Bap1 ^{tm1c(EUCOMM)Hmgu} mouse strain carries a floxed (conditional) allele of Bap1-gene and will be referred to here as Bap1 ^fl. The strain is derived from ES-cell clone HEPD0526_2_G01; allele structure is provided at https://www.mousephenotype.org/data/alleles/MGI:1206586/tm1c(EUCOMM)Hmgu, and the mice are available from EMMA at: https://www.infrafrontier.eu/search?keyword=Bap1. Specifically, loxP sites flank exons 6-12 (ENSMUSE00000121807- ENSMUSE00000121801) of the Bap1 transcript ENSMUST00000022458.10. We thank the Wellcome Trust Sanger Institute Mouse Genetics Project, its funders, and INFRAFRONTIER/EMMA (www.infrafrontier.eu) for providing this strain. Funding information is found at www.sanger.ac.uk/mouseportal and associated primary phenotypic information at www.mousephenotype.org (48–51). The strain was bred to a transgenic line expressing Cre recombinase under the control of a B cell lineage specific promoter mb1-Cre (from Prof. Michael Reth, MPI für Immunbiologie und Epigenetik, Germany) (52). The Bap1 ^Δ allele is predicted to disrupt Bap1 protein coding sequence from amino acid 126 onward, precluding the expression of full N-terminal UCH catalytic domain and all downstream domains of BAP1 protein. All lines were on the C57BL/6 genetic background. The mice were maintained under specific pathogen-free conditions. Test and control groups were bred as littermates and co-housed together. Experiments were in accordance with the guidelines of the Canadian Council on Animal Care and protocol AUP-2011-6029 approved by the McGill Animal Care Committee. Mice used in experiments were both male and female, 8-16 weeks of age, and always age-matched and sex-matched between the experimental groups. Genotyping was performed with DreamTaq DNA Polymerase (Thermo Fisher Scientific) and primers from IDT Technologies.

For competitive bone marrow transplantations, recipient B6.SJL-PtprcaPepcb/Boy mice (JAX002014, congenic for CD45.1) were irradiated with 2 doses of 4.5Gy, delivered 3 hours apart, in a RS2000 irradiator (Rad Source). Wild-type CD45.1-marked bone marrow cells were mixed in a 1:1 ratio with bone marrow cells from mice of Bap1^fl/+, Bap1^fl/+Cre, or Bap1^fl/flCre genotypes, and the mixes transplanted into three independent cohorts of recipient mice. The mice were kept on neomycin in drinking water (2g/l, BioShop) for 3 week, and analyzed at 10 and 17 weeks post-reconstitution.

Cell suspensions of mouse tissues were prepared in RPMI-1640 (Wisent) with 2% (v/v) FCS, 100μg/ml streptomycin and 100U/ml penicillin (Wisent). The cells were stained for surface-markers in PBS with 2% FCS for 20 minutes on ice, with the following antibodies: APC-conjugated anti-CD43 (S7, BD Biosciences) or anti-IgD (11-26c.2a, BioLegend); APC-eFluor 780-conjugated anti-CD45R/B220 (RA3-6B2, eBioscience) or anti-CD45.1 (A20, eBioscience); Brilliant Violet 650-conjugated anti-CD45R/B220 (RA3-6B2, BioLegend); BUV395-conjugated anti-CD43 (S7, BD Biosciences); eFluor 450-conjugated anti-CD24 (M1/69, eBioscience) or anti-CD45R/B220 (RA3-6B2, eBioscience); FITC-conjugated anti-CD23 (B3B4, eBioscience), anti-CD127/IL7Rα (A7R34, eBioscience), or anti-CD249/BP-1 (6C3, eBioscience); Pacific Blue-conjugated anti-IgD (11-26c.2a, BioLegend); PE-conjugated anti-IgM (II/41, eBioscience); PE-Cy7-conjugated anti-CD19 (6D5, BioLegend), anti-CD21/CD35 (eBio8D9, eBioscience), anti-CD93 (AA4.1, BioLegend), or anti-CD45.2 (104, BioLegend); PerCP-Cy5.5-conjugated anti-CD19 (1D3, Tonbo Biosciences), anti-CD93 (AA4.1, BioLegend), or anti-CD45R/B220 (RA3-6B2, BioLegend); biotin conjugated anti-Pre-B Cell Receptor (SL156, BD Biosciences).

Intracellular staining was performed as previously described (53, 54). The cells were fixed in 2% paraformaldehyde (PFA) in PBS with 2% FCS at 37°C for 10 minutes, and permeabilized in 90% methanol for 30 minutes on ice. The cells were then stained with Alexa Fluor 647-conjugated anti- p53 (1C12, Cell Signaling Technology) and Alexa Fluor 488-conjugated anti-H2A.X Phospho (Ser139) (2F3, BioLegend) antibodies, or FITC-conjugated anti-Ki-67 (B56, BD Biosciences) and Alexa Fluor 647-conjugated Histone H3 Phospho (Ser10) (11D8, BioLegend) antibodies, or the appropriate isotype controls. Viability Dye eFluor^® 506 (eBioscience) was used to discriminate dead cells. Annexin V PeCy7 (eBioscience) was used for detection of apoptotic cells. Compensation was performed with BD™ CompBeads (BD Biosciences). The data were acquired on FACS Canto II or BD Fortessa and analyzed with FACS Diva (BD Biosciences) or FlowJo (Tree Star) software.

Total B cells were isolated from mouse spleen via magnetic enrichment using EasySep Mouse CD19 Positive Selection Kit II (Stem Cell Technologies), according to the manufacturer’s protocols. B cell subsets were sorted from mouse bone marrow, as previously described (53). Bone marrow was flushed in PBS supplemented with 0.1% BSA and 2mM EDTA, filtered through 40μm cell-strainers, and subjected to red blood cell lysis in ACK buffer (0.15M NH₄Cl, 10mM KHCO₃, 0.1mM EDTA). The samples were stained with: biotin-conjugated antibodies against lineage markers [anti-CD3ϵ (145-2C11, BioLegend), anti-CD11b (M1/70, BioLegend), anti-CD11c (N418, BioLegend), anti-Ly-6G/Ly-6C (Gr-1) (RB6-8C5, BioLegend), anti-NK-1.1 (PK136, BioLegend), and anti-TER-119 (TER-119, BioLegend)], followed by APC-eFluor 780-conjugated Streptavidin (eBioscience), Alexa Fluor 488-conjugated anti-IgD (11-26c.2a, BioLegend), APC-conjugated anti-CD43 (S7, BD Biosciences), PE-conjugated anti-IgM (II/41, eBioscience), PE-Cy7-conjugated anti-CD19 (6D5, BioLegend), and PerCP-Cy5.5-conjugated anti-CD45R/B220 (RA3-6B2, BioLegend) antibodies. DAPI was added immediately before sorting for dead cell exclusion. Cell sorting was performed on FACSAria and analyzed with FACS Diva software (BD Biosciences).

Pre-B Colony Forming Unit (CFU) assays with primary mouse bone marrow were performed according to manufacturer’s protocols using MethoCult M3630 media (StemCell Technologies), with 1x10⁵ mouse bone marrow cells plated per dish.

Western blotting was performed as previously described (53). Magnetically isolated murine B cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Thermo Scientific). Protein concentration was measured using the BCA assay (Thermo Scientific). Protein lysate samples were boiled in Laemmli buffer and 1.25% β-mercaptoethanol (Sigma-Aldrich) before loading onto gels, alongside Precision Plus Protein Kaleidoscope standards (Bio-Rad). Upon gel-to-membrane transfer, nitrocellulose membranes (GE Healthcare) were blocked with 5% milk in TBS-T and probed with rabbit monoclonal primary antibodies against BAP1 (D7W7O, Cell Signaling Technology) and β-Actin (D6A8, Cell Signaling Technology) at 4°C overnight, followed by horseradish peroxidase (HRP)-conjugated goat anti-rabbit polyclonal IgG secondary antibody (Abcam) at room temperature for 1 hour, with TBS-T washes after each antibody-incubation. Protein bands were detected using Western Lightning Plus-ECL (PerkinElmer) and HyBlot CL autoradiography films (Harvard Apparatus Canada), according to manufacturer’s protocol.

RNA isolation from magnetically isolated B cells was carried out with the EZ-10 DNAaway RNA mini-prep kit (BioBasic) according to the manufacturer’s protocol. The RNA was quantified with Nanodrop spectrophotometry (ThermoFisher Scientific), and reverse transcribed with the M-MLV reverse-transcription kit (Life Technologies, ThermoFisher Scientific). All qPCRs were performed on a StepOnePlus instrument with Power SYBR Mastermix (Applied Biosystems, Life Technologies). The primer sequences for validating Bap1-depletion in murine B cells were: Bap1 -Fw GGA TTG AAA GTC TAC CCA ATT GAT, Bap1 -Rv CGA GCT TTA TCT GTC CAC TCC T, Actb -Fw CTA AGG CCA ACC GTG AAA AG, Actb -Rv ACC AGA GGC ATA CAG GGA CA; all primers were purchased from IDT Technologies.

Murine B cell lineage precursor cell line Ba/F3 was maintained at 0.5-2 x10⁶ cells/mL in RPMI-1640 (Wisent) with 10% Fetal Calf Serum (FCS, Wisent), 2mM L-Glutamine, 100μg/mL streptomycin, 100U/mL penicillin (Wisent), and 5% WEHI conditioned media as the source of IL-3. Ba/F3 cell lines stably expressing triple-FLAG-tagged murine BAP1 were derived as previously described (53), and maintained under 2μg/mL puromycin selection (Wisent).

Plasmid pSpCas9(BB)-2A-GFP (PX458) was obtained from Addgene (#48138). gRNAs targeting Bap1 were designed using the Zhang lab software (http://crispr.mit.edu): gRNA_Bap1_Exon4_ Chr14:32,066,155: gcaaatggatcgaagagcgc and gRNA_Bap1_Exon5_Chr14:32,066,654: ggcgtgagtggcacaagagt. The gRNAs were ligated into the BbsI (NEB) digested PX458 plasmid, and the resulting plasmids were transformed into DH5α competent cells, purified using PureLink HiPure Plasmid Filter Midiprep Kit (Invitrogen), and verified by Sanger sequencing. To introduce the constructed plasmid into Ba/F3 cells, nucleofection was performed using Amaxa Cell Line Nucleofector Kit V (Lonza) according to the manufacturer’s protocols. Single GFP⁺ cells were sorted 48 hours after the nucleofection and then expanded over two weeks to generate single cell clones. The loss of BAP1 protein expression in the cell line clones was validated via Western blotting, with the following antibodies: anti-BAP1 (D7W7O, Cell Signaling Technology), anti-β-Actin (D6A8, Cell Signaling Technology), HRP-conjugated anti-mouse-Ig (eB144, Rockland), and HRP-conjugated anti-rabbit-Ig (eB182, Rockland).

Wildtype and Bap1 ^Δ/Δ Ba/F3 cells were seeded into a 96-well plate at a density of 1 × 10⁵/mL, and incubated at 37°C overnight. Subsequently, AlamarBlue (Invitrogen, Thermo Fisher Scientific) was added at 1/10 the total volume to each well and incubated at 37°C for 4 h. Fluorescence intensity was recorded using an EnSpire Plate reader (Perkin Elmer), at the excitation wavelength of 560nm and the emission wavelength of 590nm.

Wildtype and Bap1 ^Δ/Δ Ba/F3 cells were washed with PBS, and 5 × 10⁶ cells were resuspended in 1 ml PBS containing 1 μl CellTrace™ CFSE/DMSO solution (Invitrogen, Thermo Fisher Scientific). After 10 min of incubation at 37°C, CFSE was deactivated by adding 12 mL pre-warmed PBS with 5% FBS and incubation at 37°C for 5 min, followed by centrifugation and a second wash with PBS. Subsequently, cells were resuspended in complete culture media at the concentration of 2 × 10⁵/mL, seeded in a 96-well plate at 4 x10⁴ cells per well, and incubated at 37°C overnight. The cells were counterstained with Fixable Viability Dye eFluor™ 780 (eBioscience, Thermo Fisher Scientific). Fluorescence intensity data was acquired on FACS Canto II and analyzed with FACS Diva (BD Biosciences) or FlowJo (Tree Star) software.

The protocols were as previously described (53). Briefly, RNA was isolated using the MagMAX total RNA kit (Ambion, Life Technologies), and quality assessed using Bioanalyzer RNA Pico chips (Agilent). rRNA depletion and library preparation were performed using the SMARTer Stranded RNA-Seq kit (Takara Clontech). The libraries were sequenced on an Illumina HiSeq 4000 sequencer in paired-end 50bp configuration. The high quality of sequence reads was confirmed using the FastQC tool (Babraham Bioinformatics), and low-quality bases were trimmed from read extremities using Trimmomatic v.0.33 (55). Reads were then mapped to the mouse UCSC mm9 reference assembly using TopHat v2.0.9 in conjunction with Bowtie 1.0.0 algorithm (56–58). Gene expression was quantified by counting the number of uniquely mapped reads with featureCounts using default parameters (59). We retained genes that had an expression level of minimum 5 counts per million reads (CPM) in at least 3 of the samples, and performed quantile normalization with the preprocessCore package to remove batch effects (60). TMM normalization and differential gene expression analyses were conducted using the edgeR Bioconductor package (61). Dimension reduction analysis was performed using the Partial Least Square regression method (62). Pairwise comparisons were performed between samples across different mouse genotypes. Genes with changes in expression ≥ |1.5| fold and Benjamini-Hochberg adjusted p values ≤ 0.01 across the three genotypes in either pre-B or immature B cells were considered significant. For data visualization in Integrative Genomics Viewer (IGV) (63), replicates with the same genotype were combined, and bigwig files were generated using a succession of genomeCoverageBed and wigToBigWig tools and scaled per 10 million exon-mapped reads. Gene ontology (GO) and disease ontology enrichment analyses on differentially expressed gene clusters were performed with DAVID Bioinformatics Resources 6.8 (64), and Gene Set Enrichment Analysis (GSEA) was performed in command-line using MSigDB v6.0 with default configuration and permutation within gene sets (65).

ChIP was performed as described previously (66), with minor modifications. Briefly, cells were fixed with 1% formaldehyde in the culture media for 10 minutes at room temperature, followed by addition of 0.125M of glycine to stop fixation. Nuclei were extracted with 5 minutes lysis in 0.25% Triton buffer, followed by 30 minutes in 200mM NaCl buffer. Nuclei were resuspended in sonication buffer and sonicated for twelve cycles of 30 seconds with a digital sonifier (Branson Ultrasonics) at 80%, with 30 seconds rest in cooled circulating water.

Beads immunocomplexes were prepared overnight by conjugating 40μL of Dynabeads Protein G (Invitrogen, Life Technologies) with antibodies: anti-BAP1 (Cell Signaling Technology, D1W9B, 52.8 μg), anti-FLAG M2 (Sigma, F1804, 5 μg) or anti-H2AK119Ub (Cell Signaling Technology, D27C4, 5 μg). Immunoprecipitation was performed by overnight incubation of antibody-bead matrices with sheared chromatin from the equivalent of 5x10⁶ cells. Four washes were performed with medium-stringency buffers for BAP1-FLAG ChIP, while six washes were performed with low-stringency buffers for histone ChIP and ChIP using the anti-BAP1 antibody (Cell Signaling Technology). Samples were de-crosslinked by overnight incubation at 65°C in 1% SDS buffer, and following RNaseA and Proteinase K enzymatic treatments, ChIP DNA was purified using Qiaquick PCR Cleanup kit (Qiagen).

ChIP-seq libraries were prepared using the Illumina TruSeq kit and sequenced on an Illumina HiSeq 4000 sequencer in paired-end 50bp configuration, with input DNA from the same cells sequenced as negative control. The reads were mapped to the UCSC mouse mm9 reference genome with Bowtie 1.0.0 (57), and BAP1 binding sites identified using peak detection algorithm MACS1.4 (67), with comparisons for read enrichment against control input DNA from the same cells. Normalized sequence read density profiles (bigwig) were generated with Homer tool (68) and visualized with IGV (63). Gene ontology (GO) enrichment analyses on genes associated with BAP1 ChIP-Seq binding clusters were performed on GREAT 4.0.4 (69) with Basal plus extension option, searching for genes within 2kb upstream, 2kb downstream, and 200kb in distal.

Full gene annotations with transcription start site (TSS) were obtained from the UCSC mouse mm9 reference genome. An in-house Python script was developed to load the genomic locations of ChIP-Seq binding sites and the TSS locations of RNA-Seq dysregulated genes, and search for gene TSS located within a specific distance to each ChIP-Seq binding site. RNA-Seq and ChIP-Seq datasets acquired and described in our study have been deposited into the NCBI public database and are available under GEO Submission GSE162085.

Statistical analyses used Prism 7.01 (GraphPad Inc.), with Student’s t-test for two datasets and ANOVA with Sidak’s or Tukey’s post hoc tests for multiple comparisons.

To study the role of BAP1 in the B cell lineage, a B cell specific conditional knockout mouse model was generated by crossing the Bap1-floxed mouse line Bap1 ^fl/fl to the mb1-Cre line expressing Cre-recombinase under the control of the B cell lineage specific Cd79a gene promoter, from the pre-pro-B cell stage in development (52). As expected, Bap1 ^fl/fl Cre offspring were born in normal Mendelian ratios, showed no gross morphological abnormalities, and bred normally. Loss of BAP1 transcript and protein expression was validated in B cells isolated from the spleen of Bap1 ^fl/fl Cre and control mice, using qRT-PCR and Western blotting, respectively ( Figures 1A, B ).

To assess the effects of BAP1-loss on the B cell lineage, spleens and mesenteric lymph nodes of Bap1 ^fl/fl Cre, Bap1 ^fl/+ Cre, and control Bap1 ^fl/+ mice were analyzed by flow cytometry, gating on CD19⁺ B cells. A significant reduction in the frequency of B cells was observed in the Bap1 ^fl/fl Cre relative to control mice in both tissues ( Figures 1C–E ). The absolute numbers of B cells were also significantly reduced in the spleen of Bap1 ^fl/fl Cre mice and showed a trend toward reduction in the lymph nodes ( Figures 1C–E ).

We further confirmed the depletion of CD19⁺B220^hi B2 lymphocytes in the spleen of Bap1 ^fl/fl Cre mice, while CD19⁺B220^lo cells were strongly expanded ( Supplemental Figures S1A–C ). Further analysis of the CD19⁺B220^lo B cell population identified the expected B1b (CD43⁺CD5^-) and B1a (CD43⁺CD5⁺) subsets in control mice, while in Bap1 ^fl/fl Cre mice the majority of the cells lacked B1 markers expression (CD43^-CD5^-, Supplemental Figures S1D, E ) (70, 71). Overall, we observed a mild depletion of B1b cells (CD19⁺B220^loCD43⁺CD5^-) and expansion of B1a cells (CD19⁺B220^loCD43⁺CD5⁺) in Bap1 ^fl/fl Cre mice ( Supplemental Figures S1D, E ), while the identity of the CD19⁺B220^loCD43^- B cells that are strongly expanded in Bap1 ^fl/fl Cre mice remains to be further investigated.

Splenic B cells were further analyzed to quantify transitional, follicular (FOL), and marginal zone (MZ) cell populations ( Supplemental Figure S2 ) (72). This demonstrated a significant reduction in the numbers of transitional T2-3, follicular FOLI-II, and marginal zone progenitor (MZP) B cells in Bap1 ^fl/fl Cre relative to control mice, while the numbers of transitional T1 and MZ B cells were maintained ( Supplemental Figure S2 ).

To further validate the cell intrinsic role of BAP1 in B cells, a chimeric mouse model was established. Wild-type CD45.1⁺ marked bone marrow cells were mixed in a 1:1 ratio with bone marrow cells from CD45.2⁺ mice of Bap1^fl/+, Bap1^fl/+Cre, or Bap1^fl/flCre genotypes, and the three mixes transplanted into three independent cohorts of lethally irradiated recipient mice to reconstitute their immune system ( Supplemental Figure S3A ). The recipient mice were analyzed for the relative contribution of CD45.1⁺ and CD45.2⁺ donors to the B cell lineages, by testing the blood at 10 weeks, and the bone marrow and spleen at 17 weeks post-reconstitution. Reduced contribution of the BAP1-deficient Bap1^fl/flCre donor cells to the B cell lineage was demonstrated in all the tissues ( Supplemental Figures S3B–E , Supplemental Figure S4A ), further confirming the cell intrinsic role of BAP1 in B cells.

To analyze the impact of BAP1 loss on B cell development and to pinpoint the specific developmental transitions dependent on BAP1, the bone marrow of Bap1^fl/flCre and control mice was analyzed by flow cytometry. This demonstrated a significant reduction in the frequency of B cells in the Bap1^fl/flCre relative to control mice (gated as B220⁺, Figures 2A–D ). The defect in B cell development in Bap1^fl/flCre mice was further confirmed with a colony forming units (CFU) assay, demonstrating a reduction in the numbers of pre-B CFUs in Bap1^fl/flCre bone marrow samples ( Figure 2E ). Further characterization of B cell progenitor subpopulations demonstrated a significant depletion in the frequencies and absolute numbers of large pre-B cells and mature B cells, and an expansion of Fraction B pro-B cells in Bap1^fl/flCre relative to control bone marrow ( Figures 2F, G ). No changes in the abundance of the common lymphoid progenitors, multipotent progenitors, and hematopoietic stem cells were observed (data not shown), further confirming the specificity of the Cre transgene and the phenotype to the B cell lineage. Overall, our findings indicate the cell intrinsic role of BAP1 as a regulator of B cell development, and demonstrate that BAP1 loss impacts pre-B cells, as well as transitional and mature B cell subsets.

To assess how the loss of BAP1 affects B cell development and physiology, we analyzed the effects of BAP1 loss on cell viability and cell cycle progression throughout the B cell lineage, with flow cytometry. This demonstrated a specific reduction in the viability of large pre-B cells, caused by an increase in the proportion of late apoptotic/necrotic cells within this B cell subset ( Figures 3A–B ).

To further analyze the effects of BAP1 loss on cell cycle progression, bone marrow samples from Bap1^fl/flCre and control mice were stained for Ki67 and histone H3S10p, to identify cells in G0 (Ki67^-H3S10p^-), G1-S-G2 (Ki67⁺H3S10p^-), and M (Ki67⁺H3S10p⁺) phases of the cell cycle within each B cell subpopulation. This revealed a significant increase in the fraction of G0 cells and a corresponding decrease in G1-S-G2 cells within the large pre-B cell, small pre-B cell, immature, and mature B cell subpopulations in Bap1^fl/flCre mice ( Figures 3C, D , Supplemental Figures S4B, C ), indicating impaired cell proliferation and cell cycle progression. In contrast no differences were observed in the cell cycle state of pre-pro-B and pro-B cells ( Figure 3C ). Overall, this indicated that impaired cell survival and proliferation contribute to the depletion of pre-B cells and possibly to the broader defect in B cell lineage development in Bap1^fl/flCre mice.

We further observed a reduction in the expression of IL7Rα on the surface of large pre-B cells, but not other B cell precursor subsets in Bap1^fl/flCre relative to control mice ( Supplemental Figures S5A, B ), while the expression of pre-B cell receptor (pre-BCR) on large and small pre-B cells was unchanged ( Supplemental Figures S5C, D ). Subsequent RNA-seq analyses of Bap1 ^fl/fl Cre and control pre-B cells ( Figure 4 ), showed no difference in the expression of Il7r gene or surrogate light chain gene Igll1 in Bap1 ^fl/fl Cre pre-B cells ( Table S1A, B , Vpreb1-2 did not reach >5CPM expression threshold for quantification). This indicates that the differences in IL7Rα protein expression on pre-B cell surface are likely mediated via indirect mechanisms, and are not a result of direct regulation of its encoding gene by BAP1.

Given the previously reported roles of BAP1 not only in transcriptional regulation but also in DNA repair (41), we further analyzed the B cell precursor cell populations in the bone marrow of Bap1^fl/flCre and control mice for the levels of γH2AX as a marker of DNA damage (73) and p53 protein as a marker of DNA damage response activation (74). There were no differences in the levels of γH2AX or p53 between Bap1^fl/flCre and control cells for any of the B cell precursor cell populations ( Supplemental Figure S6 ). This suggested that the role of BAP1 as a regulator of B lymphocyte development is likely independent of its functions in DNA repair and is primarily linked to its role as a transcriptional and epigenetic regulator of gene expression.

To further understand the functional impact of BAP1 loss on B cell development, RNA-Seq gene expression analysis was conducted on live pre-B and immature B cells, isolated from the bone marrow of Bap1 ^fl/fl Cre, Bap1 ^fl/+ Cre, and control Bap1 ^fl/+ mice through FACS-sorting ( Figure 4A , Table S1A, B ). Loss of Bap1 ^fl exon expression was confirmed in Bap1 ^fl/fl Cre datasets ( Supplemental Figure S7A ). Dimension reduction analysis of gene expression profiles showed clear segregation of the two cell types (PC1 - 46.5% variance) and the three genotypes (PC2 - 9.3% variance, PC3 - 5.6% variance), and indicated major transcriptional changes in the BAP1-deficient cells ( Figure 4B ). To explore the biological functions dysregulated within the transcriptome of BAP1-deficient B cells, a gene set enrichment analysis (GSEA) (65) was performed and highlighted the dysregulation of the transcriptional signatures linked to “cell proliferation”, “DNA replication”, and “cell division”, as well as many other biological processes in Bap1 ^fl/fl Cre cells ( Figure 4C , Table S1C ).

Differential gene expression analysis across the three Bap1 genotypes was performed, at fold change ≥1.5 and false discovery rate (FDR) ≤0.01. This identified 916 genes significantly dysregulated in Bap1 ^fl/fl Cre relative to control Bap1 ^fl/+ cells ( Figure 4D , Table S1A, B ). In contrast, in Bap1 ^fl/+ Cre cells the downregulation of Cd79a was the only transcriptional change to reach the significance threshold, and is likely due to the direct effect of Cre knockin on the Cd79a locus (52). This demonstrates the very limited effects of Cre or heterozygous Bap1 loss on the transcriptome, and confirms that the major transcriptional changes in Bap1 ^fl/fl Cre are due to the loss of BAP1 function.

The dysregulated genes were segregated into Clusters I-VI, based on their upregulation or downregulation status in Bap1 ^fl/fl Cre versus control samples across the two cell types ( Figure 4D ). Gene ontology (GO) enrichment analysis was performed for the genes in each cluster, to explore the biological functions dysregulated in the BAP1-deficient B cell precursors. The upregulated transcriptional signatures, represented by Clusters I-III, showed a mild enrichment for GO-terms “immune system process”, “cell migration” and “apoptotic process” ( Figure 4E , Supplemental Table S1D ). As an example the upregulated genes included several receptors for chemokines and other messengers (Cxcr7, Cxcr5, Cr2, IL2rg, Il9r, S1pr3, CD40, Ahr), some pattern recognition receptors (Tlr1, Tlr9, Naip5, Clec12a), transcriptional regulators involved in immune cell development and activation (Irf5, Irf7, Irf8, Nfkb2), both pro-survival and pro-apoptotic Bcl-family proteins (Bcl2, Bbc3), and some oncogenes (Myc). In contrast, the downregulated transcriptional signature of BAP1-deficient pre-B cells (Cluster V) showed a highly significant enrichment of GO-terms “cell cycle progression” and “DNA replication” ( Figure 4E , Supplemental Table S1D ), indicating the downregulation of genes involved in cell proliferation and cell cycle progression. This is consistent with the depletion of the highly proliferative large pre-B cell subset within the cell population ( Figures 2F, G ), but also with the previously reported functions of BAP1 in the regulation of transcriptional programs of cell cycle progression in other cell types (32, 34–36, 47). Overall, we hypothesize that a disruption in cell proliferation and cell cycle progression may contribute to the defects in B cell development in BAP1-deficiency.

To further analyze the impact of BAP1 loss on B cell survival and proliferation and the molecular mechanisms involved, Bap1 gene was targeted with CRISPR/Cas9 in the B cell precursor cell line Ba/F3. Two gRNAs were designed to target Bap1 exons 4 and 5 that encode the key N-terminal catalytic domain of the protein ( Figure 5A ). Cell clones were screened by PCR for deletions within the Bap1 locus, and six clones were selected and further analyzed by Western blotting to confirm the loss of BAP1 protein ( Figure 5B ). There was no defect in the viability of Bap1 ^Δ/Δ cell lines, as demonstrated by equivalent frequencies of live, early apoptotic, and late apoptotic/necrotic cells between the Bap1 ^Δ/Δ and control cell cultures ( Figure 5C ). The effect of BAP1 loss on the proliferation of the B cell lines was further analyzed with two approaches: a colorimetric alamarBlue assay conducted in bulk cell population, and a flow cytometric CFSE-dilution assay providing proliferation data at single cell resolution. Both approaches demonstrated a reduction in cell proliferation for all the Bap1 ^Δ/Δ clones tested, reaching statistical significance for the majority of the clones ( Figures 5D–F ). This further confirmed the role of BAP1 in the regulation of B cell proliferation, and indicated the Bap1 ^Δ/Δ Ba/F3 cell lines as a model for further analyses of the molecular mechanisms involved.

To identify the genes directly regulated by BAP1, we mapped the genome-wide DNA-binding sites of BAP1 by ChIP-Seq. The experiment was carried out in the B cell precursor cell line Ba/F3: a) using wild type cells with an anti-BAP1 antibody, and b) using cells stably expressing 3xFLAG-tagged BAP1 protein with an antibody against the FLAG epitope (38). We identified a total of 13,163 BAP1 binding sites (or peaks) across the genome, with a very high concordance between the two datasets ( Figure 6A , Supplemental Table S1E ). Ordering the BAP1-binding sites based on their distance to the nearest gene transcription start site (TSS), 74% were located within 1kb of the nearest TSS and were therefore classified as gene-proximal (≤1kb to TSS, 9,680 sites), and the remaining 26% were classified as gene-distal (>1kb to TSS, 3,483 sites) ( Supplemental Table S1E ). To gain functional insight into the transcriptional targets of BAP1 in B cells, gene ontology analysis was performed on the genes in the vicinity of each BAP1 binding site using the Genomic Regions Enrichment of Annotations Tool (GREAT) (69). This demonstrated that the genes near the gene-proximal BAP1 binding sites were highly enriched for biological process ontology terms “cell cycle” and “DNA metabolic process” ( Figure 6B , Supplemental Table S1F ), suggesting the direct involvement of BAP1 in the regulation of the transcriptional programs required for B cell proliferation.

To gain insights into the mechanisms of BAP1 transcriptional and epigenetic regulation in the B cell lineage, we performed ChIP-Seq analysis for the repressive histone mark H2AK119ub, comparing the Bap1 ^Δ/Δ and control Ba/F3 B cell precursor cell lines. Two clones of Bap1 ^Δ/Δ cells were selected, KO3 and KO4, based on full loss of BAP1 protein and a proliferation defect representing an average of that seen across all the available clones ( Figures 5D, E ). The ChIP-Seq data demonstrated that on average the genomic regions surrounding the BAP1 binding sites had low levels of histone H2AK119ub in wild type Ba/F3 cells, consistent with their predominant localization at promoters of expressed genes ( Figures 6C, D ). Importantly, the levels of histone H2AK119ub at these BAP1-bound genomic sites were significantly elevated in Bap1 ^Δ/Δ relative to wild-type cells ( Figures 6C, D , Supplemental Figure S7B ). This indicated the non-redundant role of BAP1 as a deubiquitinase for histone H2AK119ub in the B cell lineage, as previously also reported for other cell types (27–30).

To compare the roles of BAP1 in transcriptional regulation in B cells and other cell types, we consolidated our BAP1 ChIP-Seq and previously published BAP1 ChIP-Seq datasets, with macrophages and ES cells being the source of the available data (38, 75). Spearman rank correlation analysis was performed for all pairwise comparisons. The results demonstrate a considerable overlap in the genomic location of BAP1 binding across the cell types, and also highlight a subset of BAP1 binding sites potentially specific to B cells ( Figures 6E, F , Supplemental Figures S7C–D , full lists of sites available in Supplemental Table S1E ). Genes nearest to each of the “shared” and the “cell-type specific” subsets of BAP1 binding sites were further analyzed for enrichment of biological process GO-terms, demonstrating shared BAP1 binding at the genes involved in many housekeeping biological processes, including “cell-cycle”, across the different cell lineages ( Supplemental Figure S7C ).

To gain mechanistic insights into BAP1 cross-talk with other transcriptional regulators, we further consolidated our BAP1 ChIP-Seq with ChIP-Seq datasets for BAP1-associated transcriptional regulators ASXL1 (27–30), HCF1 (31–34), OGT (37, 38), and YY1 (33), for components of PRC complexes RING1B, CBX7, and EZH2 (9–11), and with a dataset for deubiquitinase USP16 (17), using data from murine B cells (where available) and also datasets from macrophages and hematopoietic progenitors (17, 38, 76, 78). A significant correlation in genomic localization was observed between BAP1 and BAP1-interacting proteins ASXL1, HCF1, YY1 and to a lesser extent OGT ( Figure 6F , Supplemental Figure S7D , Table S1E ), suggesting that BAP1 acts in cooperation with these proteins in the B cell lineage, as in other cell types (27–34, 37, 38). Weak correlation was observed between the genomic localization of BAP1 and the RING1B catalytic subunit of the histone H2A ubiquitin ligase PRC1, and no correlation was seen with the PRC2 catalytic subunit EZH2 ( Figure 6F , Supplemental Figure S7D , Table S1E ), supporting the antagonistic roles of BAP1 and PRCs in transcriptional regulation. Interestingly, a positive correlation was observed between BAP1 and USP16, indicating that these deubiquitinases may target some common genomic locations, and suggesting cooperative or complementary functions ( Figure 6F , Supplemental Figure S7D , Table S1E ).

We consolidated our ChIP-Seq and RNA-Seq datasets and identified in total 591 genes that had BAP1 binding sites in their proximity and were dysregulated in expression in BAP1 deficiency ( Supplemental Table S1G ). Repeating the analysis for each cluster of genes dysregulated in BAP1-deficient B cells, we observed a strong and significant overrepresentation of BAP1 binding sites at the genes in the downregulated Clusters IV-VI ( Figure 7A , Figures 4D, E ). In particular, the strongest overrepresentation of genes with BAP1 binding sites was seen for Cluster V, enriched for the genes involved in cell cycle progression and downregulated in expression in BAP1-deficient pre-B cells ( Figure 7A , Figures 4D, E ). Overall, this demonstrates a strong concordance between our RNA-Seq and ChIP-Seq datasets, consistent with the primary function of BAP1 as a transcriptional activator. The strong correlation between BAP1 occupancy, histone H2AK119ub levels, and gene expression suggests a direct role of BAP1 in the regulation of genes essential for normal cell cycle progression in pre-B cells.

Gene ontology (GO) analysis further confirmed that the set of genes in proximity of BAP1 binding sites and downregulated in expression in BAP1-deficient pre-B cells was strongly enriched for GO-terms “cell cycle” and “DNA replication” (Cluster V, Figure 7B , Supplemental Table S1H ). Notably, this gene set included the key genes encoding: regulators of G1/S cell cycle transition (Ccne2, Cdc25a) and the initiation of DNA replication (Cdc6, Cdc45, Cdt1); components of the core DNA replication machinery (Pola1, Pold1, Pole, Pole2, Pcna, Rfc1, Dna2, Lig1, Mcm2-7) and enzymes involved in deoxyribonucleotide synthesis (Rrm1, Rrm2); regulators of G2/M cell cycle transition (Bub1, Cdc25a, Espl1, Plk1), mitotic spindle dynamics (Haus4-5, Ndc80, Kif2c, Kif23, Kif11, Kntc1), and cytokinesis (Anln, Ect2, Prc1); key regulators of chromatin structure through the S and M phases of the cell cycle (Esco2, Ncapd2, Ncapg2, Uhrf1); and several general biomarkers of cell proliferation (Foxm1, Mki67) ( Figure 7C , Supplemental Figure S7E , Table S1G ). Overall, our data suggests a direct role of BAP1 in the regulation of the transcriptional programs of cell proliferation, linked to its requirement for cell cycle progression in pre-B cells and for normal B cell development.

To gain insights into the putative mechanisms through which BAP1 may regulate the transcriptional programs of cell cycle progression in B cell development, we assessed the levels of the repressive histone mark H2AK119ub in Bap1 ^Δ/Δ relative to control Ba/F3 B cell precursor cell lines from our ChIP-Seq data, specifically focusing on the BAP1 binding sites within the promoters Cluster V genes downregulated in expression in BAP1 deficient pre-B cells and assigned GO-terms “cell cycle”, “DNA replication”, and “cell division” (69 genes, Supplemental Table S1G ). Importantly, a significant increase in histone H2AK119ub levels was observed in Bap1 ^Δ/Δ relative to control cells at these promoters ( Figures 7D, E ). This strongly suggests that BAP1 regulates the expression of the genes essential for cell proliferation and cell cycle progression in B cell development at least in part via the deubiquitination of histone H2AK119ub.