The germinal center reaction depends on RNA methylation and divergent functions of specific methyl readers

The adaptive immune response has the capacity to generate long-lasting immunological memory that provides a rapid response to recurrent pathogen exposures (Ahmed and Gray, 1996; Tarlinton and Good-Jacobson, 2013; Cyster and Allen, 2019). The primary goal of vaccination is to induce the generation of long-lived antibody-forming cells through the activation of the humoral arm of the adaptive immune response (Corcoran and Tarlinton, 2016; Plotkin, 2008; Ise and Kurosaki, 2019). High-affinity antibodies are secreted by plasma cells (PCs), which typically originate from germinal centers (GCs), microanatomic sites in lymphoid organs where B cells proliferate extensively and accumulate somatic hypermutations (SHMs) in their Ig genes (Victora and Nussenzweig, 2012; Shlomchik and Weisel, 2012; Cyster 2010; De Silva and Klein, 2015; Berek, 1992; MacLennan, 1994; Jacob et al., 1991). The selection of high-affinity clones for preferential expansion in the GC is mediated primarily by helper T cells (Vinuesa and Cyster, 2011; Victora et al., 2010; Crotty, 2011). In this process, the degree of T cell–derived signals induces corresponding expression levels of MYC in GC B cells, which in turn activates downstream cell cycle and metabolic gene programs (Gitlin et al., 2014; Gitlin et al., 2015; Finkin et al., 2019; Calado et al., 2012; Dominguez-Sola et al., 2012; Monzón-Casanova et al., 2018; Ersching et al., 2017; Chou et al., 2016). Myc is an immediate-early gene that rapidly produces mRNA transcripts with an extremely short half-life in response to mitogenic signals (Jones and Cole, 1987; Dani et al., 1984). Although GC B cells rapidly divide, their metabolic program is very different from that of other proliferating lymphocytes (Weisel et al., 2020). Whereas activated T and B cells can consume glucose and generate acetyl-CoA through glycolysis, GC B cells primarily oxidize fatty acids in the mitochondria and peroxisomes for respiration (Weisel et al., 2020; Vander Heiden et al., 2009).

Immune cell differentiation and acquisition of effector functions depend on extensive changes in gene expression that are regulated by transcriptional and post-transcriptional machineries (Turner and Díaz-Muñoz, 2018). The addition of a methyl group at position N6 on adenosines (N⁶-methyladenosine [m⁶A]) is the most abundant post-transcriptional modification in mammalian mRNA (Shulman and Stern-Ginossar, 2020; Zaccara et al., 2019). This process is mediated by an mRNA-modifying complex that includes METTL3, a methyltransferase that catalyzes the addition of m⁶A at consensus sites (Zaccara et al., 2019; Shulman and Stern-Ginossar, 2020). The functional consequence of this complex activity is determined by m⁶A-binding proteins that interact with the modified mRNA and modulate its half-life, location, splicing, and translation rate (Shi et al., 2019; Shi et al., 2017). Typically, m⁶A modifications promote mRNA destabilization, primarily through the binding of the m⁶A-interacting proteins, YTH domain–containing family proteins 1, 2, and 3 (YTHDF1–3), in the cytosolic compartment (Ivanova et al., 2017; Du et al., 2016; Wang et al., 2014; Shi et al., 2017; Zaccara and Jaffrey, 2020). Conversely, the family of IGF2 mRNA-binding proteins (BPs) 1, 2, and 3 (aka IMP1–3) stabilize transcripts and prolong their lifetimes (Palanichamy et al., 2016; Ren et al., 2020; Dai, 2020). RNA modifications by METTL3 augment the binding capacity of IGF2BP paralogs to mRNAs, resulting in the stabilization of several methylated transcripts, including DNA replication– and cell cycle–related genes (Sun et al., 2019; Huang et al., 2018; Müller et al., 2020). IGF2BP3 is the major paralog expressed in immune cells, yet its physiological functions remain unknown.

The m⁶A molecular machinery was found to play a critical role in supporting the adaptive immune response, including in differentiation of T follicular helper cells that provide help to B cells in GCs (Li et al., 2017; Zheng et al., 2020; Bechara et al., 2021; Ma et al., 2021; Yao et al., 2021). Here, we examined how RNA methylation and m⁶A binders regulate the expression of gene programs that support the generation and persistence of GC B cells. We found that effective cell cycle progression, GC formation, and persistence depend on METTL3 functions. The m⁶A interactor, IGF2BP3, is required for GC persistence by supporting Myc transcript stabilization and downstream pathways, whereas the m⁶A reader, YTHDF2, governs the proper gene expression and function of mitochondrial respiration. Thus, integration of mRNA modulation events by distinct m⁶A readers supports genetic programs that are critical for effective GC reactions.

To examine whether METTL3 is required for generation and maintenance of the GC reaction, we crossed a Mettl3^fl/fl mouse strain to a mouse that expresses Cre recombinase under the B cell–specific CD19 promoter (abbreviated as CD19-Mettl3^fl/fl); Mettl3^fl/+ mice that express the Cre recombinase were used as controls. Since CD19 is expressed by B cells during their generation in the bone marrow (BM), we first examined whether B cell development and establishment of the mature B cell compartment is defective in the CD19-Mettl3^fl/fl mouse model. Analysis of the different B cell populations in the BM and spleen revealed only minor and insignificant alterations in B cell development in this mouse model (Fig. S1, A and B). Whereas the frequency of mature B cells was normal in these mice, reduced levels of Igs were observed in their serum under homeostatic conditions, suggesting a defect in the ability of Mettl3-deficient B cells to mount an effective immune response (Fig. 1 A). To examine whether METTL3 functions are required for the generation of GC B cells, control and CD19-Mettl3^fl/fl mice were immunized subcutaneously in the hind footpads with a hapten (4-hydroxy-3-nitrophenyl [NP]) coupled to keyhole limpet hemocyanin (KLH) in alum, and the presence of GC B cells was examined in the popliteal LNs after 7 d. Flow cytometric analysis revealed that the frequency of GC B cells was fourfold reduced in CD19-Mettl3^fl/fl mice compared with controls on day 7 of the response (Fig. 1 B). Furthermore, the frequencies of PCs and memory B cells in the LNs of the CD19-Mettl3^fl/fl mice were significantly lower compared with littermate controls (Figs. 1 B and S1 C).

To completely exclude the possibility that defects in GC formation are a result of aberrant B cell development in the BM of CD19-Mettl3^fl/fl animals (Zheng et al., 2020), we took an additional approach. CD23 is expressed on B cells in the spleen after their maturation and departure from the BM. Analysis of the B cell immune response in Mettl3^fl/fl mice that were crossed to a B cell–specific CD23-Cre mouse strain (CD23-Mettl3^fl/fl; Kwon et al., 2008) showed complete elimination of GC and PC formation 7 d after immunization (Fig. 1 C). Moreover, immunohistochemistry analysis revealed that GC structures in CD23-Mettl3^fl/fl mice were absent (Fig. 1 D). Importantly, diminished Mettl3 mRNA and protein expression was detected in B cells derived from these mice (Fig. S1, D and E). Furthermore, the observed defects were specific for activated B cells, as the frequency of mature naive B cells in LNs was intact (Fig. S1 F). To determine whether early B cell activation or GC seeding was dependent on METTL3 functions, we immunized CD19-Mettl3^fl/fl mice with PE, a multi-epitope fluorescent antigen that allows detection of rare antigen-specific B cells in a polyclonal cell population at early time points (Pape et al., 2011). Flow cytometric analysis revealed a major defect in expansion of PE-specific B cells and in up-regulation of the early activation markers, CCR6 and FAS, in CD19-Mettl3^fl/fl mice 3 d after immunization (Fig. 1, E and F; Schwickert et al., 2011). These experiments demonstrate that METTL3 is essential for initial B cell functions before GC formation.

The findings obtained using CD23-Cre and CD19-Cre mouse models suggest that METTL3 is required for early B cell activation. Overnight stimulation with anti-IgM of B cells derived from UBC^CreERT2-Mettl3^fl/fl mice and treated with tamoxifen (4OHT) showed no defects in CD86 expression, and B cells from CD23-Mettl3^fl/fl mice showed no defect in ERK phosphorylation (pERK) upon activation (Fig. S1, G and H). Yet, these cells proliferated significantly less than control B cells in vitro (Fig. S1 I). Collectively, we conclude that METTL3 plays a critical role in B cell proliferation, but not in their initial activation.

Deletion of Mettl3 in naive B cells did not allow evaluation of m⁶A functions in the GC response. Activation-induced cytidine deaminase (AID, encoded by Aicda) is expressed in B cells after they encounter cognate antigen, but before GC seeding (Roco et al., 2019; Rommel et al., 2013; Crouch et al., 2007). To examine the functional role of METTL3 after the initial activation events, we crossed Mettl3^fl/fl mice to an Aicda^Cre/+ mouse strain (AID-Mettl3^fl/fl). In these mice, in the absence of immunological challenge, serum IgG titers but not IgM titers were lower compared with controls, suggesting that effective mounting of antibody immune responses was defective (Fig. S2 A). As opposed to the observations in CD19-Mettl3^fl/fl mice, Mettl3 deletion at the time of AID expression did not affect GC seeding, and the frequencies of GC B cells, PCs, and IgG1⁺ memory cells 7 d after immunization were intact (Figs. 2 A and S2 B). However, after an additional week, the frequencies of these cell populations in AID-Mettl3^fl/fl mice were severely reduced compared with control mice (Figs. 2 A and S2 B). The proportion of class-switched IgG1⁺ cells among Mettl3-deficient and control GC B cells was similar (Fig. S2 C). Analysis of Mettl3 mRNA by quantitative RT-PCR (RT-qPCR) and TdTomato expression in GC B cells revealed effective Cre-mediated recombination in this mouse model, when analyzed 7 and 14 d after immunization (Fig. S2, D and E). Some deficiencies in molecular machineries in B cell immune responses can be detected only under competitive pressure (Zaretsky et al., 2017). To examine whether METTL3 provides a competitive advantage in the GC reaction, we generated mixed BM chimeric mice that host control- and AID-Mettl3^fl/fl–derived immune cells. In these mice, the frequency of Mettl3-deficient and -proficient B cells in the GC was similar to that of naive B cells on day 7 after immunization (Fig. 2 B), suggesting that even under selective pressure, METTL3 expression did not endow a competitive advantage at this time point. However, after an additional 7 and 14 d, the frequency of GC Mettl3-deficient B cells was strongly reduced, similarly to that observed under noncompetitive conditions (Fig. 2 B). Thus, since the control B cells did not outcompete Mettl3-deficient B cells and since the latter diminished over time regardless of competition, we conclude that the observed defects in AID-Mettl3^fl/fl mice are B cell intrinsic. Furthermore, very few Mettl3-deficient B cells were detected in chronic GCs within the Peyer’s patches (PPs) of the chimeric mice (Fig. S2 F). Since GCs in PPs are driven by multiple gut-derived antigens, we conclude that the observed defects were not specific for any single immunogen.

To determine whether the GCs in AID-Mettl3^fl/fl mice are functional and can support B cell SHM and selective clonal expansion, we first examined the presence of antigen-specific B cells in GCs by staining cells derived from immunized mice with PE-NP₂₈ (which identifies total NP-specific B cells) or PE-NP₁₂ (which identifies NP-specific B cells that express higher-affinity B cell antigen receptors [BCRs]). Flow cytometric analysis revealed that AID-Mettl3^fl/fl mice hosted slightly fewer NP-specific B cells in the GC, whereas the proportion of B cells that bind NP₁₂ was significantly reduced in these mice (Fig. 2 C). Accordingly, lower total and high-affinity NP-specific antibody titers were detected in the sera of AID-Mettl3^fl/fl mice (Fig. 2 D). To examine if Mettl3-deficiency affects clonal expansion in the GC, individual IgG1⁺ GC B cells (responding to either NP or KLH) were sorted from single LNs followed by Igh mRNA sequencing. This analysis revealed that the degree of Ig diversity within the GCs of control and AID-Mettl3^fl/fl mice was similar (Fig. 2 E). However, the typical clones that respond to NP in C57BL/6 mice (V_H186.2; Allen et al., 1987) were less expanded in AID-Mettl3^fl/fl mice (33% in control compared with 11% in AID-Mettl3^fl/fl mice; Fig. 2 E). GC B cells in AID-Mettl3^fl/fl mice showed a small (1.3-fold) but significant reduction in the number of total mutations per cell (Fig. 2 F). Nonetheless, the ratio of replacement to silent mutations was similar, suggesting that positive selection of high-affinity variants is intact in AID-Mettl3^fl/fl mice (Fig. 2 F). Accordingly, the few V_H186.2 clonal members in the Mettl3-deficient GC B cells carried a mutation in the Igh V_H186.2 gene (W33L or K59R), which serves as a marker for affinity-based selection (Allen et al., 1987; Furukawa et al., 1999).

To more thoroughly probe whether acquisition of affinity-enhancing mutations is intact in the absence of METTL3, control and AID-Mettl3^fl/fl mice were immunized, and sorted Igλ⁺ IgG1⁺ GC B cells were subjected to Igh mRNA sequencing. Analysis of the V_H186.2 mRNA transcripts revealed normal frequencies of replacement mutations that are associated with enhanced BCR affinity in Mettl3-deficient B cells (Fig. 2 G). Accordingly, calculating the ratio between total and high-affinity NP-specific IgG in the mouse serum revealed a small decrease in average antibody affinity in AID-Mettl3^fl/fl mice; however, the rate of high-affinity antibody formation was significantly slower compared with control mice (Fig. 2 D). Thus, affinity-based clonal selection is intact in Mettl3-deficient GC B cells, although clonal expansion of the typically selected clones is suboptimal.

Our findings demonstrate that METTL3 functions are required for maintenance of GC size over time. The GC is composed of two compartments: the DZ, where B cells acquire SHM and undergo clonal expansion, and the light zone (LZ), where affinity-based selection takes place (Allen et al., 2007; Victora and Nussenzweig, 2012; Allen et al., 2004). GCs of AID-Mettl3^fl/fl mice showed a higher proportion of LZ B cells (CXCR4^low CD86^high) compared with control mice at day 14 of the response but not at day 7 (Fig. 3 A). Immunohistochemical analysis of the topographical localization of GC B cells within the DZ and LZ in chimeric mice confirmed the decrease in Mettl3-deficient cells at day 14. Furthermore, the number of Mettl3-deficient B cells in the DZ was reduced compared with the control, whereas no change was observed in the LZ. This finding suggests that LZ-to-DZ trafficking is not compromised in the absence of METTL3 (Fig. S3 A). To determine which molecular mechanisms depend on METTL3 functions in specific GC sites, LZ and DZ B cells were sorted from control and AID-Mettl3^fl/fl mice and subjected to RNA-seq 7 and 14 d after immunization. Analysis of differential gene expression revealed that METTL3 plays a significant role in regulating gene expression in the DZ, while very small reproducible transcriptional differences were observed in the LZ at both time points (Figs. 3 B and S3 B). Some transcriptional heterogeneity was found within the LZ in individual mice, but was not reproduced across several mice, as shown by principal component analysis (Fig. S3 C). In Mettl3-deficient DZ B cells, 226 genes showed increased expression, whereas 100 genes showed decreased expression compared with control cells (fold-change ≥1.5, adjusted P = 0.1). Furthermore, whereas the transcriptome of control GC B cells showed expected differential gene expression between LZ and DZ, Mettl3-deficient GC B cells showed very small transcriptional differences between these compartments (Fig. 3 C). Consistently, gene set enrichment analysis (GSEA; Subramanian et al., 2005) revealed that Mettl3-deficient DZ B cells (sorted as CXCR4^high and CD86^low) did not express typical DZ genes, but rather showed increased expression of the typical LZ transcriptional program (Fig. 3 D). Although the mRNA level of Cxcr4 was decreased in Mettl3-deficient cells, no significant changes were observed at the protein level, suggesting additional mechanisms that specifically compensate for the transcriptional impairment (Fig. S3 D). Transcription of genes that regulate critical biological processes in the DZ, including cell cycle, DNA repair, and RNA metabolic pathways, were significantly reduced in Mettl3-deficient DZ B cells (Fig. 3 E). In addition, GSEA revealed that cell cycle–, oxidative phosphorylation–, and DNA repair–related genes were down-regulated, whereas genes that are part of the apoptosis pathway were up-regulated, compared with control mice (Fig. 3, F and G). Furthermore, the MYC signature, which is associated with proper GC functions and positive B cell selection, was down-regulated. Accordingly, gene expression signatures that are associated with B cell selection and GC core genes were not properly expressed (Fig. 3, F and G). Together, these results indicate that although discernible DZ and LZ compartments expressing zone-specific markers are formed independently of METTL3, this enzyme is required for proper expression of typical DZ-associated gene programs, including genes that are activated by MYC and cell cycle–related genes.

To verify our transcriptomic findings in functional assays, we sought to measure cell division and cell cycle stages in GC B cells. To this end, we pulsed immunized mice with 5-ethynyl-2′-deoxyuridine (EdU) by i.v. injection and examined the frequency of EdU⁺ cells after 2.5 h. Flow cytometric analysis revealed that Mettl3-deficient GC B cells proliferated to a lesser degree than cells derived from control mice during this short experimental period (∼5% fewer Mettl3-deficient GC B cells per 2.5 h, 17% fewer compared with controls; Fig. 4 A). Co-labeling of DNA-incorporated EdU and total DNA revealed that significantly fewer cells were detected in the S-phase of the cell cycle in Mettl3-deficient GC B cells (Fig. 4 B), and similar defects were detected in LZ and DZ compartments (Fig. S3 E). These small differences in cell cycle progression reflect a defect in proliferation capacity that can greatly affect the number of GC B cells over time.

The transcriptomic analysis revealed that genes involved in apoptosis are up-regulated in Mettl3-deficient GC B cells. To determine if the reduction in GC B cell frequencies is also a result of an increase in apoptotic events, in addition to reduced proliferation, we examined the activity of caspase-3/7. To distinguish between apoptotic and dead cells, we colabeled B cells with a DNA dye (SYTOX AADvanced [SAAD]) and a cleavable peptide that labels cells with active caspase-3/7. SAAD- and active caspase-3/7–positive signals label dead cells, whereas SAAD-negative and active caspase-3/7–positive signals label apoptotic cells. As expected, very few apoptotic cells were observed in the naive B cell compartment (Fig. 4 C). In contrast, apoptotic cells were clearly detected in the GC cell population, and their frequency in Mettl3-deficient GC B cells was threefold higher than in the control cells in both DZ and LZ (Fig. 4 C). Consistent with the lack of effect on GC size on day 7 after immunization, Mettl3-deficient B cells did not show significantly increased apoptosis at this time point, although a trend in that direction was observed (Fig. 4 D). These results demonstrate that METTL3 activity prevents excess B cell apoptosis in the GC reaction.

AID enzymatic activity inserts mutations into the Igh genes that may result in the inability of GC B cells to express a functional BCR (Mayer et al., 2017; Stewart et al., 2018). In addition, this enzyme acts as a major driver of off-target DNA damage events and chromosomal translocations in GC B cells (Casellas et al., 2016). Furthermore, AID activity induces DNA double-strand breaks, and METTL3 was suggested to play a role in the subsequent DNA repair mechanisms (Zhang et al., 2020; Daniel and Nussenzweig, 2013). To examine whether enhanced apoptosis and reduction in the frequency of GC B cells is a result of AID-mediated DNA damage or defects in the subsequent DNA repair mechanism, we generated Aicda^cre/cre Mettl3^fl/+ and Aicda^cre/cre Mettl3^fl/fl mice that lack expression of the AID enzyme. The deletion of AID did not increase the frequency of GC B cells in Aicda^cre/cre Mettl3^fl/fl immunized mice to approach that of the control mice on day 14 after immunization, and the DZ-to-LZ ratio defect remained (Fig. 4 E). These results indicate that deficiency in neither AID functions nor DNA repair of AID-mediated DNA damage can account for the decreased GC cell number in Mettl3-deficient mice. Furthermore, BCR downstream signaling was intact in Mettl-deficient GC B cells, suggesting that these cells can effectively respond to antigens in the GC (Fig. S3 F). Collectively, these results suggest that METTL3 supports B cell persistence and attenuates apoptosis in the GC, primarily through regulation of proliferation.

IGF2BPs are a family of mRNA-binding proteins that increase the mRNA lifetime of target genes and whose binding capacities are enhanced by m⁶A methylation (Palanichamy et al., 2016; Ren et al., 2020; Huang et al., 2018; Samuels et al., 2020). IGF2BP2 plays a role in regulation of mRNAs in Th17 cells through interaction with m⁶A (Bechara et al., 2021). IGF2BP3 binds the 3′-UTR of Myc mRNA and is the main paralog expressed in immune cells (Palanichamy et al., 2016). Because we found that the Myc pathway is down-regulated in Mettl3-deficient GC B cells, and because m⁶A enhances IGF2BP3 regulation of Myc mRNA (Huang et al., 2018), we sought to examine the role of IGF2BP3 in GCs. The size of the GC compartment 7 d after immunization in Igf2bp3-deficient mice was similar to that of WT mice; however, the frequency of GC B cells decreased after an additional 7 d (Figs. 5 A and S4 A). This defect was B cell specific, as a similar phenotype was observed in chimeric mice hosting both WT and Igf2bp3^−/− immune cells (Fig. S4 B). In addition, Igf2bp3-deficient B cells responded to anti-IgM stimulation in vitro, suggesting that this mRNA binder does not play a role in B cell activation (Fig. S1 I). To examine the effect of IGF2BP3 on MYC-responsive genes, GC B cells were isolated from mice 14 d after immunization and subjected to RNA-seq followed by GSEA. Significant down-regulation of MYC-responsive genes, GC-related genes, and B cell selection signatures, as well as a loss of the DZ transcriptional program, were observed in Igf2bp3-deficient mice (Fig. 5 B and Fig. S4, C and D). In contrast to the changes in gene signature observed in Mettl3-deficient B cells, we did not detect reduction in expression of genes that are associated with oxidative phosphorylation (Fig. 5 B). Collectively, we conclude that IGF2BP3 is required for adequate expression of MYC-downstream targets in GC B cells.

YTHDFs are a family of mRNA binders that interact with m⁶A, and the YTHDF2 paralog is associated with transcript destabilization and degradation (Du et al., 2016; Wang et al., 2014; Zaccara and Jaffrey, 2020). YTHDF2 is the dominant paralog in GC B cells (Fig. S4 E). To examine whether the inability of YTHDF2 to bind m⁶A in Mettl3-deficient mice also accounts for defects in B cell functions, we produced CD23-Cre Ythdf2^fl/fl and AID-Cre Ythdf2^fl/fl (CD23- and AID-Ythdf2^fl/fl) mouse strains (Ivanova et al., 2017). Specifically, the CD23-Ythdf2^fl/fl strain was used for GC size analyses, and to avoid pre-GC defects, the AID-Ythdf2^fl/fl strain was used for gene expression and subsequent functional analyses. In CD23-Ythdf2^fl/fl mice, GCs were detectable in response to immunization; however, they were smaller compared with their control counterparts in CD23-Ythdf2^fl/+ mice at day 14 after immunization (Fig. 5 C). In contrast to the results obtained in Mettl3-deficient cells, RNA-seq analyses revealed that MYC-responsive genes, the core GC genes, and gene programs associated with the selection of high-affinity clones were not altered in GC B cells derived from AID-Ythdf2^fl/fl mice, wherein Ythdf2 is deleted after early B cell activation (Figs. 5 D and S4 C). In contrast, GSEA analyses revealed severe down-regulation in the expression of genes that are associated with oxidative phosphorylation, which was also observed in Mettl3-deficient GC B cells (Fig. 5 D). These results suggest that YTHDF2 plays a role in the GC; however, as opposed to METTL3 and IGF2BP3, it does not enhance MYC-induced genetic programs.

Oxidative phosphorylation is the major respiratory pathway in GC B cells (Weisel et al., 2020). To examine if oxidative phosphorylation is intact in Ythdf2-deficient GC B cells, we used a probe (JC-1) that allows assessment of mitochondrial health by measuring the polarization of this organelle. Ythdf2-deficient GC cells stained with the JC-1 probe exhibited hyperpolarized mitochondria compared with control GC cells, which suggests a defect in the mitochondrial electron transport chain (Fig. 5 E). This defect was not observed in naive B cells in which Ythdf2 is not deleted. In contrast, analysis of metabolism in naive and activated B cells in vitro using the Seahorse system did not reveal a pronounced defect in either oxidative phosphorylation or glycolysis, suggesting that either the observed defects are specific for GC B cells in situ or that the metabolic status in cultured B cells is different from GC B cells under physiological conditions (Fig. S4 F). Collectively, we conclude that intact proliferation through MYC expression as well as mitochondrial functions depend on the m⁶A machinery in GC B cells.

The GC reaction and selection of high-affinity clones by T cells depend on MYC functions and activation of downstream gene transcription promoting clonal expansion (Calado et al., 2012; Dominguez-Sola et al., 2012). Therefore, we examined in depth how Myc expression is regulated by the m⁶A machinery. RT-qPCR analysis of sorted GC B cells revealed reduced levels of Myc mRNA in Mettl3-deficient LZ B cells, and lower expression of MYC protein was observed in LZ B cells derived from AID-Mettl3^fl/fl mice that express MYC-GFP compared with AID-Mettl3^fl/+ littermates (Fig. 6, A and B; Huang et al., 2008). We also examined Myc expression in this model at early time points using PE-specific cells and RT-qPCR. Here, we did not find a significant reduction in Myc expression, consistent with the observation that GC formation at day 7 was intact in this model (Fig. S4 G). However, lower levels of MYC protein were also observed in PE-specific B cells derived from CD23-Mettl3^fl/fl mice 3 d after immunization with PE, suggesting that METTL3 regulates Myc early during the response as well (Fig. S4 H). Consistent with the decrease in Myc levels in Mettl3-deficient GC B cells, Igf2bp3-deficient GC B cells showed reduced Myc mRNA and protein levels; in contrast, Ythdf2-deficient cells did not show a reduction in Myc mRNA levels (Fig. 6, C–E). These findings indicate that METTL3 maintains sufficient Myc mRNA and protein through IGF2BP3 functions rather than YTHDF2. To examine whether the defect in maintenance of sufficient Myc transcripts through METTL3 activity is the reason for the reduction in the frequency of GC B cells in AID-Mettl3^fl/fl mice, we generated mice that overexpress MYC in Mettl3-deficient and -proficient GC B cells (Sander et al., 2012). AID-Mettl3^fl/fl mice were bred with a mouse strain that carries the human Myc gene in the Rosa26 locus and a stop codon in front of it, flanked by LoxP sites (R26^Myc/+; Sander et al., 2012). Age-matched transgenic mice were immunized with NP-KLH, and GC size was examined by flow cytometry after 14 d. Whereas the frequency of GC B cells was smaller in AID-Mettl3^fl/fl mice compared with control mice, the AID-Mettl3^fl/fl R26^Myc/+ mice showed normal frequencies of GC cells; however, an increase in GC size was also observed in AID-Mettl3^fl/+ R26^Myc/+ mice (Fig. 6 F). Thus, METTL3 functions regulate Myc transcript dosage and expression of downstream genes; nonetheless, since overexpression of Myc also increased the frequency of GC B cells in the control mice as previously observed (Sander et al., 2012), it was impossible to determine whether suboptimal Myc expression is the only reason for the reduction in the frequency of GC cells over time.

METTL3 is a methyltransferase that catalyzes the addition of m⁶A on mRNA transcripts at consensus sites. To examine whether Myc mRNA carries m⁶A modifications in B cells, we performed immunoprecipitation of methylated mRNA (m⁶A-IP). RT-qPCR of WT in vitro–activated B cells revealed that Myc mRNA was highly enriched in the pull-down fraction compared with control unmethylated mRNA (Gapdh; Fig. 7 A). To further validate the methylation of Myc transcripts, we performed m⁶A-IP followed by high-throughput sequencing of the input and IP fractions. To assess m⁶A levels in control and Mettl3-deficient B cells (CD23-Mettl3^fl/fl), we compared the average IP over input fractions in both populations and in selected key B cell genes including Myc, and a statistically significant reduction in Myc mRNA methylation was detected in this assay (Fig. S5, A–F). M⁶A-seq on mRNA derived from Mettl3-deficient B cells yielded peak over input (POI) scores that were reduced to 60% of control (Fig. S5, A–D). The m⁶A sites that were specifically detected in control cells were enriched around m⁶A consensus motifs, 3′-UTRs, and stop codons (Fig. S5, A–D). Notably, as previously shown, Myc mRNA was found to carry methylations at sites that are close to the stop-codon region in B cells (Fig. 7 B). Interestingly, AID mRNA was also methylated; however, we did not observe changes in AID mRNA transcript levels in Mettl3-deficient GC B cells (Fig. S5 G). Thus, Myc mRNA is methylated at typical m⁶A sites in B cells.

To understand the magnitude of the contribution of direct m⁶A targets to the changes in expression of gene signatures, we examined the fraction of m⁶A targets in genes that are significantly differentially expressed in DZ GC B cells. We found 23 genes in the down-regulated fraction and 99 genes in the up-regulated fraction (Fig. S5 H). Furthermore, we analyzed the potential enrichment of m⁶A targets in GSEA hallmark gene sets. However, we did not find any significantly enriched gene sets or single genes that could explain the observed changes in gene signatures. In addition, considering the general trend of m⁶A target up-regulation upon deletion of Mettl3 (Fig. S5 I) and that many of the affected gene signatures were down-regulated, we conclude that, apart from Myc, many of the gene expression changes in the signatures observed in Mettl3-deficient GC B cells are due to indirect effects of m⁶A modification.

Next, we examined how METTL3 functions regulate Myc transcript quantity in B cells. To gain more insight into the abundance and persistence of Myc mRNA in the GC at the single-cell level, we labeled Myc transcripts by single-molecule fluorescence in situ hybridization (smFISH; Bahar Halpern and Itzkovitz, 2016). B cells in GC structures were identified by staining with Ki-67 probes and B220 antibodies in LN slides derived from immunized mice (Fig. S5 J). Quantification of Myc transcript number per B cell revealed that in control mice, 1.5-fold more single GC B cells expressed ≥2 mRNA molecules compared with the corresponding Mettl3-deficient GC B cells (Fig. 7 C). The fewer Myc transcripts observed overall and in single GC B cells can be a result of ineffective transcription through modulation of upstream regulators or enhanced mRNA degradation. To distinguish between these two possibilities, we measured the transcription rate through the visualization of Myc transcription start sites (TSSs) in GC B cells by colabeling Myc exons and introns in LN sections using a previously described protocol (Bahar Halpern et al., 2015; Bahar Halpern and Itzkovitz, 2016). The transcription rate was measured based on the number of mRNA molecules in the TSS (as described in Materials and methods). Using this analysis, we found that the rate of Myc transcription in situ in Mettl3-deficient GC B cells was similar to that in controls (Fig. 7, D and E). By further comparing the number of Myc mRNA molecules per cell to the transcription rate, we calculated the rate of degradation in both control and Mettl3-deficient cells. Since we detected fewer mRNA molecules in Mettl3-deficient GC B cells while the transcription rate was intact, we found that the degradation rate was enhanced (Fig. 7, D and E; Bahar Halpern et al., 2015; Bahar Halpern and Itzkovitz, 2016). To examine if IGF2BP3 indeed stabilizes Myc mRNA in B cells, we measured the Myc mRNA decay rate by blocking transcription with actinomycin D. We found that Myc transcript decay is accelerated in B cells that lack IGF2BP3 (Fig. S5 K). We conclude that the lower amounts of Myc mRNA observed by RT-qPCR and smFISH, as well as the defect in MYC-downstream gene expression in Mettl3-deficient GC B cells, was not a result of direct or indirect defects in Myc transcription, but rather resulted from an enhanced Myc mRNA degradation rate.

Methylation of mRNA affects various cell functions through the binding of several types of m⁶A readers (Shulman and Stern-Ginossar, 2020; Zaccara and Jaffrey, 2020; Zaccara et al., 2019). Here, we demonstrate the role of mRNA methylation and two different mRNA binders in supporting biological processes that are critical for GC formation and maintenance. Whereas Myc expression depends on IGF2BP3, oxidative phosphorylation–related genes are modulated indirectly by YTHDF2. Together, these findings portray how distinct m⁶A binders modulate two independent pathways that support critical processes required for effective GC functions.

MYC expression is induced rapidly upon B cell interaction with antigen, but after 2 d, the level of this protein is rapidly decreased, followed by its limited reexpression in the fully formed GC (Dominguez-Sola et al., 2012; Calado et al., 2012). MYC expression dynamics in B cells is consistent with our findings that demonstrate a requirement for METTL3 at the early and late stages of the response, but not during initial AID expression at the time of pre-GC events (Calado et al., 2012; Dominguez-Sola et al., 2012). In particular, Myc-inducing factors such as cognate antigen and T cell help signals are limited at later stages of the response, when METTL3 functions and IGF2BP3 are essential for maintaining the GC reaction. Myc expression levels are important for both B cell proliferation and positive selection (Dominguez-Sola et al., 2012). The frequency of Mettl3-deficient B cells in GCs declined over time independently of competition with Mettl3-expressing cells, indicating that the m⁶A machinery primarily plays a role in global B cell proliferation machineries in the GC, rather than specifically in the positive clonal selection process. In support of this conclusion, we found that the acquisition of SHM and affinity-enhancing mutations in the GC is METTL3 independent, whereas effective proliferation and proper expression of the GC core genes required METTL3 functions. In parallel, we found an increase in the frequency of apoptotic B cells in AID-Mettl3^fl/fl. Since GC B cells are intrinsically pro-apoptotic and depend on T cell help for effective proliferation and survival, we suggest that lack of sufficient expression of MYC-downstream genes can explain the increase in apoptosis in AID-Mettl3^fl/fl mice. An additional explanation for the GC B cell decline may be early GC termination; however, in chimeric mice, control B cells persisted, whereas the Mettl3-deficient B cells declined over time, suggesting an intrinsic defect rather than a global one, such as reduction in antigen availability or the presence of high-affinity antibodies that block antigens in the GC (Zhang et al., 2013). Our findings are consistent with a model wherein T cells are able to distinguish between low- and high-affinity clones in the GC independently of m⁶A modifications in B cells, while effective global and preferential clonal expansion in the GC DZ depends on m⁶A mRNA modifications including Myc methylation.

The outcome of mRNA methylation depends on interactions with reader proteins. Typically, the deletion of Mettl3 or Ythdf2 in different cell types leads to an increase in the abundance of many target transcripts (Ivanova et al., 2017; Zaccara et al., 2019; Zaccara and Jaffrey 2020). In contrast, in vitro and in vivo analyses revealed that MYC-responsive genes are down-regulated upon deletion of Mettl3 or its binding partner Mettl14, primarily as a result of destabilization of Myc transcripts (Vu et al., 2017; Cheng et al., 2019b; Weng et al., 2018; Huang et al., 2018; Zhao et al., 2020; Cheng et al., 2019a; Barbieri et al., 2017). Indeed, m⁶A pull-down experiments revealed that Myc transcripts were highly methylated in B cells. Our direct quantification of Myc mRNAs in GC B cells in situ demonstrates that the decline in transcript abundance is a result of transcript degradation rather than a reduction in expression of a transcription factor that controls transcript dosage. These findings are in line with several previous studies that show Myc destabilization in the absence of m⁶A and introduce a physiological function for this machinery in the immune response (Cheng et al., 2019b; Vu et al., 2017; Weng et al., 2018; Ma et al., 2021).

In several nonmammalian model organisms and cell lines and in a specific subset of T cells, IGF2BP paralogs enhance the stability of several mRNAs including Myc transcripts, and m⁶A has been shown to reinforce these mRNA–protein interactions (Palanichamy et al., 2016; Huang et al., 2018; Ren et al., 2020; Samuels et al., 2020; Bechara et al., 2021). Our findings add a new role to this mRNA stabilization machinery in GC B cells and demonstrate how IGF2BP3 supports cellular functions in effective establishment of antibody-mediated immunity. In contrast, maintenance of sufficient Myc transcript levels through METTL3 functions does not depend on the dominant YTHDF family member in GC B cells, suggesting that other pathways are controlled by this reader, including proper expression of genes that support oxidative phosphorylation (Weisel et al., 2020). YTHDF2 promotes the degradation of mRNAs, and in its absence, the levels of m⁶A-modified mRNAs increase (Ivanova et al., 2017; Zaccara and Jaffrey, 2020). Thus, this reader most likely modulates mitochondrial respiration–related genes in an indirect manner through the regulation of upstream events. Whether YTHDF2 and IGF2BP2 depend exclusively on direct binding to m⁶A (Huang et al., 2018) or additional indirect events, such as m⁶A-dependent changes in mRNA conformation or direct interactions with each other, requires additional investigation (Sun et al., 2019; Youn et al., 2018; Weidensdorfer et al., 2009; Zaccara et al., 2019). Expression of oxidative phosphorylation–associated genes is also regulated by HuR, an additional RNA binder that plays a role in B cell functions (Diaz-Muñoz et al., 2015). The interplay between HuR functions and YTHDF paralogs requires further investigation.

Collectively, our findings describe how the m⁶A machinery supports central processes in the adaptive immune response through distinct m⁶A readers. The study suggests that components of the m⁶A machinery may serve as targets for manipulation of the adaptive arm of the immune system in MYC-driven lymphoma. In addition, since GC cells depend on mitochondrial respiration for energy generation, drugs that target the m⁶A machinery are expected to interfere with the GC reaction and can potentially alleviate symptoms in antibody-driven autoimmunity.

Mettl3^fl/fl mice were provided by J.H. Hanna and crossed to Cre-expressing mice; B cell–specific Cd23cre mice were generated and provided by M. Busslinger (Research Institute of Molecular Pathology, Vienna, Austria). Aicda^Cre/+, Cd19-Cre, R26^{flox-stop-flox Myc}, MYC-GFP, and Igf2bp3^−/− mice were purchased from The Jackson Laboratory. Ythdf2^fl/fl were provided by Donal O’Carroll (University of Edinburgh, Edinburgh, UK). MYC-GFP–expressing mice were bred to mice carrying Aicda^Cre and Mettl3^fl/fl alleles. WT mice (C57BL/6) were purchased from Envigo. WT and transgenic mice were used at the age of 6–12 wk. To avoid age variations in experiments that included a small number of mice carrying the R26^{flox-stop-flox Myc} gene, only age-matched mice (∼8 wk) were used. All experiments on mice were approved by the Weizmann Institute Animal Care and Use Committee.

Chimeric mice were generated by irradiating host mice with 950 rad followed by injection of fresh BM cells. For the generation of mixed chimeras, CD45.2⁺ hosts were reconstituted with a mixture of BM cells derived from Aicda^Cre/+ CD45.1⁺ (30%) and Aicda^Cre/+ Mettl3^fl/fl CD45.2⁺ (70%) mice or from CD45.1⁺ (50%) and Igf2bp3^−/− CD45.2⁺ (50%) mice. Experiments with chimeric mice were performed 8 wk after BM transplantation.

NP-KLH (BioSearch Technologies) was prepared with alum and PBS at a final concentration of 0.4 mg/ml. Mice received a single administration to each footpad of 25 µl (10 µg NP-KLH). For immunizations with PE, 10 µg PE (Molecular Probes) in alum was injected.

Immunized mice were injected i.v. with 2 mg of the nucleoside analogue EdU (Molecular Probes) in PBS. After 2.5 h, popliteal LNs were dissected, and isolated LN cells were stained for surface antigens as described below, followed by EdU detection using the Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Molecular Probes) according to the manufacturer’s instructions. 7AAD was from BD Biosciences and added according to the manufacturer’s instructions.

For measurements of apoptosis and cell death, popliteal LNs from immunized mice were isolated, and single-cell suspensions were stained for GC surface antigens and analyzed for apoptotic events using CellEvent Caspase-3/7 Green Flow Cytometry Assay Kit (Molecular Cell Probes) according to the manufacturer’s instructions.

Popliteal LNs were removed, washed in cold PBS, and forced through a mesh into PBS containing 2% FCS and 1 mM EDTA to create single-cell suspensions. Cell suspensions from PPs were similarly prepared by excising PPs from small intestines prewashed with ice-cold PBS to remove fecal content. To block Fc receptors, washed cells were incubated with 2 µg/ml anti-16/32 (TruStain FcX, BioLegend) for 5–10 min before antibody staining. Cells were subsequently incubated with fluorescently labeled antibodies (Table S1) for 30 min on ice. Intracellular antibody staining was performed after fixation and permeabilization with Fixation/Permeabilization Solution Kit (BD Biosciences).

GC cells were gated as live/single, B220⁺ CD38⁻ FAS⁺, and/or GL-7⁺. DZ and LZ cells were gated as CXCR4^hi CD86^lo and CXCR4^lo CD86^hi. BM transplanted cells in chimeric mice were distinguished by staining cell suspensions with GC markers, as indicated, in addition to the congenic markers CD45.1 and CD45.2 (eBioscience and BioLegend, respectively). Antigen-specific cells in PE- and NP-KLH immunization experiments were detected by staining cell suspensions with GC markers along with PE, PE-NP12, or PE-NP28. Stained cell suspensions were analyzed using a CytoFlex flow cytometer (Beckman Coulter). For RNA-seq, cells were stained for negative markers (dump⁻: CD4⁻, CD8⁻, GR-1⁻, and F4/80⁻) in addition to GC markers, sorted directly into 40 µl Lysis/Binding buffer (Life Technologies) using a FACS ARIA II or FACS ARIA III (BD) instrument, and immediately frozen on dry ice.

For analysis of mitochondrial mass and membrane potential, single-cell LN suspensions were first stained with surface antibodies. Cells were then incubated with 0.3 µg/ml JC-1 probe (Invitrogen) or 50 nM Mitoview-Green probe (Biotium) at 37°C for 30 min and washed in 2 ml PBS before analysis by flow cytometry.

Splenic cells from CD23-Mettl3^fl/fl and control mice or UBC^CreERT2-Mettl3^fl/fl mice were stained with CellTrace Violet (Invitrogen) according to manufacturer's instructions and seeded at 1 × 10⁶ cell density in a 96-well plate. Cells were stimulated with 0–10 µg/ml LPS or anti-IgM for 3 d in a 37°C cell incubator, and CellTrace Violet dilution was assessed by flow cytometry. Half of the cells from UBC^CreERT2-Mettl3^fl/fl mice were treated with 1 µM 4OHT for 24 h and washed in fresh RPMI before stimulation.

Phosphorylation of ERK upon BCR engagement was analyzed by stimulating isolated (using anti-CD43 negative-selection beads, Miltenyi) splenic B cells with anti-IgM for 0–30-min time intervals. 200 µl of 1× 10⁶ B cell suspension was added to Eppendorf tubes; 0.4 mM H₂O₂, together with 10 µg/ml anti-IgM, was added and the sample tube was vortexed. Samples were incubated in a 37°C cell incubator for indicated time intervals (see figure legends). Cells were fixed by adding 150 µl 4% paraformaldehyde and incubating for 15 min at room temperature. All samples in an independent experiment were processed together by washing with PBS and then permeabilizing and preserving the cells with 200 µl ice-cold methanol. Permeabilized cells were incubated for a minimum of 45 min on ice or kept overnight at −80°C and then washed twice with FACS buffer (see Flow cytometry) before pERK staining. Flow cytometry was used to determine the levels of pERK staining.

mRNA from 2–5 × 10⁴ sorted cells was captured with 12 µl of oligo(dT) Dynabeads (Life Technologies), washed, and eluted at 85°C with 6.5 ml of 10 mM Tris-HCl (pH 7.5). A bulk adaptation of the massively parallel single-cell RNA sequencing protocol (MARS-seq; Jaitin et al., 2014; Keren-Shaul et al., 2019) was used to generate RNA-seq libraries for expression profiling of GC B cells from AID-Mettl3^fl/+ and AID-Mettl3^fl/fl mice 1 and 2 wk after immunization, AID-Ythdf2fl/+ and AID-Ythdf2fl/fl mice 1 wk after immunization, or WT and Igf2bp3^−/− mice 2 wk after immunization. Two to four biological replicates were included in each population. Briefly, mRNA from each sample was barcoded during RT and pooled. Following Agencourt Ampure XP bead cleanup (Beckman Coulter), the pooled samples underwent second-strand synthesis and were linearly amplified by T7 in vitro transcription. The resulting RNA was fragmented and converted into a sequencing-ready library by tagging the samples with Illumina sequences during ligation, RT, and PCR. Libraries were quantified by Qubit and TapeStation as well as by qPCR for the ACTB gene as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019). Sequencing was done using a Nextseq 75-cycle high-output kit (Illumina; paired-end sequencing). Sequencing data have been deposited into the NCBI Gene Expression Omnibus under accession no. GSE180359.

Alignment and differential expression analyses were performed using the UTAP pipeline (Kohen et al., 2019). Reads were trimmed using Cutadapt and mapped to the Mus musculus genome (UCSC mm10) using STAR (Dobin et al., 2013) v2.4.2a with default parameters. The pipeline quantifies the genes annotated in RefSeq (extended by 1,000 bases toward the 5′ edge and 100 bases in the 3′ direction). Counting of sequenced reads was done using htseq-count (Anders et al., 2015; union mode). Genes having a minimum of five unique molecular identifier–corrected reads in at least one sample were included in the analysis. Normalization of the counts and differential expression analysis was performed using DESeq2 (Love, Huber, and Anders 2014) with the following parameters: betaPrior = true, cooksCutoff = false, and independentFiltering = false. Raw P values were adjusted for multiple testing using the procedure of Benjamini and Hochberg.

Gene annotation and pathway analysis was performed using Metascape (Zhou et al., 2019). GSEA was performed using GSEA 3.0 with the GSEA-preranked tool (Mootha et al., 2003; Subramanian et al., 2005). Gene names were converted to human gene symbols, and GSEA was run with default parameters (1,000 permutations). The Molecular Signature Database hallmark gene sets were used to perform pathway enrichment analysis using a hypergeometric distribution. Gene sets for GC formation and GC B cell selection were generated from GSE110669 and GSE98778, respectively, based on the 200 most significantly up-regulated genes in each condition. The dataset used for expression profiling in GC and DZ and LZ cells was GSE93554.

RNA was extracted using Dynabeads mRNA DIRECT Purification Kit (Invitrogen) and converted to cDNA using either qScript (Quantabio) or Super Script III (Invitrogen) enzyme according to the manufacturer’s instructions with a mix of random and poly A–specific primers. For RT-qPCR reactions, a QuantStudio 5 Real-Time PCR system and Fast SYBR Green Master Mix (Applied Biosystems) were used. Primers are listed in Table S2. Stability assays were performed by incubating isolated B cells with 5 µg/ml actinomycin D in RPMI complete medium for 0–40-min intervals before lysis and RNA extraction.

Serum was collected from multiple mice in several experiments, and titers of IgM, IgG1, and IgG2 antibodies were determined by a single direct ELISA. Diluted serum was used to coat 96-well plates by incubating overnight at 4°C. Relative quantities of antibodies were detected using HRP-conjugated anti-mouse IgM, IgG1, and IgG2 (Abcam) conjugated to HRP. NP-specific ELISAs were performed by first incubating ELISA plates with 5 µg/ml NP1-9 or NP>20 BSA overnight at 4°C. Diluted serum was incubated for 2 h at room temperature before detection.

Fixed single popliteal LNs were frozen in optimum cutting temperature compound and cut into 7-µm sections. Slides were blocked and incubated with fluorescently labeled antibodies. B220-PE, Streptavidin-647, IgD-PE, and GL-7-A647 antibodies were from BioLegend. CD35-biotin antibody was from BD Biosciences. GCs were identified as IgD⁻ or IgD⁻ GL-7⁺ areas. Where the number of available fluorescent channels exceeded the number of required labels, consecutive slides were stained with separate antibody panels. Images were captured using a 20× objective and a Ti2 eclipse inverted confocal microscope (Nikon). For imaging specifications, see smFISH: Hybridization and imaging. Quantification of cell number within GC DZ and LZ compartments was done using automatically detected nuclei/spots in Imaris software (v9.5.1, Oxford Instruments). Automatic threshold selection was manually verified, and non-B220⁺ cells were removed from analysis.

Single popliteal LNs were harvested from control and AID-Mettl3^fl/fl mice and processed for flow cytometric analysis. Cell suspensions were stained for negative markers (dump⁻: CD4⁻, CD8⁻, GR-1⁻, and F4/80⁻) and gated as dump⁻ B220⁺ GL-7⁺ CD38⁻ FAS⁺ IgG1⁺ or dump⁻ B220⁺ GL-7⁺ CD38⁻ FAS⁺ IgG1⁺ IgL⁺ for isolation of VH186.2-expressing cells. Cell sorting was performed using a FACS Aria II cell sorter (BD Biosciences). For total V(D)J sequencing of Ighγ1 heavy chains, GC-derived B cells were sorted into 96-well plates containing lysis buffer (PBS containing 3 U/µl ribonuclease inhibitor and 10 mM dithiothreitol). For sequencing of Ighγ1 heavy chains from VH186.2-expressing cells, GC-derived B cells were sorted in bulk into lysis/binding buffer (Life Technologies).

Complementary DNA from single cells was generated using random primers (NEB), as previously described (Scheid et al., 2009). Ighγ1 heavy-chain sequences were amplified in a nested PCR using primers for the Ighγ1 constant region, together with a mix of primers for the variable regions (von Boehmer et al., 2016). Complementary DNA from bulk sorting was produced using qScript cDNA Synthesis Kit (QuantaBio). VH186.2 (V1-72*01) Ighγ1 heavy-chain sequences were amplified in a nested PCR using primers for the Ighγ1 constant region together with a specific primer for VH186.2 (Mayer et al., 2017; Table S1) and using high-fidelity Q5 polymerase (NEB). Amplified products were cloned (CloneJET PCR Cloning Kit, Thermo Fisher Scientific), sequenced, and analyzed.

PCR products were sequenced by Sanger sequencing, and CDR3 regions were analyzed by aligning Ig Fasta sequences against the IMGT mouse heavy gene database (September 2017) using IgBlast (v1.7.0; Ye et al., 2013). Sequence alignment was performed using SnapGene software (GSL Biotech). For a schematic representation of total Ighγ1-sequence results, sequences derived from single cells were clustered according to their CDR3 region (identical V, D, and J segment represents one clone) and presented as a fraction of total sequenced cells. Primer-derived mutations were excluded from the analysis. For schematic representation of VH186.2-sequence results, sequences were clustered according to presence/absence of W33L and K59R mutations and presented as the proportion of total sequenced cells.

Resting B cells were isolated from mouse spleens by forcing tissue through 70-µm mesh into PBS. B cells were purified from cell suspensions using αCD43 magnetic beads (magnetic-activated cell sorting) according to the manufacturer’s instructions (Miltenyi Biotec). B cell purity of ≥96% was confirmed by staining for B220 and analysis by flow cytometry. Isolated B cells were stimulated with LPS (5 mg/ml) for 1 d in vitro at 1.5 × 10⁶ cells per ml in RPMI complete medium (10% FCS, penicillin/streptomycin, L-glutamine, Hepes, and β-mercaptoethanol). Total RNA was isolated using Trizol reagent (Invitrogen), and polyadenylated RNA was purified twice using oligo(dT) beads (Life Technologies).

M⁶A-IP and RNA-seq library preparation were performed as previously described (Garcia-Campos et al., 2019). Three biological replicates of control and Mettl3-deficient samples were included. Reads were aligned to the mouse mm9 genome using STAR aligner (Dobin et al., 2013). A previously published approach was used to identify putative m⁶A sites (Schwartz et al., 2014). Each site was assigned POI and peak over median (POM) scores representing the fold-change of enrichment in the peak region over the corresponding region in the input fraction or over the median coverage in the gene, respectively, as in Schwartz et al. (2014). For identification of sites present in the control sample but absent in the Mettl3-deficient sample, we performed wo separate t tests to assess whether the POM and/or POI scores differed significantly (P < 0.05) between control and Mettl3-deficient samples. A peak was determined to be METTL3-specific if the mean POI and mean POM scores (across the triplicates) in the control samples were higher than their counterparts in the Mettl3-deficient samples and if at least one of the two associated P values was significant. For estimating the global methylation levels per sample (Fig. S5), we calculated POI and POM scores for a preassembled set of 16,321 consistently identified m⁶A sites and a set of 30,115 control sites (Garcia-Campos et al., 2019). Sequencing data have been deposited into the NCBI Gene Expression Omnibus under accession no. GSE180359.

LNs from AID-Mettl3^fl/+ and Mettl3^fl/fl mice were extracted and fixed in 4% paraformaldehyde in ultra-clean water (Sigma-Aldrich). 5 µm (exon-only slides) or 8 µm (exon and intron slides) slices were cryo-sectioned, adhered to poly-L-lysine–treated coverslips, and washed with ice-cold 70% ethanol. Coverslips were then placed in a 6-well plate and incubated with hybridization mix overnight. Hybridization mix consisted of 10% dextran sulfate, 15% formamide, Escherichia coli tRNA (1 mg/ml), 2× saline sodium citrate buffer, 0.02% BSA, and 2 mM vanadyl ribonucleoside complex with both probes (1:3,000) according to a previously described protocol (Itzkovitz et al., 2012).

Probe libraries were designed and constructed as described in Raj et al. (2008). Myc intron and exon libraries consisted of 48 probes of 20-bp length, complementary to the intronic sequence or coding sequence, respectively. Ki67 libraries consisted of 96 probes (Table S3). Hybridizations were done overnight with two to three differentially labeled probes using Cy5, Alexa Fluor 594, and TMR fluorophores. A488-conjugated antibody for B220 (BioLegend) was added to the hybridization mix and used for protein immunofluorescence. 4,6-Diamidino-2-phenylindole dye for nuclear staining was added during the washes. Images were acquired with a Nikon Ti-2 inverted fluorescence microscope equipped with a 100× oil-immersion objective and a Photometrics Prime BSI scientific complementary metal oxide–semiconductor camera using NIS-Elements AR software (v5.20.01; Nikon). The image-plane pixel dimension was 0.13 µm. Quantification was done on stacks of optical sections in which not more than a single cell was observed. Dots were automatically detected using Imaris software (v9.5.1, Oxford Instruments). Automatic threshold selection was manually verified and corrected for errors. Background dots were detected according to size and by automatically identifying dots that appeared in more than one channel (typically <1% of dots) and were removed. Such dots occasionally appeared in the surrounding non-B cells but were rare in the B220⁺ B cells. Bleed-through of transcript signal between channels was minimal. Cell segmentation was performed manually on a maximal projection of the A488 channel.

For metabolic analysis, splenic B cells were purified using negative-selection anti-CD43 magnetic beads (Miltenyi Biotec). Stimulated B cells were incubated in full RPMI with anti-IgM (10 µg/ml) overnight. Subsequently, B cells were collected, centrifuged at 300 g for 7 min, and resuspended in assay medium (unbuffered Seahorse RPMI, pH 7.4, supplemented with 10 mM glucose, 1 mM pyruvate, and 2 mM glutamine for the Mito assay or with 2 mM glutamine for the glyco assay). 50 µl of the cell suspensions was transferred to Cell-Tak–coated Seahorse XFe96 plates and spun at 120 g (0 break) for 1 min. 2.5 × 10⁵ stimulated B cells or 5 × 10⁵ unstimulated B cells were plated per well. Cell adhesion was evaluated using a light microscope. Before Seahorse run, cells were allowed to rest in a non-CO₂ incubator for 25 min at 37°C followed by addition of 130 µl preheated assay medium and additional 15-min incubation. A Seahorse XFe96 instrument (Agilent) was used for analysis. Additives for metabolic measurements were as follows: for oxygen consumption rate, oligomycin (1 µM), carbonyl cyanide-p trifluoromethoxyphenylhydrazone (1 µM), and rotenone/antimycin A (0.5 µM); for extracellular acidification rate, glucose (10 mM), oligomycin (1 µM), and 2-deoxyglucose (50 mM). Normalization was performed by flow cytometry–based cell counting.

Statistical significance was determined using GraphPad Prism v7.0 with the tests indicated in each figure.

Fig. S1 provides analysis of B cell development and early activation in METTL3-deficient B cells. Fig. S2 provides analysis of memory B cell formation, serum Ig titers, and class-switch recombination in GC B cells, as well as analysis of Cre efficiency and function of METTL3 in PPs. Fig. S3 provides immunohistochemistry analysis of DZ- and LZ-resident GC B cells in the absence of METTL3, as well as supporting results for Figs. 3 and 4. Fig. S4 provides supporting analysis of persistence of Igf2bp3-deficient GC B cells, GSEA of differential gene expression in Igf2bp3- and Ythdf2-deficient GC B cells, Seahorse metabolic analysis of ex vivo–activated Mettl3-deficient B cells, and analysis of MYC expression 3 d after mouse immunization. Fig. S5 provides additional detailed results of m⁶A-IP experiments as well as analysis of overlap between differential gene expression and m⁶A target genes. Fig. S5 provides description of GC identification in smFISH experiments and analysis of Myc mRNA stability in Igf2bp3-deficient B cells. Table S1 lists antibodies used in flow cytometry and immunohistochemistry. Table S2 lists primers used for RT-qPCR. Table S3 lists smFISH probe libraries.

This work was done with critical advice from Dr. Hadas Keren-Shaul from the Genomics Sandbox unit at the Life Science Core Facility of Weizmann Institute of Science. The graphical abstract was created with BioRender.com.

Z. Shulman is supported by European Research Council grant 101001613, Israel Science Foundation grant 1090/18, Azrieli Foundation, and the Morris Kahn Institute for Human Immunology. Z. Shulman is a member of the European Molecular Biology Organization Young Investigator Program. Z. Shulman is supported by grants from the Benoziyo Endowment Fund for the Advancement of Science, the Sir Charles Clore Research Prize, Comisaroff Family Trust, Irma & Jacques Ber-Lehmsdorf Foundation, Gerald O. Mann Charitable Foundation, and David M. Polen Charitable Trust.

Author contributions: A.C. Grenov designed and conducted the experiments, performed data analysis, and wrote the manuscript. L. Moss conducted smFISH experiments and data analysis. S. Edelheit prepared m⁶A-seq pull-downs and libraries. R. Cordiner designed and advised on m⁶A pull-down experiments. D. Schmiedel helped in the design and execution of some of the experiments. A. Biram assisted in single-cell Ig sequencing and analysis. J.H. Hanna provided Mettl3-deficient mice. T.H. Jensen and S. Schwartz designed and supervised m⁶A pull-down experiments. Z. Shulman designed experiments, supervised the study, and wrote the manuscript.