Epigenetic remodeling is required during B cell differentiation. However, little is known about the direct functions of epigenetic enzymes in Ab-secreting cells (ASC) in vivo. In this study, we examined ASC differentiation independent of T cell help and germinal center reactions using mice with inducible or B cell–specific deletions of Ezh2. Following stimulation with influenza virus or LPS, Ezh2-deficient ASC poorly proliferated and inappropriately maintained expression of inflammatory pathways, B cell–lineage transcription factors, and Blimp-1–repressed genes, leading to fewer and less functional ASC. In the absence of EZH2, genes that normally gained histone H3 lysine 27 trimethylation were dysregulated and exhibited increased chromatin accessibility. Furthermore, EZH2 was also required for maximal Ab secretion by ASC, in part due to reduced mitochondrial respiration, impaired glucose metabolism, and poor expression of the unfolded-protein response pathway. Together, these data demonstrate that EZH2 is essential in facilitating epigenetic changes that regulate ASC fate, function, and metabolism.
The humoral immune response is initiated when B cells are stimulated to differentiate into Ab-secreting cells (ASC), also known as plasma cells. Irrespective of how they are activated or the availability of T cell help, a distinct set of reprogramming events are required to terminate the B cell–fate program and initiate a new gene expression program that supports enhanced metabolism and Ab secretion (1–3). The ASC transcriptional program is enabled by the expression of transcription factors—such as Blimp-1, XBP1, and IRF4, which reinforce and support ASC transcriptional changes (2, 4)—that are coupled to a reorganization of the epigenome (5–7). Although the transcription factors and genes they regulate have been studied biochemically and genetically, little is known about the role of epigenetic modifiers in ASC function and programming.
Epigenetic modifications are dynamic during distinct stages of B cell differentiation. In both mice and humans, DNA methylation is primarily lost as B cells differentiate to ASC in response to both T cell–dependent and –independent stimuli (8–10). Deletion of the maintenance methyltransferase Dnmt1 leads to a reduction in germinal center (GC) B cells (11), but whether de novo DNA methylation is required for B cell differentiation is not known. Histone modifications, characterized by chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq), have cell type–specific patterns in naive B cells (nB), ex vivo–differentiated ASC, and in GC B cells (12–16). However, few studies have examined the role of histone-modifying enzymes using genetic approaches (5). Deletion of the histone acetyltransferase, MOZ, reduces GC B cells and skews responding B cells toward low-affinity IgM+ memory B cells (17). Additionally, treatment of mice with histone deacetylase inhibitors reduces B cell responses (18), indicating that both erasing and writing de novo epigenetic modifications is an essential process in B cell differentiation. Importantly, epigenetic modifiers are frequent targets of both activating and inactivating mutations in lymphomas (19, 20). Therefore, a full understanding of epigenetic mechanisms and targets for distinct enzymes is important to manipulate B cell differentiation and to understand the effects of therapeutics targeting these enzymes.
One of the best characterized repressive epigenetic histone modifications is the trimethylation of histone H3 at lysine 27 (H3K27me3), which is mediated by the polycomb repressive complex 2 (PRC2) (21, 22). Enhancer of zest 2 (EZH2) is the catalytic subunit of the PRC2 complex and functions as an essential transcriptional silencer (23–25). EZH2 is upregulated in pre–B cells, in which it is necessary for VDJ recombination during B cell development (26) and to repress germline Igκ transcription (27). EZH2 is expressed at low levels in quiescent, nB, but is highly upregulated in GC B cells where it facilitates cellular proliferation, protects from activation-induced cytidine deaminase (AID) off-target activity, and represses the differentiation of GC B cells into ASC (15, 16, 28, 29). EZH2 interacts with distinct sets of transcription factors, such as BCL6 in GC B cells (29) and Blimp-1 in ASC (30), to direct cell type–specific gene repression programs. B cell–specific deletion of Ezh2 leads to a loss of GC formation, thereby leading to defects in the formation of ASC (15, 16). However, no studies have directly assessed whether or how EZH2 functions in ASC.
In this study, we tested the role of EZH2 in two T-independent models of ASC differentiation, one initiated by the T-independent Ag, LPS, and the other initiated by influenza infection in the absence of CD4 T cells. We found that EZH2 was progressively upregulated in stimulated B cells, with expression peaking in ASC. In addition, B cell–specific genes gained H3K27me3 in their promoters as ASC differentiation progressed, indicating that EZH2 may repress these genes. Following immunization with the T-independent Ag, LPS, or infection with influenza virus in the absence of T cells, mice with a tamoxifen-inducible Ezh2 deletion generated fewer ASC. The EZH2-dependent defect was cell intrinsic to B cells and resulted in the enhanced expression and increased chromatin accessibility of B cell genes that gain H3K27me3 and are normally repressed in ASC, including Blimp-1 target genes and inflammatory genes. Ezh2-deficient ASC failed to upregulate oxidative phosphorylation and glycolysis pathways, as well as the unfolded-protein response (UPR), resulting in decreased secreted Ig. Together, these data demonstrate a critical role for EZH2 in the programming of T-independent B cell differentiation and define specific roles for EZH2 in regulating ASC function.
Materials and Methods
Ezh2fl/fl (022626; JAX) (31), Rosa26CreERT2 (08463; JAX) (32), CD19Cre (06785; JAX) (33), CD45.1 (002014; JAX), C57BL/6J (00664; JAX), and μMT mice (02288; JAX) (34) were purchased from The Jackson Laboratory and bred on site. CD45.2 μMT mice were bred onto the CD45.1 background. All animal protocols were approved by the Emory Institutional Animal Care and Use Committee. Mice used for experiments were between 6.5 and 10 wk old and were age and gender matched. Cre-mediated deletion was induced by the treatment of 100 μl of 40 mg/ml tamoxifen (J63509; Alfa Aesar) by daily i.p. injection for five consecutive days. For LPS experiments, 50 μg (ALX-581-008; Enzo Life Sciences) was administered i.v. and mice were analyzed 3 d after inoculation. For mixed bone marrow chimera experiments, 10 × 106 bone marrow cells from Ezh2fl/+CD45.1/2 and Ezh2fl/flRosa26CreERT2/+CD45.2 were mixed at 1:1 ratios, transferred to lethally irradiated CD45.1 hosts, and the immune system allowed to reconstitute for 6 wk. For wild-type cell division experiments, 20 × 106 splenic CD45.1 B cells were labeled with CFSE and adoptively transferred into CD45.2 μMT hosts. For competitive cell division experiments, 10 × 106 splenic B cells from Ezh2fl/+CD45.1/2 and Ezh2fl/flRosa26CreERT2/+CD45.2 mice were mixed at a 1:1 ratio, labeled with CellTrace Violet (CTV), and adoptively transferred to CD45.1 μMT hosts.
CD4+ T cell depletion and influenza infection
For depletion of CD4+ T cells, mice were treated with 200 μg anti-mouse CD4 (GK1.5, BE0003-1; Bio X Cell) by i.p. injection 3 and 1 d before infection. Mice were infected intranasally with 15,000 viral focal units of A/PR8/34 influenza virus (PR8) and analyzed 7 d later.
Magnetic enrichment procedures
B cells were enriched from splenocytes from naive mice by negative selection using CD43 microbeads (130-090-862; Miltenyi Biotec). LPS-induced ASC were enriched by positive selection of CD138+Supplemental Fig. 1A, 1B). For FACS sorting of cellular division samples in Fig. 1: following 3 d LPS inoculation, adoptively transferred CD45.1 cells were stained with CD45.1-PE (12-0453-82, clone A20; Bioscience) and immunomagnetic enrichment was performed using anti-PE microbeads (130-048-801; Miltenyi Biotec) prior to FACS sorting as previously described (8).
Flow cytometry and cell sorting
For staining, cells were resuspended at 1 × 106 cells/100 μl in FACS buffer (1× PBS, 1% BSA, 2 mM EDTA) and blocked with anti-Fc (2.4G2; Tonbo Biosciences) at a concentration of 0.25 μg/1 × 10635). Cells were stained for 30 min on ice, protected from light, and fixed with 1% paraformaldehyde. Intracellular staining was performed with the Fixation/Permeabilization Kit (555028; BD Biosciences) following the manufacturer’s protocol. Flow cytometry was performed on a BD Biosciences LSR II Flow Cytometer using FACSDiva (version 6.2). Flow cytometry data were processed by FlowJo (version 9.9.6). Sorting of nB and ASC by FACS was performed on a BD FACSAriaII at the Emory Flow Cytometry Core. The following gating strategy preceded all flow cytometry analyses presented. Cells were gated on 1) lymphocytes (forward light scatter [FSC]–area × side scatter [SSC]–area), 2) singlets (FSC-width × FSC-height and SSC-width × SSC-height), and 3) live cells (Viability Dye−). Finally, non–B cell lineage cells were removed from the analyses based on the presence of Thy1.1, F4/80, and CD11c.
DNA was extracted from purified splenic B cells following tamoxifen injection using the DNeasy Blood and Tissue kit (Qiagen). A total of 45 ng was used in a 35 cycle PCR with a three-primer design that amplified either the wild-type or deleted alleles. PCR products were resolved on a 1.5% agarose gel. The following primers were used: Ezh2.del-fwd 5′-GCTAGGCCTGCTGGTAAATA-3′, Ezh2-del-rev 5′-AGGAAATGGCAGGGTCTTTAG-3′, and Ezh2-del 5′-CAGTACAATCTCCTGTGTC-3′.
ELISA plates (52-333801301F; Evergreen Scientific) were coated with 5 μg/ml capture Ab (5300-01B; Southern Biotech) or 1 μg/ml influenza HA at 4°C overnight and blocked with 1% nonfat dry milk at room temperature for 2 h. Standard Abs and serum were bound to the plates at 4°C overnight, plates were washed, and secondary HRP-conjugated goat anti-mouse IgM or IgG Abs were added for 2 h at room temperature. Plates were developed with TMB ELISA peroxidase substrate (800-666-7625; Rockland) and the reaction was stopped with 0.2 M sulfuric acid. Plates were read at 450 nm with the Gen5 software.
Quantitative RT-PCR analysis
Total RNA was extracted from B cells and ASC using the RNeasy Mini Kit (Qiagen) and cDNA was synthesized with SuperScriptII Reverse Transcriptase (Invitrogen) as described (36). Real-time PCR was performed using SybrGreen incorporation on a BioRad CFX96 Thermocycler measuring the deleted exon of Ezh2 (Ezh2-del-fwd 5′-CAGGATGAAGCAGACAGAAGAG-3′ and Ezh2-del-rev 5′-TTGTTGCCCTTTCGGGTT-3′) and normalized to 18S rRNA (18s-fwd 5′-GTAACCCGTTGAACCCCATT-3′ and 18s-rev 5′-CCATCCAATCGGTAGTAGCCG-3′).
Tamoxifen-treated Ezh2fl/fl (control [Ctl]) and Ezh2fl/flRosa26CreERT2/+ (ERT2-Ezh2KO) CD138+ ASC were magnetically enriched 3 d following LPS inoculation. Three independent replicates were generated for Ctl and KO ASC. RNA was isolated using the RNeasy Mini Kit (Qiagen) and sequencing libraries were generated using the mRNA HyperPrep Kit with poly(A) selection beads (KAPA Biosystems) using 500 ng total RNA as input according to the manufacturer’s instructions. Final libraries were quality checked on a bioanalyzer, quantitated by quantitative PCR (qPCR), pooled at equimolar ratio, and sequenced on a HiSeq2500 using paired-end, 50-bp sequencing chemistry. Raw fastq files were mapped to the mm9 version of the mouse genome using TopHat2 (37) (version 2.0.13) with the University of California, Santa Cruz mm9 knownGene table (38) as the reference transcriptome. PCR duplicates were removed from all downstream analyses with Picard (http://broadinstitute.github.io/picard/). Reads that overlapped exons were summarized into unique EntrezID genes using the GenomicRanges (39) (version 1.22.4) package in R/Bioconductor. Genes that were not expressed at one read per million in at least three samples were discarded for low expression. Differential expression was tested using a pairwise test in edgeR (40) (version 3.12.1). For gene set enrichment analysis (GSEA), all detected genes were ranked by multiplying the sign of the fold change (FC) (+/−) by the −log10 of the p value. This ranked gene list was used as input for the GSEA preranked analysis. To determine the genes involved in protein secretion and transport, the UPR and XBP1 up gene lists were annotated using the Gene Ontology Consortium web portal (41). Genes categorized into protein transport (GO:0015031), protein localization (GO:0008104), or Golgi vesicle transport (GO:0048193) biological processes were annotated. Sequencing depth for each gene was normalized to fragments per kilobase per million using custom scripts implemented in R/Bioconductor.
Preparation of Tn5 for the assay for transposase-accessible chromatin sequencing
In-house purification of adapter-loaded Tn5 transposase was performed as previously described (42). Briefly, the pTXB1-Tn5 plasmid (60240; Addgene) was transformed into C3013 cells (C3013l; NEB) and a single colony grown in 250 ml of Luria Broth, Miller with 100 μg/ml ampicillin at 37°C until the culture reached an A600 of 0.75–0.9. To induce Tn5 expression, isopropyl β-d-1-thiogalactopyranoside was added to a final concentration of 0.25 mM for 4 h at 23°C. Cells were then pelleted at 2800 rpm for 15 min, resuspended in 15 ml HEGX (20 mM HEPES-KOH at pH 7.2, 0.8 M NaCl, 1 mM EDTA, 0.2% Triton X-100, 10% glycerol, complete with Roche protease inhibitors), and lysed with a French pressure cell (9000 lb/in2). The resulting lysate was pelleted at 15,000 rpm for 30 min, and 550 μl of 10% polyethyleneimine was added to the supernatant on a magnetic stirrer. The precipitate was removed via centrifugation at 12,000 rpm for 10 min. The supernatant was then loaded onto a 1-ml chitin column (S6651S; NEB) and washed with 20 ml HEGX. On-column loading of Tn5 with preannealed mosaic end double-stranded (MEDS) oligonucleotides was achieved by adding 60 nmol of mixed and annealed Tn5 MED-p7/-p5 (ME-p7, 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3′; ME-p5, 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3′; and ME-rev, 5′-[phos]-CTGTCTCTTATACACATCT-3′) to the column in 1.2 ml HEGX buffer. After 48 h, the column was washed with 20 ml of HEGX to remove unbound MEDS and cleavage of Tn5-MEDS was initiated by adding 1.2 ml HEGX with 100 mM DTT to the column. After 48–72 h, small elution fractions were collected and those with the highest protein concentration were pooled and dialyzed overnight versus 2× Tn5 dialysis buffer (100 mM HEPES at pH 7.2, 0.2 M NaCl, 0.2 mM EDTA, 2 mM DTT, 0.2% Triton X-100, 20% glycerol). Following dialysis, 0.6 vol of glycerol was added to make the final storage buffer (50 mM HEPES at pH 7.2, 0.1 M NaCl, 0.1 mM EDTA, 1 mM DTT, 0.1% Triton X-100, 50% glycerol). Tn5 was dispensed into 50-μl aliquots and placed at −80°C for long-term storage.
Assay for transposase-accessible chromatin sequencing
Tagmentation was performed as described in detail previously (43, 44). Briefly, 2000–4000 cells from Ezh2fl/fl (Ctl) and ERT2-Ezh2KO nB and ASC from tamoxifen-treated bone marrow chimeras 3 d following LPS inoculation were FACS sorted and tagmentation was performed using 2.5 μl of Tn5 in 1× TD Buffer (Illumina) in 25 μl total volume for 1 h at 37°C. Tagmented nuclei were lysed, DNA was purified using a double solid phase reversible immobilization-bead size selection (0.7× negative followed by 1× positive selection), and PCR amplified using Nextera Indexing Primers (Illumina) and the HiFi HotStart Polymerase (KAPA Biosystems) for 14 cycles of PCR. Final libraries were purified using a second double SPRI-bead size selection (0.2× negative followed by 1× positive selection), quantitated using the Illumina qPCR Quant Kit (KAPA Biosystems), and sequenced using 50-bp, paired-end chemistry on a HiSeq2500.
Assay for transposase-accessible chromatin using sequencing data analysis
Raw reads from the assay for transposase-accessible chromatin sequencing (ATAC-seq) were mapped to the mm9 version of the mouse genome using Bowtie (45) (version 1.1.1) with the default parameters. PCR duplicates were removed from all downstream analyses with Picard (http://broadinstitute.github.io/picard/). Enriched regions were identified using MACS2 (46) (version 220.127.116.1140616). A composite list of all identified peaks in any sample was generated using the HOMER (47) (version 4.8.2) “mergePeaks” function. The read depth for each sample was then annotated for all peaks and differential accessibility was calculated using the GLM function of edgeR (40) (version 3.12.1), controlling for the mouse each sample originated from. Principal component (PC) analysis plots were generated using all differentially accessible regions (DAR) and the vegan (version 2.4-3) package in R/Bioconductor. Genomic annotation of peaks was performed using HOMER (47) (version 4.8.2). Peaks that overlapped a promoter (+500 and −2000 bp surrounding the transcription start site) were k-means clustered using the Biganalytics package (//CRAN.R-project.org/package=biganalytics). Enriched transcription factor motifs were identified using HOMER (47) (version 4.8.2) and the findMotifsGenome.pl script. Histograms of read depth surrounding each motif were generated using custom R/Bioconductor scripts.
ChIP was performed as described previously (48). For each immunoprecipitation, 10 × 106 CD43− splenic B cells or 1 × 106 CD138+ Hoxa9 (positive) and Actb (negative) loci was tested by qPCR. The remaining ChIP–DNA was used as input for the KAPA HyperPrep Kit (KAPA Biosystems). ChIP-seq libraries were sequenced on a HiSeq2500 using 50-bp, paired-end chemistry. Raw sequencing reads were mapped to the mm9 version of the mouse genome using Bowtie (45). Uniquely mappable and nonredundant reads were used for subsequent analyses. HOMER (47) software was used for peak calling and annotation. Data were normalized to reads per peak per million (rppm) as previously described (44) using Eq. 1:For the generation of scatterplots between nB and ASC samples the rppm read depth was quantile normalized and the log2FC and the average log2 rppm between nB and ASC samples were calculated.
All sequencing data are available at the National Center for Biotechnology Information Gene Expression Omnibus database (//www.ncbi.nlm.nih.gov/geo/) under the accession number GSE103195. All code and data processing scripts are available upon request.
Extracellular flux assays
A Seahorse Bioanalyzer XFe96 instrument was used for all extracellular flux assays. A FluxPak cartridge was hydrated at least 12 h prior to running each assay with 200 μl dH2O in a 37°C non-CO2 incubator. One hour prior to each assay, dH2O was removed and 200 μl prewarmed Seahorse Calibrant solution (103059-000; Agilent) was added to all experimental wells. For the mitochondrial stress test and measurement of oxygen consumption, purified cell populations were washed in Seahorse XF Assay Media, pH 7.4 ± 0.1, supplemented with 1 mM sodium pyruvate, 2 mM l-glutamine (G7513; Sigma), and 5.5 mM glucose at 37°C. For the glycolysis stress test and extracellular acidification measurements, purified cell populations were washed in Seahorse XF Assay Media, pH 7.4 ± 0.1, supplemented with 2 mM l-glutamine. Cells were washed once in appropriate media and counted by flow cytometry using AccuCheck counting beads (PCB100; Invitrogen). Prior to each experiment, CellTak (354420; Corning) was diluted in sterile 1× PBS to a final concentration of 22.4 μg/ml, and 25 μl was added to each well of a Seahorse XFe96 cell culture plate, incubated for 20 min at room temperature. The plates were washed with 200 μl dH2O and then 400,000 cells per well were plated and incubated in a 37°C non-CO2 incubator for 45 min prior to beginning the assay. The indicated drugs were diluted in assay-specific media for injection into each port. For the mitochondrial stress test the ports were as follows: port A, oligomycin (75351; Sigma) was used at a final concentration of 1 μM; port B, carbonyl cyanide-4-(triflurome-thoxy) phenylhydrazone (C2920; Sigma) was used at a final concentration of 2.5 μM; and port C was injected a combination of Rotenone (R8875; Sigma) and Antimycin A (A8674; Sigma), each at a final concentration of 1 μM. For the glycolysis stress test the ports were as follows: port A, glucose (G8270; Sigma) was used at a final concentration of 10 mM; port B, oligomycin was used at a final concentration of 1 μM; and Port C, 2-deoxyglucose (D8375; Sigma) was used at a final concentration of 50 mM.
EZH2 is progressively upregulated during type-I, T-independent B cell differentiation
Ezh2 is progressively upregulated during ex vivo B cell differentiation in response to LPS (49). To assess its expression pattern in vivo, we transferred CFSE-labeled CD45.1 B cells into CD45.2 μMT mice and subsequently inoculated the recipients with LPS to induce differentiation (8). Distinct divisions (Div0, -1, -3, -5, -8+) were FACS isolated and gene expression was quantitated by quantitative RT-PCR (RT-qPCR). For consistency across the differentiation models, we define plasma cells that have acquired CD138 expression as ASC. Ezh2 mRNA was upregulated during the divisions, peaking in CD138+ ASC at Div8+ (Fig. 1A). EZH2 protein levels were quantified in splenic nB and 3 d after LPS stimulation in B220+GL7+CD138− activated B cells (actB) and CD138+ ASC. Similar to previous reports (15), EZH2 protein was relatively low in nB, elevated in actB, and peaked in ASC (Fig. 1B).
To determine what genes might be affected by EZH2-mediated histone methylation, we performed ChIP-seq for H3K27me3 in nB and ASC, which were enriched to >94% purity (Supplemental Fig. 1A, 1B). We found that 1623 gene promoters gained H3K27me3 in ASC (Fig. 1C), suggesting that these were targets of EZH2-mediated repression. Gene Ontology functional analysis, summarized and condensed by REVIGO (50) (Fig. 1D), indicated that these genes were enriched in pathways associated with B cell functions that are repressed in ASC, such as cell cycle (e.g., Cdkn1a), response to LPS (e.g., Nfkb1 and Tnf), and Ag presentation by MHC (e.g., H2-Ab1 and Cd74). Thus, these data indicate that the progressive upregulation of Ezh2 during B cell differentiation coincides with repression of the B cell transcriptional program in ASC.
ASC differentiation in response to type I, t-independent stimuli requires EZH2
To determine the role of Ezh2 in vivo, we crossed Ezh2fl/fl mice (31) with tamoxifen-inducible Rosa26CreERT2 mice (32). Following tamoxifen treatment, we observed genomic rearrangement of the Ezh2 locus in splenic B cells and a significant reduction in Ezh2 transcripts, and protein was observed (Supplemental Fig. 1C–E). Ezh2 is essential for B cell development (26, 51); however, it is dispensable for peripheral B cell homeostasis (26). Consistent with previous findings, following tamoxifen-induced deletion of Ezh2, a block in B cell development at the pre-B stage was observed (Supplemental Fig. 1F). However, in the short time frame following deletion, no defect was observed in the viability of peripheral splenic B cells or in the frequency of marginal zone or follicular B cells (Supplemental Fig. 1G–I).
To test the function of EZH2, ERT2-Ezh2KO and Ezh2fl/fl (Ctl) mice were treated with tamoxifen and subsequently inoculated with LPS to induce B cell differentiation. In ERT2-Ezh2KO mice, we observed a significant reduction in the frequency and absolute numbers of splenic and lymph node ASC and actB (Fig. 2A–E). These differences were not due to changes in the viability of ERT2-Ezh2KO splenocytes (Fig. 2G). Following LPS inoculation, a sharp increase in IgM titers occurred in control mice; however, ERT2-Ezh2KO mice failed to reach the same titers of serum IgM (Fig. 2H), a reduction that correlated with the decreased ASC response. IgG levels were also measured by ELISA and showed a similar reduction in ERT2-Ezh2KO mice; however, the overall levels of IgG were very low in response to LPS (data not shown).
The expression of Cre recombinase can negatively impact rapidly proliferating cells (52). To ensure that the observed phenotype was specific to deletion of Ezh2, we compared hemizygous Ezh2fl/+Rosa26CreERT2/+ (ERT2-Cre Ctl) mice with ERT2-Ezh2KO and the Cre− control mice described above. Three days after LPS inoculation, ERT2-Cre Ctl mice had similar frequencies of splenic actB and ASC as control mice and significantly more of each population than ERT2-Ezh2KO mice (Supplemental Fig. 2A, 2B). To further assess this system, the Cre recombinase expressed from the Cd19 locus was used (33) because it bypasses the developmental defects observed in the ERT2-Ezh2KO strain, allowing normal B cell development to occur (26) and facilitating B cell–specific deletion. In this study, Ezh2fl/flCD19Cre/+ (CD19-Ezh2KO), CD19Cre/+ (CD19-Cre Ctl), and Ezh2fl/fl (Ctl) mice were inoculated with LPS and differentiation tested as above. Similar to tamoxifen-induced Ezh2 deletion, CD19-Ezh2KO mice formed 50% fewer ASC and actB compared with controls (Fig. 2I–L, Supplemental Fig. 2C, 2D). Together, these data demonstrated that Ezh2 was required for efficient ASC differentiation in response to type-I, T-independent stimuli.
EZH2 controls the early T-independent burst of influenza-specific ASC
During the humoral response to protein Ags and pathogens, a burst of ASC provides an initial surge of serum Abs before T cell–dependent processes facilitate affinity maturation and class-switched B cell responses (53, 54). In response to the hapten NP, EZH2-deficient B cells displayed reduced NP-specific serum Ab titers as early as day 7 (15), suggesting defects in early ASC responses. To assess the early T cell–independent differentiation events, CD4 T cells were first depleted from CD19-Ezh2KO and control cohorts of mice using an anti-CD4 Ab and subsequently infected with PR8. CD4 T cells were fully depleted at both the time of PR8 infection and the assay end point at day 7, at which no GC B cells were detected (Supplemental Fig. 3A, 3B). At day 7, control mice exhibited expanded lymph node CD138+ ASC, whereas the frequency and absolute number of ASC were significantly reduced in CD19-Ezh2KO mice (Fig. 3A, 3B). Ag-specific, PR8-responding cells were identified using B cell tetramers against HA (35), and the frequency of HA+ ASC, HA+ actB, and total HA+ cells was significantly reduced in CD19-Ezh2KO compared with control mice (Fig. 3C–G, Supplemental Fig. 3C, 3D). Additionally, assessment of Ab levels revealed an increase in HA-specific IgM titers from day 0 to 7 in control mice but significantly reduced levels in CD19-Ezh2KO at mice at day 7 (Fig. 3H). Together, these data define an essential role for EZH2 in the formation of T-independent ASC in response to both nonprotein and protein Ags.
Cell-intrinsic requirement for EZH2 in ASC differentiation
To determine a cell-intrinsic role for Ezh2 in ASC differentiation, bone marrow from CD45.2 ERT2-Ezh2KO and CD45.1/2 control mice were mixed 1:1 and transferred into lethally irradiated CD45.1 hosts. At 4 wk, the percentages of CD45.2 and CD45.1/2 donor cells in the periphery of recipient mice were similar (Fig. 4A). After 6 wk, mice were treated with tamoxifen (as above) to induce EZH2 deletion and subsequently inoculated with LPS. Three days after LPS, both ERT2-Ezh2KO (CD45.2) and control cells were identified (Fig. 4B) and a significant reduction in CD138+ ASC in the spleen (Fig. 4C) and lymph nodes (Fig. 4D) was observed in cells transferred from ERT2-Ezh2KO mice. Additionally, the frequency of actB was significantly reduced (Fig. 4E, 4F). These data indicate that EZH2 performs an essential cell-intrinsic function during LPS-induced B cell differentiation.
EZH2 represses B cell transcription factor networks
Transcriptome profiling was performed on tamoxifen-treated ERT2-Ezh2KO or control ASC following LPS stimulation to determine the molecular consequences of Ezh2 deletion. In ERT2-Ezh2KO ASC, 1498 genes were significantly upregulated (false discovery rate [FDR] < 0.05, log2FC > 1) whereas only 79 were downregulated (Fig. 5A). Although not entirely, these data are largely consistent with a repressive function for EZH2. Confirming the tamoxifen-induced deletion in ERT2-Ezh2KO ASC, an 18-fold reduction in the transcripts across Ezh2’s deleted exons was observed compared with wild type (Supplemental Fig. 4A). Moreover, in ERT2-Ezh2KO ASC, Ezh2 was the most significantly downregulated gene by an FDR differential of 10113. GSEA (55, 56) indicated that, in the absence of Ezh2, genes involved in the regulation of inflammation and NF-κB–mediated TNF-α signaling and the p53 pathway were upregulated (Fig. 5B). During B cell development, EZH2 repressed the cell cycle inhibitor Cdkn2a and prevented p53 activation (51); thus, EZH2 may be performing a similar role in ASC.
To determine if the differentially expressed genes (DEG) were direct targets of EZH2, we compared the change in promoter H3K27me3 enrichment between nB and ASC with the change in expression between ERT2-Ezh2KO versus control ASC. We found that 923 genes were upregulated in ERT2-Ezh2KO ASC and gained promoter H3K27me3 compared with nB (Fig. 5C, 5D), indicating that these DEG are direct repression targets of EZH2. Bcl6 was among the top genes that gained promoter H3K27me3 in wild-type ASC and was increased in expression in ERT2-Ezh2KO ASC, suggesting that other B cell transcription factors could be dysregulated. GSEA analysis was performed using a set of transcription factors specifically expressed in follicular B cells compared with bone marrow ASC (57) and showed that this set of transcription factors were significantly upregulated in ERT2-Ezh2KO ASC (Fig. 5B). Other examples of genes that were significantly upregulated in ERT2-Ezh2KO ASC and demonstrated extensive gains in H3K27me3 in ASC included the developmental transcription factor Klf4, inflammatory genes Ifit3 and Lta, and cell cycle inhibitors Cdkn1a and Slfn1 (Fig. 5G, Supplemental Fig. 4B). These results indicate that EZH2 functions to epigenetically silence key B cell genes involved in inflammatory responses.
EZH2 controls chromatin accessibility in ASC
The derepression of 923 genes could be directly attributed to a failure to gain H3K27me3 in ASC. However, an additional 575 DEG gained expression and were marked by H3K27me3 in both nB and ASC, suggesting they are normally repressed in both cell types. To further understand the consequences of Ezh2 deletion, we performed ATAC-seq (43, 44) on nB and ASC from ERT2-Ezh2KO and control mice. PC analysis of all DAR revealed a large difference between nB and ASC regardless of EZH2 status defined by PC1 (Supplemental Fig. 4C), indicating that in the absence of EZH2, most differentiation-associated chromatin accessibility changes occurred normally. PC2 separated control and ERT2-Ezh2KO ASC but not nB, which contained minimal DAR (Supplemental Fig. 4D), indicating chromatin accessibility differences were primarily specific to Ezh2-deficient ASC.
Promoter accessibility of genes upregulated in ERT2-Ezh2KO ASC was annotated in nB and ASC and the patterns were categorized by k-means clustering. Three distinct patterns of promoter chromatin accessibility were observed: k1, promoters that lost accessibility between nB and ASC; k2, promoters that were minimally accessible in nB, remained such in control ASC yet gained accessibility in ERT2-Ezh2KO ASC; and k3, promoters that were inaccessible in nB and control ASC and only gained accessibility in ERT2-Ezh2KO ASC (Fig. 5E). We found that promoters of DEG in ERT2-Ezh2KO ASC were more accessible than those in control ASC. Direct comparison with control ASC revealed 1317 loci that gained accessibility in ERT2-Ezh2KO ASC compared with only 347 that decreased (Fig. 5F). Examples of chromatin accessibility changes include specific regions located in Bcl6 and Klf4 (k1), Hoxa1 (k2), and Eml6 (k3) (Fig. 5G).
The failure to repress key B cell transcription factors suggested that such factors may contribute to a unique accessibility footprint in ERT2-Ezh2KO ASC. Analysis of DNA sequence motifs within regions of increased accessibility revealed an enrichment for CTCF, ETS, OCT, and IRF family transcription factor binding sites (Fig. 6A). Although CTCF was expressed at similar levels in nB and ASC, the presence of CTCF sites within DAR suggests a potential defect in rearranging the three-dimensional genome during ASC differentiation (58). Transcription factor footprinting for the occurrence of each motif revealed unique patterns at each binding site and clear increases in accessibility surrounding each motif in ERT2-Ezh2KO compared with control ASC (Fig. 6A). Increased accessibility could be due to an increase in expression of a transcription factor family member that binds to a specific motif, such as the ETS factor SPIB that is upregulated, or due to a failure to recruit EZH2 to a region. Thus, these data depict the epigenetic derepression of loci controlled by EZH2 through increases in gene expression and chromatin accessibility at sites normally repressed in ASC.
Blimp-1–repressed genes are upregulated in the absence of EZH2
Blimp-1 can elicit gene repression through physical interactions with epigenetic modifiers (59–61), including EZH2 (30). GSEA was performed using a high-confidence set of Blimp-1–repressed target genes derived from transcriptional profiling of genetic deletions and ChIP-seq data (30). Of the 109 genes that overlapped in the data sets, 96% demonstrated higher expression in the absence of EZH2 (Fig. 6B). Examples included the ETS family transcription factor gene Spib, which was upregulated and contained one of the strongest Blimp-1 binding sites (Fig. 6C, 6D). Additionally, genes for Klf2, a transcription factor involved in B cell homing and migration (62); the TLR Tlr9; and Btg1, which functions as a negative regulator of cellular proliferation and apoptosis (63, 64); all contained Blimp-1 binding sites at regions that gained H3K27me3 in ASC and were upregulated in the absence of EZH2. These data provide an in vivo link between EZH2 and Blimp-1 in ASC and demonstrate that epigenetic modifications via EZH2 may be required for the Blimp-1–mediated gene repression program.
Deletion of Ezh2 impairs the proliferation of actB
Recent studies in B cells and B cell–derived lymphomas demonstrated that inhibition or depletion of EZH2 resulted in cell cycle arrest, impaired proliferation, and apoptosis (16, 28, 65–67). Indeed, a cell cycle gene set was one of the most downregulated gene sets in ERT2-Ezh2KO ASC (Fig. 5B). Additionally, the EZH2 target gene Cdkn1a (16), which encodes the P21 cell cycle inhibitory kinase, is upregulated in ERT2-Ezh2KO ASC (Supplemental Fig. 4B). These data suggest that the observed decrease in actB and ASC was due to an EZH2-mediated defect in the proliferative phase of B cell differentiation. To test this hypothesis, tamoxifen-treated CD45.2 ERT2-Ezh2KO and CD45.1/2 control B cells were labeled with CTV, transferred into congenically marked CD45.1 μMT hosts, and stimulated to differentiate by inoculation with LPS. After 3 d, ERT2-Ezh2KO and control B cells were analyzed for their ability to divide and differentiate (Fig. 7A). Consistent with previous results (8), eight cellular divisions were typically observed with control B cells (Fig. 7B, 7C). In contrast, ERT2-Ezh2KO B cells initially proliferated normally through three cellular divisions but were reduced in number at all subsequent divisions. The B cell activation marker GL7 was progressively upregulated in control cells through eight divisions. However, ERT2-Ezh2KO B cells failed to gain additional GL7 expression after three divisions (Fig. 7D, 7E). Analysis of differentiation kinetics through expression of CD138 revealed that B cells from control and ERT2-Ezh2KO mice both achieved seven to eight divisions before an accumulation of CD138+ ASC were observed, however the ERT2-Ezh2KO CD138+ cells were reduced in frequency (Fig. 7F, 7G). These data indicated that in the absence of EZH2, a defect in the ability to proliferate, possibly through the failure to repress negative regulators of the cell cycle, contributes to the reduced numbers of differentiated ASC.
EZH2 is required for the metabolic programming of ASC
Although a subset of ERT2-Ezh2KO B cells could undergo eight rounds of division and differentiate into ASC, the molecular phenotype of these cells indicated that they were likely dysfunctional (Fig. 5). To directly measure the Ab-secreting capacity of these cells, control and ERT2-Ezh2KO ASC—derived following LPS-induced differentiation in vivo—were purified, equal numbers cultured for 3.5 h, and Ab titers from the resulting supernatant measured by ELISA. On a per-cell basis, ERT2-Ezh2KO ASC secreted 50% fewer molecules of IgM (Fig. 8A). However, no difference in the levels of IgM mRNA transcripts (Fig. 8B) or intracellular IgM protein levels (Fig. 8C) were observed between control and ERT2-Ezh2KO ASC, indicating that the IgM deficiency may lie in protein secretion.
Consistent with this idea, GSEA indicated that the UPR and genes upregulated by the transcription factor XBP1 (68) failed to be induced in ERT2-Ezh2KO ASC (Fig. 8D). Annotation of genes involved in protein secretion and transport revealed that these genes were more highly expressed in control than in ERT2-Ezh2KO ASC, consistent with decreased IgM levels in the cultures of ERT2-Ezh2KO ASC. XBP1-mediated induction of the UPR pathway is also important for increased mitochondrial function (69), and ASC require a high metabolic capacity to support synthesis and secretion of Ab molecules (70). Indeed, GSEA revealed the expression of oxidative phosphorylation and glycolysis metabolic pathways were enriched in control but not ERT2-Ezh2KO ASC (Fig. 8E). Moreover, EZH2 regulates the mTOR pathway in follicular lymphoma (71), suggesting the reduced IgM secretion in ASC may be due to a metabolic defect.
To test the hypothesis that ERT2-Ezh2KO ASC were metabolically distinct, extracellular flux assays were performed to probe both mitochondrial respiration and glucose metabolism. The oxygen consumption rate was measured as a readout for mitochondrial respiration. At basal levels, ERT2-Ezh2KO ASC consumed significantly less oxygen than control ASC (Fig. 8F). Following inhibition of ATP synthase with oligomycin, similar declines were observed in oxygen consumption. Maximal respiration rates were also greater in control ASC compared with ERT2-Ezh2KO, as shown by the addition of carbonyl cyanide-4-(triflurome-thoxy) phenylhydrazone to the system. This suggests that control ASC are more capable of using the oxidative phosphorylation metabolic pathway than ERT2-Ezh2KO. Because metabolism is a balance between multiple metabolic pathways and the expression of the glucose transporter Glut1 was upregulated after LPS stimulation ex vivo (72), the ability of ASC to metabolize glucose was assessed. The extracellular acidification rate, which is measured by pH changes associated with excretion of lactic acid generated from pyruvate (73), can be used to measure rates of glucose metabolism. Addition of glucose to glucose-deprived ASC resulted in enhanced lactate secretion in control compared with ERT2-Ezh2KO ASC, resulting in a significantly higher ability to perform glycolysis (Fig. 8G). These data therefore define EZH2 as an essential upstream regulator of ASC metabolic potential.
In this article, we tested whether EZH2 regulated ASC differentiation using both type-I and type-II T-independent B cell responses in vivo. Consistent with other data (26), we found that EZH2 is not required for homeostasis of naive follicular and marginal zone B cells. However, immunization with T-independent Ags or protein Ags in the absence of T cell help generated poor B cell responses in Ezh2-deficient mice. The role of EZH2 was B cell intrinsic as Ezh2-deficient B cells also poorly differentiated into ASC in mixed bone marrow chimeric mice. As EZH2 is the continued target of therapeutic intervention, these data have implications for the effect of inhibitors on normal humoral immune responses.
The poor differentiation of Ezh2-deficient ASC could be due, in part, to a defect in the ability of Ezh2-deficient ASC to upregulate genes involved in the cell cycle. In fact, the top upregulated gene sets in Ezh2-deficient ASC were inflammatory and p53 pathways, with each consisting of genes that function to negatively regulate the cell cycle. Among the most induced genes within these sets was Cdkn1a, which encodes the G1/S cell cycle checkpoint inhibitor P21 (74). ChIP-seq data showed that Cdkn1a has increased H3K27me3 across its promoter and proximal upstream sequences, confirming a previously described direct role for EZH2 in its regulation (15, 16, 28). In vivo, B cells undergo approximately eight cell divisions before differentiating into ASC (8). Ezh2-deficient cells went through fewer divisions, resulting in fewer cells that acquired the actB phenotype (GL7+) and fewer CD138+ ASC, consistent with the continued activity of cell cycle inhibitors. This result is consistent with findings in multiple myeloma in which EZH2 is overexpressed and pharmacological inhibition limits cell growth (75, 76). Thus, the failure to repress cell cycle inhibitors, such as Cdkn1a, is a likely mechanism leading to decreased numbers of observed ASC.
Ezh2-deficient ASC also failed to repress the transcriptional program associated with mature B cells, including the accumulation of H3K27me3 at genes, such as Ciita which functions in MHC class II Ag processing, B cell transcription factors like Bcl6, as well as NF-κB–mediated inflammatory genes. Transcriptome profiling of discrete divisions in vivo following B cell activation with LPS revealed that the third cell division following activation was the first stage in which gene repression occurred. This division also coincided with the downregulation of NF-κB–regulated genes (8). Ezh2-defient cells displayed a proliferation defect at division three as well, suggesting that this stage may be an initial differentiation checkpoint requiring the repression of a set of genes to continue the process.
The failure to repress B cell–associated transcriptional programs was also associated with the enrichment of transcription factor binding motifs in regulatory regions of those genes. For example, in Ezh2-deficient ASC, CTCF and the ETS family transcription factor binding motifs were significantly more accessible. The chromatin landscape associated with CTCF within the MHC class II region is reorganized during B cell differentiation in response to LPS (58), and CTCF is known to function in the maintenance of GC B cells (77). It is possible that this architecture may not be properly organized in the absence of EZH2. The most differentially regulated ETS family member was Spib, which is normally silenced in ASC through direct repression by Blimp-1 (30, 78). Furthermore, genes normally repressed in both B cells and ASC and marked by H3K27me3, including Hoxa1 and Eml6, failed to remain repressed. Of note, the classification of promoter accessibility changes of genes that failed to be repressed by EZH2 revealed a category of promoters that lost accessibility normally in ASC. This indicates that chromatin accessibility data derived from ATAC-seq may not correlate with failed epigenetic processes at all loci, and that additional epigenetic mechanisms are required for full repression of the B cell transcriptome.
Deficiencies in proliferation and repression of B cell fate accounted for reduced ASC numbers; however, Ezh2-deficient ASC secreted less IgM on a per cell basis, indicating that ASC-specific pathways were dysregulated. From the adoptive transfer proliferation experiment in which splenic B cells were transferred into μMT hosts (Fig. 7), it is likely that the observed EZH2 defect lies in the marginal zone B cell compartment; however, we did not specifically assess the impact of EZH2 deletion on B1 B cells, which secrete IgM in response to LPS (79). ASC have enhanced metabolic potential and upregulate endoplasmic reticulum stress response pathways to facilitate Ab secretion (3, 8, 80). Transcriptome profiling demonstrated that both of these processes were altered in the absence of EZH2 with reduced expression of genes in the UPR, glycolysis, and oxidative phosphorylation pathways. Oxidative phosphorylation is fueled by pyruvate molecules derived from both glycolysis and fatty acid oxidation. In this study only glycolysis was assayed; therefore, we cannot rule out a role for EZH2 in regulating fatty acid metabolism. Ezh2-deficient ASC failed to upregulate XBP-1 target genes, which is required for the induction of the UPR (81). Cellular metabolism is regulated through the PI3K signaling pathway, with the mTORC1 and mTORC2 kinases representing essential nodes (3). ASC require the continued activity of mTORC1 to maintain maximal levels of Ab secretion (82). In the absence of EZH2, the ability of ASC to perform oxidative phosphorylation and metabolize glucose through the glycolysis pathway was reduced and correlated with a reduced ability to secrete Abs. These data converge with recent results describing a role for EZH2-mediated control of mTORC1 in follicular lymphoma (71) and indicate that there is an epigenetic component to the induction and/or maintenance of ASC metabolism. Therefore, Ezh2-dependent and possibly other epigenetic mechanisms are required to program ASC metabolic potential and facilitate Ab secretion.
Metabolic reprogramming following activation in immune cells is necessary for effector functions in both the myeloid and lymphoid lineages, and epigenetic enzymes require cofactors synthesized during metabolism (83). However, it is less clear how the metabolic changes are programmed and inherited through cell fate transitions. Although not essential for their survival, bone marrow ASC require the continued activity of Blimp-1 to sustain the UPR and secrete Abs (68). These results pose that EZH2 may also be continually required for these ASC functions. If ASC require the constant activity of transcription factors and epigenetic enzymes to maintain metabolic potential, this suggests that throughout their lifetime ASC may be able to adapt to nutrient availability and respond to environmental cues. These data therefore identify an Ezh2-dependent epigenetic process that facilitates ASC metabolism.
The authors have no financial conflicts of interest.
We acknowledge the members of the Boss laboratory for scientific contributions and editorial input to the project, the Emory Flow Cytometry Core for FACS expertise, the Emory Integrated Genomics Core for sequencing library quality control, and the New York University Genome Technology Center for Illumina sequencing.
This work was supported by National Institutes of Health Grants 1R01AI123733 to J.M.B.; P01 AI 125180-02 to J.M.B. and T.D.R.; T32 GM0008490 to R.R.H., A.K.K., and B.G.B.; F31 AI112261 to B.G.B.; and F31 1F31 AI131532 to R.R.H.
The sequencing data presented in this article have been submitted to the National Center for Biotechnology Information’s Gene Expression Omnibus (//www.ncbi.nlm.nih.gov/geo/) under accession number GSE103195.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- activated B cell
- activation-induced cytidine deaminase
- Ab-secreting cell
- assay for transposase-accessible chromatin sequencing
- chromatin immunoprecipitation
- ChIP sequencing
- CellTrace Violet
- differentially accessible region
- differentially expressed gene
- ERT2-Cre Ctl
- enhancer of zest 2
- fold change
- false discovery rate
- forward light scatter
- germinal center
- gene set enrichment analysis
- histone H3 lysine 27 trimethylation
- mosaic end double stranded
- naive B cell
- principal component
- A/PR8/34 strain of influenza
- polycomb repressive complex 2
- quantitative PCR
- reads per peak per million
- quantitative RT-PCR
- side scatter
- unfolded-protein response.
- Received October 23, 2017.
- Accepted November 29, 2017.
- Copyright © 2018 by The American Association of Immunologists, Inc.