-ACV (RCH), and UOCB1 empty vector (control) or SETD2+ overexpressing cells following growth in doxycycline media. Anti-SETD2 antibody used and either nonspecific band or actin was used as loading control. (B) Representative cytotoxicity curves assessed by CellTiter-Glo^® of REH (top), RCH (middle), and UOCB1 (bottom) control and SETD2+ overexpressing cell lines exposed to cytarabine, 6-mercaptopurine (6-MP), doxorubicin (Dox), and prednisolone (pred). X-axis is in Molar (M) (1x10^e). Each cell line plated in triplicate and ± bars represent standard deviation. All experiments were repeated at least 3 times.

DNA damage response components remain functional in SETD2 mutated B-ALL

Several authors have studied the role of SETD2 in DNA damage recognition and repair, finding that the H3K36me3 mark localizes proteins involved with homologous recombination (LEDGF) [42] and mismatch repair (MSH6) [43], and that SETD2 loss leads to a hypermutator phenotype [44, 45]. In B-ALL SETD2 KO lines we observed no difference in the activation of DNA damage response (DDR) pathway components when they were exposed to 6-TG over time (Figure 4(a,b)). Phosphorylation of Chk1, p53, and H2AX were increased equally in both the heterozygous and compound heterozygous 697 lines, and in KOPN-8 heterozygous lines, compared to controls (Figure 4(a,b)). Additionally, 697 KO lines exposed to 6-TG had similarly increased cycle arrest compared to controls (Figure 4(c)). Our observation of a functional DDR in B-ALL lines despite SETD2 loss is consistent with the absence of differences in chemoresistance to DNA damaging agents.

Figure 4. Neither DNA damage response nor cell cycle progression are affected by loss of SETD2 when exposed to DNA-damaging chemotherapy. (A, B) Western blot analysis of whole cell lysates collected at various timepoints after exposure to 6-thioguanine from (A) 697 and (B) KOPN-8 cell lines. (C) Cell cycle in 697 cells measured following 72–120 h of treatment with 6-TG by PI staining using flow cytometry and percentage of cells in each phase were calculated using FlowJo software watson (pragmatic) modeling. (D) 6-TG mutation rate measured by performing Whole exome sequencing on three single cell clones each from KOPN-8 control and SETD2-/+(A) pre and post 9 days of treatment. Total mutations and G to a or C to T transitions were calculated using pretreated clone as the reference.

Given that SETD2 plays an important role in the DDR, we hypothesized that LOF of SETD2 possibly leads to increased mutation frequency [43] and is therefore permissive for subsequent mutations that interfere with treatment efficacy [46,47]. To examine the presence of a hypermutation phenotype, we single-cell cloned SETD2 KO and WT KOPN-8 cells generating 3 sub clones of each, then subsequently treated them with 6-TG for 9 days. Whole exome sequencing (WES) on the treated cells vs their individual pretreated clones allowed us to determine sites of novel 6-TG induced mutations (Figure 4(d), left). We observed a low mutation rate in both groups and in particular we did not see an increase in G to A or C to T transitions (Figure 4(d), right), contrary to what was observed in the SETD2 mutant MOLM-13 lines [17].

SETD2 KO leads to modest transcriptional reprogramming and chromatin accessibility that varies across cell lines

Given that SETD2 encodes a chromatin modifier we set out to determine changes in the chromatin network upon SETD2 KO. Principal component analysis (PCA) revealed the majority of variation observed was due to inherent differences in the two cell line models as noted on PC1 (Figure 5(a)). PC2 accounted for the variance observed due to LOF of SETD2. Differential ATAC peak analysis between the KO and control lines revealed 3973 and 3618 regions with significant changes (absL2FC >0.32, FDR <0.01) in chromatin accessibility in 697 and KOPN-8 cell lines respectively. Approximately 66,650 average total peaks were called resulting in 6% and 5.4% of ATAC peaks changing upon KO. This is somewhat less than what has been observed in other cancer models [14,48]. Interestingly, we observed a greater number of genomic regions with increased accessibility upon loss of SETD2 in the 697 cell line (3571 open vs. 402 closed), as previously described [14,40], but not in the KOPN-8 cells (1469 open vs. 2149 closed). Increased ATAC peaks were primarily located in promoters and gene bodies whereas decreased peaks were concentrated in intergenic regions (Figure 5(b)), which is also consistent with what others have found [49]. Only a minority of genes with ATAC peak changes were shared between the two cell lines (6% of the total open and 3.4% of total closed) (Figure 5(c)), indicating that changes in the chromatin landscape mediated by SETD2 loss is highly cell context dependent, similar to our findings with NSD2 mutations [21]. The paradoxical increase in ATAC peaks may be explained by studies showing that SETD2 loss is associated with higher levels of H3K36me2 and H3K79me2 [50,51].

Figure 5. SETD2 loss leads to cell line specific chromatin associability and transcriptional changes. (A) Principal component analysis (PCA) of duplicate ATACseq peaks from 697 and KOPN-8 control and SETD2-/+ lines. (B) Association bar plot showing the percentage of stable, increased, or decreased ATAC peaks that lie within promoter, intergenic, or gene body regions. (C) Venn diagram showing overlap of genes assigned to nearest differential ATAC peak between 697 and KOPN-8. (D) PCA of triplicate RNAseq data from 697 and KOPN-8 control and SETD2-/+ lines. (E) Volcano plots demonstrating differentially expressed genes (abs(L2FC) >0.32, p-value <0.05) per cell line from SETD2 KO (-/+) compared to control. (F) Venn diagram showing overlap of differentially expressed genes assessed by RNAseq comparing SETD2 KO (-/+) to control.

RNAseq analysis showed minimal impact on transcriptional output (Figure 5(d,e)), especially in the 697 cell line. PCA showed KOPN-8 KO and EV control lines separated by PC2, which only accounted for 1.4% of variance, and minimal to no separation of 697 KO and EV control was observed. KOPN-8 had 2,644 differentially expressed genes (absL2FC >0.32, p < 0.05) while 697 had only 761. Again, limited overlap in gene expression changes was observed between the two cell lines (2% of the total upregulated, and 3.9% of the total downregulated) (Figure 5(f)). Pathway analysis was performed on the differential genes from each cell line individually, identifying p53 and Rap1 signaling in both (Supplemental Fig. S4).

Integration of RNAseq and ATACseq did reveal a significant correlation between changes in chromatin accessibility and gene expression in both cell lines when restricted to changes within promoters and gene bodies (Figure 6(a)). Furthermore, when we looked specifically at those differential ATAC peaks within promoter or gene body regions we observed a clear association in both lines with greater than 85% of regions with decreased accessibility showing decreased expression (Figure 6(b)), suggesting that upon loss of SETD2, areas of the genome become more inaccessible leading to decreased transcription. There was a similar trend for increased gene expression in regions of increased accessibility, although to a lesser extent, with approximately 55% of increased ATAC peaks in KOPN-8 and 35% in 697 associated with an increase in gene expression. Further cell line specific pathway analysis, limited to genes with concordant changes in gene expression and chromatin accessibility specifically within promoters and gene bodies, identified p53 signaling in 697 and Ras signaling in KOPN-8 (Supplemental Fig. S5).

Figure 6. SETD2 mediated chromatin accessibility changes correlate with gene expression changes. (A) Correlation boxplots of gene expression changes and chromatin accessibility changes (decreased, stable, increased) for each cell line. (B) Association bar plots showing fraction of genes (decreased, increased, or stable) at accessibility changes (decreased, increased, or stable).

Discussion

Relapsed leukemia is a major cause of cancer-related mortality in children. We and others have discovered mutations that are enriched at relapse and play a role in promoting resistance to one or more agents used in treatment [5,8,21]. Many investigators have reported enrichment of mutations in genes encoding epigenetic regulators at relapse implicating shifts in the epigenome as mediators of drug resistance [5,10].

SETD2 mutations are seen in 10% of patients with B-ALL and the overwhelming majority are heterozygous nonsense, frameshift or deletions resulting in loss of function [10]. SETD2 is the only known histone methyltransferase that catalyzes H3K36 trimethylation (H3K36me3) [52]. H3K36me3 is associated with transcriptional activation but is also essential for recruiting MSH6 to DNA as part of the mismatch repair pathway, and therefore SETD2 plays a dominant role in orchestrating response to DNA damaging based chemotherapeutic agents [53]. In fact, we have shown that MSH6 deletions, exclusively seen at relapse, result in resistance to thiopurines due to impaired DDR signaling [2,8].

Furthermore, many studies have suggested that SETD2 mutations may impact therapeutic responsiveness or provide a clonal advantage in various cancer models. In renal cell carcinoma [54], chronic lymphocytic leukemia [55], myelodysplastic syndrome [12] and AML [13] SETD2 mutations have an adverse prognostic impact suggesting a role in therapy evasion. Importantly, studies in preclinical models of SETD2 mutant KMT2Ar AML demonstrate cell cycle defects as well as impaired DDR and apoptosis in response to DNA damaging agents [17,18].

Thus, we had anticipated a similar role for SETD2 mutations in B-ALL. Our studies herein, however, failed to demonstrate a vital role for SETD2 in cancer progression using two preclinical models. While we confirmed CRISPR mediated loss by sequencing and used single cell clones one could argue we didn’t observe the same level of depletion compared to the MOLM-13 AML cell line (Figure 1). However, rescue experiments using cell lines with endogenous SETD2 mutations confirmed our observations with the CRISPR KO clones. Interestingly, our difficulty generating double knockouts in B-ALL cell lines naturally harboring SETD2 heterozygous mutations (Supplementary Table S1) is consistent with previous findings. Zhang et al. demonstrated complete loss of SETD2 results in embryonic lethality while conditional loss impairs hematopoietic stem cell self-renewal/competitiveness and skews differentiation [56]. Likewise, biallelic loss of SETD2 delayed leukemia onset in an MLL::AF9 AML mouse model while heterozygous loss accelerated tumor growth [17].

Under the continuous selective pressures of therapy, defects in DNA damage response can result in genomic instability and the accumulation of secondary mutations, which could also lead to increased clonal fitness. Mar et al. provide evidence for this hypothesis in their AML model by demonstrating increased GA/CT transitions following 6-TG in SETD2 KO clones. They proposed that SETD2 mutations could be used for risk stratification in acute leukemia and that targeting H3K36 demethylases could be used to increase H3K36me3 levels and restore chemosensitivity [17]. However, in our model, we did not see changes in DDR signaling or an increased mutation burden following treatment with thiopurines. Collectively these results indicate that despite lower levels of SETD2, B-ALL cells are still able to recruit components of the DDR to sites impacted by chemotherapeutic agents. Interestingly, a recent report has demonstrated a role for SETD5 in depositing H3K36me3 along gene bodies and impacting transcriptional regulation [57]. Perhaps this redundancy could explain a lack of functional differences upon LOF of SETD2 as well as the incomplete reduction in H3K36me3.

While the literature supports a role for SETD2 in cancer, a common finding is cooperation with or dependency on other oncogenic driver genes in specific cellular contexts [14,48]. SETD2 alterations are particularly enriched in KMT2Ar AML (22% compared to 6% in non KMT2Ar cases) [11] supporting cooperativity. However, we observed no phenotypic changes in our KOPN-8 KMT2Ar cell line model. SETD2 loss in an AML KMT2Ar model was shown to lead to the expected decrease in H3K36me3 but further elevation of K3K79me2. Notably the increased H3K79me2 did not result in upregulation of known KMT2Ar targets but a novel set of genes that enhanced proliferative and self-renewal capabilities [40]. Mar et al. also showed SETD2 mutations in ALL were often found with Ras mutations [10]. Our KO models, which included Ras mutated lines (697 p.G12D, KOPN-8 p.G12D), in conjunction with our SETD2 restoration models, suggests that in B-ALL either the signaling pathways are activated by other means or the cells are no longer reliant on loss of SETD2.

Although we did not observe any phenotypic changes upon loss of SETD2 in our B-ALL cell lines, our data does support the overall finding that SETD2 loss creates a more open, permissive chromatin state allowing for transcriptional changes. SETD2 loss in our model did impact the chromatin landscape but less so than previously reported in other cancer models. Xie reported 8% of ATAC peaks changed (6.3K differential sites) with SETD2 loss in a renal cell carcinoma model [14] and over 14K sites in lung cancer [48], while we report 6% and 5.4% in KOPN-8 and 697 respectively. Ninety percent of peak changes resulted in open chromatin in 697 as previously noted, but this was not the case for the KOPN-8 cell line where we saw an equal distribution of open and closed chromatin changes. Importantly, we did find a significant correlation between ATACseq changes and transcriptional output, suggesting that the cell specific chromatin landscape changes mediated by SETD2 loss are impacting gene expression. While minimal overlap was observed between the two cell lines for either chromatin or expression changes, we were able to identify common downstream signaling cascades, including both Rap1 and p53 signaling. SETD2 is known to interact with TP53 and impact p53-mediated transcriptional changes. Specifically, reduced SETD2 leads to downregulation of apoptosis related genes regulated by p53[58]. We have shown that Rap1 signaling is upregulated by mutations in NSD2 [21]. Interestingly, Ras signaling pathway was identified in our KOPN-8 model when pathway analysis was restricted to genes that showed both chromatin and transcriptional concordant changes similar to prior observations [48]. It is well documented that MAPK/Ras pathway is activated and plays a role in drug resistance in relapsed B-ALL [41], but our data would suggest that loss of SETD2 does not further impact this phenotype.

One of the strongest pieces of evidence for the role of mutations in driving drug resistance is enrichment of mutations at relapse compared to diagnosis. Previous studies have suggested an increase in mutations in genes encoding epigenetic regulators at relapse [10]. For example, Mullighan et al. showed increased CREBBP mutations [5] and we and others showed enrichment in NSD2 mutations at relapse [21]. SETD2 mutations were not analyzed in the study by Mullighan, but Xiao et al. studied adults with ALL who relapsed after allogeneic transplantation and found among six patients with SETD2 mutations, four had shared diagnosis/relapse mutations while the other two patients had a diagnosis or relapse specific mutation [59]. Mar studied 30 matched diagnosis/relapse pairs and two patients had a shared SETD2 mutation at diagnosis and relapse while in a third case a mutation was noted exclusively at relapse [10]. Li et al. studied 103 ALL diagnosis/relapse pairs and somatic alterations in NR3C1, CREBBP, NSD2, PRPS1, PRPS2, MSH2, MSH6, PMS2, NT5C2, FPGS and TP53 were enriched at relapse [60]. Defects in SETD2 were not enriched having been detected in five patients: one case with a shared diagnosis/relapse mutation, two diagnosis-specific and two relapse-specific. The lack of clear enrichment of SETD2 loss in relapsed B-ALL further supports our findings here.

In summary our data further supports that role of SETD2 loss in creating a more permissive, open chromatin landscape that works in concert with key oncogenic drivers or transcription factors in a cell context specific way. However, our results do not support a role in driving clonal evolution and relapse and therefore SETD2 is not a high priority target for novel therapeutic interventions to prevent or treat relapse in B-ALL.

Acknowledgements

The authors greatly appreciate the cell lines provided by Dr. Terzah Horton at Texas Children’s Cancer Center/Baylor College of Medicine and Dr. Benjamin Ebert at Brigham and Women’s Hospital, as well as the CRISPR plasmids from Dr. Feng Zhang and SETD2 plasmid from Dr. Sérgio De Almeida. The authors gratefully acknowledge the funding received to complete this work.

Disclosure statement

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript.

Additional information

Funding

WLC has received grants from the National Cancer Institute of Health [R01 CA140729-05], Hyundai Hope On Wheels (Hyundai Scholar Hope Grant), The Leukemia and Lymphoma Society Specialized Center for Research [7010-14], Perlmutter Cancer Center Arline and Norman M. Feinberg Pilot Grant for Lymphoid Malignancies, and the Perlmutter Cancer Center [P30 CA016087]. AT has received grants from the NCI/NIH (P01CA229086), NCI/NIH [R01CA252239], NCI/NIH [R01CA260028] and NIH/NCI [R01CA140729]. GCY and GR received funding from the Pediatric Cancer Foundation (Fellowship Training Grant). GCY received funding from Sohn Conference Foundation (Fellowship Training Grant) and Hyundai Hope on Wheels (Impact Award). We gratefully acknowledge the support of the NYU School of Medicine Cytometry and Cell Sorting Laboratory and the Genome Technology Center, which are supported by the NYU Langone Health Perlmutter Cancer Center Support Grant [P30 CA016087].