BRWD1 directs B cell epigenetic states required for germinal center initiation, maintenance and function

Abstract

Germinal center (GC) B cells, differentiated from follicular B cells following activation, can be assigned to light zone (LZ), dark zone proliferation (DZp), or dark zone differentiation (DZd) states, as compartmentalized by the antagonistic molecular programs of selection, proliferation, and somatic hypermutation respectively. In the bone marrow, an epigenetic reader, BRWD1, orchestrates the sequential stages of B cell progenitor proliferation and B cell receptor recombination. Here, we analyze mice with a conditional knockout of Brwd1 in follicular B cells (Cd23^Cre/wt;Brwd1^fl/fl) and GC B cells (Aicda^Cre/wt;Brwd1^fl/fl) to investigate the functions of BRWD1 during peripheral B cell activation. Compared to Brwd1^fl/fl mice, Cd23^Cre/wt;Brwd1^fl/fl mice show diminished GC B cell responses and class switch recombination after immunization. By contrast, Aicda^Cre/wt;Brwd1^fl/fl mice exhibit increased GC B cell proliferation but defects in affinity maturation and tolerance. Somatic hypermutation and differentiation into memory and plasma cells are intact. RNA-seq and ATAC-seq analyses of GC B cells from Aicda^Cre/wt;Brwd1^fl/fl mice indicate that BRWD1 deficiency alters the transcriptional and epigenetic identities of the LZ, DZp, and DZd states. Our results thus demonstrate that BRWD1 regulates multiple transitions along B cell developmental and activation pathways, suggesting stringent epigenetic control of GC states drives positive and negative selection.

The epigenetic reader BRWD1 is central for B cell progenitor development in the bone marrow, but its role in the adaptive immune response is not well understood. Here, the authors analyze conditional Brwd1 knockout and immunization models, finding transcriptional germinal center B cell subset identities to depend on BRWD1.

Introduction

Central to the initiation of adaptive humoral immunity in mammals is the activation and differentiation of B cells located within the follicles of secondary lymphoid organs¹. In T cell-dependent responses, activated follicular (Fol) B cells can differentiate into germinal center (GC) B cells over 4 days, a multi-stage process which includes upregulation of the transcription factor (TF) BCL6^2–7. This pathway from Fol B cell to GC B cell involves large shifts in molecular programs, including changes in TF networks, chromatin accessibility, and three-dimensional (3D) chromatin architecture^8–12. It is during this transition from Fol to GC B cells that B cells undergo class switch recombination (CSR)¹³.

Following GC entry, B cells further differentiate. While histologically the GC has two zones, a dark zone (DZ) and a light zone (LZ), we have demonstrated that the DZ contains two distinct populations of B cells that differ in their location and molecular programs¹⁴. This three-zone model organizes fundamentally different GC programs into specific cell states^15,16. First, in the LZ, CXCR4⁻CD83⁺ GC B cells are selected by capturing antigen displayed on follicular dendritic cells and presenting this antigen to T follicular helper cells (Tfh) cells^17–27. Second, in the DZ proliferation (DZp), CXCR4⁺CD83⁺ B cells undergo mitosis or apoptosis adjacent to tingible body macrophages. Third, in the DZ differentiation (DZd), CXCR4⁺CD83⁻ B cells exit the cell cycle and upregulate the mechanisms of somatic hypermutation (SHM). Transit through GC subsets is tightly regulated, with LZ selection dictating a precise and limited proliferative program in the DZp. B cells rapidly cycle through the GC until they differentiate into memory B cells (MBC) or plasma cells (PC)¹. Consistent with their functional specialization, the three GC B cell subsets have large differences in transcriptional programs and chromatin accessibility^14,28. However, the mechanisms controlling rapid transit through these three different GC B cell states are unknown.

B cell progenitors also experience large shifts in molecular programs as they develop in the bone marrow^29,30. Specifically, the epigenetic reader bromodomain and WD repeat-containing protein 1 (BRWD1) mediates the transition between proliferating large pre-B cells and small pre-B cells undergoing light chain recombination via multiple mechanisms^31–33. At the local scale, BRWD1 repositions nucleosomes relative to DNA GAGA motifs to open and close chromatin³¹. BRWD1 also mediates long-range chromatin contacts over a megabase scale by converting static to dynamic cohesin capable of loop extrusion³³. In these ways, BRWD1 coordinates the transcriptional, epigenetic, and chromatin 3D transitions necessary for B cell progenitors to stop proliferation and begin recombination³³. BRWD1 is also highly expressed in both Fol and GC B cells, yet its contributions to the cell state transitions between each population in response to antigen have not been explored.

Herein, we investigate the role of BRWD1 in peripheral B cells using Brwd1-floxed mice to delete Brwd1 in Fol and GC B cells. We demonstrate that BRWD1 is important for GC initiation and CSR in activated Fol B cells. In GC B cells, we find that loss of Brwd1 results in a striking collapse of epigenetic transitions and dysregulation of transcriptional programs within the three GC B cell subsets. These results indicate that at multiple stages, BRWD1 regulates B cell transitions between proliferation and states in which DNA is recombined or mutated to generate diversity and shape immune responses to infection.

Results

Design of a Brwd1-floxed mouse

Brwd1 is first induced in small pre-B cells during B lymphopoiesis in the bone marrow, where it mediates widespread epigenetic changes^31,32. Using RNA-sequencing (RNA-seq) from different B cell populations in WT mice, we found that Brwd1 expression was over twice as high in Fol B cells as in small pre-B cells (Supplementary Fig. 1a). Furthermore, Brwd1 expression changed across GC B cell subsets, with the greatest expression in the DZd¹⁴. This dynamic regulation of Brwd1 in Fol and GC B cells suggested that BRWD1 may mediate both transitions between and maintenance of peripheral B cells.

To study the role of BRWD1 in peripheral C57BL/6 B cells, we derived a Brwd1-floxed allele with Lox71 and Lox66 sites in opposite orientations surrounding Brwd1 exons 6, 7, and 8 (Supplementary Fig. 1b). We first crossed these mice with mice expressing Cre under the Fcer2a promoter (Cd23^Cre)³⁴. The targeting construct includes an inverse tdTomato (tdT) reporter so that following Cre recombination, exons 6, 7, and 8 are inverted and tdT is expressed (Supplementary Fig. 1c). We confirmed the presence of the Lox71 site by PCR (Supplementary Fig. 1d). Using RNA-seq, we confirmed that tdT⁺ Fol B cells in Cd23^Cre/wt;Brwd1^fl/fl mice do not transcribe exons 6, 7, and 8 of Brwd1 (Supplementary Fig. 1e). Homozygous Cd23^Cre/wt;Brwd1^fl/fl mice had greater tdT median fluorescence intensity (MFI) than heterozygous Cd23^Cre/wt;Brwd1^fl/wt mice, demonstrating that both alleles were deleted in homozygous mice (Supplementary Fig. 1f). Finally, western blot of BRWD1 confirmed the absence of the full length 263 kD protein in tdT⁺ Fol B cells from Cd23^Cre/wt;Brwd1^fl/fl mice and little but detectable expression in the corresponding tdT⁻ Fol B cells (Supplementary Fig. 1g). A truncated approximately 11 kD BRWD1 protein was detected only in tdT⁺ Fol B cells (Supplementary Fig. 1h). This truncated protein was consistent with expression of exons 1 through 5, which are before any functional BRWD1 domains.

BRWD1 is not necessary for the maintenance of Fol B cells

To study whether BRWD1 is important for the maintenance of peripheral B cells, we characterized Cd23^Cre/wt;Brwd1^fl/fl mice at the steady state. Flow cytometry demonstrated that control (Brwd1^fl/fl), Cd23^Cre/wt;Brwd1^fl/wt, and Cd23^Cre/wt;Brwd1^fl/fl mice had comparable numbers of B220⁺CD19⁺CD93⁻ mature B cells in the spleen (Supplementary Fig. 2a, b). Additionally, these mice had the same number and frequency of Fol and marginal zone (MZ) B cells (Supplementary Fig. 2c–f). Approximately 96% of Fol B cells and 82% of MZ B cells expressed tdT in Cd23^Cre/wt;Brwd1^fl/fl mice, confirming efficient Cre recombination (Supplementary Fig. 2g, h). Together, these results demonstrate that steady-state splenic B cells do not require BRWD1.

BRWD1 is important for GC initiation

Upon antigenic stimulation, Fol B cells differentiate into GC B cells. To determine whether BRWD1 is important for the differentiation of Fol B cells, we intraperitoneally immunized Cd23^Cre/wt;Brwd1^fl/fl mice with sheep red blood cells at days 0 and 5 and analyzed spleens 14 days post-immunization at the peak of the GC response (Fig. 1a, b)³⁵. The number of B220⁺ B cells was equivalent between Cd23^Cre/wt;Brwd1^fl/fl and control mice (Fig. 1c). However, the number and frequency of GL7⁺CD95⁺ GC B cells were decreased over twofold in the Cd23^Cre/wt;Brwd1^fl/fl mice (Fig. 1d, e). This decrease was observed in GC B cells within the LZ, DZp, and DZd (Fig. 1f). Over 85% of GC B cells expressed tdT in the Cd23^Cre/wt;Brwd1^fl/fl mice, demonstrating that the GC population primarily contained GC B cells with Brwd1 deleted (Fig. 1g). Cd23^Cre/wt;Brwd1^fl/fl mice had no significant change in the number of Tfh cells but did have a decreased Tfh cell frequency (Fig. 1h, i)³⁶. Furthermore, the frequency of GC B cells and Tfh cells positively correlated with one another, showing that the decreased Tfh cell frequency was consistent with the smaller GC B cell response (Fig. 1j)³⁶.

Fig. 1. Deletion of Brwd1 in Fol B cells represses GC initiation.

a Cd23^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice were immunized intraperitoneally (i.p.) with 10⁹ sheep red blood cells (SRBC) at days 0 and 5. Spleens were collected at day 14 post-immunization (Ctrl n = 6 mice, KO n = 11 mice). b Representative flow plots first gated on live B220⁺ cells. c Total number of B220⁺ B cells. Absolute number (d) and frequency (e) of GL7⁺CD95⁺ germinal center (GC) B cells. f Number of light zone (LZ), dark zone proliferation (DZp), and dark zone differentiation (DZd) GC B cells. g Proportion of GC B cells expressing tdTomato (tdT). Total number (h) and frequency (i) of CXCR5⁺PD-1⁺ T follicular helper (Tfh) cells. j Linear regression of frequency of GC B cells vs. frequency of Tfh cells. m = slope. k Representative images of spleen sections from Cd23^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice imaged at ×10 magnification and stained with GL7. (DIC = differential interference contrast, scale bar = 2000 μm, Ctrl n = 3 mice, KO n = 9 mice) l Number of GCs per spleen area. m Size of GCs as measured by area of GL7 staining. Violin plots show median and quartiles. n Cd23^Cre/wt;Brwd1^fl/fl (KO), Cd23^Cre/wt;Brwd1^fl/wt (Het), and Brwd1^fl/fl (Ctrl) mice were immunized i.p. with 10⁹ SRBCs, and spleens were harvested at day 5 (Ctrl n = 10 mice, Het n = 6 mice, KO n = 8 mice). Total number (o) and frequency (p) of GL7⁺CD95⁺ GC B cells. q Proportion of GC B cells expressing tdT. r Median fluorescence intensity (MFI) of BCL6 in GC B cells. Control data was normalized to 1 to account for voltage differences between flow cytometry experiments (Ctrl n = 9 mice, KO n = 6 mice). s Representative histogram of BCL6 intensity in Ctrl and KO GC B cells and in non-GC B220⁺CD95⁻GL7⁻ B cells. t Frequency of IgM⁻ cells after gating on tdT⁻ or tdT⁺ GC B cells. Cd23^Cre/wt;Brwd1^fl/fl (KO) mice were immunized i.p. with sheep red blood cells at days 0 and 5. Spleens were analyzed day 14 post-immunization (KO n = 5 mice). u In vitro class switch recombination (CSR) to IgG1. Splenic B cells from Cd23^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice were cultured for 96 h with lipopolysaccharides (LPS) and IL-4, and class switching was measured by flow cytometry (Ctrl n = 6 mice, KO n = 5 mice). v Representative flow plots of IgG1⁺ frequency in CD19⁺tdT⁻ or CD19⁺tdT⁺ cells from KO mice. ns = not significant. Groups were compared with a two-sided Mann–Whitney in (m), a simple linear regression in (j), a two-sided paired t-test in (t), and a two-sided unpaired t-test in all other panels. Bar plots show means ± standard deviations.

We confirmed a defect in the GC response by immunofluorescence microscopy. Spleens from Cd23^Cre/wt;Brwd1^fl/fl mice had fewer GCs per splenic area than controls (Fig. 1k, l). Additionally, GCs from Cd23^Cre/wt;Brwd1^fl/fl mice were smaller compared to controls as measured by GL7 fluorescence (Fig. 1m). Together, these results demonstrate that BRWD1 is important for a proper GC response.

This diminished GC B cell response in Cd23^Cre/wt;Brwd1^fl/fl mice could reflect either a defect in initial differentiation of Fol B cells into GC B cells or in expansion and maintenance of GC B cells. Because GCs form around day 4 post-immunization, we measured the GC response at day 5 post-immunization (Fig. 1n)^3,5. At this time point, we observed a 2.5-fold decrease in both the number and frequency of GC B cells, suggesting that BRWD1 is important for GC initiation (Fig. 1o, p). Approximately 80% of GC B cells in Cd23^Cre/wt;Brwd1^fl/fl mice expressed tdT, suggesting that GC B cells with Brwd1 deleted can populate the GC once the response is initiated (Fig. 1q).

We considered and ruled out several possibilities for this defect in GC initiation. First, the decreased GC response in Cd23^Cre/wt;Brwd1^fl/fl mice at day 5 post-immunization was not caused by Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ B cells alternatively differentiating into plasmablasts, because there was no significant difference in the number or frequency of CD138⁺B220^low plasmablasts compared to controls (Supplementary Fig. 3a–c). Next, GC B cells in Cd23^Cre/wt;Brwd1^fl/fl mice had no difference in the prevalence of apoptosis or cell death as measured by Annexin V or fluorescent inhibitors of caspase (FLICA) (Supplementary Fig. 3d–i). The proliferating subpopulation among GC B cells is marked by the highest CD83 and CXCR4 expression¹⁴. There was no difference in the proportion of proliferating cells among the CD83^highCXCR4^high population as measured by DAPI^high cells in the S, G2, or M phases of the cell cycle (Supplementary Fig. 3j). Furthermore, there was no change in the DZp to LZ ratio or the DZd to LZ ratio (Supplementary Fig. 3k, l).

We next examined the expression of BCL6 because of its role in GC B cell differentiation. During GC initiation, the GC B cell precursor population is characterized by intermediate BCL6 expression, and this population peaks between days 3 and 4 post-immunization⁵. Flow cytometry of GC B cells from Cd23^Cre/wt;Brwd1^fl/fl mice showed lower levels of BCL6 expression due to the presence of precursor cells with intermediate BCL6 expression, suggesting a block in full GC B cell differentiation (Fig. 1r, s). Thus, these results support that BRWD1 is necessary for the differentiation of Fol B cells into GC B cells.

BRWD1 facilitates class switch recombination

CSR requires changes in chromatin topology at the immunoglobulin heavy chain locus³⁷. Therefore, we investigated whether BRWD1 is important for CSR. Indeed, at day 14 post-immunization, fewer tdT⁺ GC B cells were IgM⁻ when compared with tdT⁻ cells from the same Cd23^Cre/wt;Brwd1^fl/fl mouse, suggesting that following Brwd1 deletion, B cells undergo less CSR (Fig. 1t). Next, we cultured Fol B cells in vitro with IL-4 and TGFβ to induce class switching to IgG1. tdT⁺ cells from Cd23^Cre/wt;Brwd1^fl/fl mice had decreased switching to IgG1 compared to controls (Fig. 1u, v). tdT⁻ cells from Cd23^Cre/wt;Brwd1^fl/fl mice had no change in IgG1 class switching, demonstrating that Brwd1 plays a cell-intrinsic role in CSR. Cultured Fol B cells from Cd23^Cre/wt;Brwd1^fl/fl mice had no significant change in expression of IgG1 germline transcripts compared to controls (Supplementary Fig. 4a).

Class switching requires precise chromatin loop extrusion across the immunoglobulin heavy chain locus^37,38. During B cell development, a loop across the entire locus is established that brings together the 5’ iEμ enhancer and the 3’ Igh regulatory region (Supplementary Fig. 4b), and this loop is maintained in naïve Fol B cells^38,39. Because BRWD1 regulates cohesin to modulate chromatin looping, we measured BRWD1 and components of the cohesin complex at the immunoglobulin heavy chain locus³³. Chromatin immunoprecipitation followed by sequencing (ChIP-seq) revealed that in wild-type (WT) small pre-B cells, BRWD1 binds at multiple sites within the Igh locus, with an especially strong peak at iEμ (Supplementary Fig. 4c). In WT small pre-B cells, we observed a dynamic cohesin complex containing coincident NIPBL and WAPL at the 3’ end of the locus (Supplementary Fig. 4c, highlighted green)³³. In small pre-B cells from mice with a germline Brwd1 nonsense mutation (Brwd1^mutant)³¹, the 3’ dynamic cohesin complex was lost, and instead static cohesin was found at the 3’ CTCF sites (highlighted yellow). Invariant cohesin complexes at CTCF sites were also observed throughout the locus (highlighted red). This loss of the dynamic cohesin complex was associated with decreased Hi-C contacts and decreased looping across the locus (Supplementary Fig. 4c, d). Furthermore, the loss of looping resulted in decreased Ighm expression compared to WT small pre-B cells (Supplementary Fig. 4e). Thus, BRWD1 binds at and is necessary for chromatin looping across the immunoglobulin heavy chain locus during B cell development. These data suggest that the BRWD1-dependent chromatin topology established in small pre-B cells is necessary for efficient CSR upon Fol B cell activation.

BRWD1 establishes epigenetic states for GC initiation

We next studied whether Brwd1 deletion affects the transcriptional or epigenetic states of resting Fol B cells. We first sorted tdT⁺ from Cd23^Cre/wt;Brwd1^fl/fl mice and performed RNA-seq. tdT⁺ Fol B cells clustered with Fol B cells from Brwd1^fl/fl control mice by principal component analysis (PCA) when compared with WT GC B cells (Fig. 2a). Fol B cells were also highly similar by a Pearson correlation (Fig. 2b). Only 17 genes (log₂FC > 1 or log₂FC < −1, q < 0.05) were differentially expressed between tdT⁺ Fol B cells and control Fol B cells from Brwd1^fl/fl mice (Fig. 2c). Unsurprisingly, Brwd1 expression was significantly decreased in Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells, albeit modestly, as exons not targeted by Cre recombination were still transcribed (log₂FC = −0.57, q = 2.44 × 10⁻⁶). These results demonstrate that deletion of Brwd1 in Fol B cells has minimal transcriptional effects.

Fig. 2. BRWD1 determines genomic accessibility in resting Fol B cells.

a Principal component analysis (PCA) plot of RNA-seq from Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ follicular (Fol) B cells (KO), Brwd1^fl/fl Fol B cells (Ctrl), and wild-type (WT) germinal center (GC) B cell subsets (KO Fol B n = 3 mice, Ctrl Fol B n = 3 mice). WT GC B cell RNA-seq (n = 2 for each cell population) was previously published¹⁴. (LZ light zone, DZp dark zone proliferation, DZd dark zone differentiation, PC principal component). b Pearson correlation heatmap of RNA-seq from the listed populations. c Volcano plot of differentially expressed genes (q < 0.05, log₂ fold change (FC) > 1 or log₂FC < −1) between tdT⁺ Cd23^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) Fol B cells. d PCA plot of ATAC-seq from Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells (KO), WT Fol B cells (Ctrl), and WT GC B cell subsets (KO Fol B n = 3 mice, Ctrl Fol B n = 2 mice). WT GC B cell ATAC-seq (n = 2 for each cell population) was previously published¹⁴. e Pearson correlation heatmap of ATAC-seq from listed populations. Note scale differences between (b) and (e). f Heatmap of differential accessibility peaks (q < 0.05, log₂FC > 1 or log₂FC < −1) between tdT⁺ Cd23^Cre/wt;Brwd1^fl/fl and WT Fol B cells. Chromatin accessibility by ATAC-seq of the listed populations is shown for these differentially accessible peaks. (CPM = counts per million). g Chromatin accessibility tracks at the Bcl2 locus (mm9, chromosome 1) for WT and KO Fol B cells. Differential accessibility peaks (q < 0.05, log₂FC > 1) are marked by *. h Transcription factor (TF) motifs enriched in chromatin accessibility regions up in WT or KO Fol B cells relative to a GC-normalized background generated using HOMER, which uses a one-sided Fisher’s exact test. TF motifs in the ETS family, which have similar binding sequences, are shown in green, while all other TF motifs are in orange. TF motifs enriched at p < 10⁻³⁰ are shown.

To determine whether loss of BRWD1 affects the chromatin accessibility of Fol B cells, we performed ATAC-seq. By PCA, tdT⁺ Fol B cells from Cd23^Cre/wt;Brwd1^fl/fl mice clustered separately from both WT Fol B cells and WT GC B cells (Fig. 2d). Chromatin accessibility was more different across Fol B cell populations than the differences observed in transcription (Fig. 2e, note scale).

When we compared differential chromatin accessibility peaks (log₂FC > 1, q < 0.05) between Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ and WT Fol B cells, we observed that 1956 peaks increased in WT and 2141 peaks increased in Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells (Fig. 2f, g). Interestingly, the accessibility of many of these peaks in WT Fol B cells matched the accessibility in GC B cells, suggesting that the accessibility of these sites is maintained by BRWD1 as Fol B cells differentiate into GC B cells. TF motif accessibility was altered between WT and Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells for TFs important in B cell differentiation (Fig. 2h). For example, binding sites for OCT2, which prepares and facilitates Fol to GC B cell differentiation programs, were more accessible in WT Fol B cells^12,40,41. Conversely, accessibility at ETS family binding sites, which can restrain B cell activation and differentiation programs, were increased in Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells^42–45. Accessibility at IRF4 and IRF8 binding sites, which regulate activated B cell differentiation programs, were also increased in Cd23^Cre/wt;Brwd1^fl/fl tdT⁺ Fol B cells^46,47. These results suggest that in Fol B cells, BRWD1 maintains genomic accessibility for transition to the GC state following antigen stimulation.

BRWD1 restrains GC B cell proliferation

While BRWD1 was important for GC initiation, it is also expressed in GC subsets and therefore may have additional roles once B cells have fully differentiated into GC B cells. To study this, we crossed the Brwd1-floxed mice with Aicda^Cre mice to delete Brwd1 in GC B cells approximately 4 days post-immunization when GCs have begun to form^3,8,48. We immunized Aicda^Cre/wt;Brwd1^fl/fl mice and control Brwd1^fl/fl mice and analyzed spleens after 14 days (Fig. 3a, b). We observed no significant difference in the total number of B220⁺ B cells in Aicda^Cre/wt;Brwd1^fl/fl mice compared with controls (Fig. 3c). However, there was a significant increase in the total number and frequency of GC B cells in the Aicda^Cre/wt;Brwd1^fl/fl mice (Fig. 3d, e). In the Aicda^Cre/wt;Brwd1^fl/fl mice, 80% of GC B cells had deleted Brwd1 as shown by tdT expression, and this frequency was consistent across GC B cell subsets (Fig. 3f and Supplementary Fig. 5a). tdT MFI varied across the three GC subsets, with the greatest tdT expression in the LZ (Supplementary Fig. 5b, c). The increase in GC B cell number was observed across each of the three GC subsets, with minimal changes in the relative proportion of cells within each GC subset as measured by the DZd/LZ and DZp/LZ ratios (Fig. 3g–i). Similar results were observed in heterozygous Aicda^Cre/wt;Brwd1^fl/wt mice (Supplementary Fig. 5d–g). As a control, Aicda^Cre/wt mice had the same GC B cell frequency as Brwd1^fl/fl mice, demonstrating that the GC defect observed in Aicda^Cre/wt;Brwd1^fl/fl mice was not due to Aicda haploinsufficiency (Supplementary Fig. 5h). In total, both Brwd1 deletion and Brwd1 haploinsufficiency are sufficient to alter GC B cell numbers.

Fig. 3. BRWD1 restrains proliferation in GC B cells.

a Aicda^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice were immunized intraperitoneally (i.p.) with 10⁹ sheep red blood cells (SRBC) at days 0 and 5. Spleens were collected at day 14 post-immunization (Ctrl n = 12 mice, KO n = 14 mice). b Representative flow plots first gated on live B220⁺ cells. c Total B cell number. Total number (d) and frequency (e) of GL7⁺CD95⁺ germinal center (GC) B cells. f Proportion of GC B cells expressing tdTomato (tdT). g Number of light zone (LZ), dark zone proliferation (DZp), and dark zone differentiation (DZd) GC B cells. h Ratio of DZd and LZ frequency. i Ratio of DZp and LZ frequency. Total number (j) and frequency (k) of GL7⁺CD95⁺ GC B cells in Peyer’s Patches (PP). 7 PPs throughout the small intestine were pooled per mouse (Ctrl n = 5 mice, KO n = 7 mice). l Proportion of PP GC B cells expressing tdT. T follicular helper (Tfh) cell number (m) and frequency (n). o Correlation of GC B cell and Tfh cell frequency. m = slope. p Frequency of proliferating GC B cells in the S, G2, or M phases of the cell cycle as measured by DAPI^high frequency (Ctrl n = 5 mice, KO n = 7 mice). q Frequency of proliferating CD83^highCXCR4^high GC B cells. r GC size measured by GL7 fluorescence (Ctrl n = 8 mice, KO n = 11 mice). s Number of GCs normalized by splenic area in mm². t Representative immunofluorescence imaging of spleens at ×20 magnification using GL7, anti-CD4, anti-tdT, anti-CD35, and anti-IgD. tdTomato signal outside the GC is background staining from rabbit anti-tdT and donkey anti-rabbit-IgG-AF488 (Ctrl n = 4 mice, 29 GCs; KO n = 4 mice, 35 GCs). u Ratio of LZ size to total GC size. v CD4 mean pixel intensity (MPI, arbitrary units (a.u.) per μm²) within the LZ per GC imaged. ns = not significant. Groups were compared with a simple linear regression in (o), a two-sided Mann–Whitney in (r, u, v), and a two-sided unpaired t-test in all other panels. Bar plots show means ± standard deviations. Violin plots show median and quartiles.

To study the role of BRWD1 in chronic GCs with continual antigen exposure, we examined GC B cells in the Peyer’s Patches (PP) of unimmunized mice. Aicda^Cre/wt;Brwd1^fl/fl mice had a significant increase in the total number and frequency of PP GC B cells compared with controls, and 75% of GC B cells had deleted Brwd1 as shown by tdT expression (Fig. 3j–l). Thus, BRWD1 restrains GC B cell numbers in multiple immune tissues.

We next examined the splenic Tfh cell compartment. The number and frequency of CD3⁺CD4⁺CXCR5⁺PD-1⁺ Tfh cells were significantly increased in Aicda^Cre/wt;Brwd1^fl/fl mice, and the frequency of GC B cells strongly correlated with the frequency of Tfh cells within each spleen (Fig. 3m–o and Supplementary Fig. 5i). These results are consistent with the increase in GC B cells, because increased antigen presentation to Tfh cells results in greater Tfh cell proliferative expansion³⁶. This suggests that antigen presentation in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells is intact.

To determine whether Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells were undergoing greater proliferation, we stained cells with DAPI to measure the proportion of cells in the S, G2, or M phases of the cell cycle. When gated on total GC B cells, Aicda^Cre/wt;Brwd1^fl/fl mice had a greater proportion of proliferating GC B cells (Fig. 3p). In prior work, we demonstrated that the CD83^highCXCR4^high DZp population within the GC is where GC B cells undergo mitosis, and that expression of both surface markers positively correlates with DNA content¹⁴. Gating on this CD83^highCXCR4^high DZp population showed an increase in proliferating cells in Aicda^Cre/wt;Brwd1^fl/fl mice compared to controls (Fig. 3q and Supplementary Fig. 5j, k). CXCR4⁺CD83⁻ DZd GC B cells also had greater proliferation, suggesting a blending of proliferative programs throughout the GC, although most proliferation was still confined to the DZp (Supplementary Fig. 5l). Finally, differences in GC B cell numbers were not due to differences in cell viability (Supplementary Fig. 5m). These results demonstrate that BRWD1 represses proliferation in GC B cells, consistent with BRWD1’s role as a Myc repressor in small pre-B cells³².

We further characterized GC responses by immunofluorescence microscopy of spleen sections from Aicda^Cre/wt;Brwd1^fl/fl mice. This analysis revealed that a portion of GCs in Aicda^Cre/wt;Brwd1^fl/fl mice were twice as large in cross-sectional area as GCs from control mice, with a greater spread in the distribution of GC size (Fig. 3r). There was no significant difference in the number of GCs per spleen area (Fig. 3s). GCs from Aicda^Cre/wt;Brwd1^fl/fl mice had normal architecture with distinct light and dark zones, and there was no difference in the ratio of LZ size to total GC size (Fig. 3t, u). In agreement with the Tfh flow cytometry data, there was an increase in the CD4 mean pixel intensity within the LZ (Fig. 3v). In summary, Brwd1 deficiency in GC B cells caused an increase in GC size without affecting GC architecture.

BRWD1 is not required for post-GC MBCs or PCs

GC B cells can exit GCs by further differentiating into MBCs or PCs. To determine whether BRWD1 is important for these transitions, we studied the prevalence of MBCs and PCs in the spleens of Aicda^Cre/wt;Brwd1^fl/fl mice and control Brwd1^fl/fl mice 14 days post-immunization (Fig. 4a). We observed no significant difference in the number or frequency of B220⁺CD38⁺GL7⁻IgD⁻ MBCs nor of CD138⁺B220^low PCs (Fig. 4b–g).

Fig. 4. BRWD1 is not required for GC-derived MBCs and PCs.

a Aicda^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice were immunized with sheep red blood cells (SRBC) intraperitoneally (i.p.) and boosted at day 5. Splenic cells were analyzed at day 14 post-immunization (Ctrl n = 8 mice, KO n = 10 mice). b Representative flow cytometry plot of memory B cell (MBC) gating strategy. c Number of B220⁺CD38⁺GL7⁻IgD⁻ MBCs. d Frequency of MBCs as a proportion of all B220⁺ cells. e Representative flow cytometry plot of plasma cell (PC) gating strategy. Number (f) and frequency (g) of CD138⁺B220^low PCs. h Aicda^Cre/wt;Brwd1^fl/fl (KO) and Brwd1^fl/fl (Ctrl) mice were immunized with SRBCs conjugated to 4-hydroxy-3-nitrophenylacetyl (NP-SRBC) i.p. at days 0, 5, and 56. Spleens and bone marrow were analyzed at day 61 post-immunization (Ctrl n = 3 mice, KO n = 3 mice). i GL7⁺CD38⁻ GC B cell frequency. j B220⁺CD38⁺GL7⁻IgD⁻ MBC frequency. k CD138⁺B220⁺ plasmablast frequency. l Splenic CD138⁺B220^low PC frequency. m Bone marrow (BM) CD138⁺B220^low PC frequency. ns = not significant. Groups were compared with two-sided unpaired t-tests. Bar plots show means ± standard deviations.

MBC progenitors (pre-MBC) differentiate from LZ B cells in the GC and express some MBC markers, such as CCR6^49–51. We observed a decreased frequency of pre-MBCs in the LZ of Aicda^Cre/wt;Brwd1^fl/fl mice (Supplementary Fig. 6a–c). However, due to the increase in the total number of GC B cells in these mice, the number of pre-MBCs was comparable to that of control mice (Supplementary Fig. 6d). A similar trend in pre-MBCs was observed in heterozygous Aicda^Cre/wt;Brwd1^fl/wt mice (Supplementary Fig. 6e, f). The MBC population is heterogeneous, and MBCs can differentiate from both activated B cells and GC B cells⁵². To measure GC-dependent MBCs, we examined multiple markers enriched in GC-derived MBCs, namely CD73, CD80, PD-L2, IgG1, and the absence of IgD^35,53–57. Flow cytometry for various combinations of these markers showed no difference in the GC-derived MBC subsets (Supplementary Fig. 6g–m).

To determine whether Aicda^Cre/wt;Brwd1^fl/fl mice produce long-lived PCs in the bone marrow and mount a memory recall response, we immunized mice with SRBCs conjugated to 4-hydroxy-3-nitrophenylacetyl (NP), followed by a boost immunization at day 5. On day 56, we immunized again with NP-SRBCs, and on day 61, we measured B cell populations in the spleen and bone marrow (Fig. 4h). There were no significant differences in the number or frequency of CD38⁺GL7⁻IgD⁻ MBCs, GL7⁺CD38⁻ GC B cells, CD138⁺B220⁺ plasmablasts, or CD138⁺B220^low PCs (Fig. 4i–l). Similarly, we observed no significant difference in bone marrow CD138⁺B220^low PCs (Fig. 4m). In total, these results indicate that BRWD1 is not important for differentiation of GC-dependent MBCs, PCs, or plasmablasts.

BRWD1 is necessary for optimal affinity maturation

We next explored whether BRWD1 is important for the principal functions of the GC, SHM, and affinity maturation. We immunized mice with NP bound to keyhole limpet hemocyanin (KLH), which elicits a B cell response in which most antibodies specific to NP are of the IgG1 isotype and include the V_H186.2 variable heavy chain gene segment⁵⁸ (Fig. 5a). We compared Aicda^Cre/wt;Brwd1^fl/fl mice to Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl mice to control for Aicda heterozygosity. At day 14 post-immunization, we sorted tdT⁺ GC B cells by FACS and cloned the V_H186.2 segment for sequencing⁵⁸. To prevent contamination by plasma cells, which have high levels of Ig transcripts, we used a primer specific to the Cγ1-membrane-encoding exon⁵⁸.

Fig. 5. BRWD1 enhances IgG1 affinity maturation.

a Aicda^Cre/wt;Brwd1^fl/fl (KO), Aicda^Cre/wt;Brwd1^fl/wt (Het), and Aicda^Cre/wt;ROSA26-LSL-tdTomato^fl/fl (Ctrl) mice were immunized intraperitoneally (i.p.) with 4-hydroxy-3-nitrophenylacetyl (NP) conjugated to keyhole limpet hemocyanin (KLH) in complete Freund’s adjuvant (CFA) and boosted with NP-KLH in incomplete Freund’s adjuvant (IFA) at day 5. At day 14 post-immunization, spleens were collected, and tdTomato⁺ (tdT) germinal center (GC) B cells were sorted. V_H186.2 gene segments were cloned using a Cγ1 primer and sequenced (Ctrl n = 4 mice, Het n = 5 mice, KO n = 6 mice). b Frequency of mutations within V_H186.2 in GC B cells. c Frequency of silent mutations within V_H186.2 in GC B cells. d Frequency of W33L or K59R amino acid substitutions that increase the affinity of V_H186.2 to NP. e, f Control and knockout mice were immunized with NP-KLH in CFA or with sheep red blood cells (SRBC) (KO n = 7 mice, Ctrl n = 8 mice). Sera was collected before immunization or at day 14 post-immunization and used to stain HEp-2 cells. e Representative images of HEp-2 cells with sera from a control Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl mouse, a control Brwd1^fl/fl mouse, a knockout Aicda^Cre/wt;Brwd1^fl/fl mouse immunized with SRBCs, or a knockout Aicda^Cre/wt;Brwd1^fl/fl mouse immunized with NP-KLH with CFA. Full image scale bar = 300 μm. Inset scale bar = 30 μm. f Proportion of mice with anti-nuclear antibodies. Control Brwd1^fl/fl and Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl mice are grouped. Mice immunized with either SRBCs or NP-KLH are grouped. Groups were compared with two-sided unpaired t-tests. Lines show means ± standard deviations.

Sequencing of V_H186.2 of IgG1 revealed that tdT⁺ GC B cells from Aicda^Cre/wt;Brwd1^fl/fl mice and Aicda^Cre/wt;Brwd1^fl/wt mice undergo the same rate of SHM as GC B cells from Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl control mice (Fig. 5b). The rate of silent SHM, which is not affected by selection, was also similar between Aicda^Cre/wt;Brwd1^fl/fl tdT⁺, Aicda^Cre/wt;Brwd1^fl/wt tdT⁺, and Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl tdT⁺ control GC B cells (Fig. 5c). In contrast to control GC B cells, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ and Aicda^Cre/wt;Brwd1^fl/wt tdT⁺ GC B cells had a significantly broader distribution of SHM with some clones containing a high number of mutations and some clones with none or one mutation (F test, p < 0.0001). In clones with a high amount of SHM, mutations were distributed throughout V_H186.2 (Supplementary Fig. 7a). Furthermore, the increased SHM distribution relative to controls was consistent across individual mice (Supplementary Fig. 7b). The rate of mutations within framework regions and complementarity-determining regions was the same between groups (Supplementary Fig. 7c–e). Furthermore, there was no significant difference between groups in the rate of unproductive immunoglobulin mutations (Supplementary Fig. 7f). In the WT GC cycle, there are relatively fixed, deterministic relationships between selection, proliferation, and SHM^{20,23,27,59,60}. Our results suggest that upon loss of Brwd1, these tight relationships are uncoupled, and the degree of proliferation following SHM and selection becomes stochastic.

We also measured affinity maturation by comparing the frequency of mutations known to increase affinity toward NP, including W33L and K59R⁶¹. Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ and Aicda^Cre/wt;Brwd1^fl/wt tdT⁺ GC B cells had an approximately two-fold decrease in the frequency of high-affinity mutations relative to Aicda^Cre/wt;ROSA-LSL-tdTomato^fl/fl tdT⁺ control GC B cells (Fig. 5d). These results demonstrate that BRWD1 is necessary for optimal affinity maturation of IgG1 GC B cells. Furthermore, decreased high-affinity mutations in the Aicda^Cre/wt;Brwd1^fl/wt tdT⁺ GC B cells show that the amount of BRWD1 is important for affinity maturation. Diminished affinity maturation was not due to a skewing of the expressed Ighv repertoire (Supplementary Fig. 7g).

While the antibody response to NP is mostly the IgG1 isotype, we also cloned V_H186.2 in IgM GC B cells using a Cμ primer. Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ and control IgM GC B cells underwent the same average rate of SHM (Supplementary Fig. 8a). In contrast to IgG1 GC B cells, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ and control IgM GC B cells had the same frequency of high affinity mutations (Supplementary Fig. 8b). This was despite IgG1 Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells from the same mice having low affinity maturation (Supplementary Fig. 8c). Within the GC, IgG⁺ B cells are positively selected over IgM⁺ B cells so that IgG⁺ cells eventually dominate the GC response⁶². Together, these results suggest that BRWD1 is necessary for IgG1 cell selection but not IgM.

In addition to selection for higher affinity, GCs also eliminate autoreactive clones that arise from SHM^63–66. To test GC negative selection, we collected sera from mice immunized with SRBCs or NP-KLH and tested reactivity to HEp-2 cells (Fig. 5e)⁶⁷. Prior to immunization, sera from all mice did not bind HEp-2 cells. Likewise, there was no detectable HEp-2 reactivity in control mouse serum after immunization. In contrast, sera from Aicda^Cre/wt;Brwd1^fl/fl mice immunized with either SRBCs or NP-KLH bound to HEp-2 cell nuclei, indicating self-reactivity. Approximately 40% of Aicda^Cre/wt;Brwd1^fl/fl mice demonstrated HEp-2 reactivity (Fig. 5f). These results suggest that Aicda^Cre/wt;Brwd1^fl/fl mice are beginning to break tolerance following immunization, although the self-reactive sera may not be pathologically significant.

BRWD1 is required for GC B cell transcriptional identity

We next performed RNA-seq on GC B cells from the LZ, DZp, and DZd of Aicda^Cre/wt;Brwd1^fl/fl mice and compared them with WT GC B cell subsets. PCA of these populations revealed that Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells were transcriptionally distinct from WT GC B cells (Fig. 6a, b). Comparison of WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells within the same subset revealed a large transcriptional increase within the Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ, DZp, and DZd (Fig. 6c–e). Interestingly, many of the same genes were upregulated across subsets (Fig. 6f and Supplementary Fig. 9a–f). For example, 57% of genes increased in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ cells were also increased in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp and DZd cells (Supplementary Fig. 9b). A heatmap of differential expression between all groups revealed many genes with similar expression in each Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subset (Supplementary Fig. 9g). This was especially true of Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp and DZd cells. Together, these results demonstrate that BRWD1 represses a large transcriptional program that is shared across GC B cell subsets in which Brwd1 is deleted.

Fig. 6. BRWD1 is required for GC B cell subset transcriptional identity.

a Principal component analysis (PCA) plot of RNA-seq of each germinal center (GC) subset in wild-type (WT) and Aicda^Cre/wt;Brwd1^fl/fl (KO) mice. n = 2 per cell type. Each n represents cells pooled from 20 mice. WT GC B cell RNA-seq was previously published¹⁴. (LZ light zone, DZp dark zone proliferation, DZd dark zone differentiation, PC principal component). b Heatmap of Pearson correlation of RNA-seq. (CPM = counts per million). c–e Volcano plots of differentially expressed genes (q < 0.05, log₂ fold change (FC) > 1 or log₂FC < −1) between WT and KO GC B cells from the LZ (c), DZp (d), and DZd (e). f Differentially expressed genes between WT and KO GC B cells from the LZ, DZp, or DZd. Purple lines show genes shared between groups. Dark orange signifies genes in the group that are shared with another group. Gene expression of WT and KO GC subsets at clusters generated by unsupervised K-means clustering of WT GC subsets performed previously plotted as a heatmap (g) and a box-and-whisker plot (h)¹⁴. Z-scored log₂ CPM are plotted. The box limits show the interquartile ranges (IQR), and the center lines show the medians. Maximum and minimum values (whiskers) are defined as Q3 + 1.5 × IQR and Q1 – 1.5 × IQR, respectively. i Gene set enrichment analysis (GSEA) using the Molecular Signatures Database hallmark genes sets for each GC subset. Pathways are ordered by normalized enrichment score (NES). Pathways with q < 0.01 are shown. The WT DZd had no pathways with q < 0.01.

In WT GC subsets, RNA expression clustered into 8 groups, which previously revealed a transcriptional program unique to each GC subset, most notably the DZp (Fig. 6g, h)¹⁴. Surprisingly, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp cells lost this unique transcriptional program and instead had intermediate expression between Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ and DZd cells. For example, WT DZp cells had the greatest expression of genes in clusters 5 and 6 (enriched for cell cycle genes) and the lowest expression of genes in clusters 1, 2, and 4 (enriched for lymphocyte activation) (Fig. 6g, h). Yet Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp cells had gene expression intermediate that of Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ and DZd cells for genes in these same clusters. The observation that transcriptional programs associated with cell cycle progression were diminished in DZp cells in which Brwd1 was deleted suggests that increased proliferation was due to defective selection (described below). Thus, while BRWD1 maintains the transcriptional program of all GC subsets, the transcriptional program unique to DZp cells is preferentially lost in the absence of BRWD1.

To understand the function of dysregulated genes in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, we performed a gene set enrichment analysis (GSEA) comparing WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells from each GC subset. GSEA revealed a significant enrichment of many pathways in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, including TNF signaling via NFκB and the inflammatory response (Fig. 6i and Supplementary Fig. 9h–n). Similarly, a gene ontology (GO) pathway analysis of genes significantly upregulated in WT or Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells from each GC subset also revealed shared transcriptional pathways in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, including multiple immune and inflammatory response GO pathways (Supplementary Fig. 9o). Within these pathways were many differentially expressed genes important for GC B cell biology (Supplementary Fig. 9p). Taken together, these results demonstrate that within the genes upregulated in each Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subset are inflammatory transcriptional programs.

RNA-seq revealed increased expression of the Brwd1 homologs Phip (also known as Brwd2) and Brwd3 in the Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC subsets (Supplementary Fig. 9q). These genes could partially compensate for Brwd1 deletion in Aicda^Cre/wt;Brwd1^fl/fl mice, thereby mitigating the observed phenotype.

Broadly, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC subsets had minimal changes in the expression of TFs or chromatin factors important for GC B cells, suggesting that the transcriptional changes observed above were not caused by BRWD1 altering the activity of a single TF or gene (Supplementary Fig. 10a, b). Indeed, GSEA of the BCL6, IRF8, and E2A target gene sets were not significantly enriched within WT or Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subsets (q < 0.05) (Supplementary Table 1). In Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, some TFs with high expression in the LZ (Irf4, Nr4a1, Hhex) had persistently elevated expression in the DZp, consistent with a blending of the DZp transcriptional program.

BRWD1 maintains chromatin accessibility across GC subsets

Finally, we performed ATAC-seq on LZ, DZp, and DZd GC B cells from Aicda^Cre/wt;Brwd1^fl/fl and WT mice to understand the effects of deleting Brwd1 in GC B cells. Comparison with WT GC B cells revealed that Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells were epigenetically distinct in the PCA space (Fig. 7a). In Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, differences in chromatin accessibility were diminished, as shown by PCA, where Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ samples clustered more closely together. Indeed, the biological coefficient of variation (BCV) was lower across Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC subsets (12.7%) compared to WT GC subsets (25.2%). In WT GC B cells, many accessibility peaks significantly changed during transitions between subsets; however, these changes were diminished in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells (Fig. 7b). For example, transition between the DZd and LZ involves 61,451 differential accessibility peaks in WT cells but only 15,522 differential accessibility peaks in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells. A striking example of this occurred at the Myc locus. In WT cells, accessibility peaks at Myc and downstream enhancers in the WT LZ were closed in WT DZd GC B cells. In contrast, in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cells, chromatin accessibility was essentially unchanged at the Myc locus across the Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subsets (Fig. 7c).

Fig. 7. BRWD1 maintains the distinct epigenetic states of each GC subset.

a Principal component analysis (PCA) plot of ATAC-seq of each germinal center (GC) subset in wild-type (WT) and Aicda^Cre/wt;Brwd1^fl/fl (KO) mice. n = 2 per cell type. Each n represents cells pooled from 20 mice. WT GC B cell ATAC-seq was previously published¹⁴. (LZ light zone, DZp dark zone proliferation, DZd dark zone differentiation, PC principal component). b Number of differentially regulated accessibility peaks (q < 0.05) between GC subsets in WT and KO mice. The percentage of differentially accessible peaks from all called peaks (91,563) is shown above each bar. c Chromatin accessibility tracks at the Myc locus (mm9, chromosome 15) for each GC subset in WT and KO. Myc enhancers (E1-14) were previously characterized¹⁰⁴. d Chromatin accessibility of WT and KO GC subsets at clusters generated by unsupervised K-means clustering of WT GC subsets, performed previously¹⁴. Z-scored log₂ counts per million (CPM) are plotted. The box limits show interquartile ranges (IQR), and the center lines show the medians. Maximum and minimum values (whiskers) are defined as Q3 + 1.5 × IQR and Q1 – 1.5 × IQR, respectively. e Heatmaps of differentially accessible peaks (q < 0.05) between WT and KO GC B cells from the LZ, DZp, or DZd. Z-scored log₂ CPM are plotted. f ATAC-seq footprinting in WT and KO GC B cells for CTCF, ETS1, PU.1, SPIB, AP1/IRF (AICE motif), ETS/IRF (EICE motif), and IRF/IRF (ISRE motif) at differentially accessible peaks (q < 0.05) up in WT or up in KO GC B cells.

In WT GC B cells, accessibility peaks clustered into eight groups, each with a different distribution pattern across GC subsets (Fig. 7d)¹⁴. In contrast, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC subsets had decreased variability within these groups compared to WT. Accessibility was especially dysregulated in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp and DZd cells. For example, at sites in cluster 5, WT DZd cells have low accessibility; however, Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZd cells had greater accessibility than Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ or DZp cells at these same sites. In a similar manner, for cluster 6, WT DZp cells have the greatest accessibility; however, these sites were accessible in both Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZp and DZd cells. Generally, sites uniquely accessible in WT LZ cells were also accessible in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ LZ cells (clusters 3, 4, and 8), yet these clusters also showed a blending across Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ subsets. These results demonstrate that BRWD1 is necessary for polarizing the chromatin accessibility states of GC B cells.

Comparison of WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ cells within the same GC B cell subset revealed large differences in chromatin accessibility (Fig. 7e). We identified TF motifs enriched in differentially accessible peaks in WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subsets relative to chromatin accessibility peaks that were unchanged in the corresponding WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subset (q > 0.05) (Supplementary Fig. 11). Interestingly, chromatin accessibility peaks up in WT LZ cells were enriched for RELA and SPIB motifs, and peaks up in WT DZp cells were enriched for OCT2 and ETS1 motifs. Peaks up in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZd cells were enriched for SP1. We also measured putative TF binding by ATAC-seq footprinting. Chromatin accessibility sites up in WT or Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ GC B cell subsets showed putative binding of multiple TFs including ETS1, PU.1, SPIB, and IRF, as evidenced by protection of these TF motifs in both WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ cells (Fig. 7f). Differences in chromatin accessibility around these protected TF motifs were greatest for chromatin accessibility peaks up in WT LZ and DZp cells and up in Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ DZd cells. In total, we conclude that BRWD1 is necessary for maintaining the unique transcriptional and epigenetic states of each GC subset.

Discussion

There are four stages of B cell development and activation in which the molecular necessities of diversity must be insulated from, and coordinated with, proliferation and selection¹⁵. The precise cellular states to which light chain recombination and SHM are restricted have been defined^14,31,32. In contrast, we only know the general developmental windows in which heavy chain recombination and CSR occur. BRWD1, which is first highly expressed in small pre-B cells, regulates the large epigenetic and transcriptional transitions that enable and insulate light chain recombination, CSR, and SHM from proliferation^31,32. Therefore, BRWD1 is essential for most of the defining molecular events of B-cell adaptive immunity.

In GC B cells, BRWD1 maintained subset epigenetic identity. Indeed, there was a striking collapse of unique accessible TF binding sites toward a common state among Aicda^Cre/wt;Brwd1^fl/fl GC B cell subsets. This epigenetic blending did not precisely map to transcriptional changes. There was a blending of transcriptional programs, especially those associated with proliferative programs, across Aicda^Cre/wt;Brwd1^fl/fl GC subsets. However, prominent inflammatory gene signatures were also induced in all Aicda^Cre/wt;Brwd1^fl/fl GC subsets. The exact mechanism of this latter induction is not clear. Type I interferon signaling in B cells can inhibit affinity maturation, suggesting that the inflammatory gene signatures in Aicda^Cre/wt;Brwd1^fl/fl GC B cells may contribute to the observed defect in affinity maturation⁶⁸. While Brwd1 deletion altered expression and accessibility of multiple regulators of MBC and PC differentiation, these molecular changes were not sufficient to induce phenotypic differences in the production of these cell types.

These profound molecular defects were associated with very specific functional defects. Indeed, BRWD1 was only necessary for GC selection while rates of SHM and effector cell production were normal. These observations are consistent with GC evolution. GCs and efficient affinity maturation convergently evolved only in endothermic mammals and birds approximately 200 million years ago⁶⁹. In contrast, SHM arose approximately 450 million years ago in ectothermic cartilaginous fish, which can also generate MBCs and PCs^69–71. In cartilaginous fish, SHM functions to diversify antibody responses and only modestly enhances affinity^72,73.

In addition to enhancing selection, GCs might have evolved in response to the more rapid cell proliferation of endotherms⁶⁹. Indeed, we postulate that with higher proliferative rates, insulation between SHM and mitosis via specialization of GC subsets became an evolutionary imperative to diminish the risk of neoplastic transformation. BRWD1 also mediates Igk recombination, and indeed Brwd1 homologs can be found in cartilaginous fish, where immunoglobulin-like genes first developed⁷⁴. Therefore, as GCs evolved, it is likely that BRWD1 was adapted for new functions.

While Aicda^Cre/wt;Brwd1^fl/fl GCs had normal SHM rates, the distribution of SHM in GC B cells was greater, such that some clones had few if any mutations while others had ten or more mutations at 14 days post-immunization. Normally, there is a stringent relationship between SHM, selection, and proliferation, with most B cells proliferating once or twice following LZ selection^{20,23,27,59,60}. In contrast, our data are consistent with Aicda^Cre/wt;Brwd1^fl/fl GC B cells proliferating chaotically following selection.

GC spatial organization into an LZ and DZ requires CXCR4 expression by B cells, which recognizes the ligand CXCL12 expressed by DZ reticular cells⁷⁵. However, disruption of either CXCR4 or its ligand CXCL12 only mildly affects SHM or affinity maturation despite the loss of spatial separation between the LZ and DZ^76,77. These results suggest that GC subset molecular identity is more important than spatial polarization for GC function.

Deletion of BRWD1 in Fol B cells suggests that it facilitates molecular transitions rather than maintaining cell states. Indeed, BRWD1 did not substantially contribute to the transcriptional state of Fol B cells. Rather, it prepared the epigenome by both opening binding sites for TFs necessary for GCs while repressing ETS1 binding sites, a TF that can inhibit B cell differentiation^42–45. These results are consistent with prior work in B cell development, where BRWD1 mediates transitions between cell states by priming cells for the next stage of differentiation^31,32. Indeed, our data support a general model that throughout the B cell lineage the opening and closing of TF motifs precede cellular differentiation⁷⁸.

In addition to preparing the epigenetic landscape for GC differentiation, BRWD1 enabled CSR by remodeling the 3D organization of the immunoglobulin heavy chain locus. BRWD1 binds to topologically associating domain (TAD) boundaries, where it converts static to dynamic cohesin to initiate intra-TAD loop extrusion³³. The immunoglobulin heavy chain locus had the hallmarks of cohesin conversion, including accumulation of static cohesin (RAD21 and SMC3) at TAD boundaries and loss of dynamic cohesin (co-occurring RAD21, SMC3, NIPBL, and WAPL) in Brwd1^mutant cells³³. However, BRWD1 appeared to convert cohesin at the opposite TAD boundary to which it was bound. This suggests that BRWD1 can convert distant cohesin on the same chromosome if this static cohesin is within genomic distances accessible by diffusion³⁷.

In both B cell development and the GC, BRWD1 represses Myc³². In GC B cells, we observed changes in chromatin accessibility at Myc but not changes in RNA expression following Brwd1 deletion in bulk-sorted GC B cells, suggesting that additional factors determine Myc expression in positively selected cells. Furthermore, the global genetic programs regulated by BRWD1 in B cell development and the GC are not identical. An obvious example is that BRWD1 mediates Igk contraction in small pre-B cells yet contributes to CSR in the periphery. BRWD1 is recruited to specific histone modifications downstream of signaling pathways^30,31. Therefore, it is likely that signaling and developmental contexts dictate where BRWD1 is recruited in the genome and which TADs and genes it acts on.

Considering BRWD1’s essential role in chromatin loop extrusion, it was surprising that Cd23^Cre/wt;Brwd1^fl/fl and Aicda^Cre/wt;Brwd1^fl/fl mice still formed GCs³³. One possibility is that Brwd1 homologs Phip and Brwd3 compensated for Brwd1 loss, although these homologs may lack functional domains unique to Brwd1. Alternatively, our results may reveal the degree to which loop extrusion is necessary for cell differentiation. Disruption of loop extrusion by depleting cohesin in embryonic stem cells only modestly affects transcription^79,80. However, loop extrusion by cohesin is required for many promoter-enhancer interactions, especially those occurring over longer distances in differentiated cells^81,82. Our results suggest that BRWD1-mediated changes in chromatin topology allow GC B cell subsets to fully polarize as they rapidly cycle through the GC³³.

The Brwd1-floxed mouse efficiently removed exons 6 through 8. A limitation of our study is that the Brwd1-floxed mouse produced a truncated protein predicted to be approximately 13 kD. This truncated protein corresponded with the expected size of Brwd1 exons 1 through 5, which do not encode any functional domains, and was likely nonfunctional.

In conclusion, our results show how BRWD1 is necessary for GC integrity. Deletion of Brwd1 in GC B cells resulted in major transcriptional and epigenetic disruption, which had specific effects on GC function, namely proliferation, affinity maturation, and tolerance (Supplementary Fig. 12). While we previously proposed a three-zone model of the GC that illuminates the transcriptional and epigenetic variation throughout the GC cycle¹⁴, our data here validate this model and demonstrate that the molecular distinctions between the three GC zones are essential for GC function.

Methods

Mice

Brwd1-floxed mice were generated by Ingenious Targeting Laboratory. A targeting vector was designed containing a Lox71 site 5’ of Brwd1 exon 6, a neomycin-resistance cassette flanked by flippase recognition target (FRT) sites between Brwd1 exons 7 and 8, an inverse tdTomato reporter with a bovine growth hormone polyadenylation sequence (BGHpA), and a Lox66 site 3’ of Brwd1 exon 8. The Lox72 and Lox66 sites were in opposite orientations to one another. The targeting vector was introduced by electroporation to C57BL/6 embryonic stem cells containing flippase. Resulting cells were microinjected into Balb/c blastocysts. Resulting chimeras with a high percentage of black coat color were mated to C57BL/6 WT mice to generate germline neomycin-resistance cassette-deleted mice. PCR and sequencing were used to confirm the deletion of the neomycin-resistance cassette, the presence of the tdTomato cassette, and the presence of the Lox66 site. PCR was used to confirm the absence of the flippase transgene and the presence of the Lox71 site.

Routine genotyping of the Brwd1-floxed mice was done with the LOX1 and SDL2 primers, which cover the 5’ Lox71 site (Supplementary Table 2). All single-stranded DNA primers used in this study were synthesized by Integrated DNA Technologies. PCR conditions were 94 °C for 2 min, then 35 cycles of 94 °C for 30 s, 60 °C for 30 s, and 72 °C for 1 min, followed by 72 °C for 2 min. The WT amplicon is 389 bp, and the floxed amplicon is 431 bp.

Wild-type C57BL/6J mice (strain #000664), Aicda^Cre mice (strain #007770, C57BL/6J background), and Ai14 Rosa-CAG-LSL-tdTomato-WPRE mice (strain #007914, C57BL/6J background) were purchased from Jackson Laboratory. Cd23^Cre mice (C57BL/6J background) were obtained from the laboratory of Jayanta Chaudhuri (Memorial Sloan Kettering)³⁴. Experimental and control mice were co-housed in the University of Chicago barrier animal facilities. The facility light cycle was 12 h of light and 12 h of dark. The temperature was 68–79 °F, and the humidity was 30–70%. Permission to perform mouse experiments was granted by the University of Chicago Institutional Animal Care and Use Committee (protocol number 71577). Mice were used at 6–12 weeks of age. Female and male mice were used for all experiments, and experimental mice were sex-matched. Mice were euthanized by carbon dioxide asphyxiation followed by cervical dislocation.

Immunizations

Mice were immunized intraperitoneally with 10⁹ sheep red blood cells (SRBC, Lampire Biological Laboratories, Cat. 7249007) in 1X phosphate-buffered saline (PBS, MP Biomedicals Cat. 1960454) and boosted with 10⁹ SRBCs 5 days later. Alternatively, mice were immunized intraperitoneally with 200 μl of 4-hydroxy-3-nitrophenyl-acetyl (NP)-keyhole limpet hemocyanin (KLH) (1 mg/ml, valency of 27 or 33, Biosearch Technologies, Cat. N-5060) in a 1:1 ratio with Complete Freund’s Adjuvant (CFA, Sigma-Aldrich, Cat. F5881)⁵⁸. Mice were boosted with NP-KLH in a 1:5 ratio with Incomplete Freund’s Adjuvant (IFA, Sigma-Aldrich, Cat. F5506) at day 5. Mice were also immunized intraperitoneally with 10⁹ SRBCs conjugated with NP by incubating NP-Osu (Biosearch Technologies, Cat. N-1010) in 0.15 M NaHCO₃ with SRBCs. For this experiment, mice were boosted with NP-SRBCs at days 5 and 56 post-immunization.

Flow cytometry

Spleens were harvested, and cells were resuspended in PBS with 3% (v/v) fetal bovine serum (FBS, Sigma, Cat. F4135). Erythrocytes were lysed with ACK lysing buffer (Lonza, Cat. BP10-548E). Splenocytes were passed through a 100 μm cell strainer (Fisher Scientific, Cat. 22-363-549) to obtain single-cell suspensions. Peyer’s Patches were incubated in media with 1 mg/ml Collagenase VIII (Sigma-Aldrich, Cat. C2139) at 37 °C, 250 rpm for 20 min and passed through a 100 μm cell strainer (Fisher Scientific, Cat. 22-363-549). Fc block was done with anti-mouse-CD16/CD32 (2.4G2, BD Biosciences, Cat. 553141, 1:50). Cells were stained with fixable viability dye eFluor506 (eBioscience, Cat. 65-0866-14, 1:1000).

Fol B cells were stained with anti-B220-APC/Cy7 (RA3-6B2, BioLegend, Cat. 103211, 5 μg/ml), anti-CD19-PerCP/Cy5.5 (1D3, BD Biosciences, Cat. 561113, 5 μg/ml), anti-CD93-BV421 (AA4.1, BD Biosciences, Cat. 563806, 5 μg/ml), anti-CD23-PE/Cy7 (B3B4, BD Biosciences, Cat. 562825, 5 μg/ml), and anti-CD21-APC (7G6, BD Biosciences, Cat. 558658, 5 μg/ml). GC B cells were stained with anti-B220-AF488 (RA3-6B2, BioLegend, Cat. 103225, 5 μg/ml), GL7-PerCP/Cy5.5 (BioLegend, Cat. 144610, 5 μg/ml), anti-CD95-BV421 (Jo2, BD Biosciences, Cat. 562633, 5 μg/ml), anti-CD83-PE/Cy7 (Michel-19, BioLegend, Cat. 121518, 20 μg/ml), and anti-CXCR4-APC (2B11, Invitrogen, Cat. 17-9991-82, 20 μg/ml). Cells were also stained with various combinations of anti-CD138-APC (281-2, BioLegend, Cat. 142506, 5 μg/ml), anti-CD38-PE/Cy7 (90, BioLegend, Cat. 102718, 5 μg/ml), anti-IgD-PE/Cy7 (11-26c, Invitrogen, Cat. 25-5993-82, 5 μg/ml), anti-IgD-BV421 (11-26c.2a, BioLegend, Cat. 405725, 5 μg/ml), anti-IgM-APC (II/41, BD Biosciences, Cat. 550676, 5 μg/ml), anti-B220-APC (RA3-6B2, BioLegend, Cat. 103211, 5 μg/ml), anti-BCL6-PE/Cy7 (K112-91, BD Biosciences, Cat. 563582, 1:20), Annexin V-FITC (BD Biosciences, Cat. 556420, 1:20), FxCycle Violet (Invitrogen, Cat. F10347, 1:1000), and the FAM FLICA poly caspase kit (Bio-Rad, Cat. ICT092, 1:5). Cells were fixed with the Foxp3/Transcription factor staining buffer kit (Invitrogen, Cat. 00-5523-00). Tfh cells were stained with anti-CXCR5-BV421 (L138D7, BioLegend, Cat. 145512, 5 μg/ml), anti-PD-1-PerCP/Cy5.5 (RMP1-30, BioLegend, Cat. 109119, 5 μg/ml), anti-CD4-APC (RM4-5, BioLegend, Cat. 100515, 5 μg/ml), and anti-CD3-FITC (145-2C11, BD Biosciences, Cat. 553062, 5 μg/ml).

For flow cytometry of MBC subsets, cells were stained with anti-CD38-BUV496 (90/CD38, BD Biosciences, Cat. 741090, 3 μg/ml), anti-IgM-BUV615 (II/41, BD Biosciences, Cat. 751479, 3 μg/ml), anti-PD-L2-BUV737 (TY25, BD Biosciences, Cat. 741889, 5 μg/ml), anti-CCR6-BV421 (29-2L17, BioLegend, Cat. 129818, 1 μg/ml), anti-IgD-BV480 (11-26c.2a, BD Biosciences, Cat. 566106, 0.3 μg/ml), anti-CD80-BV650 (16-10A1, BioLegend, Cat. 104732, 5 μg/ml), anti-IgG1-BV711 (A85-1, BD Biosciences, Cat. 565786, 0.3 μg/ml), anti-CD95-BV786 (Jo2, BD Biosciences, Cat. 740906, 1 μg/ml), anti-B220-AF488 (RA3-6B2, BioLegend, Cat. 103225, 0.75 μg/ml), GL7-PerCP/Cy5.5 (BioLegend, Cat. 144610, 1 μg/ml), anti-CD83-PE/Cy7 (Michel-19, BioLegend, Cat. 121518, 3 μg/ml), anti-CXCR4-APC (2B11, Invitrogen, Cat. 17-9991-82, 3 μg/ml), anti-CD138-APC-R700 (281-2, BD Biosciences, Cat. 565176, 2 μg/ml), anti-CD73-APC-Fire750 (TY/11.8, BioLegend, Cat. 127222, 2 μg/ml), and fixable viability dye eFluor506 (eBioscience, Cat. 65-0866-14, 1:3000) with Brilliant Stain Buffer Plus (BD Biosciences, Cat. 566385).

Flow cytometry was done with a LSRFortessa (BD Biosciences) running FACSDiva version 8.0.2, and spectral flow cytometry was done with a Cytek Aurora (Cytek Biosciences). Flow cytometry data were analyzed with FlowJo version 10.8.1 (BD Biosciences). Total splenocytes were counted either with a hemocytometer or by adding 10 μm latex particles (Beckman Coulter, Cat. 15-655-123) to samples before flow cytometry analysis. The absolute numbers of different cell populations were calculated by multiplying the number of total splenocytes by the frequencies of the parent populations, including the lymphocyte frequency, the live cell frequency, and the parent populations noted within the gating strategies. Flow cytometry gating strategies are shown in Supplementary Figs. 13–17. Antibody clones, vendors, catalog numbers, lot numbers, dilutions or concentrations, and research resource identifiers are listed in Supplementary Table 3.

Cell sorting

Fol B cells were sorted by staining with anti-B220-BV421 (RA3-6B2, BioLegend, Cat. 103240, 1.25 μg/ml), anti-CD93-FITC (AA4.1, Invitrogen, Cat. 11-5892-82, 5 μg/ml), anti-CD21-APC (7G6, BD Biosciences, Cat. 558658, 5 μg/ml), and anti-CD23-PE/Cy7 (B3B4, BD Biosciences, 562825, 5 μg/ml). Enrichment of GC B cells for fluorescence-activated cell sorting (FACS) was done using magnetic-activated cell sorting (MACS)⁸³. GC B cell enrichment was done with anti-CD43-biotin (S7, BD Biosciences, Cat. 553269, 2.5 μg/ml), anti-CD11c-biotin (N418, BioLegend, Cat. 117304, 1.6 μg/ml), anti-IgD-biotin (11-26c.2a, BioLegend, Cat. 405734, 2 μg/ml), streptavidin microbeads (Miltenyi Biotec, Cat. 130-048-101), and LS columns (Miltenyi Biotec, Cat. 130-042-401). Cells were stained with the GC B cell antibody panel described above, and FACS was performed on a FACSAria Fusion (BD Biosciences) running FACSDiva version 8.0.2.

Western blot

Fol B cells were sorted and lysed in RIPA buffer (150 mM NaCl, 10 mM Tris, pH 7.4, 1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, 0.4 mM Na vanadate, 10 mM Na pyrophosphate) with protease inhibitor (Sigma-Aldrich, Cat. P8340) and 1 mM PMSF (Sigma-Aldrich, Cat. 93482) at a concentration of 2 × 10⁶ cells/30 μl. Samples were heated for 5 min at 100 °C, then run on a NuPage 3–8% tris-acetate gel (Invitrogen, Cat. TA03810) or a 4–20% tris-glycine gel (Invitrogen, XP04200). Proteins were transferred for 45 h at 85 mA constant amperage at 4 °C to a PVDF membrane (Invitrogen, Cat. LC2005) and immunoblotted with anti-BRWD1 (rabbit polyclonal, ThermoFisher, Cat. PA5-88688, 1:500 dilution) and anti-ACTIN (BA3R, ThermoFisher, Cat. MA5-15739, 1:1000 dilution) followed by anti-rabbit IgG HRP (donkey monoclonal, Cytiva, Cat. 45-001-276, 1:25,000 dilution) or anti-mouse HRP (sheep monoclonal, Cytiva, Cat. 45-000-679, 1:50,000 dilution), respectively.

Microscopy

On day 14 post-immunization, spleens were harvested, transferred to OCT medium (Fisher HealthCare, Cat. 4585), and frozen on dry ice in 2-methylbutane. Frozen spleens were sectioned at 6 μm. For immunofluorescence microscopy, sections were fixed with 4% paraformaldehyde, permeabilized with 1X NP40 Permeating Solution (Boston Products, Cat. BM-686) in PBS, and Fc receptors were blocked with anti-mouse-CD16/CD32 (2.4G2, BD Biosciences, Cat. 553141, 1:200) in 10% normal donkey serum (Jackson ImmunoResearch, Cat. 017-000-121).

For Cd23^Cre/wt;Brwd1^fl/fl mice, spleens were collected day 14 post-immunization. Tissue was stained with GL7-AF647 (BioLegend, Cat. 144606, 1:20) in 5% normal donkey serum (Jackson ImmunoResearch, Cat. 017-000-121). Images were acquired on a Leica Stellaris 8 laser scanning confocal microscope (Leica Microsystems) at ×10 magnification.

For Aicda^Cre/wt;Brwd1^fl/fl mice, tissue was stained with GL7-AF647 (BioLegend, Cat. 144606, 1:20), anti-CD4-AF594 (GK1.5, BioLegend, Cat. 100446, 1:20), anti-CD35-BV421 (8C12, BD Biosciences, Cat. 740029, 1:20), anti-IgD-BV711 (11-26c.2a, BD Biosciences, Cat. 564275, 1:20), anti-tdTomato (polyclonal, Rockland, Cat. 600-401-379, 1:50), followed by donkey anti-rabbit-IgG-AF488 (polyclonal, Invitrogen, Cat. A21206, 1:1000) in 5% normal donkey serum (Jackson ImmunoResearch, Cat. 017-000-121). The GL7 channel and differential interference contrast (DIC) were collected at ×10 magnification, and all channels were collected for individual GCs at ×20 magnification.

Image analysis

Images were visualized and analyzed with Fiji version 2.9.0⁸⁴. For quantification of GC size, GL7 images were Gaussian filtered (sigma = 10), and an intensity threshold and size exclusion threshold were used to remove background signal.

For mean pixel intensity (MPI) image analysis, images collected at ×20 magnification were normalized to the 99^th percentile for each channel. For a given channel, each image in the dataset was standardized through histogram matching to ensure similar contrast throughout the dataset. GCs were segmented by applying a Gaussian filter (sigma = 5) to the GL7 channel, then adjusting contrast and thresholding. The binary mask created by this segmentation algorithm was separated by instance using the regionprops function in the sci-kit image python library⁸⁵. After identifying discrete objects (i.e., GCs) from the binary segmentation mask, small objects were rejected using a size filter, and GCs at the image boundaries were rejected to ensure each GC was only measured once. The GC area was computed from the resulting segmentation masks. The LZ was segmented by applying a similar Gaussian filtering and thresholding algorithm to the CD35 channel. Resulting LZ segmentations were masked using the GC mask defined above to ensure that the segmented LZ was fully encompassed within the GC segmentation. MPI (pixel value sum/segmented area) was calculated for each marker within the regions defined by the GC and LZ masks.

In vitro class switch recombination

Splenic cells were harvested as above. B cells were enriched by incubating cells with anti-CD43-biotin (S7, BD Biosciences, Cat. 553269, 2.5 μg/ml) and streptavidin microbeads (Miltenyi Biotec, Cat. 130-048-101) followed by magnetic enrichment with LS columns (Miltenyi Biotec, Cat. 130-042-401). For class switching to IgG1, cells were cultured in RPMI 1640 with L-Glutamine media (Gibco, Cat. 11875093) with 10% FBS (Sigma, Cat. F4135), 0.1% β-mercapto-ethanol (Gibco, Cat. 21985-023), and 1% penicillin-streptomycin (Gibco, 15140-122) at 37 °C, 5% CO₂. For class switching to IgG1, 25 μg/ml lipopolysaccharides (Sigma-Aldrich, Cat. L2630) and 10 ng/ml IL-4 (R&D Systems, Cat. 404-ML) were added to the media.

After 72 h, RNA was isolated using an RNeasy Mini kit (Qiagen, Cat. 74104). IgG1 germline transcripts were measured by qPCR using primers to the Iγ1 and Cγ1 regions and Power SYBR Green PCR Master Mix (Applied Biosystems, Cat. 4367659). Data were normalized by the ΔΔCt method using Actin as the housekeeping gene.

After 96 h, class switching was measured by flow cytometry. Cells were stained with anti-CD19-APC/Cy7 (1D3, BD Biosciences, Cat. 557655, 5 μg/ml), anti-IgD-PE/Cy7 (11-26c, Invitrogen, Cat. 25-5993-82, 5 μg/ml), anti-IgM-APC (II/41, BD Biosciences, Cat. 550676, 5 μg/ml), anti-IgG1-FITC (RMG1-1, BioLegend, Cat. 406606, 5 μg/ml), and fixable viability dye eFluor506 (eBioscience, Cat. 65-0866-14, 1:1000).

Somatic hypermutation

RNA was extracted from sorted GC B cells using the RNeasy Mini Kit (Qiagen, Cat. 74104). cDNA was generated using the SuperScript III First-Strand Synthesis System for RT-PCR kit (Invitrogen, Cat. 18080051) and the mCγ1-cDNA or Cμ-cDNA primer. A semi-nested PCR was performed using the V186.2-leader and V186.2-nested primers with the Cγ1-PCR or Cμ-PCR primers⁵⁸. PCR products were purified with the MinElute PCR Purification Kit (Qiagen, Cat. 28006). Cloning was performed with the TOPO TA Cloning kit (Invitrogen, Cat. 450641) and One Shot TOP10 chemically competent cells (Cat. C404004). DNA from colonies was isolated with the PureLink Quick Plasmid Miniprep Kit (Invitrogen, Cat. K210011). DNA was sequenced with the M13 Forward primer.

The vector sequence was trimmed with 4Peaks (Nucleobytes). Alignment and mutations were analyzed with IMGT/V-QUEST⁸⁶. Duplicate sequences from the same mouse, sequences that did not align most closely to V186.2, and sequences with an uneven distribution of mutations were not used. Mutations in CDR3 introduced by junctional diversity were not counted to measure SHM.

Anti-nuclear antibody test

Mouse sera was diluted 1:100 in PBS and incubated with Kallestad HEp-2 cell line slides (BioRad, Cat. 26103). Slides were stained with donkey anti-mouse-IgG-AF647 (Invitrogen, Cat. A31571, 1:1000) and Hoechst 33342 (Invitrogen, Cat. H3570, 1:1000). Immunofluorescence microscopy was done with a Dragonfly 200 confocal microscope (Andor).

RNA-seq

Total RNA was isolated using a RNeasy Mini kit (Qiagen, Cat. 74104). Stranded poly-adenylated RNA libraries were prepared by the University of Chicago Genomics Facility before sequencing on the NovaSeq-X-Plus (Illumina). Raw reads were aligned to the reference genome mm9 in a splice-aware manner using STAR⁸⁷. Gene expression was quantified using featureCounts⁸⁸ against UCSC genes, with Ensembl IG genes from mm10 converted to mm9 coordinates with UCSC liftOver.

ATAC-seq

Cells were washed with PBS, then lysed with ATAC lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% IGEPAL CA-630). Resulting nuclei were then incubated with the tagment DNA TDE1 enzyme and buffer kit (Illumina, Cat. 20034210). Libraries from purified samples were made with the Nextera XT DNA library preparation kit (Illumina, Cat. 15055293).

Raw reads were aligned to the reference genome mm9 using BWA MEM⁸⁹. Apparent PCR duplicates were removed using Picard MarkDuplicates (https://github.com/broadinstitute/picard/releases/tag/2.11.0).

For ATAC-seq, read alignments were first adjusted to account for TAC transposon binding: +4 bp for +strand alignments, −5 bp for −strand alignments. The open chromatin enrichment track was generated by first creating a bedGraph from the raw reads using bedtools genomecov⁹⁰, then converted to bigWig using UCSC tool bedGraphToBigWig. Tracks were normalized by the sum of alignment lengths over 1 billion. Open chromatin peaks were called using Macs2 with no model set and no background provided⁹¹. Peaks with a score greater than 5 were retained. To quantitatively measure changes in epigenetic enrichment, we first identified empirical regulatory elements based on the peak calls obtained from each sample in the analysis. Peaks were merged into a uniform set of regulatory elements using bedtools merge⁹⁰. Enrichment levels for each regulatory element were then quantified with featureCounts⁸⁸.

Differential expression of RNA-seq and ATAC–seq

Differential expression statistics (fold-change and p value) were computed using edgeR on raw expression counts obtained from quantification (either genes or peaks)⁹². Pairwise comparisons were computed using exactTest, and multigroup comparisons using the generalized linear modeling capability in edgeR. In all cases, p values were adjusted for multiple testing using the FDR correction of Benjamini and Hochberg. Output data is given in Supplementary Data 1–4.

Next-generation sequencing analysis

RNA-seq and ATAC-seq data were visualized with the Integrated Genome Browser⁹³. Heatmaps were generated with ComplexHeatmap and Circlize in R, and plots were made with ggplot2^94,95. Metascape data portal was used for pathway analyses and for RNA-seq circle plots⁹⁶. Hypergeometric optimization of motif enrichment (HOMER) was used to perform known transcription factor binding motif analyses using the FindMotifsGenome function either with a random GC-normalized background or with unchanged chromatin accessibility peaks (q > 0.05) as background using the -bg function to generate enrichment and p values⁹⁷. Gene set enrichment analysis (GSEA) was performed using the Molecular Signatures Database hallmark genes sets^98,99. Hi-C data was visualized using Juicebox¹⁰⁰. Hi-C arc plots were generated with HiCcompare by normalizing Hi-C matrices using the hic_loess function and calculating differences using the hic_compare function¹⁰¹. Then the plotBedpe function from Sushi. R was used to plot differences with p < 0.05¹⁰².

TF footprinting analysis

TF footprinting analysis was performed with the following TF motifs and corresponding JASPAR Matrix IDs: CTCF (MA0139.2), ETS1 (MA0098.4), SPIB (MA0081.3), PU.1 (MA0080.7), and E2A (MA0091.2). The ETS/IRF EICE, IRF/IRF ISRE, and AP1/IRF AICE motifs were used with the letter probability matrices given in Supplementary Data 5⁴⁷. Differentially accessible regions between WT and Aicda^Cre/wt;Brwd1^fl/fl tdT⁺ cells in each GC subset (q < 0.05) were obtained separately for up- and down-regulated chromatin, and a motif search in each of these regions was performed using FIMO¹⁰³. TAC cut site signals from normalized enrichment tracks were aggregated around the motif hits for each set of peaks.

Statistical analysis

Statistical analyses were performed with GraphPad Prism. Bar graphs are displayed as the mean ± standard deviation. Significance, as defined by p value or false discovery rate (q value), is defined in the figure legends. All measurements were taken from distinct samples as described in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

N.E.W., M.M., and M.R.C. conceived and designed experiments. N.E.W. performed and analyzed most of the experiments. D.E.K., M.L.V., Y.M.Y., and N.E.W. performed flow cytometry. J.A. and N.E.W. performed immunofluorescence microscopy. M.D. and N.E.W. analyzed immunofluorescence microscopy images. M.L.V. performed western blotting. M.A. and N.E.W. performed in vitro class switch recombination experiments. J.V. assisted with SHM experiments. D.E.K. and N.E.W. performed RNA-seq and ATAC-seq. M.M.-C., D.E.K., M.M., and N.E.W. analyzed next-generation sequencing data. M.M.-C. and M.M. analyzed the Hi-C data. N.E.W. wrote the first draft of the manuscript. N.E.W. and M.R.C. edited the manuscript.

Peer review

Peer review information

Nature Communications thanks Jeremy Boss and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by National Institutes of Health (NIH) grants T32 GM007281 (to N.E.W. and M.R.C.), F30 AI174324 (to N.E.W.), R01 AI143778 (to M.R.C.), CZ Biohub Innovation Award (to M.R.C.), and R01 AI150860 (to M.R.C. and M.M.). M.M.-C. is supported in part by NCATS through grant UL1TR002003. The Cytometry and Antibody Technology Core Facility and the Genomics Facility receive financial support from the Cancer Center Support Grant P30CA014599.

Data availability

The RNA-seq and ATAC-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE264164. This study also analyzed WT GC B cell RNA-seq and ATAC-seq data from GEO accession code GSE133743¹⁴ and small pre-B cell Hi-C and ChIP-seq data from GEO accession code GSE221519³³. All data are included in the Supplementary Information or available from the authors, as are unique reagents used in this article. The raw numbers for charts and graphs are available in the Source Data file whenever possible. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77293-7.