Modulation of Schwann Cell Homeostasis by the BAP1 Deubiquitinase

Abstract

Schwann cell programming during myelination involves transcriptional networks that activate gene expression but also repress genes that are active in neural crest/embryonic differentiation of Schwann cells. We previously found that a Schwann cell-specific deletion of the EED subunit of the Polycomb Repressive Complex (PRC2) led to inappropriate activation of many such genes. Moreover, some of these genes become re-activated in the pro-regenerative response of Schwann cells to nerve injury, and we found premature activation of the nerve injury program in a Schwann cell-specific knockout of Eed. Polycomb-associated histone modifications include H3K27 trimethylation formed by PRC2 and H2AK119 monoubiquitination (H2AK119ub1), deposited by Polycomb repressive complex 1 (PRC1). We recently found dynamic regulation of H2AK119ub1 in Schwann cell genes after injury. Therefore, we hypothesized that H2AK119 deubiquitination modulates the dynamic polycomb repression of genes involved in Schwann cell maturation. To determine the role of H2AK119 deubiquitination, we generated a Schwann cell-specific knockout of the H2AK119 deubiquitinase Bap1 (BRCA1-associated protein). We found that loss of Bap1 causes tomacula formation, decreased axon diameters and eventual loss of myelinated axons. The gene expression changes are accompanied by redistribution of H2AK119ub1 and H3K27me3 modifications to extragenic sites throughout the genome. BAP1 interacts with OGT in the PR-DUB complex, and our data suggest that the PR-DUB complex plays a multifunctional role in repression of the injury program. Overall, our results indicate Bap1 is required to restrict the spread of polycomb-associated histone modifications in Schwann cells and to promote myelin homeostasis in peripheral nerve.

Graphical Abstract

INTRODUCTION

Schwann cells play an important role in the stability, integrity, and function of the peripheral nervous system. In their terminally differentiated state, the myelin formed by Schwann cells enables saltatory conduction (1), and most cases of inherited neuropathy are caused by impaired Schwann cell function (2–4). Several studies have identified molecular determinants of a homeostatic phase upon myelin maturation, and several mouse mutants of neuropathy-associated genes have demonstrated progressive myelin deformations in adult nerve that eventually impact peripheral nerve function (5–13). Loss of myelin homeostasis can be prompted by alterations in key signaling pathways involved in neuregulin regulation, and also has been observed when epigenetic regulators are disrupted.

In previous studies, we tested the role of the Polycomb Repressive Complex 2 (PRC2), which forms repressive histone H3K27 trimethylation (H3K27me3), by creating a Schwann cell-specific knockout of the EED subunit of PRC2. We observed disruptions in myelin stability with a progressive hypermyelination of small caliber axons (14). Moreover, gene expression analysis revealed that EED had a role in maintaining repression of a significant proportion of the injury program that is normally activated only after peripheral nerve injury (14–16). The injury-induced reprogramming of mature Schwann cells allows transformation to pro-regenerative cells that support growth and survival of axons, clear myelin debris by autophagy/phagocytic responses, and form elongated Bungner bands to pave the way for nerve regeneration, before eventually myelinating any regenerated axons (17). Several transcription factors (e.g. c-JUN) are important activators of injury response genes (17–20). In addition, activation of the injury program was associated with reduction of repressive H3K27me3 on these injury genes (14, 15). Overall, these studies indicated that Polycomb repression was important to suppress the injury program in mature Schwann cells (16). In addition, this provides additional evidence that the homeostatic phase of myelination depends upon repression of JUN and the injury program (21, 22).

Polycomb repression is established by coordinate actions of PRC2 and PRC1, which catalyzes monoubiquitination of histone H2A (H2AK119ub1) through RING1A/B subunits (23). The canonical forms of PRC1 complexes are epigenetic readers of histone H3K27me3 formed by PRC2, leading to further repression through histone H2A ubiquitination and chromatin compaction (24). Accordingly, H2AK119ub1 is often associated with H3K27me3, and we found that H2AK119ub1 also decreases upon activation of polycomb-repressed injury genes (25). While PRC1-mediated repression was originally thought to be secondary to PRC2-dependent H3K27me3, there can be independent regulation of PRC1 complexes to establish H2AK119ub1 that triggers H3K27 methylation (23, 26, 27). Therefore, regulation of H2AK119ub1 can play a coequal or even primary role in establishing and maintaining polycomb repression.

The activity of the PRC1 and PRC2 complexes is opposed by epigenetic eraser proteins that can remove polycomb-associated histone modifications, such as the JMJD3/KDM6B demethylase for H3K27, and the BAP1 deubiquitinase for H2AK119 (28). We have previously tested the role of H3K27 demethylases in Schwann cell gene regulation. The Schwann cell-specific knockout of Jmjd3/Kdm6b and Utx/Kdm6a had little effect on Schwann cell development, other than some modest perturbations in PRC2-regulated genes. In addition, nerve injury responses in Schwann cells were temporally delayed but ultimately little affected at later stages of nerve injury (25).

We therefore decided to determine the role of the BAP1 deubiquitinase in Schwann cells. BAP1 is part of the PR-DUB (Polycomb Repressive-Deubiquitinase) complex, containing also O-Linked N-Acetylglucosamine (GlcNAc) Transferase (OGT) (29, 30). Based on the loss of H2AK119ub1 on injury-induced genes (25), we originally hypothesized that BAP1 may regulate polycomb repression of injury genes. Accordingly, we created a Schwann cell-specific deletion of the H2AK119 deubiquitinase, BAP1 (31). Loss of BAP1 led to impaired myelin homeostasis with tomacula and premature activation of the injury program, suggesting that writers and erasers of polycomb repression play mutually reinforcing roles in gene regulation and myelin homeostasis.

METHODS

Mouse Colony.

Animal experiments were performed according to protocols approved by the University of Wisconsin, Madison. Bap1 floxed mice [Jackson Laboratories 031874] were maintained on the C57BL/6 genetic background and mated to mP0TOTA-Cre (Mpz-cre) (32). The Bap1tm1c(EUCOMM)Hmgu mouse strain carries a floxed (conditional) allele of Bap1, which is referred to as Bap1 fl/fl. Derivation of the strain is described at http://www.informatics.jax.org/strain/MGI:6149727. Specifically, loxP sites flank exons 6–12 of the Bap1 gene. We thank EUCOMM (European Conditional Mouse Mutagenesis Program) for making this strain available. the Bap1 fl/fl and Bap1 fl/fl/Mpz-cre genotypes are referred to as control and Bap1 cKO, respectively. Genotyping employed primers for Bap1: TGGAAGGTAGTTGATTTTAGGC and GTGCACAGCCATCAGTGACT, and for the cre transgene: GAGTGATGAGGTTCGCAAGA-3’ and CTACACCAGAGACGGAAATC-3’

Morphometric quantification of myelination.

Freshly dissected sciatic nerves were immersion fixed in a solution of 2.5% glutaraldehyde, 2.0% paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7.4, overnight at 4°. The nerves were then post-fixed in 1% osmium tetroxide in the same buffer for 2 h at room temperature. The nerves were dehydrated in a graded ethanol series, and then further dehydrated in propylene oxide and embedded in Epon resin. Ultrathin transverse sections were either contrasted with Reynolds lead citrate and 8% uranyl acetate in 50% ethanol or stained with toluidine blue. Images were obtained either with a Philips CM120 electron microscope with an AMT BioSprint side-mounted digital camera at the UW Medical School Electron Microscope Facility or with Nikon Ti2 microscopy. Quantification for g ratios was performed using NIS-Elements 4.0. Three mice per genotype were analyzed, and statistical analyses were evaluated using unpaired t-tests. Axons with myelination abnormalities (tomacula, in/outfolding) were not selected for g ratio quantitation.

Immunofluorescence.

Freshly dissected nerves were embedded in Tissue-Tek OCT compound (Sakura Finetek) and snap frozen with liquid nitrogen. Longitudinal or transverse cryostat sections (10 μm) were air-dried for 5 min and fixed in 4% paraformaldehyde for 10 min. The sections were then blocked in PBS containing 5% donkey serum/1% BSA/0.5% Triton-X 100 for 1 hr at room temperature. Primary antibody incubation was performed overnight at 4 ° in PBS containing 5% donkey serum/1% BSA/1% Triton-X 100 and secondary incubation was performed in PBS at room temperature for 1 hr. Hoechst 33342 (1:5,000 in PBS, Sigma) was applied to stain nuclei for 1 min. Three 4 min washes were performed in PBS after fixation and blocking, and in PBS containing 0.1% Tween20 after primary antibody incubation and nuclear staining. After coverslips were mounted using Fluoromount-G^™ (SouthernBiotech), sections were examined on a Nikon A1R confocal and quantitated by both Columbus imaging software and manual curation.

For KI67 staining, tissue sections were blocked for 45 min with 2.5% donkey serum diluted in PBS, and then incubated with an anti-KI67 antibody, diluted 1:500 in blocking buffer, at 4° overnight. Tissue sections were washed with PBS-T (PBS with 0.05% Tween 20) three times, incubated with Alexa Fluor 488-conjugated secondary antibody (Invitrogen, A11008) at 25° for 45 min, and washed with PBS-T three times. Mounting medium containing DAPI (Electron Microscopy Sciences, 1798550) was applied, and tissue sections were imaged with a Nikon Eclipse Ti2 inverted microscope.

Western blot.

Freshly dissected nerves were snap frozen with liquid nitrogen and crushed. The nerves were then homogenized in lysis buffer (50 mM Tris–HCl at pH 6.8, 10% glycerol, 2% SDS, 10% β-mercaptoethanol, 50 mM NaF, 1 mM Na3VO4 and Protease Inhibitor Cocktail (Sigma, P8340) using a motorized pellet pestle. Cells in culture were homogenized in 3x lysis buffer. After a 15 min incubation in ice, lysates were boiled at 95 ° for 3 min and centrifuged at 4 ° for 15 min. Subsequently, supernatants were collected and subjected to SDS-PAGE. After transfer to polyvinylidene fluoride membrane, proteins were blocked in TBST containing 5% nonfat dry milk for 1 hr at room temperature. Primary and Secondary antibody incubations were performed in TBST containing 5% BSA (Sigma, A7906) at 4 ° for overnight and at room temperature for 1 hr, respectively. Three 5 min-washes were performed in TBST after the incubations. The membranes were scanned and quantitated with the Odyssey Infrared Imaging System (Li-Cor Biosciences). Antibodies are listed in Table 1.

Table 1.

List of Antibodies Used.

Antibodies	Catalog number	Company
SOX10	AF2864	R & D Systems
Ki67	Ab16667	Abcam
BAP1	13271	Cell Signaling Technology
ACTB	#AC004	Abclonal
IgG	12–370	Millipore
H2AK119ub1	8240	Cell Signaling Technology
H3K27me3	AM39155	Active Motif
H3	14269	Cell Signaling Technology
c-JUN	9165	Cell Signaling Technology
PTEN	9188	Cell Signaling Technology

Quantitative RT-PCR.

RNA was isolated from sciatic nerves using the Trizol/chloroform RNA extraction protocol following purification with RNA Clean and Concentrator kit (Zymo). To prepare cDNA, 0.5–1 μg of total RNA was used from each sample. qRT-PCR and data analysis were performed as described previously (33). qPCR was performed in duplicate per sample. Primers for Bap1 expression were: CTGGGCTCTCGTTGAACTACTCA and CCTCATCAGGGCCCTTCAC.

RNA seq.

Using control and Bap1 cKO mice at 8 weeks of age (n=6/group), RNA was purified from sciatic nerves, and 1000 ng total RNA was sent to Genewiz (South Plainfield, NJ) for library preparation after PolyA selection and Illumina sequencing (Illumina HiSeq 2×150bp). Reads were aligned to GRCm38/mm10 genome using the STAR aligner (34). Data were analyzed using DESeq2 (35) to determine differentially regulated genes between wild type and Bap1 KO mice (p-value < .05). RNA-seq data are deposited in NCBI GEO under accession number GSE197149.

Chromatin Immunoprecipitation.

Six freshly dissected mouse sciatic nerves per condition (Bap1 fl/fl/Mpz-cre and Bap1 fl/fl) were minced in 1% formaldehyde for 10min then quenched for 10 min with 0.125M glycine. Samples were incubated in LB1 for 10min, washed with LB2, then bead-homogenized using the Next Advance Bullet Blender for 5min with speed 12 (33). Samples were centrifuged to pull down nuclei-containing pellet. Pellet was resuspended in LB3 containing 0.1% SDS. Samples were sonicated via Bioruptor (Diagenode) on high for 2×15min cycles (30sec on/off). Sonicated chromatin was incubated with 5ug of target antibody. A 50 ul aliquot of Dynabead Protein G slurry was used to pull down the protein-antibody complex. A 10% aliquot of remaining chromatin was saved for input analysis. Immunoprecipitations were washed three times in RIPA buffer and then eluted at 65 °C in reverse cross-linking buffer (50 mM Tris, 10 mM EDTA, 1% SDS). ChIP-DNA was purified using the Cell Signaling DNA Purification Kit (CST #14209S). Antibodies used are in Table 1. Sample libraries were prepared from ChIP DNA, and respective inputs, using the Illumina TruSeq ChIP Sample Preparation Kit at the UW Biotechnology Center DNA Sequencing Facility. After sequencing on the Illumina Novaseq 6000, FastQ files were generated by bcl2fastq v2.20.0.422. H2AK119ub1 and H3K27me3 reads were mapped to the reference genome mm10 using Bowtie2. BAM files were filtered using the NGSUtils suite for mapped and properly paired reads. Peak tracks were generated using MACS2 callpeak relative to an input file. Bedgraph files were generated using the genome coverage tool in the bedtools suite. The data have been deposited in NCBI GEO GSE211544.

Bioinformatic Analysis.

The ChIP-seq data sets from mouse and similar data sets (GSE159265) of H2AK119ub1 in rat peripheral nerve (25) were used to create the heat maps for the deregulated genes. The data matrix for each heatmap and read density plots were generated using the EAseq suite (36) from filtered bam files based on ChIP-seq data sets and then clustered and visualized. Called peak annotations were created using ChIPSeeker (37). The gene ontology enrichment analyses were conducted with PANTHER and GSEA software to identify the enriched biological processes based on respective upregulated and downregulated lists of Bap1 cKO genes. (38). For comparative analysis in transcriptomes, the statistical significance for the overlap between potential gene lists was tested at Nemaotes bioinformatic tools which utilizes the Fisher’s Exact Test.

RESULTS

Conditional inactivation of Bap1 in Schwann cells

To test the role of BAP1 in regulating Schwann cell development and maturation, we developed a Schwann cell specific conditional knockout (cKO) of Bap1 using the Mpz-cre driver, which is active in embryonic Schwann cell development at E13.5-14.5 (32). The loxP sites surround exons 6–12 of Bap1 including part of the catalytic UCH domain (Figure 1A) (39). The deletion results in a frameshift 3’ of the deletion, leading to loss of the C-terminal sequence. The mouse knockout was validated by quantitative RT-PCR analysis of sciatic nerve cDNA using primers located in exons 7/8 in the deleted region, showing loss of Bap1 expression in the conditional knockout (Figure 1A,B). The decrement in Bap1 expression is consistent with the proportion of Schwann cells found in sciatic nerve. We also further confirmed the loss of BAP1 protein through immunostaining with an antibody targeting an epitope in the C-terminus of BAP1. Coexpression of BAP1 and SOX10, a marker of Schwann cells (40), was absent in the knockout whereas there are cells positive for such markers in control samples (Figure 1C, triangles).

Figure 1. Schwann-cell specific knockout of Bap1.

A. The Bap1 conditional allele contains loxP sites flanking exons 6–12, which encode part of the catalytic domain. B. The qRT-PCR analysis of RNA extracted from the control and BAP1 cKO sciatic nerves was performed using primers within the excised regions. P<0.05 unpaired t-test, n=3/group. C. Immunofluorescence analysis of the longitudinal sections from the wildtype and BAP1 cKO sciatic nerves was performed using the indicated antibodies. n=3 for control and n=3 for cKO. Error bars indicate standard deviation.

Myelin abnormalities in the Bap1 knockout

Mice with a Schwann cell-specific knockout of Bap1 were overtly normal with no evident motor impairment after weaning. To determine if Bap1 loss affects Schwann cell morphology, we performed toluidine blue staining of semi-thin sections to analyze the nerve ultrastructure (Figure 2A). At 6 weeks, the sciatic nerves of mutant mice were examined and many axons were observed to contain tomacula, and myelin infolding/outfolding (Figure 2B,C). Bap1 cKO mice have a significantly greater prevalence of abnormal myelination compared to control (Figure 2C). Myelin thickness was assessed using the g-ratio, which is the axon diameter divided by outer diameter of myelin sheath (33). The overall g-ratio was not significantly different in the Bap1 cKO (Figure 2D). A binned analysis also showed no significant change in g ratios in any size range (not shown). However, the distribution of axon diameters was skewed with a notable deficit of the higher caliber axons (>4 micron) and an elevated number of smaller diameter axons (Figure 2E). The average median axon diameter for control was 2.97 microns (n=3), compared to 2.5 for the Bap1 cKO (p=0.03, 1-tailed t-test). We separately tested the overall distribution of axons in control vs. Bap1 cKO using the Mann-Whitney test, which yields a highly significant change in distribution (p<10⁻¹⁴), although the effect size is small.

Figure 2. Myelin morphology in the Bap1 cKO.

A. Electron micrographs of the sciatic nerves at 6 weeks of Bap1 cKO mice and littermate controls. B. The highlighted fibers display abnormal morphologies, including tomacula (arrow), and outfolding (arrowhead) and infolding (asterisk). C. Quantitation shows increased occurrence of abnormal myelin profiles (including tomacula and in/outfolding) in the Bap1 cKO compared to the control (p<0.05, unpaired t-test, n=3/group). D. For g-ratio analysis, the axon and outer diameters of myelinated fibers were measured on randomly selected fibers per genotype. The overall g ratios were not significantly different between the Bap1 cKO and control. E. The bar graph shows the distribution of axon diameters in control vs. Bap1 cKO. The average median diameter for control was 2.97 microns (n=3), compared to 2.5 for the Bap1 cKO (p=0.03, 1-tailed t-test). A Mann-Whitney test of the overall distribution of axons in control vs. Bap1 cKO yields a highly significant change in distribution (p<10⁻¹⁴). Data: n=3 per genotype, 717 WT vs 745 cKO axons were counted.

We also assessed femoral motor nerve at 4 months. While there was a clear thinning of myelin around large diameter axons (Figure 3), there was significant hypermyelination of smaller diameter axons (Figure 3C). The overall g ratio was reduced (0.62 in Bap1 cKO vs. 0.71 in control, n=4, p<0.0005, t-test). As seen in sciatic nerve, the distribution of axon diameters was shifted to an even greater degree, with the median diameter being reduced to 3.7 microns in the Bap1 cKO compared to 6.85 in the control (p<0.0005). Additionally, numerous amyelinated axons were observed in the Bap1 cKO, which are rarely observed in wild type nerves. Quantitation of total myelinated axons in femoral motor nerve showed a reduction in Bap1 cKO mice/nerves compared to control (avg. 337 vs 506, n=4, p<0.0005).

Figure 3. Myelin morphology in the femoral motor nerve.

A. Toluidine blue-stained semithin sections of femoral motor nerve of 4 month old Bap1 cKO mice and littermate controls. B. For g-ratio analysis, axon and myelinated fiber diameters were measured on all femoral motor nerve axons per genotype. The overall g ratio was reduced (0.62 in Bap1 cKO vs. 0.71 in control, n=4, p<0.0005, t-test). C. The bar graph shows the distribution of axon diameters in control vs. Bap1 cKO. The median axon diameter is reduced to 3.7 microns in the Bap1 cKO compared to 6.85 in the control (p<0.0005). The total number of myelinated axons is reduced in Bap1 cKO mice/nerves compared to control (avg. 337 vs 506 per nerve, n=4, p<0.0005).

Derepression of injury genes in the BAP1 knockout

Given the myelin defects observed in the Bap1 cKO, RNA-seq of sciatic nerve was used to compare Bap1 cKO nerves and controls (n=6) at ~6 weeks of age. As shown in the volcano plot, there were widespread changes in gene expression (Figure 4A), and we identified 2446 upregulated and 2788 downregulated genes (p<0.05, FDR corrected, Supplementary Table 1). Gene Set Enrichment Analysis (Figure 4B) indicated that cholesterol and sterol biosynthesis and pathways are downregulated, whereas ribosome/polysome pathway genes are elevated by the loss of BAP1. The RNA-seq analysis did not suggest any increase in other cell populations such as macrophages. An independent gene ontology analysis via PANTHER (Supplementary Figure 1) revealed categories of upregulated genes associated with proliferation and gliogenesis (38). Enriched categories for downregulated genes included processes of Schwann cell differentiation and lipid synthesis (Supplementary Figure 1). The increase in cell cycle-associated genes revealed by RNA-seq suggested that there may be changes in proliferation. While there are normally very few proliferative cells in mature nerve, we did find an apparent increase in KI-67-positive cells in the Bap1 cKO (Supplementary Figure 4).

Figure 4. RNA-seq analysis of Bap1 cKO mice.

A. RNA-seq analysis was performed in 6 week old Bap1 cKO and littermate controls (n=6/genotype). Volcano plots indicate increased and decreased genes in the Bap1 cKO relative to control after filtering for >2-fold change and <0.05 p value. B. GSEA enrichment plots summarize some of the major biological and cellular processes that are affected in the Bap1 cKO.

For further analysis, we used a >2-fold cutoff, yielding 708 upregulated genes and 604 downregulated genes in the Bap1 cKO (Supplementary Table 2). Many of the downregulated genes were already expressed at relatively low levels in control and then repressed further in the Bap1 cKO, but the downregulated Schwann cell genes include β-tubulin 4a (Tubb4a), Contactin 1 (Cntn1), Claudin 11 (Cldn11), Dual specificity phosphatase 26 (Dusp26), fatty acid elongase (Elovl4), and Hoxb2. Hox (homeobox) genes are commonly regulated by Polycomb repression. (16). In the downregulated dataset, there was modest downregulation of several myelin genes with Mbp being the most significant. Other genes are consistent with the ontology analysis that indicates downregulation of lipid biosynthetic genes.

Polycomb repression typically represses expression of transcription factors of other cell lineages (41–43). Schwann cells and oligodendrocytes share the ability to myelinate axons, but their regulatory networks are quite distinct (44) and two oligodendrocyte-specific transcription factors (Myrf and Nkx6.2) are induced in the Bap1 cKO, along with MYRF target genes in oligodendrocytes (Dct, Adamts4, Mog, Omg, Klk6, Ptgds, Rab33a) (45, 46). Since MYRF is regulated by SOX10 (44), this would suggest that BAP1 is required to prevent inappropriate expression of MYRF and oligodendrocyte gene networks in Schwann cells.

Polycomb-associated histone modifications in BAP1-regulated genes

To determine if BAP1 regulates polycomb histone modifications in Schwann cells, we also measured the levels of H2AK119ub1 by Western blot of sciatic nerve of the Bap1 conditional knockout and found significantly increased levels of both histone marks (Figure 5A,B). Elevated H2AK119ub1 is consistent with the loss of H2A deubiquitinase activity, and increased levels of H3K27me3 likely reflects that PRC1 activity can stimulate H3K27me3 deposition by PRC2 (27, 47, 48).

Figure 5. Loss of Bap1 causes redistribution of H2AK119ub1.

A. Western blot analysis of lysates from the sciatic nerves in control and mutant was performed using the indicated antibodies. B. The quantification of western blot results indicating the relative expression of the histone modifications normalized to total histone H3. Individual data points are shown (p<0.05, n=3, t-test). C. ChIP-seq analysis of H2AK119ub1 was performed in control and Bap1 cKO sciatic nerve, and heatmaps of control and knockout show density of H2AK119ub1 reads centered on the control peak locations. The heatmaps indicate decreased read density of H2AK119ub1 at these sites in the Bap1 cKO. D. The ChIP-seq analysis of H2AK119ub1 in Bap1 cKO nerve shows an increased number of peaks compared to the control. Using the expanded Bap1 cKO peak set, the heatmaps of wildtype and knockout H2AK119ub1 reads centered on knockout peaks show accumulation of H2AK119ub1 at many of the Bap1 cKO peaks compared to control.

Other studies have determined that BAP1 regulates the distribution of polycomb histone modifications in studies of cultured cells (49, 50)., but there has been limited analysis of BAP1 function in specific cell types in vivo. We then determined the genomic distribution of H2AK119ub1 and H3K27me3 in the control and Bap1 conditional knockout mice, as we had done previously in rat sciatic nerve (14–16, 25). Peak calling for H2AK119ub1 in the control sample identified >35,000 peaks, and the heat maps over these peaks showed a decreased read density in the Bap1 cKO (Figure 5C). In contrast, peak calling in the Bap1 cKO sample identified a much larger number (>60,000 peaks) of H2AK119ub1 occupied sites (Figure 5D). The increased number of peaks is reflected by the apparent increase of H2AK119ub1 read density at many of these sites in the Bap1 cKO compared to control. Therefore, BAP1 loss leads to a profound redistribution of the H2AK119ub1 modification from normal sites to an enlarged set of novel sites. Further analysis of read distribution in a binned analysis confirmed that the normal sites of H2AK119ub1 deposition were diminished in the Bap1 cKO, but there was a redistribution of reads to the expanded peak set in the Bap1 cKO (Supplementary Figure 2)

Changes in H2AK119ub1 can affect H3K27me3 deposition by PRC2 (27, 47, 48). A similar analysis of H3K27me3 (Figure 6A) showed that control peaks (12325) are diminished in the Bap1 cKO, but peak calling in the cKO sample indicates sites that H3K27me3 becomes redistributed to other genomic locations in the Bap1 cKO, resulting in a similar increase in the total number of H3K27me3 peaks (19308). The shift of H3K27me3 reads to an expanded peak set in the Bap1 cKO was confirmed by distribution plot analysis (Supplementary Figure 3).

Figure 6. The H3K27me3 histone modification is redistributed similar to H2AK119ub1 in the Bap1 knockout.

A. ChIP-seq analysis of H3K27me3 was performed in control and Bap1 cKO sciatic nerve, and heatmaps of wildtype and knockout H3K27me3 reads centered on wildtype peaks show the loss of H3K27me3 at these sites in Bap1 cKO nerve. B. Peak calling in the Bap1 cKO ChIP-seq analysis of H3K27me3 identified an increased number of peaks compared to control. Heatmaps of wildtype and knockout H3K27me3 reads centered on Bap1 cKO peaks show increased read density of H3K27me3 in the expanded peak set compared to control.

We wanted to determine if deregulated genes in the Bap1 cKO from the RNA-seq analysis were associated with polycomb repression, and more than a third of the upregulated genes are associated with H3K27me3 and/or H2AK119ub1 peaks. Specifically, at least 216 upregulated genes in the Bap1 cKO are associated with both or either of histone marks, which includes many key nerve injury genes (e.g. Shh, Gdnf, and Runx2) which had been found to be regulated by Eed (15, 16, 25). In addition, 187 downregulated genes in the Bap1 cKO are also associated with H2AK119ub1 and/or H3K27me3, suggesting that the redistribution of H2AK119ub1 may strengthen polycomb repression of some genes in the Bap1 knockout.

BAP1 regulation of nerve injury genes.

Within the upregulated dataset, we observed several significant nerve injury genes that were prematurely activated in our analysis of PRC2 function in the Eed cKO (16), including Shh, Gdnf, Fgf5, Hmga2, Vgf, and Igfbp2. Upregulated genes include growth factors (Neuregulin 1 and Betacellulin) (51), cytokines (Cxcl10), and transcription factors (Tfap2a, Runx2, and Pou3f1) (14, 33, 52). Since our previous studies had indicated that loss of EED led to upregulation of a significant proportion of the nerve injury program in Schwann cells (14–16), we focused on the overlap between the nerve injury genes and the polycomb histone marks: H3K27me3 and H2AK119ub1 (Figure 7A). Our previous RNA-seq studies identified 1855 injury genes that are >2-fold induced in sciatic nerve (1 and 7 d post-transection gene, Supplementary Table 3) (16). Using the called peaks from the histone modification analysis, we found that around 844 out of 1855 injury-induced genes are associated with either or both histone marks. 327 genes have promoter-proximal peaks for both H2AK119ub1 and H3K27me3. There are 182 injury genes with called peaks for H2AK119ub1 alone and 335 genes with H3K27me3 alone.

Figure 7. Association of nerve injury genes with polycomb-associated histone modifications.

A. The pie chart indicates injury-induced genes in sciatic nerve as determined by RNA-seq analysis. The proportion of these genes associated with either or both polycomb-associated histone modifications (H2AK119ub1 and/or H3K27me3) is indicated by shading. B. The overlap of injury-induced genes and those genes with altered regulation in the Bap1 cKO is shown. C. Profiles of H2AK119ub1 from the ChIP-seq analysis of control and Bap1 cKO nerve show localization at the transcription start sites for two injury-induced genes: Gdnf and Nrg1 type 1.

Focusing on the set of 844 injury genes associated with H3K27me3 and/or H2AK119ub1, a significant set are affected by BAP1 cKO. At least 70 Bap1-dependent upregulated genes that are induced by nerve injury are also associated with either or both histone marks. Several of them are prominent nerve injury genes including Gdnf, Igfbp2, Pou3f1, Fgf5, Shh, Runx2, and Hmga2 (53, 54). In contrast, only 10 injury-induced genes associated with either or both histone marks are downregulated in the Bap1 cKO (Figure 7B). Figure 7C shows the peaks of H2AK119ub1 in the Gdnf and Nrg1 type 1 promoters, which are diminished in the Bap1 cKO.

Comparison of the roles of BAP1 and OGT

BAP1 interacts with OGT as subunits of the PR-DUB complex (29, 30), and a Schwann-cell specific Ogt KO mouse model exhibited deficient remyelination after injury that was traced to the premature activation of JUN in uninjured nerve (21, 55). The Ogt mutant mouse exhibited a significant number of tomacula and thinner myelination of larger caliber axons, with eventual loss of myelinated axons, which is similar in several respects to the Bap1 cKO. We determined if a common PR-DUB target gene set could be defined by comparative analysis of genes regulated by OGT and BAP1, we compared the RNA seq data of the Bap1 cKO to the published RNA-seq analysis of the Ogt cKO (21). Out of 1298 upregulated genes with significant p-values in Ogt-KO datasets (Supplementary Table 2), there are 153 matches with upregulated Bap1 cKO genes and 54 matches with downregulated Bap1 cKO genes (Figure 8C).

Figure 8. c-JUN is not affected in the Bap1 knockout.

A. Western blot analysis of lysates from the sciatic nerves in control and mutant was performed using the indicated antibodies. B. Quantification of c-JUN protein expression is shown relative to ACTB in bar graph. Data: n=3/group Individual data points are shown. C. The overlap of Bap1-regulated and OGT-regulated genes is shown.

Interactions of polycomb-regulated genes with the JUN-regulated network of injury genes.

A significant proportion of dysregulated target genes in the Ogt cKO are unchanged in the Bap1 cKO. An important feature of the Ogt knockout is the activation of JUN in uninjured nerve (21), as expression of JUN can drive demyelination (22). While JUN is a major regulator of nerve injury program in Schwann cell (20), JUN is normally not expressed at high levels prior to injury. It is possible that part of the induction of injury genes could be mediated by JUN, since there is a slight 1.7-fold induction of the c-Jun transcript in our RNA-seq analysis of the Bap1 knockout. Therefore, we examined the JUN by western blot but observed no significant difference in JUN expression between genotypes at 6 weeks of age (Figure 8A,B).

It is still possible that JUN activity is stimulated post-transcriptionally, and RNA-seq indicates that Fos, a component of AP-1, is upregulated 2.8-fold. Therefore, we compared the RNA seq profile of the BAP1 cKO to RNA-seq analysis of a mouse model with Schwann cell-specific JUN overexpression (22). Using this data set, there are only 11 common genes that are upregulated in both BAP1 cKO and the JUN overexpression model, such as Shh, S100a4, Btc, and Cxcl10 (Supplementary Table 5). The overlap between upregulated Bap1 cKO and c-JUN cKO genes is significant (p < 2e-10). While there is overlapping regulation of injury genes by JUN and polycomb repression, the evidence indicates that they also play independent roles since we do not see induction of the entire JUN target gene network in Schwann cells. Therefore, we conclude that BAP1’s active role in regulating nerve injury genes is not mediated by JUN. There is also an overlap of JUN-regulated genes and those that were found to be deregulated in the EED cKO, but we also did not detect measurable induction of JUN in the EED cKO (16).

DISCUSSION

Since premature activation of the Schwann cell injury program has been shown to adversely impact the formation of stable myelin (21, 22), several mechanisms have been identified that repress injury genes in mature Schwann cells. One of these mechanisms, Polycomb repression, is regulated by diverse complexes that impinge on this type of regulation. In particular, there are multiple assemblies of PRC2 and PRC1 complexes that can play unique roles in establishing polycomb repression in different genes (28, 56). Our previous study of the Eed knockout in Schwann cells revealed roles in the regulation of myelin homeostasis and also in repression of the injury program (14–16). PRC2 regulation in Schwann cells has been tied to interactions with CTCF and the ACTL6A-containing BAF complex (57, 58) and loss of PRC2 regulation is a critical step in formation of malignant peripheral nerve sheath tumors in Neurofibromatosis (16, 59). However, PRC2 regulation is intertwined in several respects with PRC1, and we had found changes in PRC1-mediated H2A ubiquitination in Schwann cells after injury (25).

To understand how dynamics of H2AK119 ubiquitination affect polycomb repression in Schwann cells, we focused on the BAP1 deubiquitinase, which is a component of PR-DUB complex that also includes other subunits like OGT and ASXL1/2 (30, 60). Loss of BAP1 in Schwann cells causes an abnormal myelin morphology that is evident at six weeks. Specifically, there are increased levels of tomacula and abnormal myelin, and some large caliber axons have thinner myelin in the mutant mouse, although the overall g ratio is unaffected at this time point. Analysis of femoral motor nerve at 4 months showed a more severe phenotype, with a shift of axon distribution to smaller calibers with thicker myelin, accompanied by a reduction in the total number of myelinated axons. Overall, these phenotypes are similar to those seen in the Schwann cell-specific knockout of Eed (14), although somewhat more severe, indicating an important role of BAP1 in myelin homeostasis.

An RNA-seq analysis of sciatic nerve revealed significant changes in gene regulation. Importantly, we found that several key nerve injury genes are prematurely activated in Bap1 cKO mice including Shh, Gdnf, Runx2, Hmga2, Cdkn2a, and Fgf5, which were also activated in the Eed ckO (16), but this analysis revealed a broader range of polycomb-repressed genes, many of which are associated with H2AK119 ubiquitination. The gene expression changes are similar in several respects with previous analysis of a Schwann cell knockout of OGT, which interacts with BAP1 in the PR-DUB complex. The Schwann cell-specific Ogt-cKO leads to elevated JUN activity, resulting in the premature activation of nerve injury genes and a significant degree of tomacula and thinner myelination of larger caliber axons. Specifically, OGT suppresses the nerve injury response by catalyzing O-glcnacylation of the JUN protein (21, 55), and one of the striking findings of the OGT analysis was that the failed regeneration after injury could be rescued by deletion of one allele of Jun (21).

Since BAP1 and OGT are part of the PR-DUB complex, it was possible that loss of Bap1 could phenocopy the myelination defects seen in Ogt-KO (at 3 months), and the myelin defects showed some similarities as noted above. A comparison of the RNA-seq data sets revealed that 153 out of 708 upregulated genes in Bap1 cKO overlaps with those of Ogt-KO genes suggesting that there is some level of functional overlap, but there are also quite significant differences in the transcriptomic effect of BAP1 loss. Some of these 153 genes are Schwann cell nerve injury genes, which could explain the abnormal myelination in Bap1 cKO mouse model. However, Jun was only slightly elevated (1.7-fold) in the RNA-seq analysis. Moreover, we did not observe any change in JUN protein in the Bap1 cKO. Using Ogt-KO RNA seq data (21), there were 11 matches out of 28 upregulated JUN target genes with Bap1 cKO genes. Therefore, we conclude that BAP1 and OGT appear to play overlapping but functionally distinct roles within a multifunctional PR-DUB complex by maintaining polycomb repression of the nerve injury program and inhibiting JUN, respectively. These roles are mutually reinforcing since there is considerable overlap of JUN-regulated and polycomb-regulated genes in the nerve injury response (16). Accordingly, OGT is less tightly associated with BAP1 compared to other PR-DUB subunits and can participate in other complexes (29, 60).

Several of the deregulated genes could cause the myelination defects in Bap1 mutant Schwann cells. One gene that was derepressed in both the Eed and Bap1 cKO is IGF-binding protein 2 (Igfbp2), which was shown to increase PI3 kinase signaling (14). Another leading candidate is the Neuregulin 1 type I (Nrg1) gene, since previous studies have found that efficient remyelination of injured nerves requires the de novo activation of type I NRG1 in denervated Schwann cells (61). We had found that H3K27me3 mediates repression of Nrg1 and there is increased expression of the type I Nrg1 transcript in the uninjured Eed cKO nerves (16). Our data also indicate the presence of H2AK119ub1 on the Nrg1 type 1 promoter (Figure 7). A recent study showed that overexpression of NRG1 type 1 is responsible for hypermyelination of small caliber axons and the onion bulb pathology that has been observed in CMT1A neuropathy (62). Another potential mechanism could involve BAP1 regulation of the PTEN lipid phosphatase, since a Schwann cell-specific deletion of Pten caused hypermyelination (8, 63). It has been reported that BAP1 could repress Pten at the transcriptional and post-transcriptional levels (64, 65). However, Pten levels were not elevated in our RNA-seq analysis and Western blot showed no change in PTEN protein levels (data not shown).

Since PRC1 and PRC2 regulation are connected at multiple levels, we compared the deregulated genes with those obtained in our previous RNA-seq profiling of a Schwann cell-specific deletion of Eed (16). Although we had speculated that deletion of Bap1 could result in strengthened Polycomb repression, we observed derepression of many polycomb-regulated genes; 105 upregulated Bap1-KO genes are in common with those upregulated in the Eed conditional knockout (Supplementary Table 4). Additionally, there are 52 downregulated genes that are regulated by EED. It was unexpected that deletion of the BAP1 deubiquitinase as a polycomb modification eraser would have a similar phenotype and overlapping gene expression changes with inhibition of the PRC2 writer complex in the Eed cKO (14). Several cell culture studies have characterized the role of BAP1 and divergent results have been obtained. In some cases, loss of BAP1 leads to increased polycomb repression which is associated with increased levels of H2AK119ub1 (60, 66). In contrast, other studies have shown that loss of BAP1 leads to loss of polycomb repression (49, 50). This has been attributed to a function of BAP1 to safeguard transcriptionally silent genes by preventing accumulation of extragenic H2AK119 and H3K27me3 modifications that would titrate out polycomb repression if allowed to spread (49, 50). Accordingly, the Drosophila homolog of BAP1, calypso, was originally classified as a polycomb group gene (31). These mechanisms are highly relevant to the observed mutation of BAP1 in several types of cancer since both gain- and loss-of-function models of polycomb repression can contribute to malignancy (49, 56, 66, 67). In our case, we found both increased levels of H2AK119ub1 and H3K27me3 in peripheral nerve, and also a larger number of called peaks for H2AK119ub1 and H3K27me3 in our ChIP-seq analysis of Bap1 mutant nerve. Our study is unique in that we have analyzed the effect of Bap1 on polycomb repression in post-mitotic Schwann cells in vivo compared to previous studies in cell culture models. However, based on the earlier studies (49, 50), our data are consistent with a model in which derepression of some polycomb-associated genes is due to inappropriate accumulation of H2AK119ub1 and H3K27me3 at new sites across the genome.

Numerous pro-regenerative genes in Schwann cells are repressed by both H2AK119ub1 and H3K27me3 and their basal expression can be virtually undetectable in intact, healthy peripheral nerve. Previous studies had suggested that removal of polycomb repression is required for induction of nerve injury genes such as Shh, and Gdnf (14–16). PRC2 is commonly known to work cooperatively with PRC1, but recent studies had indicated that PRC1-mediated repression can be mediated independently of PRC2 (28). Moreover, it has emerged that phenotypes of PRC1-related phenotypes are often more severe than PRC2 mutants (68), and the Bap1 knockout phenotype is more severe than either the Eed or the Jmjd3/Utx knockouts in Schwann cells (25). Therefore, future experiments will be used to determine if deubiquitinase(s) play a role in regulation of injury genes in Schwann cells.

Data availability:

The data that support the findings of this study are openly available in NCBI Gene Expression Omnibus at https://www.ncbi.nlm.nih.gov/geo/, accession number GSE197149 and GSE211544.