HnRNP A1 and A2B1 enforce Ezh2 mRNA splicing to promote germinal center B cell responses Authors: Zhijian Zhu, Rui Zhang, Haoran Kang, Meiyuan Chen, Hengjun Huang, Jing Wang, Ling Guo, Yuxing Li, Wenjing Chen, Limin Meng, Ruisi Wang, Ling Li, Kong-Peng Lam, Shengli Xu, Xijun Ou Published: 2026 ABSTRACT AbstractHeterogeneous nuclear ribonucleoproteins (hnRNP) are key regulators of gene expression, yet the physiological functions of the highly homologous hnRNP A1 and A2B1 in B cells remain unclear. Here, we conditionally delete Hnrnpa1 and Hnrnpa2b1 in mouse B cells and find that loss of hnRNP A1 and A2B1 compromises the germinal center (GC) reaction during T cell-dependent immune responses. Loss of hnRNP A1 and A2B1 impairs GC B cell proliferation and high-affinity antibody production. Mechanistically, hnRNP A1/A2B1 bind UAG-rich motifs in Ezh2 pre-mRNA to promote Ezh2 exon 14 inclusion, thereby preserving EZH2 catalytic activity; in their absence, Ezh2 exon 14-skipping produces a catalytically inactive EZH2 isoform (Ezh2Δ14) with diminished repression of the cell cycle inhibitor Cdkn1a. The resulting CDKN1A accumulation restricts B cell proliferation, while Cdkn1a deletion partially rescues the GC defects in B cells deficient for hnRNP A1 and A2B1. Our findings thus uncover an hnRNP A1/A2B1–EZH2–CDKN1A axis that integrates RNA splicing with epigenetic regulation of B cell immunity.Subject terms: Germinal centres, Epigenetics in immune cells, B cells Introduction Germinal centers (GC) are specialized microanatomical structures within secondary lymphoid organs such as lymph nodes and the spleen. Within the GC, B cells are activated and undergo robust proliferation and affinity maturation—a process driven by iterative cycles of clonal expansion, somatic hypermutation of immunoglobulin (Ig) variable genes, and clonal selection. This leads to the generation of antigen-specific, high-affinity antibodies, contributing to both immediate and long-term immune protection1,2. The GC response is tightly regulated by a coordinated multi-layered regulatory network, including transcriptional, epigenetic, and post-transcriptional programs3–8. RNA-binding proteins (RBP) represent key regulators of gene expression at both the transcriptional and post-transcriptional levels3. Among them, heterogeneous nuclear ribonucleoproteins (hnRNP) constitute a large family of classical RBPs that are abundantly expressed in mammalian cells. This family comprises approximately 30 members, including 20 major types designated hnRNPs A through U, which perform diverse functions such as transcriptional regulation, maintenance of RNA stability, and alternative splicing (AS)9–15. Previous studies have demonstrated that several hnRNPs play important roles in B cell physiology. For example, the hnRNP L/LL subgroup regulates B cell fate decisions by controlling c-Myc and E2F-driven transcriptional programs, which are essential for cell cycle progression, and by modulating the alternative splicing of Ptprc, which encodes the phosphatase CD45, a central regulator of B cell receptor signaling9–11. During terminal B cell differentiation, hnRNP LL promotes plasma cell formation and antibody secretion by suppressing Bcl6 and enhancing global RNA processing12. In addition, hnRNP L and hnRNP K play an important role in driving IgA class switching recombination13. hnRNP I (also known as PTBP1) supports B cell proliferation by coordinating metabolic reprogramming—such as PKM isoform switching—and regulating the expression of key cell cycle genes, including Ccnd2 and Myc14,15. More recently, hnRNP F has been shown to play an important role in the GC B cell response by regulating the alternative splicing of Cd408. HnRNP A1 and hnRNP A2B1 are two highly homologous paralogs among the most abundantly expressed and widely distributed members of the hnRNP family. Structurally, both proteins contain two N-terminal RNA recognition motifs for sequence-specific RNA binding and a C-terminal glycine-rich domain that facilitates protein-protein interactions16. Functionally, they act as key regulators of AS programs associated with cell proliferation. HnRNP A1 promotes exon inclusion in genes such as Fas (apoptosis regulation), Atp7b (copper transport), and Cdk2 (cell cycle control), while hnRNP A2B1 regulates the splicing of Caprin2 (Wnt signaling), Tcf12 (lymphoid development), Fgfr3 (growth factor signaling), and Prkrir (antiviral response)17–19. However, their physiological roles during the GC response remain to be elucidated. In this study, we demonstrate that hnRNP A1 and hnRNP A2B1 play overlapping yet substantial roles in regulating B cell activation and humoral immune responses. Their combined ablation in B cells impairs GC B cell formation and T cell-dependent (TD) antibody production, defects driven by the loss of the hnRNP-controlled EZH2 isoform switch and the subsequent derepression of Cdkn1a. These findings identify an hnRNP A1/A2B1–EZH2–CDKN1A cascade as a checkpoint in humoral immunity, with implications for understanding post-transcriptional regulation of epigenetic machinery in immune cell differentiation and for potential therapeutic targeting in antibody-mediated disorders. Results B cell development is largely normal in CD19-Cre-mediated hnRNP A1/hnRNP A2B1 double knockout mice hnRNP A1 and hnRNP A2B1 belong to the hnRNP A/B subfamily, which also includes hnRNP A0 and hnRNP A3. Within this subfamily, hnRNP A1 and A2B1 are the most evolutionarily conserved, highly expressed, and extensively studied members16. HnRNP A1, A2B1, and A3 exhibit significantly higher sequence conservation compared to hnRNP A0. Previous studies have demonstrated complementary expression patterns between hnRNP A1 and A2B1, suggesting potential functional redundancy20. To investigate the roles of hnRNP A1 and A2B1 in B cells, we first examined their expression profiles. RNA-seq analysis revealed that hnRNP A1 and A2B1 are the two most abundantly expressed hnRNP family members in splenic B cells, while hnRNP A3 exhibits minimal expression (Supplementary Fig. 1A). Further analysis using the ImmGen database showed significantly elevated expression of hnRNP A1 and A2B1 in early B cells and GC B cells compared to other B cell subsets (Supplementary Fig. 1B). In addition, the protein levels of hnRNP A1, hnRNP A2B1, and hnRNP A3 dynamically change across different stages of B cell development, with hnRNP A1 and hnRNP A2B1 being expressed at significantly higher levels in Fr.C/C’ B cells and GC B cells than at other stages (Supplementary Fig. 1C). Public dataset mining also confirmed their specific upregulation in GC B cells relative to naïve B cells, which we further validated by qRT-PCR in sorted GC and non–GC B cells (Supplementary Fig. 1D, E). These findings suggest that hnRNP A1 and A2B1 may play important roles in B cell development and/or differentiation. To assess the role of hnRNP A1 and A2B1 in B cell development, we crossed mice carrying loxP–flanked (floxed) Hnrnpa1 or Hnrnpa2b1 alleles with Cd19Cre/+ mice, which express Cre recombinase from the pro-B cell stage onward, to generate B cell-specific hnRNP A1 single knockout (A1-SKO: Hnrnpa1f/f Cd19Cre/+), hnRNP A2B1 single knockout (A2B1-SKO: Hnrnpa2b1f/f Cd19Cre/+), and hnRNP A1/A2B1 double knockout (DKO: Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+) mice (Supplementary Fig. 1F). RT-PCR confirmed the successful insertion of loxP sites into Hnrnpa1 and Hnrnpa2b1 alleles (Supplementary Fig. 1G). qRT-PCR and immunoblotting further demonstrated the efficient deletion of Hnrnpa1 and Hnrnpa2b1 in splenic B cells from the respective SKO and DKO mice (Supplementary Fig. 1H, I). Interestingly, we observed compensatory upregulation of hnRNP A1 in hnRNP A2B1-deficient B cells (Supplementary Fig. 1I). These findings suggest potential functional compensation between the two proteins. Next, we examined B-cell development in these mice by flow cytometry. Overall, B cell development appeared largely intact in the absence of hnRNP A1 and/or A2B1. There were no significant differences in the percentages or absolute numbers of B cell subsets in the bone marrow (BM) or spleen of A1-SKO, A2B1-SKO, or DKO mice compared with wild-type controls (Supplementary Fig. 2A, B, Fig. 1A–J). These results indicate that CD19-Cre-mediated deletion of hnRNP A1 and A2B1 does not substantially impair B cell development. B cell development is largely normal in CD19-Cre-mediated hnRNP A1/hnRNP A2B1 double knockout mice. A–F Flow cytometry analysis of B cell development in the bone marrow of Hnrnpa1f/f Hnrnpa2b1f/f (WT), Hnrnpa1f/f Cd19Cre/+ (A1-SKO), Hnrnpa2b1f/f Cd19Cre/+ (A2B1-SKO), and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) mice using Hardy’s gating strategy. Absolute cell numbers were calculated based on total cell counts obtained from a single mouse femur. A, B Representative flow cytometry plots (A) and quantification of the frequency and number (B) of total B cells in the bone marrow. C–F Representative plots (C, E) and corresponding quantification (D, F) showing the number of B cells at distinct developmental stages in the bone marrow. G–J Flow cytometry analysis of B cell subsets in the spleens of WT, A1-SKO, A2B1-SKO, and DKO mice using a standardized gating and quantification strategy. G, H Representative plots (G) and quantification (H) of the frequency and number of splenic B cells. I, J Representative plots (I) and quantification (J) of follicular and marginal zone B cells in the spleen. Sample sizes were as follows: WT, A2B1-SKO, and DKO (n = 5); A1-SKO (n = 4) (panels B, D, F, H, I). Each symbol represents an individual mouse (B, D, F, H, I). All data are shown as mean ± SEM and are representative of three independent experiments (B, D, F, H, I). Statistical significance was determined using one-way ANOVA followed by Dunnett’s multiple comparisons test (B, D, F, H, I). To specifically investigate the roles of hnRNP A1 and A2B1 in early B lymphopoiesis, we used Mb1Cre/+ mice, which drive Cre expression earlier than Cd19Cre/+. We generated Hnrnpa1f/f Mb1Cre/+ mice (denoted as A1-sKO), Hnrnpa2b1f/f Mb1Cre/+ mice (denoted as A2B1-sKO) and Hnrnpa1f/f Hnrnpa2b1f/f Mb1Cre/+ (denoted as dKO). Flow cytometric analysis revealed that the percentages and absolute numbers of total B220+ B cells in the BM of A1-sKO and A2B1-sKO mice were comparable to those in WT controls (Supplementary Fig. 3A, B). In contrast, simultaneous deletion of both hnRNP A1 and A2B1 caused a severe reduction, with dKO mice retaining only approximately one-third the number of BM B cells compared with WT mice (Supplementary Fig. 3A, B). A more detailed examination of early B cell subsets based on Hardy’s classification demonstrated that the absolute numbers of CD249– CD24– pre-pro-B (Fr. A), CD249– CD24+ early pro-B (Fr. B), CD249+ CD24low late pro-B (Fr. C), and CD249+ CD24high large pre-B (Fr. C’) cells in A1-sKO and A2B1-sKO mice were comparable to those in WT mice (Supplementary Fig. 3C, D). However, the absolute numbers of large pre-B (Fr. C’) cells in hnRNP A1/A2B1 dKO mice were significantly decreased compared to WT mice (Supplementary Fig. 3C, D). Similarly, the B220+ CD43– IgM– IgD– small pre-B (Fr. D) and B220+ CD43– IgM+ IgD– immature B (Fr. E) subsets were unaffected in A1-sKO and A2B1-sKO mice but were drastically reduced in hnRNP A1/A2B1 dKO mice (Supplementary Fig. 3E, F). Together, these results show that Mb1-Cre-mediated deletion of both hnRNP A1 and A2B1 reveals a redundant role for these proteins in early B cell development. Because the primary focus of this study is antibody immune responses, which involve more mature B cells, we used the CD19-Cre-mediated knockout mice for the subsequent experiments. HnRNP A1/A2B1 DKO mice have a defective TD antibody response To evaluate the role of hnRNP A1 and A2B1 in the TD antibody response, we immunized the WT, SKO, and DKO mice with TD antigen 4-hydroxy-3-nitrophenyl-acetyl (NP) hapten conjugated to Keyhole Limpet Hemocyanin (KLH). Mice were immunized via intraperitoneal injection of NP-KLH/Alum, and antibody-secreting cells were assessed by ELISPOT (enzyme-linked immunoSpot) assay 10 days post-immunization (Fig. 2A). A1-SKO or A2B1-SKO mice showed a comparable number of IgM+ or IgG1+ antibody-secreting cells to WT mice. In contrast, hnRNP A1/A2B1 DKO mice manifested a significant reduction in the formation of IgG1 antibody-secreting cells (Fig. 2B, C), suggesting that hnRNP A1 and hnRNP A2B1 play overlapping yet non-redundant roles in the formation of IgG1+ antibody–secreting cells. hnRNP A1 and A2B1 are required for TD antibody response. A–C Hnrnpa1f/f Hnrnpa2b1f/f (WT), Hnrnpa1f/f Cd19Cre/+ (A1-SKO), Hnrnpa2b1f/f Cd19Cre/+ (A2B1-SKO), and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) mice were immunized with NP-KLH/Alum. On day 10 post-immunization, spleens were collected and analyzed by ELISPOT to quantify NP20-specific IgM and IgG1 antibody-secreting cells. A Schematic of the immunization protocol. B Representative ELISPOT images showing NP20-specific IgM and IgG1 antibody-secreting cells in the spleens of WT, A1-SKO, A2B1-SKO, and DKO mice. C Quantification of antibody-secreting cell numbers from the ELISPOT assay on day 10 post-immunization (WT, A2B1-SKO, and DKO: n = 7; A1-SKO: n = 5). p < 0.0001 ( = 8.99e-6). D–K WT and DKO mice were immunized with NP-KLH/Alum, and sera were collected at multiple time points to assess NP-specific antibody titers by ELISA. D Schematic of the tail vein blood collection schedule. E–H Titers of NP20-specific IgM (E), IgG1 (F), IgG3 (G), and IgG2b (H) antibodies in WT and DKO mice at indicated time points (n = 7). p < 0.0001( = 9.02e-5). I–K Ratios of NP2- to NP20-binding antibodies in the sera of WT and DKO mice on days 14, 21, and 28 post-immunization (n = 7). Each symbol represents an individual mouse (C, I–K). Data are shown as mean ± SEM. Data in (B, C) are representative of three independent experiments; data in (E–K) are from a single experiment. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparisons test (C) or Mixed-effects analysis followed by Sidak’s multiple comparisons test (E–H) or two-way ANOVA followed by Sidak’s multiple comparisons test (I–K). Further analysis by ELISA (enzyme-linked immunosorbent assay) revealed that double deficiency of hnRNP A1 and hnRNP A2B1 markedly impaired the antibody production of various IgG isotypes, including IgG1, IgG2b, and IgG3 (Fig. 2D–H). During the GC response, B cells undergo somatic hypermutation and affinity maturation in the dark zone (DZ) and light zone (LZ) of the GC. We found that the secretion of NP2+ high-affinity antibodies was significantly reduced in hnRNP A1/A2B1-deficient mice compared to controls, as evidenced by a notable decrease in the NP2/NP20 binding ratio (Fig. 2I–K), suggesting impaired generation of high-affinity antibodies. These results together suggest that hnRNP A1 and hnRNP A2B1 are important for the TD antibody response. HnRNP A1/A2B1 DKO mice manifest a compromised GC B cell response Next, we investigated the antibody response in hnRNP A1/A2B1 DKO mice in detail by examining the GC B cell responses, which are known to contribute to the generation of antigen-specific, Ig heavy chain class-switched, high–affinity antibodies upon TD antigen challenge1. Flow cytometric analysis showed that the percentage and absolute number of B220+ CD95+ GL-7+ GC B cells in the spleen of A1–SKO, A2B1–SKO mice at day 10 post-immunization with NP-KLH/Alum were moderately decreased compared to WT controls, though the difference was not statistically significant (Supplementary Fig. 2C, Fig. 3A, B). In contrast, hnRNP A1/A2B1 DKO mice exhibited a severe reduction both in the percentage and absolute number of GC B cells (Fig. 3A, B), suggesting that hnRNP A1 and hnRNP A2B1 contribute to GC B cell responses. hnRNP A1 and A2B1 contribute to GC B cell development. A–H Mice were immunized with NP-KLH/Alum and analyzed on day 10. A, B Representative flow cytometry plots and quantification of Germinal center (GC) B cells frequency and number in Hnrnpa1f/f Hnrnpa2b1f/f (WT), Hnrnpa1f/f Cd19Cre/+ (A1-SKO), Hnrnpa2b1f/f Cd19Cre/+ (A2B1-SKO), and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) mice (WT, A2B1-SKO, and DKO: n = 8; A1-SKO: n = 7). C, D Representative plots and quantification of IgG1⁺ memory B cells in WT and DKO mice (WT, A2B1-SKO, and DKO: n = 8; A1-SKO: n = 7). E, F Representative plots and quantification of NP⁺ antigen-specific and IgG1⁺ class-switched B cells in Hnrnpa1f/f Hnrnpa2b1f/f (WT) and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (DKO) mice (n = 7). G, H Representative plots and quantification of plasma cells in WT and DKO mice (WT: n = 3; DKO: n = 4). I–N Mice were immunized with NP-KLH/Alum and analyzed on day 5. I, J Representative plots and quantification of early GC B cell frequency and number in the spleen (n = 6). p < 0.0001( = 3.20e-5). K, L Representative flow cytometry and statistical graphs of BCL6+ early GC B cells (n = 5). M, N Proliferation of early GC B cells assessed by BrdU incorporation on day 5 post-immunization (n = 6). Each symbol represents an individual mouse (B, D, F, H, J, L, N). Data are shown as mean ± SEM. Data in (B, D, F, J, N) represent three independent experiments; data in (H, L) represent two independent experiments. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparisons test (B, D), unpaired two-tailed Student’s t-test (F, H, J, L) or paired two-tailed Student’s t-test (N). We further characterized GC B cells in hnRNP A1/A2B1 DKO mice in detail by flow cytometry (Supplementary Fig. 2D, E). We found that IgG1+ memory B cells (Fig. 3C, D), NP-specific IgG1+ GC B cells (Fig. 3E, F), and CD138+ TACI+ plasma cells (Fig. 3G, H) were all significantly reduced in hnRNP A1/A2B1 DKO mice compared to the WT mice. These findings suggest that hnRNP A1 and hnRNP A2B1 contribute to terminal B cell differentiation during the GC responses. To elucidate the mechanism underlying this reduction, we examined the initial formation of nascent GC. On day 5 post-immunization with NP–KLH, both the percentage and absolute number of B220+ Fas+ CD38- GC B cells were significantly lower in DKO mice compared to WT mice (Supplementary Fig. 4A, Fig. 3I, J). Furthermore, the frequencies and absolute numbers of CD138+ TACI+ plasmablasts, NP+ IgG1+ B cells, and NP+ IgG1+ memory B cells were also markedly reduced in DKO mice (Supplementary Fig. 4B–G). These results demonstrate that the absence of hnRNP A1 and hnRNP A2B1 impairs the early B cell activation and differentiation processes, which in turn hinders the generation and maturation of the GC compartment. BCL6 is the master transcription factor required for GC B cell commitment and for supporting their interactions with follicular helper T cells21,22. Consistent with the observed defects in early activation, we found that BCL6 expression was significantly reduced in early GC B cells from DKO mice (Fig. 3K, L). This reduction further indicates a profound failure in the initiation of the GC program. Within the GC, GC B cells undergo robust proliferation and tightly regulated programmed cell death to enable affinity-based selection23. To determine whether the reduction in GC B cells was due to defects in proliferation, survival, or both, we first assessed proliferation using a bromodeoxyuridine (BrdU) incorporation assay on day 5 post-immunization. BrdU incorporation in early-GC B cells from DKO mice was significantly lower than in WT controls (Fig. 3M, N), whereas cell survival was not affected (Supplementary Fig. 4H, I). Interestingly, at the peak of the GC response on day 10 post-immunization, only a slight and statistically non-significant reduction in proliferative capacity was observed in DKO mice (Supplementary Fig. 4J, K). Consistently, caspase activity measured by the CaspGLOW assay revealed no significant difference in the percentage of apoptotic GC B cells between WT and DKO mice at this time point (Supplementary Fig. 4L, M). Similarly, plasmablast proliferation was impaired in DKO mice at early stages (Supplementary Fig. 4N, O), while apoptosis rates remained comparable to WT controls (Supplementary Fig. 4P, Q). Given that early plasmablasts primarily arise from extrafollicular differentiation pathways, the impaired proliferation observed in DKO mice suggests that hnRNP A1 and A2B1 may regulate early B cell proliferative programs. Together, these findings indicate that hnRNP A1 and hnRNP A2B1 play an important role in early B cell proliferative responses, which may contribute to the observed defects in GC B cell responses, without significantly affecting cell survival. Loss of hnRNP A1/A2B1 promotes Ezh2 exon 14 skipping and isoform switching Previous studies have established that hnRNP proteins exert diverse, cell-type-dependent functions beyond AS, including the regulation of RNA stability, translational control, and m⁶A recognition24,25. However, in the context of B cell biology, accumulating evidence indicates that several hnRNP family members—such as hnRNP F, hnRNP I, and hnRNP L—primarily exert their important functions through AS regulation of key transcripts governing B cell differentiation and function8,9,14. We therefore hypothesized that the GC defects observed in A1/A2B1-deficient mice arise from aberrant splicing of functionally essential target genes. To test this, we performed transcriptomic analysis on splenic naïve B cells from DKO and control mice, followed by AS event analysis using rMATS, a widely used tool for detecting differential alternative splicing from replicate RNA sequencing data26. A range of splicing alterations were identified in DKO B cells, including 63 alternative 3’ splice site (A3SS) events, 112 alternative 5’ splice site (A5SS) events, 157 mutually exclusive exon (MXE) events, 91 retained intron (RI) events, and 1037 skipped exon (SE) events (Fig. 4A). Notably, exon skipping and exon inclusion were the most affected AS types, with 236 exon-skipping and 801 exon-inclusion events detected in the mutant B cells (Fig. 4A, Supplementary Data 1). hnRNP A1 and A2B1 regulate alternative splicing of Ezh2 in B cells. A Changes in five types of alternative splicing events upon hnRNP A1/A2B1 knockout. Scatter plots show exon inclusion level differences between Hnrnpa1f/f Hnrnpa2b1f/f (WT) and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) naïve B cells. Each dot represents a significantly altered splicing event (|inclusion level difference| > 20%, P < 0.05). B Lollipop plot showing the top 10 most significant exon-skipping events (ranked by P-value) in naïve B cells from DKO mice. The height of each lollipop indicates the inclusion level difference, and the circle size reflects statistical significance. C RNA expression changes of the top 10 most significant exon-skipping target genes (ranked by P-value) in follicular and Germinal center (GC) B cells from WT mice. Circle size and color intensity represent changes in FPKM values. Data from GEO: GSE15907. D–H Alternative splicing of Ezh2 following hnRNP A1/A2B1 deletion. D Sashimi plot illustrating exon 14 splicing changes in Ezh2 in WT and DKO naïve B cells, showing both the full-length isoform and the exon 14-skipped isoform (Ezh2Δ14). E Quantification of splice junction reads spanning exons 13–14, 14–15, and 13–15 in WT and DKO naïve B cells (n = 3), data in (E) are from a single experiment (RNA-seq). F Schematic and primer design for detecting full-length (Ezh2-FL) and exon 14-skipped (Ezh2Δ14) isoforms. G Semi-quantitative PCR showing the effect of single and double hnRNP knockouts on Ezh2 exon 14 skipping in WT and DKO naïve B cells, with accompanying gel image and the ratio of Ezh2Δ14 to Ezh2-FL. The data shown are representative images from two independent experiments. H Quantification of Ezh2-FL and Ezh2Δ14 expression normalized to β-actin in WT and DKO naïve B cells (n = 2), the data shown here represent one representative set of statistical results from two independent experiments. I, J Western blot analysis detecting the abundance of H3K27me3 in B cells from WT, A1-SKO, A2B1-SKO, and DKO mice after 48 h of stimulation with anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL). The data shown are representative images from two independent experiments. K Immunoblot analysis of H3K27me3 levels in WT and DKO B cells. Cells were pre-stimulated with anti-CD40 and anti-IgM and subsequently reconstituted with an empty vector (Vector), full-length EZH2 (Ezh2-FL), or the exon 14–skipped isoform (Ezh2Δ14). The data shown are representative images from two independent experiments. Each symbol represents an individual mouse (E, H). All data are shown as mean ± SEM. Results are representative of at least two independent experiments (H). Statistical significance was assessed using an unpaired two-tailed Student’s t-test (E). Among the top 10 genes most affected by increased exon skipping (Fig. 4B), Ezh2 emerged as a strong candidate given its well-established role in B cell proliferation and GC responses27,28. Notably, Ezh2 remained among the most significantly altered targets even under stringent statistical filtering for both effect size and significance (Supplementary Fig. 5A, Supplementary Data 2). Mechanistically, EZH2 promotes GC B-cell proliferation by transcriptionally repressing the cell-cycle inhibitor CDKN1A (p21)27–29, a function that closely aligns with the proliferation defects observed in our DKO model. Supporting this functional link, Ezh2 expression is markedly upregulated in GC B cells compared with follicular B cells (Fig. 4C), with particularly high expression in AP4+ GC B cells undergoing positive selection (Supplementary Fig. 5B, C). Notably, the expression pattern of Ezh2 across GC B cell subsets closely parallels those of Hnrnpa1 and Hnrnpa2b1 (Supplementary Fig. 5D), suggesting a potential regulatory relationship in which hnRNP A1 and A2B1 may influence Ezh2 splicing and B cell proliferation. Detailed analysis of Ezh2 AS revealed a pronounced exon 14 skipping in DKO naïve B cells, whereas the splicing of other exons remained unaffected (Fig. 4D, E). Exon 14 skipping generates the Ezh2Δ14 isoform, which was detected by semi-quantitative RT-PCR using AS-Ezh2-F and AS-Ezh2-R primers (Fig. 4F). In WT naïve B cells, full-length Ezh2 (Ezh2-FL) predominates over Ezh2Δ14, with the Δ14/FL ratio typically below 0.5. Notably, this pattern was reversed in DKO naïve B cells, where the Δ14/FL ratio exceeds 1.2 (Fig. 4G, H). In contrast, single deletion of Hnrnpa1 or Hnrnpa2b1 had no detectable impact on Ezh2 AS (Supplementary Fig. 5E). To further investigate whether Ezh2 splicing was similarly altered in GC B cells, we sorted GC B cells from mice immunized with NP-KLH/Alum for 7 days. Consistent with our earlier findings in naïve B cells, single deletion of Hnrnpa1 or Hnrnpa2b1 did not significantly increase Ezh2 exon 14 skipping. However, concurrent deletion of both genes markedly enhanced exon 14 skipping in DKO GC B cells (Supplementary Fig. 5F, G), indicating that hnRNP A1 and hnRNP A2B1 function redundantly to maintain exon 14 inclusion. Since exon 14 skipping disrupts the CXC domain, which is required for DNA interaction and H3K27me3 catalytic activity—we next assessed global H3K27me3 levels. We hypothesized that hnRNP A1/A2B1 may regulate B cell activation and differentiation, potentially through modulation of EZH2 alternative splicing. To model B cell activation in vitro, we stimulated CD19⁺ splenic B cells with anti-CD40 and anti-IgM. We observed that H3K27me3 abundance was significantly reduced only in DKO B cells, but not in the single KO B cells (Fig. 4I, J). The protein abundance of H3K27me3 in unstimulated B cells was so low that it was difficult to determine whether any differences existed (Supplementary Fig. 5H). Currently, no antibody specifically recognizing the peptide corresponding to EZH2 exon 14 is available, making it impossible to detect the protein expression of the EZH2–FL and EZH2Δ14 isoforms by immunoblotting. Nevertheless, isoform-specific qRT-PCR analysis further demonstrated differential expression patterns of Ezh2–FL and Ezh2Δ14 (Supplementary Fig. 5I), despite unchanged total Ezh2 mRNA levels, as shown by RNA-seq analysis (Supplementary Fig. 5J). Furthermore, the expression and splicing patterns of other PRC2 core components–SUZ12, EED, and EZH1–remained unaltered in DKO B cells (Supplementary Data 3, Supplementary Fig. 5J, K). To conclusively determine whether the loss of H3K27me3 resulted directly from Ezh2 exon 14 skipping, we performed rescue experiments by reintroducing either Ezh2–FL or Ezh2Δ14 into DKO B cells. Re-expression of Ezh2-FL fully restored H3K27me3 levels, whereas Ezh2Δ14 failed to rescue the defect (Fig. 4K). Collectively, these findings demonstrate that hnRNP A1 and A2B1 are required for proper Ezh2 splicing, which is essential for maintaining the H3K27me3 epigenetic landscape in activated B cells. hnRNP A1 and A2B1 directly bind to U(A)AG motifs in Ezh2 pre-mRNA to promote exon 14 inclusion Next, we investigated the molecular mechanisms by which hnRNP A1/A2B1 regulates Ezh2 AS. We first performed native RNA immunoprecipitation (RIP) assays using an hnRNP A1 or hnRNP A2B1-specific antibody. The results showed significant enrichment of Ezh2 mRNA in RNA precipitated with anti-hnRNP A1 or anti-hnRNP A2B1 antibody compared to control samples, indicating specific binding of hnRNP A1 and hnRNP A2B1 to Ezh2 transcripts (Fig. 5A). hnRNP A1 and A2B1 directly bind U(A)AG motifs in Ezh2 pre-mRNA to promote exon 14 inclusion. A RIP-qPCR analysis showing enrichment of Ezh2 transcripts associated with hnRNP A1 or hnRNP A2B1 in splenic B cells from WT mice, with B cells from Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) mice used as a control for RIP (n = 2). Data are shown as mean ± SD from a single experiment. Statistical analysis was performed using an unpaired two-tailed Student’s t-test. B RBPmap prediction of hnRNP A1 and hnRNP A2B1 binding sites in Ezh2. The analysis used exon 14 and part of intron 14 of Ezh2 as input. Underlined regions denote potential shared binding sites for hnRNP A1 and A2B1. C–E Construction and analysis of an Ezh2 minigene to assess exon 14 skipping upon site-directed mutation of predicted binding sites. C Schematic of the Ezh2 minigene and the design of mutations at predicted binding sites (M1, M2, and M3). Additional constructs include double (M1 + M3) and triple (M1 + M2 + M3) mutants. D, E WT and DKO B cells were infected with retroviruses carrying minigene constructs (WT, M1, M3, M1 + M3, and M1 + M2 + M3). 48 h after infection, GFP⁺ B cells were sorted by flow cytometry and subjected to RT-PCR analysis of Ezh2 splicing isoforms. The gel image shows the relative abundance of Ezh2Δ14 and Ezh2-FL isoforms. The data shown are representative images from two independent experiments. F–I RNA pull-down assay to evaluate binding of wild-type and mutant Ezh2 pre-RNAs to hnRNP A1 and A2B1. F Schematic of the RNA pull-down workflow. B cells were purified using CD19 magnetic beads, nuclear extracts were prepared, and biotin-labeled RNAs were pulled down using Streptavidin Magnetic Beads. G Biotin-labeled RNAs synthesized for pull-down: wild-type intron 14 RNA (I14–WT–Biotin), reverse complementary RNA (I14–Antisense–Biotin), triple mutant (I14–M1 + M2 + M3–Biotin), and M1-only mutant (I14–M1–Biotin). H Immunoblot analysis of proteins pulled down with biotin-labeled RNAs (I14–M1 + M2 + M3–Biotin), detecting hnRNP A1 and A2B1. I RNA pull-down assay was performed using the biotin-labeled RNA (I14–M1–Biotin), followed by immunoblotting to detect the precipitated hnRNP A1 and hnRNP A2B1 proteins. The data shown are representative images from two independent experiments. J–M RNA EMSA to determine direct binding of hnRNP A1 and A2B1 to wild-type or mutant Ezh2 pre-mRNA. J EMSA with hnRNP A1 and wild-type vs. M1 + M2 + M3 mutant Ezh2 pre-RNA. K EMSA with hnRNP A2B1 and wild-type vs. M1 + M2 + M3 mutant Ezh2 pre-RNA. L EMSA with hnRNP A1 and wild-type vs. M1 mutant Ezh2 pre-RNA. M EMSA with hnRNP A2B1 and wild-type vs. M1 mutant Ezh2 pre-RNA. The data shown are representative images from two independent experiments. All data are shown as mean ± SEM. Results shown are representative of three independent experiments (A). Statistical significance was assessed using an unpaired two-tailed Student’s t-test (A). To identify hnRNP A1/A2B1 binding sites within the Ezh2 locus, we analyzed the sequences near exon 14 and its upstream intron. Previous studies have demonstrated that hnRNP A1 and A2B1 preferentially bind UAG-rich motifs near splice sites to regulate inclusion or skipping30,31. We identified several UAG motifs near the 5’ splice site of intron 14, suggesting their potential involvement in splicing regulation. RBPmap analysis, which is commonly used for motif analysis and predicting the binding sites of RNA-binding proteins32, confirmed a strong enrichment of predicted binding sites within these UAG-enriched sequences (Fig. 5B). To evaluate the functional relevance of these motifs, we conducted minigene splicing assays. A WT Ezh2 minigene was constructed, containing exons 13, 14, and 15, along with partial sequences of introns 13 and 14 (Fig. 5C). Mutant minigenes were generated with UAG motif disruptions in three regions (M1, M2, M3) (Fig. 5C). WT or DKO mouse splenic B cells were infected with retroviral particles carrying either the WT or mutant minigenes, and the successfully transduced GFP+ B cells were sorted by flow cytometry before analyzing splicing outcomes by RT-PCR. In WT B cells, the WT minigene predominantly produced the full-length Ezh2 transcript. However, mutation of the M1 region significantly increased the expression of the Ezh2Δ14 isoform and elevated the Δ14/FL ratio (Fig. 5D). In contrast, mutation of M3 alone had no appreciable effect, whereas combined mutations (M1 + M2 and M1 + M2 + M3) further exacerbated exon 14 skipping (Fig. 5D). When these constructs were transfected into DKO B cells, a similar pattern was observed. Notably, generation of the Ezh2Δ14 isoform was consistently more pronounced in the DKO background than in WT B cells under all experimental conditions (Fig. 5E). Together, these results demonstrate that the UAG motifs, particularly those within the M1 region, function as cis-acting elements that cooperate with hnRNP A1 and A2B1 to ensure proper inclusion of Ezh2 exon 14. We further employed RNA pull-down assays using biotinylated RNA fragments (I14–WT–biotin, I14–M1 + M2 + M3–biotin, I14–M1–biotin) to confirm specific binding of hnRNP A1/A2B1 to the wild–type intron 14 sequence of Ezh2 (Fig. 5F, G). Western blot analysis showed that I14–WT efficiently pulled down hnRNP A1/A2B1, while the antisense (I14–Antisense–biotin) and mutant fragments (I14–M1 + M2 + M3–biotin and I14–M1–biotin) exhibited drastically reduced binding (Fig. 5H, I). Electrophoretic mobility shift assays (EMSA) further validated direct interactions. His-tagged hnRNP A1/A2B1 proteins caused a shift in the migration of I14–WT–biotin RNA, which was partially reversed by the addition of excess unlabeled WT RNA (I14–WT), but not by mutant RNA (I14–M1 + M2 + M3 or I14–M1) (Fig. 5J–M). Collectively, these results confirm that hnRNP A1 and A2B1 regulate Ezh2 AS by specifically binding to UAG motifs within its intron 14 to promote exon 14 inclusion. Disruption of these interactions—either by loss of hnRNP A1/A2B1 or mutation of their binding sites—shifts the splicing toward the Ezh2Δ14 isoform, compromising EZH2 methyltransferase activity. Impaired epigenetic repression of Cdkn1a in the absence of hnRNP A1 and A2B1 We next investigated the key downstream targets of the impaired hnRNP A1/A2B1–EZH2 pathway that contribute to the severe GC B cell defects observed in DKO mice. To this end, we first stimulated splenic B cells with anti-IgM and anti-CD40 in vitro to mimic in vivo B cell activation in GCs, followed by RNA-seq analysis (Fig. 6A, Supplementary Data 4). We also analyzed a previously published H3K27me3 ChIP-seq dataset of B cells treated with EZH2 inhibitor GSK343 (GSE45982). By overlapping the genes upregulated in stimulated DKO B cells with those exhibiting reduced H3K27me3 levels, we identified Cdkn1a as one of the most promising direct targets of EZH2 affected in hnRNP A1 and A2B1 (Supplementary Fig. 6A). qRT-PCR analysis further corroborated elevated Cdkn1a RNA levels in activated B or early GC B cells lacking hnRNP A1 and A2B1, while Cdkn1a expression in the naïve B cells remained unchanged (Fig. 6B). Consistently, CDKN1A protein levels were significantly higher in activated DKO B cells than in WT controls (Fig. 6C, D). Knockout of hnRNP A1 and A2B1 abrogates EZH2-mediated epigenetic silencing of CDKN1A. A Volcano plot showing differentially expressed genes identified by RNA-seq in Hnrnpa1f/f Hnrnpa2b1f/f (WT) and Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO) B cells stimulated with anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL) for 48 h. Differentially expressed genes are defined as those with |log2FC| > 1 and p < 0.05. CDKN1A is significantly upregulated in DKO B cells. Differential gene expression analysis for RNA-seq data was performed using DESeq2, which models count data based on the negative binomial distribution. Statistical significance was assessed using two-sided tests. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg method to control the false discovery rate (FDR). Genes with an adjusted P value (FDR) < 0.05 were considered significantly differentially expressed. B qRT-PCR analysis of CDKN1A expression in naïve B cells, B cells stimulated with anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL) for 48 h, and early Germinal center (GC) B cells isolated 5 days after NP-KLH/Alum immunization (naïve B: n = 6; anti-CD40 and anti-IgM: n = 6; early-GC: n = 5). C, D Western blot analysis of CDKN1A protein expression in unstimulated and 48-h-stimulated B cells (n = 3). The data shown are representative images from three independent experiments. E, F ChIP-qPCR analysis of Cdkn1a locus enriched by anti-H3K27me3 antibody in B cells from WT and DKO mice after 48 h of stimulation with anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL). E A schematic representation of the Cdkn1a gene showing exons and the putative H3K27me3-enriched regions, P1 and P2. F ChIP-qPCR analysis validated the enrichment of H3K27me3 at the Cdkn1a genomic loci in WT and DKO B cells. Data are presented as fold enrichment relative to the WT IgG control. p < 0.0001(p = 2.06e-6). G, H CFSE dilution assay assessing proliferation of WT and DKO B cells following 72 h of anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL) stimulation (n = 4). G Representative flow cytometry histograms showing CFSE dilution. H Quantitative analysis of CFSE dilution reflecting proliferation capacity (n = 4; two independent experiments). Each symbol represents an individual mouse (B, D, F, H). Data are shown as mean ± SEM. Results shown are representative of at least three independent experiments (B, D, F, H). Statistical significance was determined using an unpaired two-tailed Student’s t-test (B, D, H) or one-way ANOVA followed by Tukey’s multiple comparisons test (F). We also assessed Cdkn1a regulation at the chromatin level using a chromatin immunoprecipitation (ChIP) assay targeting H3K27me3. Our results revealed a marked decrease in H3K27me3 enrichment at the Cdkn1a locus in DKO B cells compared to WT cells (Fig. 6E, F), indicating impaired epigenetic repression of Cdkn1a in the absence of hnRNP A1 and A2B1. CDKN1A is a cyclin-dependent kinase inhibitor that blocks cell cycle progression at the G1 phase28,33. Consistent with the impaired proliferative capacity observed in early GC B cells from the DKO mice, our CFSE labeling assay demonstrated compromised proliferation of hnRNP A1/A2B1 DKO B cells after 48 h of in vitro stimulation (Fig. 6G, H). This defect aligns with the drastically diminished GC B cell proliferation observed in the DKO mice after NP-KLH immunization (Fig. 3M, N). To further determine whether the upregulation of CDKN1A resulted from the loss of EZH2 catalytic activity caused by aberrant splicing, we reintroduced Ezh2-FL or Ezh2Δ14 into DKO B cells and sorted GFP⁺ B cells for analysis. Re-expression of Ezh2-FL partially reversed the transcriptional upregulation of CDKN1A, whereas Ezh2Δ14 failed to rescue this defect (Supplementary Fig. 6B). Notably, this pattern closely paralleled the changes in H3K27me3 levels observed following reconstitution with Ezh2-FL or Ezh2Δ14 (Fig. 4K). Collectively, these findings suggest that loss of hnRNP A1 and A2B1 significantly compromises EZH2-mediated epigenetic silencing of CDKN1A, resulting in its aberrant upregulation. In turn, elevated CDKN1A impairs the proliferation of early activated B cells and thereby influences downstream GC B cell responses. HnRNPA1/A2B1–EZH2-CDKN1A regulatory cascade controls GC B cell differentiation Our in vitro studies suggest a regulatory cascade involving hnRNPA1/A2B1, EZH2, and CDKN1A in controlling the formation and differentiation of GC B cells. To further validate this in vivo, we generated hnRNP A1/A2B1/CDKN1A triple-knockout (TKO) mice by crossing Cdkn1a−/− mice with Hnrnpa1fl/fl Hnrnpa2b1fl/fl Cd19Cre/+ mice (Supplementary Fig. 7A). Effective Cdkn1a deletion in TKO B cells was confirmed by qRT-PCR (Supplementary Fig. 7B). We then immunized the mice with NP-KLH/Alum and evaluated the GC B cell response by flow cytometry analysis at day 10 post-immunization. Both the frequency and absolute number of B220+ CD95+ CD38− GC B cells in TKO mice were significantly higher than in hnRNP A1/A2B1 DKO mice and largely restored to WT levels (Fig. 7A, B). hnRNP A1 and A2B1 rely on the EZH2-CDKN1A axis to regulate GC B cell development. A, B Representative flow cytometry plots (A) and quantification (B) showing the proportion of Germinal center (GC) B cells in the spleens of Hnrnpa1f/f Hnrnpa2b1f/f (WT), Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ (double knockout, DKO), CDKN1A-SKO, and triple knockout (TKO, Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre+/− Cdkn1a−/−) mice 10 days after NP-KLH/Alum immunization (WT: n = 10; DKO: n = 11; CDKN1A-SKO: n = 9; TKO: n = 12). C–E Bone marrow chimera experiments. C Schematic of the experimental design. Mixed bone marrow chimeras were generated by reconstituting recipients with approximately 50% CD45.2⁺ donor cells from WT, DKO, or TKO mice together with approximately 50% CD45.1⁺ WT cells. Following immunization with NP-KLH/Alum, the frequency of GC B cells was analyzed by flow cytometry. D Representative flow cytometry plots of CD45.2⁺ GC B cells from one of two independent experiments. E Quantification of the frequency of CD45.2⁺ GC B cells (WT: n = 7; DKO: n = 7; TKO: n = 5). Data are pooled from two independent experiments. In all bar graphs, bars represent the mean, and dots represent individual mice. p < 0.0001 (WT vs DKO, p = 7.34e-8; DKO vs TKO, p = 2.94e-6). F, G ELISPOT analysis of NP-specific IgG1⁺ antibody-secreting cells in the spleens of WT, DKO, CDKN1A-SKO, and TKO mice on day 10 post NP-KLH/Alum immunization. F Representative ELISPOT images. G Quantification of NP-specific IgG1⁺ antibody-secreting cells (WT: n = 10; DKO: n = 11; CDKN1A-SKO: n = 9; TKO: n = 12). p < 0.0001 (WT vs DKO = 4.73e-7, DKO vs TKO = 8.74e-5). H, I CFSE dilution assay assessing the proliferation of splenic B cells from WT, DKO, CDKN1A-SKO, and TKO mice after 4 days of anti-CD40 (1 μg/mL) and anti-IgM (2 μg/mL) stimulation. H Representative flow cytometry histograms. I Quantitative analysis of CFSE dilution (WT: n = 5; DKO: n = 5; CDKN1A-SKO: n = 4; TKO: n = 5). Each symbol represents an individual mouse (B, E, G, I). Data are shown as mean ± SEM and are representative of at least three independent experiments (B, E, G, I). Statistical significance was determined by one-way ANOVA followed by Tukey’s multiple comparisons test (B, E, G) or Mixed-effects analysis followed by Tukey’s multiple comparisons test (I). To determine whether the observed rescue associated with CDKN1A deficiency is B-cell–intrinsic, we generated 1:1 mixed bone marrow chimeras. Specifically, CD45.1⁺/CD45.1⁺ WT donor cells were mixed with CD45.2⁺/CD45.2⁺ WT (Hnrnpa1fl/fl Hnrnpa2b1fl/fl), DKO (Hnrnpa1fl/fl Hnrnpa2b1fl/fl Cd19Cre/+), or TKO (Hnrnpa1fl/fl Hnrnpa2b1fl/fl Cd19Cre/+ Cdkn1a−/−) donor cells to establish chimeras. The contribution of CD45.2⁺ cells to the GC B cell compartment was subsequently analyzed. Consistent with our previous observations, the competitive fitness of DKO cells was compromised, with a markedly reduced contribution to the GC compartment. This reduction was even more pronounced than that observed in non-chimeric mice, likely reflecting the competitive environment of mixed bone marrow chimeras. In contrast, TKO cells contributed to the GC B cell compartment at levels comparable to those of WT controls (Fig. 7C–E, Supplementary Fig. 7C, D). Collectively, these results suggest that CDKN1A deficiency alleviates the proliferative defects observed in hnRNP A1/A2B1-deficient B cells, which may contribute to the partial restoration of germinal center–associated B cell responses. Similarly, ELISPOT assays showed that the defect in NP-specific IgG1 antibody-secreting cells formation observed in DKO mice was fully rescued in TKO mice (Fig. 7F, G). Additionally, CFSE labeling assays on in vitro-stimulated B cells revealed that the severe proliferation defect seen in hnRNPA1/A2B1 DKO B cells was reversed by Cdkn1a deletion. TKO B cells exhibited significantly improved proliferation compared to DKO B cells and were largely comparable to WT controls (Fig. 7H, I). Together, these findings suggest the existence of an hnRNP A1/A2B1-EZH2-CDKN1A regulatory cascade that influences B cell proliferation and germinal center–associated processes. Discussion In this study, we conditionally deleted hnRNP A1 and hnRNP A2B1, the two highly expressed and homologous hnRNP family members, and found that the double-knockout mice showed a significant decrease in the secretion of class-switched, high-affinity antibodies. We attribute this defect to an impaired GC response resulting from early alterations in B cell activation and proliferation. Early GC B cell proliferation was profoundly compromised in DKO mice. However, by day 10 post-immunization, the proliferation defect was largely resolved (Supplementary Fig. 4J, K), and GC B cell numbers became largely comparable between the DKO and WT mice at later time points (Supplementary Fig. 8A, B). Interestingly, although Hnrnpa1 and Hnrnpa2b1 mRNA levels were markedly reduced in day 5 GC B cells from DKO mice, their expression became indistinguishable from WT levels by days 10 and 14 post-immunization (Supplementary Fig. 8C, D). We therefore hypothesized that a subset of activated B cells, particularly early GC B cells in DKO mice, may escape Cre-mediated deletion of Hnrnpa1, Hnrnpa2b1, or both. These single-sufficient or double-sufficient “escapee” cells would possess a proliferative advantage over their double-mutant counterparts, enabling them to undergo robust expansion and become the predominant GC B cell population in DKO mice at later time points, such as day 14 post-immunization. While the GC B-cell compartment in DKO mice recovers in cell number by day 14, likely due to the expansion of “escapee” clones that bypassed Cre-mediated deletion, this late-stage cellular normalization does not translate into a corresponding recovery of antibody affinity. One possible explanation is that an early proliferative bottleneck imposes a lasting constraint on the quality of the response. The marked defect observed at day 5 has limited the recruitment and expansion of an optimal pool of founder clones during the initial phase of the GC reaction. Consistent with previous models of B-cell competition, early affinity-dependent entry and competitive fitness within the nascent GC can strongly influence the composition and quality of the later antibody repertoire34,35. Thus, although escapees may later restore GC cellularity, they may arise from a restricted or skewed founder pool and therefore may not fully compensate for the early loss of high-quality clones36. In addition to this early bottleneck, the quality of the GC response may be further affected by altered RNA processing programs downstream of hnRNP A1/A2B1. Our RNA-seq analysis revealed that hnRNP A1/A2B1 regulates the alternative splicing of several genes associated with c-Myc-driven biosynthetic and proliferative programs, including Pkm, Tyms, and Eif4h. Given the central role of c-Myc in promoting re-entry of positively selected B cells into the dark zone, dysregulation of these downstream effectors may reduce the biosynthetic fitness required for iterative GC cycling and selection. This may contribute to the persistently reduced serum NP2/NP20 ratios observed for IgG1 and IgG3. By contrast, the trend toward normalization in IgG2b by day 21 may reflect subclass-specific differences in the relative contribution of early versus later GC output, class-switch kinetics, plasma-cell generation, or selection pressures, although the present data do not allow these possibilities to be resolved definitively. Importantly, the predominance of escapees at later time points presents a major technical limitation for mechanistic interpretation. Because bona fide DKO GC B cells are rare by day 14, it is not feasible in this model to directly assess their intrinsic affinity maturation capacity, for example, through analysis of VH186.2 somatic hypermutation. Therefore, our data support a defect in high-affinity antibody production at the serum level, but do not allow a definitive conclusion regarding GC B-cell-intrinsic affinity maturation at later stages. Future studies using inducible deletion systems may help distinguish the direct effects of hnRNP A1/A2B1 loss on GC B cells from the confounding contribution of escapees and may also clarify the basis of the isotype-specific recovery patterns observed here. To identify the molecular drivers underlying the GC phenotype in hnRNP A1/A2B1–deficient B cells, we prioritized candidate targets based on the magnitude of splicing changes, junction support, and the predicted functional impact. Among these, Ezh2 exon 14 emerged as a leading candidate due to its robust skipping in DKO cells and the consequent disruption of the SET catalytic domain, which is consistent with the established requirement for EZH2 in GC B cell proliferation and maintenance27,28. Nevertheless, given the broad regulatory scope of hnRNP A1 and A2B1, additional splicing events likely contribute to the complex cellular defects observed. Although systematic interrogation of these targets was beyond the scope of the current study, future pathway-level analyses and perturbation screenings will be essential to define the full spectrum of hnRNP-dependent splicing programs that shape the GC response. The high sequence identity between hnRNP A1 and hnRNP A2B1 suggests potential functional redundancy, strongly supported by our genetic data showing that Ezh2 exon 14 skipping occurs exclusively in the DKO setting. To determine whether this redundancy reflects obligate physical cooperation or independent compensation, we examined RNA-binding properties using EMSA. Although both proteins bound the Ezh2 exon 14 RNA probe, co-incubation did not produce synergistic binding or formation of a stable higher-order heterocomplex (Supplementary Fig. 9). These findings argue against an obligate heterodimerization model. Instead, they support a dosage-dependent compensation model, in which hnRNP A1 and A2B1 bind overlapping cis-regulatory elements. In this framework, the presence of either paralog provides sufficient RNA-binding activity to maintain exon inclusion, and splicing regulation is compromised only when the combined pool of both factors falls below a required threshold. Previously, SF3B3 (splicing factor 3B subunit 3), a core component of the U2 snRNP and SF3B complex involved in branch point recognition, spliceosome assembly, and dynamic regulation37, was shown to be crucial for Ezh2 exon 14 inclusion and for suppressing tumorigenesis in renal cancer38. Similarly, SNRPB, a core member of the Sm protein family in U1, U2, U4, and U5 snRNPs that stabilizes interactions between snRNAs and pre-mRNA, has been reported to facilitate Ezh2 exon 14 inclusion and inhibit the progression of liver cancer39. It is well established that hnRNPs and the spliceosome collaboratively regulate pre-mRNA splicing through competitive binding and dynamic interactions to balance splicing precision and flexibility. HnRNPs act as “regulatory switches”, promoting or repressing spliceosome binding at specific splice sites, while the spliceosome functions as the “execution machinery” that catalyzes the splicing reaction40–42. Based on these insights and our findings, we initially speculated that hnRNP A1/A2B1-mediated regulation of Ezh2 exon 14 inclusion in B cells may involve interaction with these spliceosome components. However, RNA pull-down combined with mass spectrometry using Ezh2 pre-mRNA in B cells did not detect significant binding of SF3B3 or SNRPB (Supplementary Data 5). These findings suggest that, in B cells, SF3B3 and SNRPB may not directly bind and may play a less prominent role in regulating exon 14 inclusion. The differential contributions of various factors to AS likely reflect their cell- and tissue-specific expression. For example, the cell-type-specific regulation of CD44 is known to involve spliceosome component ESRP1 and hnRNP family member PTBP143. One of the key findings of this study is that hnRNP A1 and A2B1 bind to the same UAG-rich region in Ezh2 pre-mRNA to promote Ezh2 exon 14 inclusion, which is essential for maintaining methyltransferase activity of PRC2. In tumor cells, the exon 14-skipped isoform Ezh2∆14 lacks H3K27me3 catalytic activity and may function as a dominant-negative inhibitor of residual full-length EZH2. In renal cell carcinoma, EZH2∆14 has been shown to interfere with PRC2 components EED and SUZ12, thereby suppressing the catalytic activity of EZH238. We speculate that a similar antagonism may occur in B cells, where EZH2∆14 could counteract the function of full-length EZH2, an intriguing possibility that warrants further investigation. Beyond the central regulatory node of Ezh2 exon 14 inclusion, loss of hnRNP A1/A2B1 reprograms a broader splicing landscape in B cells. This includes diverse AS events—such as intron retention (Skiv2l), alternative 5’‑splice‑site selection (Tyms), mutually exclusive exon selection (Pkm, Gas5), and differential exon inclusion (Ncf4, Traf7)—affecting genes that govern metabolism, nucleotide synthesis, RNA surveillance, and immune signaling14,44–47. Collectively, these findings support a hierarchical regulatory framework: hnRNP A1/A2B1‑dependent control of Ezh2 splicing constitutes the core regulatory hub that sustains GC proliferative capacity, while a spectrum of auxiliary splicing alterations forms a secondary layer that fine‑tunes cellular fitness, metabolic adaptation, and differentiation trajectories. To further delineate the direct regulatory scope of these hnRNP factors, unbiased approaches such as enhanced CLIP sequencing (eCLIP-seq) are warranted. Although B–cell–specific binding landscapes remain to be fully defined, analysis of publicly available hnRNP A1 eCLIP datasets from human K562 cells48 indicates that binding sites are predominantly enriched within intronic and splice-site proximal regions, with comparatively lower occupancy in UTRs or intergenic regions. This distribution is consistent with a primary role in splicing regulation. Future studies employing CLIP-based approaches in primary GC B cells will be essential to comprehensively map the RNA-binding landscape and to uncover additional regulatory layers governed by these hnRNPs. Regulation of AS by hnRNP A1 and A2B1 is highly context-dependent and often conforms to the established ‘RNA maps’ principle, whereby functional outcomes are determined by the binding position relative to the regulated exon18,49. In our dataset, loss of these factors frequently results in increased exon inclusion (e.g., in Ggnbp2 and Dennd4a), consistent with their canonical roles as splicing repressors that bind exonic or proximal intronic silencers. In contrast, Ezh2 exon 14 displays the opposite pattern, with increased skipping in DKO cells. We propose a position-dependent model to explain this directionality: in WT cells, hnRNP A1/A2B1 likely binds to UAG-rich motifs within flanking intronic regions to stabilize exon definition or suppress competing splice sites, thereby promoting exon 14 inclusion. Loss of these factors removes this regulatory support, shifting the splicing equilibrium toward the exon-skipped isoform. High-resolution CLIP-seq in primary B cells will be necessary to directly map these binding interactions and confirm how positional occupancy dictates splicing outcomes. In summary, our findings establish hnRNP A1 and hnRNP A2B1 as critical regulators of EZH2 pre-mRNA alternative splicing. By ensuring proper exon 14 inclusion, these factors preserve EZH2 catalytic activity and sustain H3K27me3-mediated repression of the cell-cycle checkpoint gene Cdkn1a, thereby promoting B cell proliferation and robust GC responses. Collectively, our study identifies an hnRNP A1/A2B1–EZH2–CDKN1A splicing–epigenetic axis that mechanistically links RNA processing to chromatin remodeling and provides insight into the post-transcriptional control of adaptive immunity. Methods Mice and immunization Mouse sources are listed in Supplementary Table 1. All mice were on a C57BL/6 J background and maintained under specific pathogen-free conditions at the Laboratory Animal Research Center of Southern University of Science and Technology (12 h light/dark cycle; 20–26 °C; ~70% humidity). Both male and female mice aged 6–12 weeks were euthanized. Experimental and control animals were co-housed, and experiments were not specifically designed to distinguish between sexes. All procedures were approved by the Institutional Animal Care and Use Committee of Southern University of Science and Technology (IACUC approval no. SUSTech-JY202504007). Mice were euthanized by CO₂ inhalation at experimental endpoints. Cd19Cre/+ mice (B6.129P2(C)-Cd19tm1(cre)Cgn/J, RRID:IMSR_JAX:018958) and Mb1Cre/+ mice (B6.C(Cg)-Cd79atm1(cre)Reth/EhobJ, strain #: 020505) were obtained from The Jackson Laboratory. Hnrnpa1 floxed mice (Cat. No. T013507) were purchased from GemPharmatech and crossed with Cd19Cre/+ mice to generate conditional knockout mice with deletion of exons 2–11 (Supplementary Fig. 1F). Hnrnpa2b1 floxed mice (Cat. No. S-CKO-11602) were obtained from Cyagen and crossed with Cd19Cre/+ mice to generate conditional knockout mice with deletion of exons 2–6 (Supplementary Fig. 1F). Cdkn1a knockout (KO) mice (Cat. No. NM-KO-190774) were purchased from Shanghai Model Organisms Center, Inc. For bone marrow chimera experiments, CD45.1 mice (Cat. No. NM-KI-210226) were obtained from Shanghai Model Organisms Center, Inc. For T cell–dependent responses, 100 μg NP₃₀-KLH (Biosearch Technologies, N-5060-25) was mixed 1:1 (v/v) with alum (Thermo Scientific 77161 or InvivoGen vac-alu-50) and injected intraperitoneally. For bone marrow chimera generation, recipient mice were irradiated (9 Gy) and immediately provided with 0.5 g/L neomycin (Aladdin, N109017)-supplemented drinking water. Within 4 h post-irradiation, bone marrow cells from CD45.1⁺/CD45.1⁺ donors were mixed at a 1:1 ratio with cells from CD45.2⁺/CD45.2⁺ WT (Hnrnpa1f/f Hnrnpa2b1f/f), DKO (Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+), or TKO (Hnrnpa1f/f Hnrnpa2b1f/f Cd19Cre/+ Cdkn1a−/−) mice after red blood cell lysis (BioLegend, 420301). A total of 4 × 106 cells were injected intravenously into the recipients (CD45.1⁺/CD45.1⁺ or CD45.1⁺/CD45.2⁺). Six weeks later, recipient mice were immunized with NP-KLH/alum and analyzed by flow cytometry 10 days post-immunization. RNA isolation, RT-PCR and qRT-PCR RNA was extracted using the TRIzol reagent kit (Takara, 9108), followed by reverse transcription with the Hifair AdvanceFast 1st Strand cDNA Synthesis Kit (Yeasen, 11139ES60). qPCR was performed using the Hieff UNICON Universal Blue qPCR SYBR Green Master Mix (Yeasen, 11184ES60). RT-PCR was performed using the 2 × Rapid Taq Master Mix (Vazyme, P222-01). PCR products were analyzed by agarose gel electrophoresis, and band intensity was quantified using ImageJ (V1.53e). The primers used are listed in Supplementary Table 2. Flow cytometry Bone marrow cells were harvested from the femora of mice, while splenocytes were obtained by mechanically dissociating the spleens. The resulting cell suspensions were filtered through a 70 μm cell strainer (BIOFIL JET, CSS013070) and resuspended in PBS containing 5% fetal bovine serum (FACS buffer). Single-cell suspensions were incubated with 1 mL of red blood cell lysis buffer (Biolegend, 420301) on ice for 4 min, followed by the addition of 4 mL FACS buffer and centrifugation. The supernatant was discarded. For surface marker staining, antibodies were diluted 1:100 and incubated with cells on ice in the dark for 15 min. Cells were then washed with 500 μL FACS buffer and finally resuspended in FACS buffer containing 0.5 μg/mL DAPI for flow cytometric analysis. For intracellular BCL6 staining, cells were fixed and permeabilized using Foxp3/Transcription Factor Staining Buffer Set (eBioscience, 00-5523-00) according to the manufacturer’s instructions, followed by incubation with BCL6-PE antibody (eBioscience, 12-5453-82). For in vivo BrdU analysis, mice were intraperitoneally injected with 1 mg BrdU (Sigma, B5002-1G) and euthanized 2 h later. Splenocytes were collected and stained with surface marker antibodies and DAPI. Cells were then fixed and permeabilized according to the BrdU Staining Kit for Flow Cytometry protocol (eBioscience, 8812-6600-42), followed by incubation with DNase I (300 μg/mL, Sigma, D4513) at 37 °C for 1 h. FITC-conjugated anti-BrdU antibody (1:50 dilution, BD Pharmingen, 559619) was used to stain the cells at room temperature for 25 min. After washing with FACS buffer, cells were resuspended for flow cytometric analysis. For CaspGLOW apoptosis assays (BioVision, 180–100; Thermo, 88-7003-42), apoptosis-associated caspase activity was detected using the FITC-conjugated active caspase inhibitor VAD-FMK. Mouse bone marrow single-cell suspensions were incubated with 1 μL VAD-FMK per 106 cells at 37 °C for 1 h. Cells were then washed once with wash buffer (centrifuged at 425 × g for 5 min) and resuspended for subsequent surface marker staining prior to flow cytometric analysis, following the CaspGLOW Apoptosis Detection Kit instructions. For CFSE staining, magnetically sorted splenic B cells (2 × 107) were labeled with 1 μL of 5 mM CFSE (Biolegend, 423801) at room temperature in the dark for 20 min. Cells were washed with 5 mL RPMI 1640 medium containing 10% fetal calf serum, centrifuged to remove the supernatant, and resuspended at a density of 106 cells/mL for plating. Cells were then stimulated with 1 μg/mL anti-CD40 (BD, 553721) and 2 μg/mL anti-IgM (JacksonImmunoResearch Laboratories, 115-006-075). Splenic B cells were isolated using the EasySep Mouse CD19 Positive Selection Kit II (STEMCELL Technologies, 18000) or CD43 (Ly-48) MicroBeads (Miltenyi Biotec, 130-049-801) according to the manufacturer’s protocols. For flow cytometric sorting of GC B cells (B220+ CD38− GL-7+), splenic single-cell suspensions were first stained with biotin-conjugated GL-7 antibody, followed by incubation with Streptavidin MicroBeads (Miltenyi Biotec, 130-048-101) at 4 °C for 20 min. Unbound beads were washed away, and GL-7+ cells were purified using LS separation columns (Miltenyi Biotec, 130-042-401). After two washes, the LS columns were removed from the magnet, and 5 mL MACS buffer was used to elute GL-7+ cells with the plunger. Purified GL-7+ cells were stained with appropriate flow cytometry antibodies and sorted using a BD FACSAria III system (BD Biosciences). Flow cytometric analysis was performed on a CytoFLEX system (Beckman Colter), and cell sorting was conducted using the FACSAria III system (Becton Dickinson). Data were acquired using CytExpert software (Beckman Colter, v2.3) and analyzed with FlowJo v10.8.1. Antibodies and reagents used in this study are listed in Supplementary Table 1. Cell culture and B-cell stimulation in vitro HEK 293 T cells were a kind gift from Professor Yan Li at the School of Life Sciences, Southern University of Science and Technology. The cells were cultured in high-glucose DMEM (Hyclone, C11995500BT) supplemented with 10% fetal bovine serum (FBS, ExCell Bio, FSP500), 100 U/mL penicillin, and 100 μg/mL streptomycin (Hyclone, SV30010) in 6-well plates or 100 mm culture dishes (NEST Biotechnology, 704002) under standard conditions (37 °C, 5% CO₂). For in vitro B cell stimulation, splenic B cells were purified using anti-CD19 magnetic beads (STEMCELL TECHNOLOGIES, 18000). A total of 2 × 106 purified splenic B cells were cultured in RPMI-1640 medium (Gibco, C11875500BT) supplemented with 10% FBS (ExCell Bio, FSP500), 100 U/mL penicillin, and 100 μg/mL streptomycin (Hyclone, SV30010). B cells were stimulated with LPS (20 µg/mL, Sigma, L2630) for retroviral infection. For transcriptomic analysis, splenic B cells were further stimulated in vitro with anti-CD40 antibody (1 µg/mL, BD Biosciences, 553721) and anti-IgM (2 µg/mL, Jackson ImmunoResearch, 115-006-075). Plasmid construction The overexpression plasmids used in this study were constructed based on the MIGR1 vector50, while plasmids for protein purification were constructed in the pET28a vector (a gift from Dr. Zhongmin Liu at the School of Life Sciences, Southern University of Science and Technology). Briefly, target DNAs were amplified from the cDNA library of mouse splenic B cells using KOD-Plus-neo (TOYOBO, KOD-401) and inserted into the MIGR1 vector using the ClonExpress MultiS One Step Cloning Kit (Vazyme, C113-02). All inserted sequences were verified by Sanger sequencing. The primers used are listed in Supplementary Table 2. Transfection and retrovirus transduction For retroviral transduction, HEK 293 T cells were co-transfected with equal amounts of MIGR1 or MIGR1-based plasmids (Ezh2-FL, Ezh2Δ14, minigene) and the pCL-ECO retroviral packaging plasmid using Lipofectamine 3000 transfection reagent (Thermo Scientific, L3000015) or jetPRIME transfection reagent (Polyplus, 101000046). Viral supernatants were collected and filtered through a 0.22 μm filter (JET, FPE204030). The viral supernatants were then added to culture medium containing 20 μg/mL LPS (Sigma, L2630-10MG) and 6 μg/mL polybrene (Sigma, TR-1003-G), followed by mixing with CD19⁺ splenic B cells. For overexpression experiments involving Ezh2-FL or Ezh2Δ14, viral supernatants were supplemented with anti-CD40 (1 µg/ml, BD Biosciences, 553721) and anti-IgM (2 µg/ml, Jackson ImmunoResearch, 115-006-075). The target cells were then centrifuged at 425 × g for 1.5 h at 37 °C. After a single round of transduction, the cells were cultured for an additional 48 h, followed by flow cytometric sorting to isolate GFP⁺ B cells for subsequent experiments. ELISA NP-specific antibodies were captured using NP2-BSA or NP20-BSA coated onto 384-well flat-bottom plates (Thermo Scientific, 464718). Endogenous serum antibodies were captured using goat anti-mouse immunoglobulins. After three thorough washes with PBS, the plates were blocked with 2% BSA for 2 h at room temperature. Diluted serum samples collected at different time points before and after NP-KLH immunization were then added and incubated for 1 h. The bound antibodies were then incubated with horseradish peroxidase (HRP)-conjugated anti-IgG1 (SouthernBiotech, 1071-05), anti-IgG2b (SouthernBiotech, 1091-05), anti-IgG3 (SouthernBiotech, 1101-05), anti-IgA (SouthernBiotech, 1040-05), or anti-IgM (SouthernBiotech, 1021-05) antibodies diluted 1:1000, and incubated for 1 h at room temperature, followed by three washes with PBST. The signal was developed using a TMB substrate (BioLegend, 421101), and the optical density at 450 nm (OD450) was measured using a microplate reader. Data analysis was performed using Gene5 Software (BioTek, V2.08). Antibodies used in the ELISA assay are listed in Supplementary Table 1. ELISPOT assay NP20-BSA was diluted in PBS to a final concentration of 10 μg/mL, and 1 μg of NP20-BSA was added per well in a MultiScreen HTS IP Filter 96-well plate (Merck Millipore, MSIPN4W50). The plate was incubated overnight at 4 °C. After washing with PBS, each well was blocked with 100 μL of PBS containing 2% BSA at room temperature for 2 h. Subsequently, cell suspensions were added, followed by incubation at 37 °C with 5% CO₂ for 2 h. After several washes with PBST, biotin-conjugated anti-IgG1 (SouthernBiotech, 1144-08) and anti-IgM (SouthernBiotech, 1140-08) antibodies were diluted 1:1000 in PBS containing 1% BSA, and 100 μL was added to each well. The plate was incubated at room temperature for 1 h, followed by washing with PBST. Streptavidin-alkaline phosphatase (streptavidin-AP, SouthernBiotech, 7100-04) was diluted 1:1000 in PBS containing 1% BSA, and 100 μL was added per well, with incubation at room temperature for 30 min. After washing with PBST, 100 μL of ELISPOT substrate solution (mabtech, 3650-10) was added per well, and the plate was developed at room temperature for 8 min. The reaction was stopped by discarding the substrate solution, followed by five washes with distilled water. The plate was air-dried at room temperature before imaging and quantification. Immunoblotting B cells were purified by magnetic bead isolation or fluorescence-activated cell sorting (FACS). Cells were lysed in RIPA buffer (Sigma, R0278) supplemented with protease inhibitors (Roche, 589279100), and protein supernatant was collected after centrifugation. Samples were mixed with SDS loading buffer (Yeasen, 20315ES20), boiled at 100 °C for 10 min, separated by 10% SDS-PAGE, and transferred onto PVDF membranes (Millipore, IPVH00010). Membranes were blocked in 5% skim milk in TBST for 1 h at room temperature and then incubated overnight at 4 °C with primary antibodies diluted in blocking buffer. After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies (TransGen, HS101-01 or HS201-01) for 1 h at room temperature. Following another wash with TBST, signals were detected using ECL substrate (Millipore, WBKLS0100) and visualized on a Tanon 5200 imaging system. Primary antibodies used were: β-actin (1:5000, Proteintech, 20536-1-AP), hnRNP A1 (1:1000, Santa Cruz Biotechnology, sc-32301), hnRNP A2B1 (1:1000, Santa Cruz Biotechnology, sc-53531), EZH2 (1:1000, ABclonal, A5743; A19577), H3K27me3 (1:1000, ABclonal, A22396), CDKN1A (1:1000, Cell Signaling Technology, 2947). RIP-qPCR 4 μg of hnRNP A1 or hnRNP A2B1 antibodies were first immobilized on 75 μL of Protein G magnetic beads (Cell Signaling Technology, 9006S) at room temperature for 1 h, and then washed with NT-2 buffer (250 mM Tris-HCl pH 7.4, 750 mM NaCl, 5 mM MgCl2, and EDTA-free Protease Inhibitor Cocktail) and resuspended in NET-2 buffer (20 mM EDTA pH 8.0, 1 mM DTT, 200 units/mL RNase OUT -supplemented NT-2 buffer). Mouse splenic B cells were lysed, and the protein supernatant was collected. The lysate was incubated overnight at 4 °C with the hnRNP A1- or hnRNP A2B1-Protein G bead complex, while 10% of the lysate supernatant was retained as the “Input” control. After washing with the NT-2 buffer, the immunoprecipitated complex was digested with proteinase K (Yeasen, 10409ES03). RNA was purified using RNAiso Plus (Takara, 9108) following the manufacturer’s instructions, which included chloroform addition, phase separation, and isopropanol precipitation. Both IP and Input RNA samples were reverse-transcribed into cDNA. Quantitative PCR was performed using Hieff UNICON Universal Blue qPCR SYBR Green Master Mix (Yeasen, 11184ES60). The percentage of Input was calculated as: 2−(Average Ct[RIP] − (Average Ct[Input] − log 2 (Input Dilution Factor))). The primers used in RIP-qPCR assay are listed in Supplementary Table 2. The specific primer sequences were as follows: Ezh2-RIP-qF, CTAAGGAGTTTGCTGCTGCTCTTACTG; Ezh2-RIP-qR, GGTCTGCTACTGTTATTCGGAAGTCTT. RNA pull-down and mass spectrometry Biotin-labeled RNA oligonucleotides were synthesized by GENEWIZ. A total of 250 pmol of biotin-labeled RNA was immobilized on 25 μL of streptavidin magnetic beads (MCE, HY-K0208) after extensive washing. The RNA-bead complex was then incubated with nuclear extracts from mouse splenic B cells in a low-salt buffer at 4 °C for 4 h. After five washes with low-salt buffer (20 mM Tris-HCl pH 7.5, 200 mM NaCl, 6 mM EDTA, 5 mM potassium fluoride, 5 mM β-glycerophosphate, 2 mg/mL protease inhibitor cocktail), the immunoprecipitated complex was eluted using SDS-PAGE loading buffer and heated at 98 °C for 10 min. Western blot analysis was performed to detect hnRNP A1 and hnRNP A2B1. Protein mass spectrometry was outsourced to Bioprofile Technology (Shanghai), and target proteins were identified based on intensity values from EZH2-sense and EZH2-antisense datasets. EMSA The pET28a vectors encoding mouse hnRNP A1 and hnRNP A2B1 were transformed into E. coli Rosetta (DE3) cells for protein expression, which was induced overnight with 0.5 mM IPTG (coolaber, CI6621). Bacterial cells were harvested and lysed by high-pressure homogenization, and the target proteins were purified from the supernatant using BeaverBeads His-tag protein purification beads (Beaver, 70501-5). The interaction between EZH2 RNA and hnRNP A1/A2B1 was assessed using the BersinBio RNA-EMSA Kit (BersinBio, Bes5107) following the manufacturer’s instructions. Each reaction contained 12.5 nM biotin-labeled RNA and 200 ng of purified protein. After separation by PAGE, the samples were transferred onto a nylon membrane (Beyotime, FFN10; Millipore INYC00010), followed by UV crosslinking for 10 min. The membrane was blocked with blocking solution (Beyotime, GS009B) for 15 min, incubated with streptavidin-HRP (SouthernBiotech, 7105-05) for 30 min, and washed several times with washing buffer (Beyotime, GS009W). Signal detection was performed using Immobilon Western HRP substrate (Millipore, WBKLS0100). Chromatin Immunoprecipitation (ChIP) assay The assay was performed following a standard protocol51. Briefly, CD19⁺ B cells were stimulated with anti-CD40 (1 μg/mL, BD Biosciences, 553721) and anti-IgM (2 μg/mL, Jackson ImmunoResearch, 115-006-075) for 48 h, followed by crosslinking with 1% formaldehyde at 25 °C for 15 min. Crosslinking was quenched by the addition of 0.125 M glycine at 25 °C for 5 min, after which the cells were washed with PBS and lysed on ice for 10 min in IP buffer (150 mM NaCl, 50 mM Tris-HCl [pH 7.5], 5 mM EDTA, 0.5% NP-40, 1.0% Triton X-100, 0.2% SDS, and 1 × protease inhibitor). Genomic DNA was then fragmented by sonication to shear chromatin into 200–1000 bp fragments. The sheared chromatin was incubated overnight at 4 °C with either anti-H3K27me3 antibody or IgG control antibody, followed by capture with ChIP-grade Protein G magnetic beads (CST, 9006S) and extensive washing. DNA was purified using the MiniBEST Universal Genomic DNA Extraction Kit (TaKaRa, 9765) according to the manufacturer’s instructions. Quantitative PCR was performed using Hieff qPCR SYBR Green Master Mix (Yeasen, 11184ES60) on a qTOWER3 real-time PCR system. Primer sequences are listed in Supplementary Table 2, and the specific primers used were as follows: ChIP-P21-F1, ACATCCATTCCACTCGGAACCTC; ChIP-P21-R1, GCGACGTAGGGGACAGAGG; ChIP-P21-F2, GCCTATTTCCCCCTCCTCAC; ChIP-P21-R2, GTTTCGGGAGGTGGGTCTGTTG. RNA-seq and data processing Splenic CD19+ B cells from wild-type (WT) and hnRNP A1/A2B1 double-knockout (DKO) mice were isolated using magnetic bead sorting. Cells were either stimulated with anti-CD40 (1 μg/mL, BD Biosciences, 553721) and anti-IgM (2 μg/mL, Jackson ImmunoResearch, 115-006-075) for 48 h or left unstimulated. Total RNA was extracted using RNAiso Plus (Takara) and submitted to GENEWIZ (Suzhou, China) for library preparation and sequencing. For library preparation, total RNA was quantified, and poly(A)-tailed mRNA was isolated and fragmented. Libraries were constructed using the NEBNext Ultra RNA Library Prep Kit for Illumina. Sequencing was performed on the Illumina NovaSeq 6000 platform using 150 bp paired-end (PE150) reads. For data analysis, raw sequencing quality was assessed using FastQC (v0.10.1), and reads were filtered with Cutadapt (v1.9.1) to obtain clean reads. Processed reads were aligned to the GRCm39.107 (Ensembl) mouse genome using HISAT2 (v2.0.1). Gene expression levels were quantified using HTSeq (v0.6.1), and differentially expressed genes (DEGs) were identified using DESeq2 (v1.6.3) with the criteria of fold change ≥ 1 and adjusted p ≤ 0.05. Gene Ontology (GO) enrichment analysis was performed using the ClusterProfiler R package (v3.14.3), and gene set enrichment analysis (GSEA) was conducted for all differentially expressed genes. Alternative splicing events were analyzed using rMATS turbo (v4.1.0), and alternative splicing visualization was performed using Integrative Genomics Viewer (IGV) based on the aligned BAM files. Quantification and statistical analysis All results are presented as the mean ± standard error of the mean (SEM), unless otherwise specified. Statistical significance between groups was determined using the Student’s t-test when comparing two groups or one-way ANOVA when analyzing more than two groups. The specific statistical test used is indicated in each figure legend. Statistical analyses were performed using GraphPad Prism software version 8 (GraphPad Software). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Dataset 1 Supplementary Dataset 2 Supplementary Dataset 3 Supplementary Dataset 4 Supplementary Dataset 5 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgments The authors acknowledge the assistance of the Southern University of Science and Technology Laboratory Animal Center for the care of mice and the Core Research Facilities for training and access to the flow cytometer. Author contributions X.O. and S.X. conceived the study; Z.Z., S.X., and X.O. designed the research and wrote the paper; Z.Z. and R.Z. performed most of the experiments and data analysis; H.K. and M.C completed part of the bone marrow chimera mouse experiments as well as the in vitro B cell reconstitution experiments. H.K. and L.G. carried out the Protein purification and EMSA assay; H.H. performed RNA pull-down; J.W. performed ELISA; Y.L. and W.C. performed ELISPOT assay; M.C., L.M., R.W., and L.L. performed animal work. K.P.L. provided insightful instructions. Peer review Peer review information Nature Communications thanks Wendy Béguelin, who co-reviewed this paper with Pengda Chen, 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 the National Key R&D Program of China (2023YFA1800100, to X.O.), Shenzhen Fundamental Research Program (Grant No. JCYJ20240813094400001, to X.O.), National Natural Science Foundation of China (Grant No. 32470938, to X.O.), and the core fund of Singapore Immunology Network (SIgN), Agency for Science Technology and Research (A*STAR), Singapore (to S.X. and L.K.P.). Data availability RNA-seq data for Hnrnpa1 and Hnrnpa2b1 expression across B cell subsets were obtained from the Immunological Genome Project (ImmGen; https://www.immgen.org/). Comparative RNA expression analyses between naïve and GC B cells were derived from GSE81999. Expression data of Ezh2 in GC LZ and DZ MYC⁺ AP4⁺ versus MYC⁻ AP4⁻ subsets were retrieved from GSE80669. Additional expression profiles for Hnrnpa1, Hnrnpa2b1, and Ezh2 in DZ B cells, plasmablasts, and three GC B cell fractions—Fraction1 (BCL6low, CD69high), Fraction2 (BCL6high, CD69high), and Fraction3 (BCL6high, CD69low)—were obtained from GSE109732. RNA-seq data following EZH2 inhibitor (GSK343) treatment and integrated H3K27me3 target analyses in GC B cells were retrieved from GSE45982. The RNA-seq datasets (raw and processed) of WT and hnRNP A1/A2B1 double knockout splenic B cells, including those stimulated with anti-CD40 and anti-IgM for 48 h, have been deposited in the NCBI Gene Expression Omnibus under accession number GSE298944. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE52 partner repository with the dataset identifier PXD077818. 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. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Zhijian Zhu, Rui Zhang. These authors jointly supervised this work: Shengli Xu, Xijun Ou. Contributor Information Shengli Xu, Email: xu_shengli@a-star.edu.sg. Xijun Ou, Email: ouxj@sustech.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74159-w. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Dataset 1 Supplementary Dataset 2 Supplementary Dataset 3 Supplementary Dataset 4 Supplementary Dataset 5 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement RNA-seq data for Hnrnpa1 and Hnrnpa2b1 expression across B cell subsets were obtained from the Immunological Genome Project (ImmGen; https://www.immgen.org/). Comparative RNA expression analyses between naïve and GC B cells were derived from GSE81999. Expression data of Ezh2 in GC LZ and DZ MYC⁺ AP4⁺ versus MYC⁻ AP4⁻ subsets were retrieved from GSE80669. Additional expression profiles for Hnrnpa1, Hnrnpa2b1, and Ezh2 in DZ B cells, plasmablasts, and three GC B cell fractions—Fraction1 (BCL6low, CD69high), Fraction2 (BCL6high, CD69high), and Fraction3 (BCL6high, CD69low)—were obtained from GSE109732. RNA-seq data following EZH2 inhibitor (GSK343) treatment and integrated H3K27me3 target analyses in GC B cells were retrieved from GSE45982. The RNA-seq datasets (raw and processed) of WT and hnRNP A1/A2B1 double knockout splenic B cells, including those stimulated with anti-CD40 and anti-IgM for 48 h, have been deposited in the NCBI Gene Expression Omnibus under accession number GSE298944. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE52 partner repository with the dataset identifier PXD077818. 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.