Cell division is an essential component of B cell differentiation to Ab-secreting plasma cells, with critical reprogramming occurring during the initial stages of B cell activation. However, a complete understanding of the factors that coordinate early reprogramming events in vivo remain to be determined. In this study, we examined the initial reprogramming by IRF4 in activated B cells using an adoptive transfer system and mice with a B cell–specific deletion of IRF4. IRF4-deficient B cells responding to influenza, 4-hydroxy-3-nitrophenylacetyl–Ficoll, and LPS divided but stalled during the proliferative response. Gene expression profiling of IRF4-deficient B cells at discrete divisions revealed IRF4 was critical for inducing MYC target genes, oxidative phosphorylation, and glycolysis. Moreover, IRF4-deficient B cells maintained an inflammatory gene expression signature. Complementary chromatin accessibility analyses established a hierarchy of IRF4 activity and identified networks of dysregulated transcription factor families in IRF4-deficient B cells, including E-box binding bHLH family members. Indeed, B cells lacking IRF4 failed to fully induce Myc after stimulation and displayed aberrant cell cycle distribution. Furthermore, IRF4-deficient B cells showed reduced mTORC1 activity and failed to initiate the B cell activation unfolded protein response and grow in cell size. Myc overexpression in IRF4-deficient cells was sufficient to overcome the cell growth defect. Together, these data reveal an IRF4–MYC–mTORC1 relationship critical for controlling cell growth and the proliferative response during B cell differentiation.
This article is featured in Top Reads, p.
Akey component of the adaptive immune response is the generation of Ab by Ab-secreting plasma cells (ASC). Upon Ag encounter, quiescent naive B cells become activated and rapidly proliferate, and a subset differentiates to ASC. One essential component of B cell differentiation to ASC is cell division (1–4). Culturing purified B cells and blocking cell division prevents the generation of ASC (3). However, the number of cell divisions does not exclusively determine ASC formation. This has led to a stochastic model of differentiation that describes population-level immune responses and accounts for heterogeneity in cell fates among responding cells, such as whether they will continue to divide, die, or differentiate (1, 5–7). One molecular determinant that contributes to such heterogeneity is the expression levels of MYC (4, 8–12). MYC levels are influenced by immune stimulation and serve as a division-independent timer to control the proliferative capacity of responding cells (4, 10, 13). IRF4 is another factor that contributes to heterogeneity at the population level (14–17). During the initial stages of B cell activation, high IRF4 expression biases cells toward the ASC fate (14, 17). Notably, initial IRF4 expression levels are influenced by the intensity of immune stimulation, and IRF4hi cells are among the first to divide (18). Indeed, proliferation is reduced in IRF4-deficient B cells stimulated ex vivo (15, 16, 19); however, the impact of IRF4 on in vivo B cell proliferation is unknown. Furthermore, the timing, scope, and mechanism by which IRF4 contributes to control the proliferative response remains undefined.
Cell division is tightly linked to ASC formation, with transcriptional and epigenetic reprogramming (20–23) occurring as the cells divide (17, 24–26). As such, each cellular division represents a distinct stage during B cell differentiation, with ASC formation occurring after at least eight cell divisions (17, 25, 27, 28). Cell-extrinsic signals can impact the specific division in which differentiation occurs, but the molecular programming events leading to ASC remain the same (17). Many essential ASC programming events (29) are initiated during the early stages of B cell activation and are progressively reinforced in subsequent divisions (24, 25, 27). For example, ASC formation requires a metabolic shift from glycolysis to oxidative phosphorylation (OXPHOS), and the OXPHOS program is increasingly established across cell divisions (24). Additionally, ASC differentiation requires activation of the unfolded protein response (UPR), an essential stress response needed during increased protein production (30, 31). Although canonically considered to be induced in newly formed ASC, recent work indicates that activated B cells (actB) upregulate an array of UPR-affiliated genes. This process is controlled by mTORC1 prior to Ab production and before XBP1 activity (32), a known regulator of UPR in ASC (33, 34). Moreover, single-cell RNA-sequencing (RNA-seq) of actB uncovered an IRF4-dependent bifurcation event that committed a portion of actB to ASC during the early stages of B cell activation (17). Thus, although recent work has highlighted critical early reprogramming events in actB, the timing and extent to which the above factors, and others, remains to be fully understood and integrated.
In this study, we aimed to understand the IRF4-dependent division-coupled reprogramming events that occur during the initial stages of B cell differentiation. Using an in vivo model of B cell differentiation (25), we found that IRF4-deficient B cells begin to divide normally but stall during the proliferative response. To assess the timing and scope of IRF4-dependent reprogramming, IRF4-sufficient and -deficient B cells at discrete divisions were sorted for RNA-seq and the assay for transposase accessible chromatin sequencing (ATAC-seq) (35, 36). RNA-seq revealed that early upregulation of gene sets critical for ASC formation was dependent on IRF4. These included MYC target genes and genes important for OXPHOS. Indeed, IRF4-deficient B cells failed to fully upregulate Myc and displayed altered cell cycle distribution. The activity of mTORC1 was also reduced, resulting in an inability of IRF4-deficient B cells to undergo cell growth and initiate the UPR (32). ATAC-seq identified hundreds of differentially accessible regions (DAR) and established a hierarchy of IRF4 activity, with AP-1:IRF (AICE) motifs active during early divisions and ETS:IRF (EICE) motifs active in later divisions. Together, these data create a road map defining the role of IRF4 during the earliest stages of B cell differentiation in vivo and reveal a critical role for IRF4 in controlling cell growth and maintaining the proliferative response.
Materials and Methods
Mice and adoptive transfers
Cd19Cre (JAX; 006785) (37) and Irf4fl/fl (JAX; 000664) (16) mice were purchased from The Jackson Laboratory and bred to generate Cd19Cre/+Irf4fl/fl. CD45.2 μMT (JAX; 008100) (38) were bred onto the CD45.1 background to obtain CD45.1 μMT mice (17). All experimental animals were between 7–12 wk of age, and sexes were equally represented. For adoptive transfers, naive splenic CD43− B cells were magnetically isolated using the B cell isolation kit (Miltenyi Biotec; 130-090-862) and LS columns (Miltenyi Biotec; 130-042-40). Isolated B cells were stained with CellTrace Violet (CTV) (Life Technologies; C34557) per the manufacturer’s protocol and resuspended in sterile PBS (Corning; 21-040-CV) before transferring 15 × 106 B cells into a disparate congenic μMT host. At 24 h posttransfer, host mice were challenged i.v. with 50 μg LPS (Enzo Life Sciences; ALX-581-008), intranasally with 0.1 LD50 influenza A/HK-X31 (X31), or i.v. with 50 μg 4-hydroxy-3-nitrophenylacetyl (NP)–Ficoll (Biosearch Technologies; F-1420-10). For influenza infections, mice were anesthetized with vaporized isoflurane (Patterson Veterinary; 07-893-1389) before X31 administration. Experimental mice were euthanized via carbon dioxide asphyxiation in accordance with American Veterinary Medical Association guidelines. All procedures were approved by the Emory Institutional Animal Care and Use Committee.
Flow cytometry and sorting
Cells were resuspended at 1 × 106/100 μl in FACS buffer (1× PBS, 1% BSA, and 2 mM EDTA), stained with Fc Block (Becton Dickinson [BD]; 553141), and Ab–fluorophore conjugates for 15 and 30 m, respectively, and then washed with 1 ml of FACS. For adoptive transfers when NP-Ficoll or X31 was used, CD45.2 transferred cells were enriched prior to Ab staining using anti-CD45.2–allophycocyanin or anti-CD45.2-PE followed by magnetic enrichment using anti-allophycocyanin (Miltenyi Biotec; 130-090-855) or anti-PE (Miltenyi Biotec; 130-097-054) microbeads. The following Ab–fluorophore conjugates and stains were used: B220-PE-Cy7 (BioLegend; 103222), B220-A700 (BioLegend; 103232), BrdU–allophycocyanin (BioLegend; 339808), c-MYC-PE (Cell Signaling Technology; 14819), c-MYC–Alexa Fluor 647 (Cell Signaling Technology; 13871), CD11b–allophycocyanin-Cy7 (BioLegend; 101226), CD138-BV711 (BD; 563193), CD138–allophycocyanin (BioLegend; 558626), CD45.1-FITC (Tonbo Biosciences; 35-0453-U500), CD45.1-PE (BioLegend; 110708), CD45.1–allophycocyanin (BioLegend; 110714), CD45.1–allophycocyanin-Cy7 (Tonbo Biosciences; 25-0453-U100), CD45.2-PE-Cy7 (BioLegend; 109830), CD45.2-PerCP-Cy5.5 (Tonbo Biosciences; 65-0454-U100), CD45.2-PE (Tonbo Biosciences; 50-0454-U100), CD45.2–allophycocyanin (BioLegend; 109814), CD90.2–allophycocyanin-Cy7 (BioLegend; 105328), F4/80–allophycocyanin-Cy7 (BioLegend; 123118), Fas-PerCP-Cy5.5 (BioLegend; 152610), GL7-eFluor 660 (Thermo Fisher Scientific; 50-112-9500), GL7-PerCP-Cy5.5 (BioLegend; 144610), GL7-PE-Cy7 (BioLegend; 144620), IRF4-PerCP-eFluor710 (Thermo Fisher Scientific; 46-9858-82), Ki67–allophycocyanin (BioLegend; 652406), pS6-PE (Cell Signaling Technology; 5316), Rabbit mAb IgG XP Isotype-Alexa Fluor 647 (Cell Signaling Technology; 2985), Rabbit mAb IgG XP Isotype-PE (Cell Signaling Technology; 5742), Zombie Yellow Fixable Viability Kit (BioLegend; 423104), Zombie NIR Fixable Viability Kit (BioLegend; 423106), CTV (Life Technologies; C34557), and 7-aminoactinomycin D (7AAD) (BioLegend; 420404). For all flow cytometry analyses involving adoptive transfers, the following general gating strategy was used: lymphocytes were gated based on side scatter area/forward scatter area (FSC-A), single cells by forward scatter height/forward scatter width or forward scatter height/FSC-A, live cells based on exclusion of Zombie Yellow or Zombie NIR Fixable Viability Kit, and the markers CD11b, F4/80, and CD90.2 to remove non–B cells. All flow cytometry were performed on an LSR II, LSRFortessa, or LSR FACSymphony (BD) and analyzed using FlowJo v9.9.5, v10.5.3, or v10.6.2. Cell sorting was performed at the Emory Flow Cytometry Core using a FACSAria II (BD) and BD FACSDiva software v8.0.
Cell cycle analysis and intracellular staining
In some adoptive transfers, hosts were injected with 800 μg BrdU (BioLegend; 423401) i.v. 1 h prior to euthanasia. For some ex vivo experiments, 5 μg BrdU/ml of culture was added 1 h prior to harvest. Staining of BrdU, Ki67, and 7AAD was achieved using the Phase-Flow BrdU Cell Proliferation Kit (BioLegend; 370704), substituting anti-BrdU for anti-Ki67 when desired. Intracellular pS6 staining was accomplished following BD’s two-step protocol using BD Phosflow Fix Buffer I (BD; 557870) and BD Phosflow Perm Buffer III (BD; 558050). As a negative control for intracellular pS6, cultured cells were treated with 200 nM of rapamycin (Sigma-Aldrich; R8781) for 2 h prior to staining. Intracellular staining of MYC and IRF4 was performed using the FIX & PERM Cell Permeabilization Kit (Thermo Fisher Scientific; GAS003) or the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific; 00-5523-00), per the manufacturer’s protocol.
Ex vivo B cell differentiation
Isolated B cells were cultured at a concentration of 0.5 × 106 cells/ml in B cell media (RPMI 1640 supplemented with 1× nonessential amino acids, 1× penicillin/streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, 10% heat-inactivated FBS, and 0.05 mM 2-ME) containing 20 mg/ml Escherichia coli O111:B4–derived LPS (Sigma-Aldrich; L2630), 5 ng/ml IL-5 (BioLegend; 581504), and 20 ng/ml IL-2 (BioLegend; 575406), as previously described (39). Additional LPS (10 μg/ml), IL-5 (2.5 ng/ml), and IL-2 (10 ng/ml) were added to the cultures every 24 h for the duration of the time course. In some experiments, cells were cultured with LPS alone or with LPS and 10 ng/ml IL-4 (R&D Systems; 404-ML).
Retroviral production and transduction
Retrovirus was prepared as previously described (40). Briefly, Platinum-E cells were transfected at 70–80% confluency on 10-cm plates with 4 μg pCL-Eco (41) and 6 μg of either pMSCV-pBabeMCS-IRES-RFP (Addgene; 33337) or pMSCV-Myc-IRES-RFP (Addgene; 35395) (42) using 40 μl TransIT-293 (Mirus Bio; MIR2700). Cell media (antibiotic-free DMEM supplemented with 10% heat-inactivated FBS) was replaced with High-BSA cell media (DMEM supplemented with 10% heat-inactivated FBS and 1g/100 ml BSA) 18 h after transfection. Retrovirus was harvested 24 and 48 h later, filtered through a 0.45-μm membrane, and concentrated using 5× PEG-it Viral Precipitation Solution (System Biosciences; LV825A-1). Transduction of B cells was performed 12–24 h after stimulation via spinfection at 800 g for 1 h.
One million cells were resuspended in 600 μl of RLT Buffer (QIAGEN; 79216) containing 1% 2-ME and snap frozen in a dry ice–ethanol bath for RNA isolation. Lysates were thawed and subjected to QIAshredder homogenization (QIAGEN; 79656) and then total RNA isolation using the RNeasy Mini Ki (QIAGEN; 74104). RNA was reverse transcribed using SuperScript II Reverse Transcriptase (Invitrogen; 18064014). cDNA was diluted 1 μg/100 μl, and quantitative PCR (qPCR) was performed on a CFX96 Instrument (Bio-Rad Laboratories) using SYBR Green incorporation. Primers used included: 18S-forward 5′-GTAACCCGTTGAACCCCATT-3′ and 18S-reverse 5′-CCATCCAATCGGTAGTAGCCG-3′ and MYC-forward 5′-CGATTCCACGGCCTTCTC-3′ and MYC-reverse 5′-TCTTCCTCATCTTCTTGCTCTTC-3′. All primers were purchased from Integrated DNA Technologies.
RNA-seq and data analysis
For all samples, 1000 cells were sorted into 300 μl of RLT buffer (QIAGEN; 79216) containing 1% 2-ME and snap frozen in a dry ice–ethanol bath. RNA isolation was achieved using the Quick-RNA Microprep Kit (Zymo Research; R1050). Isolated RNA was used as input for the SMART-seq v4 cDNA synthesis kit (Takara Bio; 634894), and 400 pg of cDNA was used as input for the NexteraXT Kit (Illumina). Final libraries were quantified by qPCR and bioanalyzer traces, pooled at equimolar ratios, and sequenced at the New York University Genome Technology Center on a HiSeq 4000.
Raw sequencing data were mapped to the mm10 genome using STAR v.2.5.3 (43). Duplicate reads were identified and removed using PICARD (http://broadinstitute.github.io/picard/). The Bioconductor package edgeR v3.24.3 (44) was employed to determine differentially expressed genes (DEG), which were defined as having an absolute log2 fold change of ≥1 and a false discovery rate (FDR) of ≤0.05. All detected transcripts were ranked by multiplying the sign of fold change (+/−) by −log10 of the p value, and gene set enrichment analysis (GSEA) (45) was performed on this ranked gene list. All t-distributed stochastic neighbor-embedded (t-SNE) projections were generated using “Rtsne” v 0.15 (https://github.com/jkrijthe/Rtsne). Clustering and heatmap analysis were achieved using “heatmap3” (https://github.com/cdschar/heatmap).
ATAC-seq and data analysis
For each sample, 10,000 cells were sorted into FACS buffer, and the ATAC-seq was performed. Tn5 transposase (Tn5) preparation and library generation was previously described (23). Briefly, cells were centrifuged at 500 g for 10 min at 4°C. The supernatant was removed, and cells were resuspended in 25 μl of Tn5 tagmentation reaction (2.5 μl Tn5, 12.5 μl 2× tagmentation buffer [20 mM TAPS-NaOH (pH 8.1), 10 mM MgCl2, and 20% dimethylformamide], 2.5 μl 1% Tween-20, 2.5 μl 0.2% digitonin, and 5 μl of molecular grade water). Resuspended samples were incubated at 37°C for 1 h. Cells were then lysed by adding 25 μl lysis buffer (300 mM NaCl, 100 mM EDTA, 0.6% SDS, and 2 μl 10 mg/ml proteinase K) and incubated for 30 min at 40°C. Transposed DNA was isolated using AMPure XP SPRI beads (A63880) by adding 0.7× volumes to remove high m.w. DNA and then 1.2× volumes to positively select for low m.w. DNA. Tagmented DNA was eluted in 15 μl elution buffer (EB) (QIAGEN; 19086) and amplified using Nextera indexing primers (Illumina) and KAPA HiFi polymerase (Roche; KK2601). Final libraries were sequenced at the New York University Genome Technology Center on a HiSeq 4000.
Raw sequencing data were mapped to the mm10 genome using Bowtie v1.1.1 (46). Peaks were called using MAC2 v 2.1.0 (47) and annotated to the nearest gene using HOMER v4.8.2 (48). Reads per peak million normalization was performed for all samples as previously described (36). The Bioconductor package edgeR v3.24.3 (44) was used to determine DAR, which were defined as having an absolute log2 fold change of ≥1 and an FDR of ≤0.05. Motif analysis was performed using the HOMER program findMotifsGenome.pl (de novo results). For plotting the rank value of transcription factors, enriched transcription factor motifs were ranked according to their p value and normalized by the total number of enriched motifs found for a given sample. Resulting values were z-score normalized and motifs binned according to their DNA binding domain family.
All statistical analyses were achieved by using R/Bioconductor v3.5.2, Microsoft Excel v16.36 or v16.48, and GraphPad Prism v6.0c, v8.4.1, or 8.4.3. The p values <0.05 were considered significant. For RNA- and ATAC-seq significance, a combination of FDR and fold change was used to designate DEG and DAR.
All sequencing data generated in this study have been deposited in National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession code GSE173437 (GSE173435 for ATAC-seq and GSE173436 for RNA-seq).
IRF4-deficient B cells responding to LPS in vivo stall during the proliferative response
Cell division is one of the earliest events following B cell activation; however, a complete understanding of factors that control or maintain the proliferative response remain to be determined. Recent work identified an IRF4-dependent bifurcation event in the earliest stages of B cell activation (17). Cells along the IRF4-dependent branch upregulated gene sets critical for proliferation, indicating IRF4 may be important for controlling the proliferative response in vivo. To explore if IRF4 impacted cell proliferation during B cell differentiation, an in vivo adoptive transfer model was applied (25). In this study, splenic naive B cells from CD45.2+Cd19+/+Irf4fl/fl (Ctrl) or CD45.2+Cd19Cre/+Irf4fl/fl (IRF4cKO) mice were isolated, labeled with CTV, and transferred to CD45.1+ μMT hosts. After 1 d, host mice were challenged with the type I T cell–independent Ag LPS, and cell division and differentiation were determined via CD138 expression (49, 50) in a time course covering 3 d (Fig. 1A). At 24 h, no division was observed for Ctrl and IRF4cKO cells, indicating a similar delay before initiating the proliferative response (Fig. 1B). At 48 h, both Ctrl and IRF4cKO cells began to divide, and the majority of responding cells were observed in divisions 2–4. A modest difference in IRF4cKO B cells in divisions 0–1 was observed at this time point (Fig. 1B, 1C). At 60 h, Ctrl were distributed in all cell divisions (0–8), with a subset differentiating after reaching or exceeding division 8. Comparatively, IRF4cKO cells accumulated in divisions 2–4, with few cells observed in divisions 5 and 6 (Fig. 1B, 1C). Strikingly, whereas more than half of Ctrl cells accumulated in division 8 at 72 h, the cell division pattern for cells from IRF4cKO largely remained the same as their 60-h time point, indicating that the IRF4cKO cells stalled during the proliferative response (Fig. 1B, 1C). Indeed, the mean division number (MDN) (51) for Ctrl cells increased by ∼2 divisions from 60 to 72 h, whereas the MDN for IRF4cKO cells was unchanged (Fig. 1D). This proliferative defect was also reflected in the reduced frequency of IRF4cKO cells detected in host spleens at 72 h (Fig. 1E, 1F). Importantly, staining for the proapoptotic marker annexin V revealed no differences in apoptosis or necrosis at 72 h in vivo (Supplemental Fig. 1). Furthermore, no differences in homeostatic proliferation were observed in mice that received Ctrl or IRF4cKO B cells but no LPS (Fig. 1B). It is also important to note that the vast majority of the splenic cells transferred divided at least once to LPS stimulation, indicating that nearly all B cells and not just a subset were responding in vivo. Proliferation defects were also observed when C57BL/6J mice were used as hosts (Supplemental Fig. 1). These data indicate IRF4 controls the proliferative capacity of B cells in response to LPS immune challenge.
IRF4-deficient B cells exhibit a proliferation defect to T-independent and -dependent Ags
To determine whether IRF4 controls the proliferative response to other stimuli, adoptive transfers were performed followed by challenge with the type II T-independent Ag NP-Ficoll or the T-dependent Ag X31. Five days post–NP-Ficoll and 6 d after X31 challenge, host mice were sacrificed, and cell division and differentiation were assessed by flow cytometry (Fig. 2A). Because NP-Ficoll and X31 stimulate Ag-specific B cells that represent a small portion of the population, the majority of Ctrl and IRF4cKO cells remained undivided for both stimulation conditions (Fig. 2B). For NP-Ficoll, Ctrl cells were distributed in all cell divisions 1–8, and a subset of cells that reached or surpassed division 8 differentiated (Fig. 2B, 2C). Similar results were observed following X31 challenge and independent of whether the transferred cells were recovered in the mediastinal lymph node or the spleen (Fig. 2B, 2D). Interestingly, CD138+ ASC were observed at division 8 for all three Ag conditions for Ctrl cells. Comparatively, cells from IRF4cKO were mainly distributed in the first few divisions for both stimulation conditions, with very few IRF4cKO B cells detected after division 4 and almost none reaching division 8 and forming ASC (Fig. 2B–D). Taken together, these data indicate IRF4 plays a critical role in controlling the proliferative response to type II T-independent and early T-dependent Ag responses.
IRF4-deficient B cells display altered cell cycle distribution
To better understand the proliferative defect observed above, the role that IRF4 played with respect to cell cycle was investigated. CTV-labeled Ctrl and IRF4-deficient B cells were adoptively transferred into µMT mice and recovered 72 h post–LPS stimulation. Cells were stained with Ki-67 and 7AAD to distinguish the frequency of cells in each phase of the cell cycle at discrete divisions (52) and analyzed by flow cytometry. These data revealed that in the final detectable divisions, IRF4cKO cells accumulated in G0/G1 with a corresponding decrease in cells found in the G2/M (Fig. 3A, 3B). This was in stark contrast to Ctrl cells, which revealed more cells in S and G2/M at the same divisions. This indicates that the cell cycle was significantly perturbed in B cells from IRF4cKO in these final divisions (Fig. 3A, 3B). To better understand the proliferative defect observed in IRF4cKO cells in vivo, the frequency of actively proliferating cells by BrdU incorporation was examined after IRF4cKO cells had stalled. Appreciably, a lower frequency of BrdU+ IRF4cKO compared with Ctrl cells was observed (Fig. 3C, 3D). BrdU+ IRF4cKO cells were also distributed proportionally to the total population. In contrast, BrdU+ Ctrl cells were largely distributed in division 8 (Fig. 3C). Thus, IRF4 is critical for cell cycle control and maintaining the proliferative response.
Cell division–coupled IRF4-dependent transcriptional reprogramming
B cell differentiation to ASC requires considerable transcriptional rewiring that consists of progressive cell division–based reprogramming events (25). To determine the impact of IRF4 on this process, Ctrl and IRF4cKO cells were sorted from divisions 0, 1, 3, 4, 5, and 6 as determined by CTV dilution (Fig. 4A) and subjected to RNA-seq analyses. Comparing gene expression profiles for Ctrl and IRF4cKO cells in the same division revealed hundreds of DEG that increased or decreased expression in IRF4-deficient B cells, indicating that IRF4 functions to repress and activate gene expression programs, even in the earliest stages of actB reprogramming (Fig. 4B). This activity is consistent with previous work, demonstrating that a significant increase in IRF4 levels occurs after the first cell division (17, 18). After successive divisions, IRF4cKO B cells became progressively transcriptionally divergent compared with Ctrl cells (Fig. 4B). Hierarchical clustering of samples reflected this divergency with Ctrl and IRF4cKO samples in divisions 0 and 1 clustering by gene expression and divisions 3–6 clustering by IRF4 status (Fig. 4C). t-SNE projections of gene expression data from all samples indicated major cell division–dependent transcriptional reprogramming events that were dependent on IRF4 and predominately in divisions 3–6. (Fig. 4D). Collectively, IRF4cKO are transcriptionally distinct by division 3 and continue to diverge through subsequent divisions. Thus, cell division–based IRF4-dependent reprogramming occurs during the initial stages of B cell differentiation.
To determine the transcriptional programs dependent on IRF4, GSEA (45) was performed for DEG that increased or decreased expression in IRF4cKO cells in divisions 3–6. IRF4cKO B cells progressively failed to induce gene sets important for cell division, metabolism, and signaling (Fig. 4E, 4F). This consisted of genes critical for glycolysis and OXPHOS, which are critical metabolic programs for actB and ASC, respectively (24, 53) (Fig. 4E, 4F). Enzymes that failed to be induced and are critical for glycolytic metabolism included Ldha (54) and Aldoa (55) (Fig. 4G). Additionally, mTORC1 signaling and MYC target genes failed to be induced in IRF4cKO cells and included genes that promote cell proliferation, such as Kpna2 (56) and Plk1 (57, 58) (Fig. 4G). Notably, the cell cycle was significantly perturbed in IRF4cKO cells in the divisions in which MYC target genes were the most dysregulated (Fig. 3A, 3B). These data are consistent with reports that reduction of Myc impacts G1–S transition of the cell cycle (59–61). Gene sets that failed to be repressed consisted of those involved in cytokine and cell signaling, such as the inflammatory response, and reflect previous reports that IRF4-deficient B cells progress down a reprogramming path whose gene expression program reflects cells responding to inflammatory stimuli (17). Collectively, these data suggest that early metabolic and proliferative programs essential for cell growth and division are dependent on IRF4.
Previous work showed that IRF4 was required for the proper regulation of AID gene expression (15, 16) and that LPS induction results in class-switch recombination to IgG3 (62). Thus, Aicda (encoding AID) was examined in Ctrl and IRF4cKO B cells. The results showed that Aicda began to be upregulated in division 3 Ctrl B cells but failed to be induced to the same levels in IRF4cKO samples (Fig. 4G). Consistent with the role of AID in class-switch recombination, fewer transcripts matching IgG3 H chain (Ighg3) were observed for IRF4cKO B cells compared with Ctrl samples (Fig. 4G). These data support work indicating that class-switch recombination occurs early following B cell activation (63) and that IRF4 coordinates this process through proper regulation of Aicda.
ATAC-seq reveals a hierarchy of IRF4 activity
To identify regions that change chromatin accessibility during B cell differentiation upon deletion of Irf4, paired ATAC-seq (64) data derived from the above divisions was analyzed to reveal IRF4-specific regulatory activities and IRF4-dependent transcription factor networks that impact B cell differentiation. Comparison of Ctrl and IRF4cKO cells in discrete divisions identified hundreds of DAR, with a progressive increase in DAR occurring after the first cell division and more than 700 DAR by divisions 5 and 6 (Fig. 5A). These differences were also reflected in t-SNE spatial projections (Fig. 5B) and indicated that similar to RNA-seq, chromatin accessibility differences occurred predominately in divisions 3–6 (Fig. 5A, 5B). Collectively, these data support the notion that IRF4-dependent reprogramming occurs progressively beginning during the initial stages of B cell differentiation and that the chromatin landscape of IRF4cKO B cells is markedly distinct by division 3.
To gain a better understanding of the transcription factor networks dependent on IRF4, the top 10 enriched DNA sequence motifs in division 6 DAR were determined and matched to known putative transcription factor binding motifs using HOMER (48). Because enrichment p values are dependent on the number of DAR, each transcription factor motif was rank normalized based on the significance at each division, and the change in rank score across the divisions was plotted, revealing how motif accessibility was altered across the divisions. Motifs enriched in regions that decreased accessibility in IRF4cKO cells (down DAR) included known IRF4 DNA binding motifs (14, 15, 18, 65), such as the core IRF motif (GAAA), AICE (66), and EICE (67, 68) (Fig. 5C). Interestingly, this revealed a hierarchy among heterodimeric IRF4 binding sites (69), with AICE more highly ranked in early divisions and EICE motifs more highly ranked in later divisions. DAR in proximity of known IRF4 target genes (70), such as Itm2c for EICE motifs and Gpcpd for AICE motifs, reflected this hierarchy of activity (Fig. 5D). These data support the kinetic control of IRF4 activity (18, 27), as well as previous work implicating the timing of IRF4 in conjunction with the AP-1 transcription factor BATF in early cell fate decisions during B cell differentiation (17). Other transcription factors enriched in down DAR in the final divisions included RUNX and E-box binding bHLH family members (Fig. 5C).
Among regions that increased accessibility in IRF4cKO (up DAR), TBOX family members were more highly ranked in early divisions compared with subsequent divisions (Fig. 5C). Notably, the TBOX family member TBET supports ASC formation through repression of the inflammatory gene expression program (71), which was progressively upregulated in IRF4cKO cells (Fig. 4E, 4F). RUNX and ETS family members were most highly ranked in the final divisions, suggesting that these transcription factors are playing roles at both regions, gaining and losing accessibility as the cells differentiate (Fig. 5C). Collectively, these data demonstrate the timing of IRF4-dependent reprogramming, establish a hierarchy of IRF4 activity that occurs at early and late cell divisions, and identify transcription factor networks dependent on IRF4.
IRF4-deficient B cells fail to upregulate MYC
Recent work described MYC as a cell division timer during lymphocyte differentiation, with division cessation occurring when MYC levels fell below a critical threshold (12). We reasoned that Myc may be dysregulated in IRF4-deficient B cells because IRF4cKO cells 1) stalled during the proliferative response to LPS (Fig. 1); 2) accumulated in the G0/G1 phase of the cell cycle (Fig. 3A, 3B); 3) progressively failed to induce MYC target genes (Fig. 4E, 4F); and 4) E-box binding bHLH family members were enriched in down DAR in divisions in which MYC target genes were the most dysregulated (Fig. 5C). In fact, IRF4cKO cells were progressively enriched for genes dysregulated in MYC-deficient B cells stimulated with LPS and IL4 (11), further supporting the notion that MYC programming is altered in IRF4cKO cells (Fig. 6A). To begin to explore the relationship between Irf4 and Myc, we investigated their expression kinetics during the initial stages of B cell activation. This was accomplished by culturing Ctrl B cells ex vivo with LPS, IL-2, and IL-5 and collecting samples for intracellular staining of IRF4 and MYC in a time course within the first 24 h after activation (Fig. 6B). Consistent with previous reports (72, 73), MYC was induced within 2 h following immune stimulation, and the peak of MYC expression occurred prior to 24 h (Fig. 6B). Comparatively, IRF4 was detected 8 h after stimulation and gradually increased until 24 h (Fig. 6B). Thus, MYC induction occurs prior to IRF4 expression, but changes to MYC expression happen concomitantly with IRF4 induction (Fig. 6B) and prior to the first cell division.
To determine if Myc was reduced in IRF4-deficient B cells, Ctrl and IRF4cKO cells were cultured ex vivo with LPS, IL-2, and IL-5 to initiate the pathway to ASC (39), and expression was analyzed by quantitative RT-PCR before and 24 h after stimulation. Whereas no differences in Myc levels were detected prior to stimulation, a significant reduction was observed at 24 h (Fig. 6C). Similar observations were detected by intracellular staining of MYC, which confirmed that whereas MYC levels were increased over naive B cells, IRF4cKO cells failed to upregulate MYC to the same level as Ctrl cells (Fig. 6D, 6E). These data are consistent with previous reports following PMA/ionomycin treatment of IRF4-deficient and -sufficient B cells (74). The observed differences in MYC expression are likely caused by transcription of Myc and not because of alterations in MYC protein stability (75) (Supplemental Fig. 2). Thus, these data support the concept that whereas MYC is induced prior to IRF4, IRF4 is required for full expression of MYC during B cell activation (74).
To explore whether MYC overexpression could rescue the cell division defect of IRF4-deficient B cells, Ctrl and IRF4cKO cells were again cultured ex vivo with LPS, IL-2, and IL-5 and transduced with retrovirus expressing MYC-RFP or control RFP. Overexpression of Myc significantly improved the proliferation capacity of cells, and this improvement was greater for B cells from IRF4cKO than Ctrl (Fig. 6F, 6G). However, although IRF4cKO cells exhibited a greater proliferative gain upon MYC overexpression compared with Ctrl cells, full cell division capacity was not restored as Ctrl B cells transduced with control RFP still displayed greater proliferative capacity. Collectively, these data suggest that IRF4cKO B cells fail to fine-tune the levels of Myc during the initial stages of B cell activation, which impact the overall cell division pattern and are consistent with the observation that IRF4cKO B cells begin to divide normally but stall in the middle of the proliferative response (Fig. 1). Indeed, analysis of IRF4 chromatin immunoprecipitation sequencing (70) revealed IRF4 binding at two of four known regulatory elements of Myc in murine actB (76) (Supplemental Fig. 4). However, Myc overexpression alone does not fully restore the division capacity of IRF4cKO B cells, indicating additional deficiencies are contributing to the proliferative defect.
Previous reports suggested that deficiencies in IRF4 could be compensated ex vivo by culturing cells with IL-4 and LPS (15). To explore this in the current system, which limited the IRF4 deficiency to mature B cells, IRF4cKO and Ctrl B cells were cultured with LPS alone, LPS and IL-4, or LPS, IL-2, and IL-5 (Supplemental Fig. 3). In some cultures, BrdU was added to label cells in S phase. The data show that the addition of IL-4 improved the proliferative capacity and frequency of actively dividing IRF4cKO cells, although full proliferation was not restored (Supplemental Fig. 3). Intriguingly, ex vivo cultures showed that IRF4cKO B cells displayed increases in cell death (15), irrespective of the stimulation. Thus, cytokine signaling may regulate additional/distinct proliferative programs (Supplemental Fig. 3).
IRF4-deficient B cells exhibit reduced mTORC1 activity and are unable to initiate the UPR
Activation of the mammalian target of rapamycin (mTOR) is essential for promoting biosynthetic processes necessary for cell growth and division (77). Importantly, ablation of mTORC1 activity impacted the proliferative effects of MYC overexpression in murine tumor cells (78), indicating there is significant cross-talk between the two signaling cascades (79–82). Recent work indicated mTORC1 coordinates an early B cell activation UPR, in which a subset of UPR-affiliated genes are upregulated independent of XBP1 (32), a known driver of the UPR (34, 83). Interestingly, although Ctrl B cells gradually upregulated the B cell activation UPR as early as division 3, IRF4cKO cells failed to initiate the program to the same levels (Fig. 7A, 7B). Indeed, genes associated with mTORC1 signaling progressively failed to be induced in IRF4cKO B cells (Fig. 4E, 4F). Collectively, these data implied that mTORC1 activation may be dysregulated in IRF4-deficient B cells. To test for mTORC1 activity, Ctrl and IRF4cKO cells were cultured ex vivo with LPS, IL-2, and IL-5 for 48 h, and intracellular staining for phosphorylation of the canonical mTORC1 substrate S6 (pS6) was performed. Strikingly, whereas the majority of B cells from Ctrl exhibited high amounts of pS6, most IRF4cKO cells contained pS6 levels similar to cultures in which mTORC1 activity was blocked following treatment with rapamycin (Fig. 7C, 7D). Consistently, proliferating IRF4cKO cells also failed to increase in cell size compared with Ctrl B cells at 48 h post-LPS in vivo (Fig. 7E, 7F). Intriguingly, this reduction in cell size was rescued via overexpression of Myc in IRF4cKO cells cultured ex vivo (Fig. 7G). Thus, IRF4cKO B cells exhibit a defect in mTORC1 activity that impacts the ability of cells to increase in cell size that is overcome with Myc overexpression. Thus, these data support the role of mTORC1 in upregulating an early B cell activation UPR, assign the cell division in which this process occurs, and implicate IRF4 in this process.
This study establishes the timing and extent of IRF4-dependent reprogramming instructed in the initial stages of B cell differentiation in vivo and ascribes a role for IRF4 in controlling cell growth and proliferation. Using multiple Ag model systems, IRF4-deficient B cells divided initially but stalled during the proliferative response. Characterization of the proliferative defect revealed fewer actively dividing cells and abnormal cell cycle distribution. B cells lacking IRF4 maintained an inflammatory gene signature but failed to induce critical actB and ASC gene expression programs, including metabolic pathways (glycolysis and OXPHOS), MYC target genes, and mTORC1 signaling. Reduced Myc expression and mTORC1 activity contributed to the cell division and growth defect following stimulation. Additionally, IRF4-deficient B cells failed to induce the B cell activation UPR, which relies on mTORC1 (32). Thus, we define the cell division–coupled IRF4-dependent reprogramming events that occur in the initial stages of B cell activation and identify an IRF4–MYC–mTORC1 relationship that impacts cell growth and proliferation.
The role of MYC as a division-independent timer to regulate lymphocyte proliferation has been described (4, 12). In this model, the combination and strength of stimuli determine the amount of MYC initially generated. This serves as a timer to regulate the overall number of cell divisions, or a cell’s division destiny, with division cessation occurring when MYC levels fall below a critical level (12). Analyzing the cell division kinetics of IRF4-deficient B cells responding to LPS revealed that they can initiate cell division appropriately but stall in the middle of the proliferative response. Applying the MYC dilution model, IRF4-deficient B cells fall below the MYC threshold sooner, which caused the observed stalling. Indeed, IRF4-deficient B cells displayed reduced MYC levels 24 h after stimulation. Interestingly, MYC expression is not dependent on cell division (12), but we found progressive dysregulation of MYC target genes in IRF4-deficient B cells, implying that other factors reinforce MYC programming throughout the cell divisions. Importantly, both Irf4 expression and Myc induction levels are dependent on the strength of signaling (12), irrespective of whether the stimulus is from BCR (18) or TLR (84). Furthermore, IRF4 binding to the Myc promoter has been reported (74, 85). Although no differences in chromatin accessibility (Supplemental Fig. 4) were observed in IRF4-deficient B cells at known regulatory elements of Myc (76), this is likely due to the timing in which the samples were collected or compensatory effects of IRF8 (86, 87), which often binds to the same sites. MYC expression was induced prior to IRF4; however, dynamic changes in MYC expression occur simultaneously with induction of IRF4 and prior to the first cell division. Collectively, these data support the concept that IRF4 may serve as a rheostat in B cells to regulate the overall proliferative response by fine-tuning Myc expression levels prior to the onset of cell division. Indeed, a similar role for IRF4 has been noted in CD8 T cells in which IRF4 serves as a molecular rheostat of TCR affinity. Similar to our observations, IRF4-deficient CD8 T cells can initiate proliferation but fail to maintain clonal expansion (88), suggesting that IRF4 may play a similar role in controlling the proliferative response in T cells.
Differentiating actB undergo an IRF4-dependent bifurcation event that commits a portion of actB to an ASC fate (17). Additionally, differentiating actB use mTORC1 to anticipate Ab synthesis by upregulating UPR-affiliated genes (32). We demonstrate that IRF4-deficient B cells display reduced mTORC1 activity and fail to initiate the B cell activation UPR. Thus, actB anticipation of Ab synthesis and secretion is a component of ASC fate commitment and programmed during the initial stages of B cell activation. Our gene expression data indicate that this process occurs as early as division 3 during B cell differentiation, with reduced expression of UPR-affiliated genes in IRF4-deficient B cells. Interestingly, the interplay between mTORC1 and IRF4 has been noted, with mTOR inhibition negatively impacting IRF4 expression (89–91). In this study, IRF4 also impacts mTORC1 activity, suggesting the existence of a positive IRF4–mTORC1 feedback loop that impacts actB reprogramming. MYC is central to this regulatory network as MYC overexpression in IRF4-deficient B cells restores cell growth. mTOR may impact IRF4 transcription by effecting downstream transcription factors or by directly impacting IRF4 protein translation or stability (92).
Occupancy of IRF4 at composite motifs is dependent on its concentration and availability of binding partners (15). IRF4 levels are increased as the cells divide and ultimately sustained at high levels in ASC (17, 18). In contrast, IRF8 levels are decreased as B cells differentiate, allowing for IRF4 to more readily partner with transcription factors and establish the IRF4-dependent gene expression program (87). In IRF4cKO cells, differentiating cells showed changes in accessibility surrounding composite motifs. Previously, ATAC-seq data in wild-type differentiating B cells suggested that EICE motifs were most accessible in early dividing actB and that AICE sites became increasingly accessible as IRF4 levels increased during the division-coupled differentiation process (27). In the absence of IRF4, this program is altered. In regions that decreased accessibility, AICE motifs were the most affected motifs in early divisions (divisions 3 and 4), whereas EICE motifs were most highly ranked at later divisions (divisions 5 and 6). Although both motifs are affected at all divisions, this analysis pinpoints specific divisions and differentiation stages in which IRF4 cooperates with AP-1 or ETS factors to establish differentiation programs, suggesting a hierarchy of IRF4 activity. Consistent with these data, single-cell analysis of LPS-responding B cells showed that IRF4 was required for BATF (an AP-1 family member) targets as early as division 3, suggesting that IRF4 may be BATF’s partner in AICE motifs at the early stages of B cell differentiation to ASC (17). Furthermore, IRF4 binding at AICE motifs largely occurs at newly established accessible regions (14). Taken together, these data indicate that these reprogramming steps occur at divisions 3 and 4.
The cell division requirement needed for ASC formation in vivo following LPS (17, 25, 27) and NP-Ficoll (17) stimulation has been described. We observed similar cell division requirements for adoptive transfers using Ctrl B cells and add that ASC formation occurs after cells reach or exceed division 8 following stimulation with the T-dependent Ag influenza X31. As this analysis was performed at day 6 following infection, it is unlikely that the generation of ASC at this time point involve a full germinal center reaction. However, Ag-specific ASC can be observed at this time point (93). These data suggest that the timing of division-coupled reprogramming events needed for ASC differentiation are similar for T-independent Ags and the early differentiation process that occurs with T-dependent Ags. Studying the cell division requirement of T-dependent ASC formation at later time points is complicated by the dynamics and selection pressures of the germinal center reaction and increased cell divisions (94, 95).
Together, these data indicate IRF4 coordinates cell growth and the proliferative response during B cell differentiation. We demonstrate that part of the mechanism involves regulation of Myc and mTORC1 activity. Indeed, the relationship between MYC and mTORC1 has been noted, with mTORC1 controlling MYC translation (96) and MYC-driven tumorigenesis dependent on mTORC1 (78, 81). Both factors converge to control protein production and cell growth. MYC controls the expression of translation initiation factors needed for increased protein synthesis (80), and mTOR controls their activity (97). In this study, IRF4cKO cells displayed reduced mTORC1 activity and were unable to increase in cell size as they divided. However, the deficiency in cell growth was overcome by overexpression of Myc, suggesting that this aspect of MYC/mTOR relationship is dependent on Myc expression. RNA-seq analyses showed IRF4-deficient B cells failed to induce MYC target genes and mTORC1 signaling by division 3, and these gene sets became progressively dysregulated as the cells divided. Thus, reprogramming events needed for continued cell growth and proliferation occur during the initial cell divisions during B cell differentiation and are coordinated by IRF4, MYC, and mTORC1.
We thank the Boss and Scharer laboratories for scientific contributions and critical reading of the manuscript, Royce Butler for mouse colony maintenance and husbandry, Tian Mi for bioinformatic assistance, Sakeenah L. Hicks for preparation of sequencing libraries, Dr. Chaoran Li for retroviral reagents (gifted Plat-E cells and pCL-Eco) and protocols, Dr. Troy D. Randall for influenza A/HK-X31 viral stocks, Drs. Susanne Heinzel and Philip Hodgkin for discussions regarding intracellular staining of MYC, the Emory Flow Cytometry Core for FACS isolation of cells, and the Emory Integrated Genetics and Computational Core for Bioanalyzer and sequencing library quality control.
This work was supported by National Institute of Allergy and Infectious Diseases Grants AI123733 and AI125180 to J.M.B., AI148471 to C.D.S., and AI138391 to M.J.P. This work was also supported by National Institute of General Medical Sciences Grant GM0008490 to A.K.K. and J.R.R.
The sequences presented in this article have been submitted to the National Center for Biotechnology Information’s Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE173437, GSE173435, and GSE173436.
The online version of this article contains supplemental material.
Abbreviations used in this article
activated B cell
Ab-secreting plasma cell
assay for transposase accessible chromatin sequencing
differentially accessible region
differentially expressed gene
false discovery rate
forward scatter area
gene set enrichment analysis
mean division number
mammalian target of rapamycin
t-distributed stochastic neighbor-embedded
unfolded protein response
The authors have no financials conflict of interest.