IRF4 Haploinsufficiency Impairs Affinity Maturation

Abstract

The germinal center (GC) reaction is a coordinated and dynamic ensemble of cells and processes that mediate the maturation and selection of high-affinity GC B cells (GCBs) from lower affinity precursors and ultimately results in plasma (PC) and memory cell fates that exit the GC. It is of great interest to identify intrinsic and extrinsic factors that control the selection process. The transcription factor IRF4, induced upon BCR and CD40 signaling, is essential for the acquisition of PC and GCB cell fates. We hypothesized that beyond this early requirement, IRF4 continuously operates at later phases of the B cell response. We show that IRF4 is expressed in GCBs above levels seen in resting cells and plays a role in efficient selection of high-affinity GCBs. Halving Irf4 gene copy number in an antigen-specific murine B cell model, we found that antigen presentation, isotype switching, GC formation and zonation, somatic hypermutation rates, and proliferation were comparable to cells with a full Irf4 allelic complement. In contrast, Irf4 haploinsufficient GCBs exhibited impaired generation of high affinity cells. Mechanistically, we demonstrate sub-optimal Blimp-1 regulation amongst high affinity Irf4 haploinsufficient GCBs. Furthermore, in cotransfer settings, we observed a marked disadvantage of Irf4 haploinsufficient cells for GC entry, evidential of ineffective recruitment of T cell help. We propose that, analogous to its role in early GC entry, IRF4 continues to function in the late phase of the antibody response to promote productive Tfh interactions and to activate optimal Blimp-1 expression during GC selection and affinity maturation.

Introduction

T-dependent antibody responses generate differentiated plasma cell (PC), Germinal Center B cell (GCB), and memory B cell progeny from activated antigen specific clonal precursor B cells¹. Critical cell intrinsic and extrinsic mechanisms allocate PC and GCB fates to control the quantity and composition, respectively, of antibody to mediate immune protection. IRF4 has emerged as a critical cell fate determinant of PC and GCB^2–4. Specifically, the PC gene program promotes antibody secretion and thus antibody quantity. Alternatively, the GCB gene program promotes somatic hypermutation (SHM) of the antibody VDJ heavy chain and VJ light chain genes to generate structural variants of which a few are subsequently selected for, termed affinity maturation. Selected clones differentiate into PC and/or memory B cells and exit the germinal center (GC); clones with high affinity mutations are enriched in the PC compartment. Entry, selection, and exit of GCBs are coordinated by T follicular helper cells (Tfh)⁵.

Upon antigen recognition BCR signaling initiates mitogenesis, reprogrammed bioenergetics, and chemokine-dependent migration to the B-T zone border⁶. Subsequent antigen processing and presentation generates peptide-MHCII (pMHC) complexes that are presented to likewise antigen-activated cognate Tfh cells⁷. In fact, both BCR signal strength and pMHC density have been shown to limit B cell clonal entry into the GC^8,9. These events are integrated during prolonged B-T interactions that involve adhesion¹⁰, stabilization¹¹, and costimulation¹². Productive cell conjugates lead to CD40:CD40L and ICOS:ICOSL signaling¹³, B cell proliferation and GCB differentiation.

The GC is classically composed of two zones – the light and the dark zones (LZ/DZ), defined by chemokines and guidance cues for zonal recirculation^1,14. Interestingly, recent studies suggest greater complexity and have identified multiple distinct GCB populations within the DZ, including actively proliferating cells (gray zone, GZ/DZp), cells undergoing AID-dependent SHM (DZd), as well as others in the LZ^15–17. Upon cessation of SHM, proliferation, and pMHC degradation¹⁸, clones migrate to the LZ to test mutated BCRs^19–21. Competition for antigen on Follicular Dendritic Cells (FDCs), BCR signaling, and subsequent pMHC presentation to Tfh cells all play critical roles in selection of the fittest mutated clones. Interestingly, we note that this sequence of events parallel that of B-T conjugate interactions that initiate the GC response. If pMHC density is above a certain threshold, strong CD40 signals^22,23 lead to sustained Foxo1^24–26, c-Myc^27–29, AP4 transcription factor expression³⁰ and DZ re-entry, proliferation, differentiation, and exit as high affinity PC or as memory B cells. In contrast, lower pMHC density causes DZ re-entry and additional mutagenesis before retesting in the LZ, termed cyclic re-entry^21,31,32.

The IRF4 transcription factor, an immediate-early gene downstream of BCR signaling³³, is cell autonomously essential for PC and GCB differentiation^2–4,34. Mechanistic studies highlight the role of IRF4 levels in this process; low levels directly induce Bcl6 upregulation, GCB differentiation, and inhibition of PC fate whereas high levels directly induce Prdm1 (Blimp-1) upregulation, PC differentiation, and suppression of GCB fate^2,4,35. Interestingly, GCBs with conditionally deleted Irf4 exhibit somewhat lower SHM, suggesting that IRF4 plays roles in GCBs beyond regulating Bcl6 and GCB fate³. Furthermore, tight regulation of IRF4 expression is important for normal GCB function, as elevated IRF4 protein results in precocious PC differentiation and decreased generation of high affinity cells³⁶.

Given the mechanistic parallels between GC entry and GCB selection, and prior mechanistic studies that demonstrate the importance of IRF4 levels during priming, we hypothesized that IRF4 activity is limiting for efficient GC selection. Herein, we show that GCBs express a basal level of IRF4 and that IRF4 is required for efficient selection of high affinity GCBs in part by upregulating ideal levels of Blimp-1. Furthermore, Irf4 haploinsufficient cells are competitively disadvantaged for Tfh help and GC entry. We propose that efficient affinity maturation depends on optimal levels of IRF4 and Blimp-1 expression, driven by successful recruitment of Tfh help during competitive Tfh cell interactions.

Materials and Methods

Mice, immunizations and injections

In all experiments recipient mice were WT CD45.1^+/+. Our colony of SW_HEL mice³⁷ were bred with Irf4^−/− mice³⁸ to generate desired genotypes. CD45.1⁺CD45.2⁺ Irf4^+/+ cells were generated by breeding CD45.1^+/+ with CD45.2^+/+ SW_HEL mice. Irf4^+/+ or Irf4^+/− SW_HEL mice were bred to Prdm1:YFP mice³⁹ and were CD45.2^+/+. Matched genders of donor cells and host mice were used in the experiments and no effect of gender was observed. Mice were housed in specific pathogen-free conditions and were used and maintained in accordance with the Institutional Animal Care and Use Committee guidelines of UC Davis.

In all adoptive transfer experiments, numbers of SW_HEL B cells from the spleen of donor mice were quantified by flow cytometry and a total spleen mixture containing 5×10⁴ SW_HEL cells were adoptively transferred with 2×10⁸ sheep red blood cells (SRBC; Lampire Biological Laboratories) conjugated to either HEL^WT or HEL^2X protein in PBS, i.v., as previously described³⁷. For experiments analyzing the day 1 response, 5×10⁵ SW_HEL cells were adoptively transferred i.v., into mice immunized with 75 μg of HEL-IEα in Imject Alum at a 1:1 ratio with PBS (Thermo Fisher).

For S-phase determination, mice were injected with either 1mg of EdU diluted in PBS i.v., 4 hours prior or 2 mg of BrdU diluted in PBS i.v., 2 hours prior to splenic harvest.

HEL^2X-IEα production

The HEL^2X-IEα fusion gene was subcloned into a pRMHa3 copper-inducible expression vector, and this plasmid was used to stably transfect D. melanogaster S2 cells. Transfectants were identified by induction with copper sulfate and the presence of HEL^2X-IEα in supernatants by western blotting using anti-HEL polyclonal antibodies (Rockland Immunochemicals). Secreted HEL-IEα was affinity purified on a nickel column (Bio-Rad NGC Quest 10 Chromatography System).

Flow cytometry

RBC-depleted splenic cell suspensions were prepared and washed in isotonic buffer⁴⁰ containing sodium azide and 1% FBS (TC buffer) before extracellular staining in PBS containing sodium azide, 1% FBS FC buffer, Brilliant Stain Buffer (BD) and rat serum (2% v/v; Jackson ImmunoResearch). Detection of total SW_HEL cells and high affinity HEL^3X-binding cells was conducted as previously described by incubating cells with non-conjugated HEL antigens prior to detection with a fluorchrome-conjugated HyHel9 antibody³⁷. For some experiments in the single transfer setting, host CD45.1^+/+ cells were depleted using an immuno-magnetic bead approach consisting of CD45.1-biotin antibody (Biolegend), Streptavidin magnetic beads (Miltenyi), and MACS LS columns (Miltenyi) prior to HEL^3X incubation. Detection of intracellular markers (such as IRF4) was performed by fixing and permeabilizing cells with Fix/Perm staining kit (Invitrogen). Flow cytometry was conducted on a BD Fortessa or BD Symphony and analysis performed using FlowJo software (Tree Star, Inc.). Detection of EdU was performed using the Click-iT EdU Plus kit from ThermoFisher (C10636) as per the manufacturer’s instructions. Detection of BrdU was performed using a BrdU Monoclonal Antibody from ThermoFisher in conjunction with a BrdU Permeabilization Buffer Plus kit and protocol from BD. To assess absolute cell number, AccuCount Blank Particles (Spherotech) were used.

Flow Cytometry Antibodies and Clones

B220 (RA3–6B2), Bcl6 (K112–91), CD4 (RM4–5), CD45.1 (A20), CD45.2 (104), CD86 (GL1), Fas/CD95 (Jo2), IgG1 (A85–1), Ly108 (13G3), PDL1/CD274 (M1H5), and Streptavidin were from BD. CCR6/CD196 (29–2L17), CD3 (145–2C11), CD38 (90), CD4 (GK1.5), CD40 (3/23), CD8 (53–6.7), CD80 (16–10A1), GL7 (GL7), ICOSL (HK5.3), IRF4 (3E4), MHCII:I-A/I-E (M5/114.15.2), PDL2/CD273 (TY25), and SLAM/CD150 (TC15–12F12.2) were from BioLegend. CXCR4 (2B11/CXCR4), ICAM (YN1/1.7.4), and YAe/Ea 52–68 (eBioY-Ae) were from eBioscience. BrdU (MoBU-1) and Active Caspase-3 (C92–605) were from Thermo Fisher. HyHel9, specific for HEL (UCSF Monoclonal Antibody Core), was conjugated in-house using an Alexa Fluor 647 Antibody Labeling Kit (Thermo Fisher).

VDJ Sequencing

Post-immunization, GC-resident HEL^WT-specific or HEL^3X-specific donor B cells were individually sorted into a 96-well plate containing digest buffer (100 μl 10x Taq PCR reaction buffer (no MgCl₂), 50 μl Proteinase K (10 mg/ml in H₂O), 10 μl 10 mM EDTA, 10 μl Tween 20 (10% solution), in final volume 1 ml H₂O,³⁷ using a Astrios EQ high speed cell sorter (Beckman Coulter). The VDJ region of single sorted SW_HEL B cells was amplified by PCR and Sanger-sequenced at the UC Davis DNA Technologies and Expression Analysis Core on an ABI Prism 3730 Genetic Analyzer with ABP Prism 3730 Data collection software (v3.0) and ABI Prism DNA Sequencing Analysis Software (v5.2). Chromatogram analysis was performed on Chromas software (v2.6.4, Technelysium Pty Ltd), and multiple alignment and mutation analysis of sequences was conducted using UGENE software (v1.28, Unipro) and Excel (Microsoft).

Cell preparation for scRNA-seq

For multi cell sorting prior to scRNA-sequencing, cells were stained as described in the flow cytometry section. Two CD45.1 mice were cotransferred with a 1:3 mixture of CD45.1⁺CD45.2⁺ Irf4^+/+ and CD45.2⁺ Irf4^+/− SW_HEL cells and immunized with HEL^2X-SRBC. On day 10 of the response, CD45.1 expression was distinguished using CD45.1-biotin and TotalSeq-A0951 PE Streptavidin (Biolegend) and equivalent numbers of splenic CD45.1⁺CD45.2⁺ Irf4^+/+ and CD45.2⁺ Irf4^+/− Fas^hiCD38^lo HEL-specific donor B cells were bulk sorted into a single polypropylene 5mL tube containing PBS using an Astrios EQ high speed cell sorter (Beckman Coulter). Viability was confirmed using a Countess (ThermoFisher) prior to loading onto the 10X Genomics Chromium fluidic system using 10X Chromium Single Cell 3’ GEM, Library & Gel Bead Kit v3 for single cell RNA extraction and processing. Sequencing was performed on a NovaSeq instrument (Illumina).

Informatics for scRNA-seq

Raw fastq files were processed with the 10x Genomics software Cell Ranger v. 3.1.0⁴¹ to perform alignment, filtering, barcode demultiplexing, and counting of UMI (unique molecular indices), using the 10X Genomics supplied mm10–3.0.0 genome and annotation. The CD45.2 (Irf4^+/−) and CD45.1/CD45.2 (Irf4^+/+) identity of the cells in the population was determined based on the frequency of the CITE-seq-encoded anti-CD45.1 biotin-streptavidin-PE-barcode (See Supplementary Fig. 2E). The two genotypes were separated by defining the breakpoint between distributions. Preprocessing, clustering, and visualization of single-cell gene counts was conducted using the R package Seurat, version 3.1.4⁴², in R version 3.6.1 (https://www.r-project.org/). Sequence counts from both experiments were pooled, log transformed, normalized, and batch corrected, then clustering was performed and t-SNE dimension-reduction coordinates were calculated on the top 100 principal components of the normalized, batch-corrected data. Identification of cluster markers was conducted using Wilcoxon rank sum tests as implemented in the Seurat function FindAllMarkers, and differential expression between groups was similarly conducted using Wilcoxon rank sum tests as implemented in the Seurat function FindMarkers. Enrichment analyses were conducted by using Fisher’s exact test to compare the proportion of genes in each custom set that was expressed in more than 50% of cells of a given cluster to what would be expected under random chance. Signature gene lists reported in Victora et al.⁴³ Supplementary Table 4e (derived from GSE38696) were used for cluster enrichment of LZ_up, DZ_up, LMP1_up, CD40_up, Myc_up, and Mitosis. Memory signature enrichment analysis used GSE89897 from Laidlaw et al.⁴⁴. The scRNA-seq data has been deposited at Gene Expression Omnibus with the GSE185188 identifier.

In vitro stimulation of B cells

Spleens of CD45.1⁺ CD45.2⁺ Irf4^+/+ or CD45.2^+/+ Irf4^+/− mice were processed and proportion of B cells determined by flow cytometry. Equivalent numbers of Irf4^+/+ or Irf4^+/− B cells were then added to a 96-well plate and B cells were cultured in RPMI medium (10% vol/vol FCS, 2 mM glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 20 mM HEPES buffer (all from Invitrogen), 1 mM Sodium Pyruvate, 1 mM nonessential amino acids (Gibco), and 2 μM β-mercaptoethanol (Invitrogen)) and stimulated with IgD dextran (0.2 μg/ml, Fina Biosolutions, LLC).

In vitro HEL^2X-IEα presentation

After B cell enrichment by MACs (Miltenyi), equal numbers of Irf4^+/+ or Irf4^−/− cells were co-cultured and number-matched to ensure each genotype had 37,500 SW_HEL cells seeing the antigen. Irf4^+/+ cells were labeled with CFSE to differentiate from Irf4^-/−. Cells were stimulated with an 11x titration series of HEL^2X-IEα, from 5 μg to 0.04 μg (and a 0 μg control) or HEL (0.2 μg) and incubated in 96 well flat bottom plates for 24 hours.

Statistical Analysis

Statistical analyses of parametric unpaired t-tests, paired t-tests, and Mann-Whitney U tests were performed using Prism software.

Results

Development of Irf4^+/− SW_HEL B cell system.

Expression levels of IRF4 regulate activated antigen specific B cell fate choices. Given the requirement for IRF4 during antigen specific priming and GC formation as well as the observation of diminished mutation frequency in Irf4-conditionally deleted polyclonal GCBs³, we hypothesized that IRF4 haploinsufficiency in responding B cells would compromise affinity maturation in the GC. We utilized a previously published VDJ knock-in mouse model (SW_HEL) in which B cells express the VDJ from the HyHEL10 mAb, specific for hen egg lysozyme (HEL). In this system, small numbers of CD45.2⁺ SW_HEL cells are adoptively transferred into CD45.1 mice, which upon immunization with HEL conjugated to SRBC (HEL-SRBC), antigen-specific B cell responses can be tracked and functionally interrogated³⁷. SW_HEL B cells are identified by flow cytometry using the CD45 congenic system as well as binding the HEL^WT antigen. In these experiments, we used an engineered form of HEL (HEL^2X) composed of 2 amino acid substitutions, R73E and D101R, that decrease its affinity for the HyHEL10 mAb by ~240 fold compared to wild type HEL⁴⁵. SW_HEL mice were bred to mice bearing the Irf4 null allele to generate SW_HEL Irf4^+/+ and Irf4^+/− mice that are CD45.2⁺. Advantageously, this monoclonal system enables the analysis of specific gene mutations on B cell fate and function during the antibody response without the confounder of cell competition in polyclonal systems.

Irf4^+/− SW_HEL B cells form normal GCs.

Either Irf4^+/+ or Irf4^+/− SW_HEL B cells were adoptively transferred into CD45.1⁺ congenic hosts and immunized with HEL^2X-SRBC (Fig. 1A). SW_HEL B cells activated and expanded well regardless of IRF4 gene copy number and sustained the response at intermediate (day 10) and late (day 15) timepoints, however, there were fewer Irf4^+/− SW_HEL cells at day 5 which corresponded to a decrement of PC frequencies as determined by downregulation of B220⁴⁵ (Fig. 1B, Supplementary Fig. 1A). This observation was corroborated using a BAC Prdm1:YFP reporter approach (see below and Fig. 6A). In addition, class switch recombination (CSR) to IgG1 was equally unaffected (Fig. 1C). Furthermore, GCB frequencies and numbers were unaffected as a function of Irf4 gene copy number, regardless of whether they were quantitated by Fas⁺CD38^-, Fas⁺GL7⁺ (Fig. 1D, Supplementary Fig. 1B), or by expression of the GC master transcription factor, Bcl6 (Fig. 1E). Due to the important roles of the DZ, LZ, and GZ in affinity maturation, we analyzed whether halving Irf4 gene copy number affected these cell states. Irf4^+/− SW_HEL GCBs exhibited comparable DZ, GZ, and LZ frequencies to wild type SW_HEL GCBs (Fig. 1F). Additionally, the proportions of DZ, GZ, and LZ antigen-specific cells in S phase, as measured by EdU incorporation, were indistinguishable (Fig. 1G). Finally, the expression of levels of key receptors that mediate GC entry and biology appear unchanged between the two genotypes (Supplementary Fig. 1C). These results indicate that to all appearances, Irf4 hemizygosity does not affect the ability of antigen specific B cells to differentiate into GCBs nor maintain the GC cell state over time.

Figure 1: Irf4^+/− SW_HEL GCB differentiation and function are normal.

A) Schematic of experimental design and SW_HEL gating strategy. 50,000 CD45.2⁺ Irf4^+/+ or Irf4^+/− SW_HEL B cells were mixed with HEL^2X-SRBC and adoptively transferred i.v. into CD45.1⁺ hosts. Spleens were harvested at days 5, 10, or 15 for analysis. Antigen-specific responding donor B cells were identified as CD45.2⁺CD45.1^-CD3^-B220⁺HEL^WT+. Flow plots represent sequential gating strategy. B) Absolute numbers of Irf4^+/+ or Irf4^+/− SW_HEL B cells, days 5, 10, and 15 post immunization. C-E) Representative flow cytometry plots, frequency, and absolute cell counts of C) IgG1-switched, D) CD38^loFas, E) BCL6⁺, Irf4^+/+ or Irf4^+/−SW_HEL B cells at the indicated time points after immunization. F) Representative flow cytometry plots and frequency of CD38^loFas⁺ DZ (CXCR4⁺CD86^-), GZ (CXCR4⁺CD86⁺), or LZ (CXCR4^-CD86⁺) Irf4^+/+ or Irf4^+/−SW_HEL GCBs at the indicated times after immunization. G) Frequency of EdU incorporation after a 4-hour pulse in Irf4^+/+ or Irf4^+/− SW_HEL GCBs within the DZ, GZ, and LZ sub-compartments gated as in F). B-G) Bar graphs represent mean ± SD with dots representing individual mice. Solid circle, Irf4^+/+ and open square, Irf4^+/− SW_HEL cells. Experiments in B–F are from 16 mice in 4 experiments performed while G is from 18 mice in 4 experiments performed; contour plots are concatenated files from all mice of a given group in a given experiment. Unpaired t tests were performed to determine significance. *** p ≤ 0.001; ns, not significant (p > 0.05)

Figure 6: Irf4 haploinsufficiency results in suboptimal Blimp-1 expression in high affinity cells.

A-D) Cells from Irf4^+/+ or Irf4^+/− Prdm1:YFP SW_HEL mice were transferred into recipient mice and immunized as in Fig. 1A prior to harvesting splenocytes at days 5 or 10 for analysis. Donor-specific responding B cells were identified as CD45.2⁺CD45.1^-CD3^-B220⁺HEL^WT+. A-B) Representative flow cytometry plots, frequency, and gMFI of YFP expression on A) total SW_HEL or B) total SW_HEL GCBs, identified by CD38^loFas⁺. C-D) Frequency and gMFI of YFP expression in C) total SW_HEL GCB DZ or LZ subsets or high affinity SW_HEL GCB DZ or LZ subsets (gated as in Fig. 1F). D) Cells from Irf4^+/+ or Irf4^+/− SW_HEL mice were transferred into recipient mice and immunized as in Fig. 1A prior to harvesting splenocytes at days 10 or 15 for analysis. Total SW_HEL, high affinity SW_HEL, or IgG1⁺ high affinity SW_HEL cells were gated for LZ GCBs as in Fig. 1F and the frequency and gMFI of CCR6 is depicted. Bar graphs represent mean ± SD or mean ± SEM (gMFI graphs) with dots representing individual mice. Experiments in A are from 17 mice in 4 experiments performed, experiments in B-C are from 10 mice in two experiments performed, experiment in D is from 10 mice in 1 experiment performed; contour plots are concatenated files from all mice of a given group in a given experiment. Unpaired t tests were performed to determine significance. ** p ≤ 0.01, *** p ≤ 0.001, ns, not significant (p > 0.05).

GCBs express low levels of IRF4.

These initial findings were surprising and prompted us to confirm a reduction in IRF4 levels in Irf4^+/− SW_HEL cells and GCBs. For this quantitative analysis, we co-transferred equal numbers of CD45.1⁺CD45.2⁺ Irf4^+/+ and CD45.2⁺ Irf4^+/− SW_HEL B cells into CD45.1⁺ hosts (Fig. 2A). Flow cytometric analysis of IRF4 protein expression at days 5 and 10 revealed a consistent 10–20% reduction in IRF4 gMFI in total SW_HEL cells as well as Bcl6-expressing GCB Irf4^+/− HEL^WT-specific B cells compared to that of Irf4^+/+ (Fig. 2B). Importantly, IRF4 expression in host resting B cells was lower than activated Irf4^+/− SW_HEL cells, demonstrating that the Irf4^+/− SW_HEL cells experienced BCR-dependent IRF4 upregulation (Fig. 2B). Polyclonal host GCBs, which are Irf4^+/+, also expressed higher levels of IRF4 compared to the Irf4^+/− SW_HEL GCBs. Further, BCR stimulation in vitro using an anti-IgD dextran agonistic reagent, resulted in roughly two-fold less induced IRF4 protein expression in Irf4^+/− compared to Irf4^+/+ cells (Supplementary Fig. 1D). Together, these results demonstrate that Irf4^+/− SW_HEL GCBs undergo BCR-dependent IRF4 induction that is reduced in an Irf4 gene copy-dependent manner.

Figure 2: IRF4 expression in hemizygous cells and in GCBs.

A) Schematic of experimental design and donor gating strategy. Irf4^+/+ (CD45.1⁺CD45.2⁺) and Irf4^+/− (CD45.2⁺) SW_HEL cells were transferred at a ratio of 1:1 (total 50,000 cells), mixed with HEL^2X-SRBC, and adoptively transferred i.v. into CD45.1⁺ hosts. Spleens were harvested at days 5 or 10 for analysis. Representative gating strategy is shown to identify Irf4^+/+ and Irf4^+/− SW_HEL responders by allotypic markers. Identification of B220⁺HEL^WT+ cells is performed as in Fig. 1A. B) IRF4 geometric mean fluorescence intensity (gMFI) of total SW_HEL cells (left) or BCL6⁺ GCB SW_HEL cells (host CD38⁺ are not Bcl6⁺) (right) at the indicated time points after immunization. Lines linking dots denote measurements from a single mouse. At day 10, solid line indicates gMFI values from experiment 1, dotted line indicates gMFI values from experiment 2. C) IRF4 gMFI in Irf4^+/+ total SW_HEL B cell subsets at the indicated time points after immunization. Naïve host CD38⁺ B cells (CD45.2^-CD45.1⁺B220⁺CD38⁺Fas^-), GC: SW_HEL GCBs (CD45.2⁺CD45.1^-B220⁺HEL⁺CD38^loFas⁺), M: SW_HEL memory B cells (CD45.2⁺CD45.1^-B220⁺HEL⁺CD38^hi), PC: SW_HEL PCs (CD45.2⁺CD45.1^-B220^loHEL⁺CD38⁺). Lines linking dots denote measurements from a single mouse. Experiments in B (Day 5) and C are from 5 mice in 1 experiment performed, while experiments in B (Day 10) are from 8 mice in 2 experiments performed. Paired t tests were performed to determine significance. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001, ns, not significant (p > 0.05)

In these experiments, we noticed that basal IRF4 expression in Irf4^+/+ SW_HEL GCBs was higher than that of CD38⁺ resting B cells. Because this contradicted earlier results measured using immunohistochemistry⁴⁶, we quantified IRF4 expression as a function of resting CD38⁺ B cells, GCB, memory, and PC cell states by flow cytometry (Fig. 2C). As expected, PCs expressed the highest levels of IRF4 protein, ≥10-fold higher than that of resting B cells. GCBs express twice as much IRF4 protein compared to resting B cells. Together, these data indicate that GCBs express IRF4; this new basal level is greater than that of resting B cells but less than that of PCs.

Irf4 haploinsufficiency impairs affinity maturation.

To test the functionality of affinity maturation in Irf4^+/− GCBs, we analyzed the generation of high affinity cells using a system developed by the Brink lab³⁷. Specifically, immunization with low affinity HEL drives the formation of GCBs exhibiting a canonical SHM pattern that can be enumerated by flow cytometry by virtue of binding to sub-saturating amounts of HEL^3X. HEL^3X contains an additional amino acid substitution (R21Q) compared to HEL^2X, which results in a ~40 fold diminished affinity compared to HEL^2X to the parental unmutated SW_HEL BCR^45,47. Using this system, analysis of Irf4^+/+ SW_HEL cells at day 10 revealed that approximately 30% of IgG1-switched GCBs bound HEL^3X (Fig. 3A). In contrast, analysis of similarly treated Irf4^+/− SW_HEL cells resulted in roughly 10% of IgG1-switched GCBs binding HEL^3X (Fig. 3A). We confirmed specificity of HEL^3X binding using a separate cohort of mice that were immunized with HEL^WT and known to not develop⁴⁷ GCBs capable of binding subsaturating amounts of HEL^3X. The reduction in HEL^3X-binding cells was not specific to IgG1-switched B cells because analysis of total B220⁺ SW_HEL GCBs that contain other isotypes revealed a similar decrement (Fig. 3B). We asked whether this reduction was due to reduced BCR abundance; however, based on HEL^WT- or anti-BCR staining we saw no difference in SW_HEL BCR gMFI on GCBs at day 10 or 15 (Supplementary Fig. 1E). At day 15, proportions of HEL^3X-binding GCBs increase. Interestingly, we observed comparable HEL^3X-binding frequencies between Irf4^+/+ and Irf4^+/− -switched or total SW_HEL GCBs at this later time point. Despite compromised early affinity maturation, the ratio of LZ:DZ populations among high affinity cells was equivalent in Irf4^+/+ and Irf4^+/− at both days 10 and 15 (Fig. 3C). Further, the defect in efficient affinity maturation at day 10 was not due to cell proliferative capacity; comparable proportions of Irf4^+/+ and Irf4^+/− high affinity cells were in S-phase in all GC compartments (Fig. 3D). Together, these results demonstrate delayed affinity maturation efficiency in Irf4 haploinsufficient GCBs.

Figure 3: Irf4 haploinsufficiency impairs affinity maturation.

Irf4^+/+ or Irf4^+/− SW_HEL B cells were transferred into recipient mice and immunized as in Fig. 1A prior to harvesting splenocytes at days 10 or 15 for analysis. A-D) Donor-specific SW_HEL GCBs were identified as CD45.2⁺CD45.1^-CD3^-B220⁺CD38^loFas^hi or CD45.2⁺CD45.1^-CD3^-B220⁺CD38^loFas^hiIgG1⁺ and high affinity cells as those that bound subsaturating amounts of HEL^3X. A-B) Mice immunized with either HEL^2X-SRBC or HEL^WT-SRBC and representative flow cytometry plots, frequency, and absolute cell counts of A) IgG1⁺, and B) B220⁺ high affinity SW_HEL B cells were assessed. C) Ratio of high affinity Irf4^+/+ or Irf4^+/− in the LZ and DZ compartments, identified as in Fig. 1F. D) Frequency of EdU incorporation after a 4-hour pulse in high affinity SW_HEL GCB subsets. A-D) Bar graphs represent mean ± SD with dots representing individual mice. Experiments in A–C are from 15 mice in 3 experiments performed, and experiment D is from 14 mice in 3 experiments performed; contour plots are concatenated files from all mice of a given group in a given experiment. Unpaired t tests were performed to determine significance. ** p ≤ 0.01, **** p ≤ 0.0001, ns, not significant (p > 0.05)

Irf4 hemizygous GCBs exhibit comparable rates of SHM.

The Aicda gene, which encodes the AID protein, has been shown to be regulated by IRF4 activity^2,3. To determine whether attenuated affinity maturation in Irf4^+/− GCBs is due to issues with AID-induced SHM, single SW_HEL GCBs were sorted on days 10 and 15 and the heavy chain VDJ region was sequenced. For both genotypes, the number of DNA mutations increases over time, with an average of 3 mutations/cell on day 10, to 5 mutations/cell on day 15, indicating comparable rates of SHM (Fig. 4A). In both Irf4^+/+ and Irf4^+/− GCBs, mutations are dispersed across the entirety of the VDJ region and the vast majority of mutations are within AID hotspots (Supplementary Fig. 2A and not shown), confirming equivalent AID activity. Furthermore, the DNA mutations were not silent because similar proportions of amino acid substitutions per cell were observed (Fig. 4B). Despite these findings, the proportions of cells with the canonical high affinity mutations⁴⁷, Y53D and/or Y58F, are reduced in Irf4^+/− GCBs on day 10 (Fig. 4C) and are consistent with diminished HEL^3X-binding frequencies (Fig. 3A–B). In contrast, by day 15, proportions of high affinity mutations are similar between genotypes, consistent with HEL^3X-binding frequencies (Fig. 3A–B). In addition to sorting and sequencing from total SW_HEL, we sorted and sequenced from the HEL^3X-binding populations to enrich for somatically mutated cells and observed the same trends (Fig. 4D–F). We note that, in some samples, the enrichment was somewhat inefficient, likely due to the position of the gate during the sort. Together, we observe impaired generation of high affinity mutations in Irf4^+/− SW_HEL GCBs at early time points despite efficient rates of SHM.

Figure 4: Irf4 hemizygous GCBs exhibit comparable rates of SHM.

A-F) Cells were transferred into 3 recipient mice per group and immunized as in Fig. 1A prior to harvesting splenocytes at days 10 or 15 for sorting. Individual SW_HEL GCB clones were sorted by FACS on days 10 and 15 post immunization and the heavy chain VDJ was sequenced. A-C) Total SW_HEL donor-specific responding GCBs were identified as CD45.2⁺CD45.1^-CD3^-B220⁺HEL^WT+CD38^loFas⁺. A) Number of DNA mutations per clone, B) Number of amino acid mutations per clone, and C) Proportion of clones with high affinity mutations Y53D and Y58F, number in pie represents total number of clones sequenced. D-F) Analysis as in A-C for GCBs that bind subsaturating amounts of HEL^3X (high affinity cells). A,B,D,E) Each dot represents an individual clone, with the median indicated as a black line. Mann-Whitney U tests were performed to determine significance. C,F) Fisher’s exact tests were performed to determine significance; analysis based on total high affinity mutations. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001, ns, not significant (p > 0.05).

GC subpopulations are intact despite Irf4 haploinsufficiency.

Cumulatively, the impaired affinity maturation of Irf4^+/− SW_HEL cells cannot be accounted by GCB phenotype, rates of SHM, or apparent proliferative abilities. Recent observations suggest deeper divisions in GCB cell subset identity beyond the foundational LZ/DZ paradigm^15–17, raising the possibility that Irf4^+/− SW_HEL B cell state transitions contribute to impaired affinity maturation. To address this, we performed droplet-based scRNA-seq of Irf4^+/+ and Irf4^+/− SW_HEL GCBs. To enable a direct comparison of cells within the same GC microenvironments, we co-transferred cells of each genotype into the same host in a 3:1 ratio of Irf4^+/− to Irf4^+/+ SW_HEL B cells; the results of the cotransfer are described in more detail in a following section. Equal numbers of Irf4^+/+ and Irf4^+/− SW_HEL GCBs from 2 mice were separately sorted and processed into two scRNA-seq runs. Viable cells were processed on the 10x Genomics Chromium platform and RNA-seq libraries were generated for a total of 8360 cells comprised of GCBs of both genotypes. We utilized CITE-seq, Cellular Indexing of Transcriptomes and Epitopes by Sequencing, to resolve the SW_HEL cells of the respective genotypes that differed in expression of CD45.1 (see methods)⁴⁸. An average of 139,000 reads/cell and 900 genes/cell were identified from both experiments. Importantly, sequence counts of the CITE-seq barcode yielded a bi-modal distribution clearly separating cells from each genotype (Supplementary Fig. 2E). Sequence counts from both experiments were pooled, log transformed, normalized, and batch corrected, then clustering was performed and t-SNE dimension-reduction coordinates were calculated on the top 100 principal components of the normalized, batch-corrected data (Fig. 5A). While 10 cell clusters were initially identified, three were removed because they represented T, myeloid/dendritic cell, and non-MHCII expressing lineages (data not shown); the genes distinguishing each cluster are depicted in Supplementary Figure 3A. Within the remaining 7 clusters, there was comparable representation of cells from each genotype except for cluster 4, which comprised of 15% Irf4^+/− cells (Fig. 5B). Cells were also projected as a function of the cell cycle and consistent with the EdU metabolic labelling analysis, we observed equal representation of cells of each genotype in each cell cycle phase (Supplementary Fig. 3B). Gene set enrichment analysis was used to assign cell state identity of each cluster. Specifically, Fisher’s exact test was used to compare the proportion of genes in each custom set that was expressed in more than 50% of cells of a given cluster to what would be expected under random chance (Fig. 5C). As with recently published analyses, this resulted in the resolution of multiple GCB subsets that corresponded to LZ, DZ, GZ, and LZ-like populations. Differential gene expression within each cluster was next used to determine whether the cell state was differentially programmed by Irf4 haploinsufficiency. This analysis, at the sequencing depth attained, retrieved few significant differentially expressed genes not known to be relevant for GC biology and only in clusters 0,1,2, and 4 (Supplementary Table I). Together, this analysis suggests that impaired affinity maturation of Irf4 haploinsufficient cells is not attributable to disadvantaged representation of GCB cell subcompartments, except perhaps a pre-memory compartment (cluster 4).

Figure 5: GC subpopulations are intact despite Irf4 haploinsufficiency.

Irf4^+/+ or Irf4^+/− SW_HEL B cells were co-transferred into recipient mice at a 1:3 ratio and immunized as in Fig. 2A prior to sorting GCBs at day 10 (identified as in Fig. 1D) and droplet-based scRNA-seq analysis of 8360 cells in two separate adoptive transfers. CITE-seq using a barcoded Streptavidin-PE reagent bound to anti-CD45.1 biotin was used to discriminate the Irf4^+/+ and Irf4^+/− genotypes among sequenced cells (see Methods). A) t-SNE dimension-reduction coordinates were calculated on the top 100 principal components of the normalized, batch-corrected data. B) Number of cells in each cluster that are Irf4^+/+ (black bars) or Irf4^+/− (gray bars). C) Enrichment of indicated gene signatures within each cluster calculated using Fisher’s exact test.

Irf4 haploinsufficiency results in suboptimal Blimp-1 expression in high affinity cells

To determine the basis for impaired affinity maturation in Irf4^+/− GCBs, we sought to determine whether the high affinity cells exhibited diminished proportions of presumptive pre-PC^49,50. To identify pre-PC cells, we utilized Prdm1:YFP mice³⁹ that encode a YFP-modified BAC allele of genomic Prdm1 (the gene encoding the PC fate determinant, Blimp-1) and generated SW_HEL Blimp-1YFP reporter mice with either half or the full complement of Irf4 gene copies. As previously shown, expression of YFP indicates activation of Blimp-1 expression and marks cells with a PC and pre-PC phenotype. At the peak of the extrafollicular PC response, day 5, Irf4^+/− SW_HEL cells generated 2–3 fold fewer YFP-expressing PCs compared to Irf4^+/+ SW_HEL cells (Fig. 6A) as previously predicted. By day 10, the frequency of YFP⁺ cells markedly contracted (Fig. 6A). Interestingly, the intensity of YFP expression, as measured by geometric mean fluorescence intensity (gMFI), was decreased amongst Irf4^+/− PCs compared to Irf4^+/+ counterparts at day 5 (Fig. 6A). To specifically quantify newly selected YFP⁺ pre-PCs^49,50, we gated on SW_HEL GCBs and in the zonal compartments (Fig. 6B–C). Similar to what was observed in total SW_HEL B cells at days 5 and 10, we observed a marked decrease in Irf4^+/− YFP⁺ SW_HEL GCBs (Fig. 6B). As previously shown, the YFP⁺ GCBs of both genotypes were enriched in the DZ compartment and rare in the LZ (Fig. 6C). Strikingly, we observed a decrease in YFP intensity among Irf4 haploinsufficient high affinity HEL^3X-binding SW_HEL DZ GCBs (although not in total SW_HEL DZ GCBs). Together, these results confirm predictions regarding IRF4 cell concentrations and the generation of PCs, as well as revealed a dose response relationship between IRF4 amounts and Blimp-1 levels. Furthermore, the results raise the possibility that suboptimal expression of Blimp-1 attenuates the generation of high affinity GCBs during affinity maturation.

Similarly, as precocious generation of presumptive pre-memory cells could dampen affinity maturation dynamics, we determined the proportions of CCR6-expressing LZ GCBs⁵¹. This analysis revealed comparable frequencies amongst total and high affinity HEL^3X-binding SW_HEL LZ GCBs (Fig. 6D). Thus, Irf4 haploinsufficiency does not alter the generation of GCB memory precursor cells.

Irf4^+/− SW_HEL cells fail to recruit optimal T cell help when in a competitive environment.

Our cumulative findings demonstrate impaired affinity maturation and Blimp-1 expression levels among presumptive positively-selected HEL^3X-binding DZ GCBs despite comparable rates and targeting of AID-dependent SHM. We reasoned that a competition experiment between Irf4^+/+ and Irf4^+/− SW_HEL cells would illuminate which check point(s) is critical for efficient affinity maturation when haploinsufficient for Irf4. For this analysis, we performed a co-transfer of equal numbers of Irf4^+/+ and Irf4^+/− SW_HEL B cells; Irf4^+/− cells were CD45.2, Irf4^+/+ cells were CD45.1/CD45.2 and the host mouse was CD45.1 (Fig. 2A) and determined whether Irf4^+/− day 10 GCBs might be disadvantaged upon HEL^2X-SRBC immunization. Strikingly, we found that Irf4^+/− GCBs were ~10 fold outcompeted (Fig. 7A). Moreover, adjustment of the Irf4^+/− to Irf4^+/+ SW_HEL cell ratio to 3:1 also resulted in outcompeting Irf4^+/+ GCBs (Supplementary Fig. 2B). To determine the timing of Irf4 haploinsufficient cell failure, we measured their frequencies in a 2:1 Irf4^+/−:Irf4^+/+ transfer over time (Fig. 7B). Importantly, we observed a strong depletion of responding Irf4^+/− SW_HEL cells and their differentiated progeny by day 5 of the response that was most pronounced for the PC fate, demonstrating the dominance of two copies of Irf4 compared to one in the response (Fig. 7C). At day 5, we estimated cell proliferation by metabolic labeling with the BrdU thymidine nucleoside analog and cell death by detection of cleaved activated Caspase 3 using a specific mAb. The analysis demonstrated comparable incorporation of BrdU and frequencies of cells undergoing Caspase 3 cleavage (Fig. 7D). Interestingly, at later time points the proportions of Irf4^+/− GCBs declined at a slower rate (Fig. 7B) without a noticeable effect on a particular zonal subset (data not shown). Impaired generation of HEL^3X-binding cells was also observed in this setting (Supplementary Fig. 2C). Thus, we conclude that a large fraction of Irf4 haploinsufficient cells is disadvantaged for T cell help; however, those that successfully compete proliferate and survive with near-comparable frequencies. Interestingly, this experiment highlights the dependency of PC for the full complement of Irf4 alleles.

Figure 7: Irf4^+/− SW_HEL cells fail to recruit optimal T cell help when in a competitive environment.

Irf4^+/+ or Irf4^+/− SW_HEL B cells were co-transferred into recipient mice and immunized as in Fig. 2A prior to analysis at the indicated time points. Donor-specific responding Irf4^+/+ cells were identified as CD45.2⁺CD45.1⁺CD3^-B220⁺HEL^WT+ and donor- specific responding Irf4^+/− cells were identified as CD45.2⁺CD45.1^-CD3^-B220⁺HEL^WT+. A) Frequency of Irf4^+/+ and Irf4^+/− total SW_HEL cells at day 10, 1:1 initial cotransfer ratio. B) Frequency of Irf4^+/+ and Irf4^+/− total SW_HEL cells or total SW_HEL GCBs at indicated time points, 1:2 initial cotransfer ratio. C-D) Flow cytometry analyses of 1:2 Irf4^+/+:Irf4^+/− cotransfer at day 5. C) Proportion of GCB, PC, or memory populations (identified as in Fig. 2C legend) among total SW_HEL donors. D) Frequency and gMFI of BrdU incorporation or active Caspase 3 among total SW_HEL or total SW_HEL GCB cells at day 5. E) Flow cytometry analyses of 1:1 Irf4^+/+:Irf4^−/− cotransfer and HEL-IEα s.c. immunization (day 1.5 post transfer and 24 hours post immunization). Frequency and gMFI of Y-Ae on total SW_HEL B cells. Lines linking dots denote measurements from a single mouse. Experiment in A is from 10 mice in 2 experiments performed, and experiments in B-D are from 5 mice in one experiment performed. Experiment in E is from 8 mice in 2 experiments performed. Paired t tests were performed to determine significance. *** p ≤ 0.001, **** p ≤ 0.0001, ns, not significant (p > 0.05).

Clonal selection of specified GCBs is T cell dependent⁹, suggesting that Irf4 haploinsufficient cells compete poorly for T cell help. It has been shown that dendritic cells require IRF4 for efficient antigen presentation⁵², raising the possibility that Irf4^+/− B cells may be presenting less pMHC during this T cell checkpoint. To test this possibility, we developed a fusion protein antigen, HEL-IEα, comprised of HEL^2X, IEα_46–74, OVA_327–339, and a 6XHIS tag. The terminal tag enabled protein purification from supernatants of engineered D. melanogaster S2 cells. The pOVA component is irrelevant for these experiments. Thus, recognition and internalization of HEL-IEα by SW_HEL cells would yield pMHC complexes, comprised of IEα:IA^b, detectable by the Y-Ae mAb⁵³. Irf4^+/+ and Irf4^+/− SW_HEL cells were transferred into congenic hosts in a 1:1 ratio, immunized with HEL-IEα s.c., and their expression of IA^b and Y-Ae were quantified one day later. Interestingly, cells of both genotypes expressed comparable MHCII and Y-Ae density and frequency (Fig. 7E), suggesting that Irf4 haploinsufficiency does not compromise antigen presentation in the B cell lineage. In fact, similar in vitro experiments suggest that complete Irf4 deficiency does not control antigen presentation efficiency by B cells (Supplementary Fig. 2D).

Discussion

The prevailing model for GCB selection upon entry into the LZ from the DZ involves competition for antigen displayed on FDC and subsequent processing and presentation of pMHC complexes for Tfh cells¹. Limiting numbers of Tfh^19,21,54, their ability to sense pMHC density²¹, and a subsequent CD40 / ICOSL feed forward loop¹³ promotes the fittest GCBs for DZ re-entry, mitosis, and differentiation. A subset of these cells have also induced the Myc and IRF4 transcription factors^27,28. Here we reveal an Irf4 haploinsufficiency in selection of high affinity cells and their ability to upregulate optimal Blimp-1 expression. In contrast, GCB attributes of proliferation (Fig. 1G, 3D), phenotype (Fig. 1D, S1B), zonal compartmentalization (Fig. 1F, 3C), subset distribution observed in scRNA-seq analysis (Fig. 5), class switch recombination (Fig. 1C) and rates of SHM (Fig. 4) are unaffected by Irf4 gene copy reduction. Interestingly, when in competition, Irf4 heterozygous cells are substantially disadvantaged at the Tfh-mediated GCB entry checkpoint (Fig. 7B), suggesting that (1) diminished recruitment of Tfh signals and (2) subsequently dampened upregulation of high levels of Blimp-1 expression (Fig. 6C) underlies impaired affinity maturation of Irf4 hemizygous GCBs. We note that these two key findings precisely position IRF4 activity at the known IRF4 steps of receipt of signals from selecting T cells and execution of the PC gene program.

These data extend previous observations and reveal new biology of IRF4 in GCBs. First, Cγ1-Cre mediated deletion of Irf4 during the polyclonal response to NP-KLH revealed a trend towards fewer W33L mutations events³, suggesting impaired affinity maturation. Second, ERT2-Cre mediated deletion of Irf4 using Tamoxifen treatment during peak GCB responses showed normal GC functionality²³ consistent with the data herein that IRF4 is dispensable for most GC biology. Interestingly, this study also showed that Irf4^+/− GCBs are impaired in acquiring a pre-PC phenotype although affinity maturation was not addressed. This observation is consistent with the impaired Blimp-1 expression in Irf4 haploinsufficient cells presented in this work (Fig. 6). Third, augmented IRF4 expression caused by inactivation of the Cbl ubiquitin ligases in GCBs results in precocious PC differentiation accompanied by reduced acquisition of high affinity conferring somatic mutations³⁶. These results, together with the observation that positively selected cells upregulate IRF4, suggest a model whereby levels of induced IRF4 expression are functionally limiting for dictating the fate of cells that have acquired high affinity-conferring somatic mutations and have successfully recruited Tfh selecting signals. In contrast to previous findings that reported negligible expression of IRF4 in GCBs⁴⁶, we found that IRF4 levels in GCBs were above that seen in resting CD38⁺ B cells (Fig. 2C), suggesting that GCBs express an augmented basal level of IRF4 expression. The role of this new basal level is unclear given the lack of a GC phenotype in our system and upon tamoxifen-induced deletion²³; however, we suggest that this basal level may counter the degradation rate imposed by Cbl activity in order to facilitate the acquisition of an IRF4 high cell state during positive selection and pre-PC differentiation.

We observed Irf4 haploinsufficiency for the generation of somatically mutated high affinity B cells during the onset of selection which was less pronounced at later time points, demonstrating that a two-fold reduction in Irf4 gene copies impairs affinity maturation (Fig. 3A, B). Sequence analysis revealed that the acquisition of the canonical Y53D and Y58F mutations were initially decreased amongst HEL-binding GCBs but not so at the later time point, despite comparable somatic mutation rates throughout (Fig. 4C, F). A possibility we entertained was that cells from the later time point were enriched for superadded high-affinity conferring somatic mutations that might function as a second site suppressor to enhance IRF4 inducibility by more intense BCR signaling. However, the sequence analysis did not support this hypothesis, suggesting that other pathways seemingly function to augment affinity maturation at later time points. Furthermore, we did not observe enhanced frequencies of CCR6 expressing pre-memory cells amongst Irf4^+/− high affinity GCBs⁵¹ (Fig. 6D), excluding the possibility that Irf4^+/− cells were pre-disposed to a memory cell fate trajectory, preventing their accumulation due to premature exit from the cell cycle. In contrast, we found a pronounced Irf4 haploinsufficient role for PCs and putative high affinity pre-PCs which manifests as a reduction in frequencies and YFP intensities, the latter as estimated with a BAC-encoded fluorescent protein reporter system (Fig. 6A–C). This suggests arrested and incomplete transition to a progenitor PC state that is perhaps accompanied with decelerated cell cycle progression⁵⁵. In addition, this implies that the process of selection is coupled to PC fate decisions. Perhaps the lack of a difference in Blimp-1 YFP intensity amongst total GCB DZ cells reflects the proposition that not all such cells exhibit PC potential⁵⁰. A deficit in high affinity pre-PCs is supported by our other observations that extrafollicular PCs depend on the full complement of Irf4 alleles, either when alone or in competition. Importantly, these observations also demonstrate, for the first time, that IRF4 not only activates Blimp-1 expression but also sets its levels. Together we propose that Irf4^+/− cells are inherently capable of acquiring high affinity conferring somatic mutations yet are defective in leveraging selection signals into productive PC fate outcomes. Future experiments utilizing Blimp-1 hemizygous SW_HEL cells may resolve and/or extend this interpretation. Finally, we speculate that the apparent normalization of high affinity cell frequencies at later time points (Fig. 3A,B) continues to reflect an Irf4 haploinsufficient deficit and reflects an accumulation of cells unable to exit, rather than accumulating and exiting. This will be resolved when the rate of the latter can be reliably measured.

The defect in upregulating Blimp-1 to high levels may constitute part of a two-step (or more) defect of Irf4 hemizygous cells. Specifically, when Irf4 wild type and heterozygous cells are in competition, the latter are disadvantaged at stages thought to be dependent on the receipt of T cell help. Interestingly, MHCII levels⁸ (Fig. S1C), pMHC density⁹ (Fig. 7E, S2D), ICAM density¹⁰ (Fig. S1C), and proliferation (Fig. 1G, 7D), parameters known to compromise this selection checkpoint, are not impaired, suggesting that other, remaining to be identified factors, compromise Irf4 haploinsufficent cells. The disadvantaged state is propagated, albeit to a lesser extent after GC commitment, leading us to propose that this feature of Irf4^+/− cells functions during selection of high affinity cells and lies upstream of the defect in upregulating optimal Blimp-1 levels. As Blimp-1 is a target gene of IRF4, it is likely that part of the problem in selection is in the execution of the Irf4 gene regulatory network. However, selection of high affinity GCBs also depends upon successful competition with limiting numbers of selective Tfh. Thus, whether Irf4 haploinsufficient cells are coincidentally defective in sensing T cell signals upon limited interactions with Tfh remains to be investigated. Our findings raise the possibility that stochastic and/or regulated dampening of IRF4 expression plays a role in maintaining GCB clonal diversity^56,57. Given the identification of IRF4 loss of function mutations in people⁵⁸, our work uncovers a B cell specific haploinsufficient effect of Irf4 gene copy loss during antibody responses.

Key Points.

• IRF4 is limiting for efficient production of high affinity GCBs

• Irf4 haploinsufficiency causes sub-optimal regulation of Blimp-1 expression in GCBs

• Irf4 haploinsufficient cells are competitively disadvantaged for T cell help