Restriction of memory B cell differentiation at the germinal center B cell positive selection stage

Germinal center (GC) positively selected B cells are primarily fated for GC expansion and plasma cell formation. This study reveals that memory B cell differentiation is restricted in positively selected B cells by the MYC–MIZ1 transcriptional repressor complex.


Introduction
The germinal center (GC) is an antigen-and T cell-dependent reaction in which B cells undergo affinity maturation and differentiation (De Silva and Klein, 2015;Victora and Nussenzweig, 2012). In GCs, B cells cyclically migrate between an area called the dark zone (DZ), which is enriched for proliferating cells and where somatic hypermutation occurs, and an area called the light zone (LZ), in which B cells retrieve antigen from follicular (FO) dendritic cells (FDCs) through their B cell receptor (BCR) and present that antigen to T cells (Allen et al., 2004;Kepler and Perelson, 1993;Victora et al., 2010). T cell help, including CD40L-CD40 engagement, positively selects a fraction (∼5-20%) of LZ B cells, and our work and that of others showed that positive selection critically involves induction of MYC to license cell cycle, after which cells migrate back to the DZ, leading to GC expansion (Calado et al., 2012;Dominguez-Sola et al., 2012;Finkin et al., 2019;Luo et al., 2018;Schwickert et al., 2011). More recently, it was shown that positively selected LZ B cells (LZ MYC + cells) are further composed of plasma cell (PC) precursors and that these also express Myc (Ise et al., 2018).
In addition to expansion in the GC and PC differentiation, LZ B cells also differentiate into memory B cells (MBCs). MBCs are key for long-term protection from reinfection, but how their fate is specified is poorly understood. MBC differentiation was thought to be an unregulated process (Inoue et al., 2018;Smith et al., 2000). Studies have shown, however, that MBCs have, in general, lower antigen affinity compared with LZ B cells fated for GC expansion and PC differentiation (De Silva and Klein, 2015;Shinnakasu et al., 2016;Weisel et al., 2016). Recently, it was found that LZ B cells expressing high levels of the gene encoding the transcription factor BACH2 are favored for MBC differentiation (Shinnakasu et al., 2016) and that quiescent LZ B cells are enriched for MBC precursors (Laidlaw et al., 2017;Suan et al., 2017;Wang et al., 2017). MYC is critically required for cell cycle entry of LZ MYC + cells, and these cells are primarily fated for GC expansion and PC differentiation (Calado et al., 2012;Dominguez-Sola et al., 2012;Ise et al., 2018). We therefore raised the question whether MYC activity in LZ MYC + cells restricts MBC differentiation.
We found that at the positive selection stage GC B cells mostly coexpress MYC and MIZ1. The absence of MYC-MIZ1 complexes impaired cell cycle entry of LZ MYC + cells, reducing GC expansion in a CDKN1A-independent manner, and interfered with PC formation. Notably, derepression of MIZ1 target genes led to a gene expression profile (GEP) alike that of MBCs, and mice lacking MYC-MIZ1 complexes had increased MBC differentiation. We propose that the transcription factors MYC and MIZ1 form a module that regulates the fate of positively selected GC B cells.

Positively selected GC B cells mostly coexpress MYC and MIZ1
We first assessed the expression of Miz1 in GC B cell subpopulations, including LZ MYC + cells, using publicly available data . In contrast to Myc, which was strongly induced in LZ MYC + cells, the expression of Miz1 in these cells was similar to that of LZ B cells negative for MYC (LZ MYC neg cells; Fig. S1 A). Next, we performed immunofluorescence of spleens of wild-type mice at 10 d after immunization with sheep RBCs (SRBCs) and delineated GCs (IgD neg ), DZ and LZ (FDC depleted and enriched, respectively), and MYC expression to identify positively selected GC B cells. We found MIZ1 to be mostly coexpressed with MYC ( Fig. 1, A and B). Collectively, these data indicated that MIZ1 expression is primarily regulated at a posttranscriptional level. MYC expression in GC B cells is synergistically induced by BCR and CD40 coengagement (Fig. S1 B; Luo et al., 2018). However, we found that BCR and CD40 coengagement was insufficient to induce MIZ1 expression in GC B cells stimulated in vitro (Fig. S1,C and D). Thus, although MYC and MIZ1 coexpression occurs in LZ MYC + cells, the requirements for their induction are not identical.
MIZ1 target genes are up-regulated in the absence of MYC-MIZ1 complexes To investigate if MIZ1 target genes are regulated by MYC-MIZ1 complexes in positively selected GC B cells, we used a genetically modified mouse strain (Myc VD ) that carries a MYC mutant encoded at the endogenous Myc locus (Herold et al., 2002;Saba et al., 2011). In these mice, replacement of a valine at position 394 by aspartic acid (V394D, Myc VD ) abrogates MYC-MIZ1 interaction without interfering with the binding of MYC to its obligatory partner, MAX, and hence MYC transcriptional activation (Fig. 2 A;Herold et al., 2002;Saba et al., 2011;Walz et al., 2014). We first investigated the expression of MYC in Myc VD . We immunized Myc VD and wild-type mice (Myc WT ) with SRBC and determined 10 d later the fraction of LZ MYC + cells using intracellular stain and flow cytometry. Myc VD contained a slight but significantly increased fraction of LZ MYC + cells compared with Myc WT (Fig. 2, B and C), whereas MYC expression levels were identical between genotypes (Fig. 2 D). Thus, the absence of MYC-MIZ1 complexes did not impair the induction of MYC nor its expression level.
To investigate if the absence of MYC-MIZ1 complexes altered gene expression we FACS-purified LZ B cells of Myc VD and Myc WT and performed RNA sequencing (RNA-seq) followed by bioinformatic analysis. We first asked whether the GEP of LZ MYC + cells was altered in the absence of MYC-MIZ1 complexes. For that, we used publicly available RNA-seq datasets  and generated signatures of genes down-regulated (LZ MYC + vs. LZ MYC neg DOWN) or up-regulated (LZ MYC + vs. LZ MYC neg UP) in LZ MYC + compared with LZ MYC neg cells. Using gene set enrichment analysis (GSEA), we found that the Myc VD LZ GEP was significantly enriched for the "LZ MYC + vs. LZ MYC neg DOWN" gene signature and that the Myc WT LZ GEP was significantly enriched for the "LZ MYC + vs. LZ MYC neg UP" gene signature ( Fig. 2 E). These data indicated that the absence of MYC-MIZ1 complexes in LZ MYC + cells profoundly altered their GEP, possibly due to the up-regulation of MIZ1 target genes repressed by MYC-MIZ1 complexes in those cells.
To identify differentially expressed genes between LZ Myc VD and LZ Myc WT that are direct targets of MYC-MIZ1 complexes, we generated and analyzed MYC and MIZ1 chromatin immunoprecipitation sequencing (ChIP-seq) in mouse B cells. In agreement with the absence of the repressive activity of MYC-MIZ1 complexes, the LZ Myc VD GEP was significantly enriched for the expression of MIZ1 target genes that are bound by MYC compared with Myc WT (Fig. 2 F). We found 60 MIZ1 target genes bound by MYC that were significantly up-regulated in the LZ of Myc VD compared with that of Myc WT (Fig. S1 E). A fraction of these genes (21 genes) enriched for the Gene Ontology (GO) biological processes related to signal transduction, response to stress, homeostasis, and gene expression (Fig. 2 G). These included genes encoding known tumor suppressors, namely Arhgef1, a RhoA-specific guanine nucleotide exchange factor frequently lost in GC B cell-derived lymphomas (Muppidi et al., 2014); the cytokine-regulated Gadd45g, required in hematopoietic stem cell differentiation and lineage selection (Lu et al., 2001;Thalheimer et al., 2014); and the U3 ubiquitin ligase Sh2b3 that regulates JAK2 stability, which is frequently lost in acute lymphoblastic leukemia (Lv et al., 2017;Perez-Garcia et al., 2013). Genes involved in signal transduction included Itpr1, a Ca 2+ channel required for normal B cell development and function (Tang et al., 2017); Map2k7, which directly activates c-JUN and is known to have an antiproliferative activity in B cells (Sasaki et al., 2001;Tournier et al., 1997); the AKT substrate Akt1s1, which negatively regulates mTOR activity (Sancak et al., 2007); and Vav2, which is critical for humoral immune responses and B cell maturation (Doody et al., 2001). Genes involved in gene expression included Myh9, which regulates BCRmediated antigen acquisition and B cell activation (Hoogeboom et al., 2018); and multiple transcription factors, including Bcl11a, which is essential for B cell development and repressed by BLIMP1 during PC differentiation (Minnich et al., 2016;Yu et al., 2012); Rela, which was shown to be required for GC-derived PC formation and more recently also for MBC differentiation (Heise et al., 2014;Koike et al., 2019); and Mafk, which encodes for the transcription factor MAFK with which BACH2 heterodimerizes to bind target genes in B cells (Huang et al., 2014;Oyake et al., 1996). In summary, we found increased expression of multiple MIZ1 target genes in the absence of MYC-MIZ1 complexes. These data, together with the described expression pattern of MYC and MIZ1 ( Fig. 1), supported a role for MYC-MIZ1 complexes in LZ MYC + cells.
MYC-MIZ1 complexes are required for cell cycle entry of LZ B cells We next tested whether MYC-MIZ1 complexes have a role in the GC reaction at the cell population level. Myc VD had smaller spleens compared with Myc WT ; however, the proportion of mature B cells was identical between genotypes (Fig. S1 F). 10 d after SRBC immunization, we observed a significant reduction in the fraction and number of GC B cells in Myc VD compared with Myc WT (Fig. 3, A and B). Histological analysis showed a significant reduction in the size of GC clusters in Myc VD compared with Myc WT (Fig. 3, C and D). However, the number of GC foci was identical between genotypes (Fig. 3 D). We also observed a small but significant increase in the LZ/DZ proportion in Myc VD compared with Myc WT (Fig. 3, E and F), whereas the proportion of IgG1-and IgM-expressing GC B cells was similar between genotypes (Fig. S1 F). These data showed that MYC-MIZ1 complexes were required for GC expansion, but not GC formation. In agreement, analysis of an earlier time-point after SRBC immunization (day 5), when GC clusters are formed but before massive expansion (Calado et al., 2012), Myc VD and Myc WT displayed a similar fraction of GC B cells (Fig. 3 G). In accordance with the knowledge that LZ MYC + cells contain PC precursors we found a reduced fraction and number of PCs in Myc VD , which was proportional to the reduction observed for GC B cells (Fig. S1, G-I). These phenotypes did not seem to arise because of impaired T cell help. First, MYC expression levels, which are in part regulated by CD40 signaling (Luo et al., 2018), were similar between genotypes (Fig. 2, B-D). Second, the number of PD1 high CXCR5 high CD4 + T cells, which mostly represents GC T FO helper cells (Tfh cells), was proportional to that of GC B cells in Myc VD and Myc WT (Fig. S1, J and K). And third, genes up-regulated by CD40 signaling were found to be enriched in the Myc VD LZ GEP compared with Myc WT (Fig. S1 L).
We next tested whether absence of MYC-MIZ1 complexes had an impact on the survival of GC B cells. However, we did not observe statistically significant differences in apoptosis between genotypes, although a trend toward a reduced fraction of active caspase 3 + cells was noticeable in the LZ of Myc VD compared with Myc WT (Fig. 3 H). Suggesting an altered cell cycle profile, GSEA of the GO term "CELL_CYCLE_ARREST" gene signature revealed a significant enrichment in the Myc VD LZ GEP compared with Myc WT , whereas, in contrast, the GO term "G1-S_PHASE_TRANSITION" gene signature was enriched in Myc WT LZ GEP compared with Myc VD (Fig. 3 I). To determine whether the activity of MYC-MIZ1 complexes was required for the cell cycle of GC B cells, we performed 5-ethynyl-29-deoxyuridine (EdU) pulse experiments. We found impaired cell cycle engagement of Myc VD LZ B cells compared with Myc WT , whereas the fraction of DZ B cells engaged in cell cycle was identical between genotypes (Fig. 3, J and K). The cell cycle defect was not due to altered mTORC1 activity (Ersching et al., 2017), as LZ MYC + cells of Myc VD and Myc WT displayed similar levels of phospho-S6 kinase (Fig. 3, L and M). We also did not observe altered expression of cyclin genes Ccnd2 (a target of MYC; Bouchard et al., 1999;Calado et al., 2012) and Ccnd3 in the absence of MYC-MIZ1 complexes (Cato et al., 2011;Peled et al., 2010;Fig. S2 A). Notably, the expression of the MIZ1 target genes Cdkn1a, Cdkn2a, and Cdkn2b, known to be repressed by MYC-MIZ1 complexes in cancer cells (Walz et al., 2014), were also similar between the LZ of Myc VD and Myc WT (Fig. S2 B). In agreement, ablation of Cdkn1a was insufficient to rescue the Myc VD phenotype (Fig. 4, A-D), including the cell cycle defect of LZ B cells (Fig. 4 E). The expression of Cdkn1b was, however, increased in the LZ of Myc VD compared with that of Myc WT (Fig. S2 C), but neither we (Fig. S1 E) nor others (Walz et al., 2014) identified Cdkn1b as a direct MIZ1 target.
We next asked if the absence of MYC-MIZ1 complexes similarly altered the cell cycle of non-GC B cells in vitro. We first tested whether stimulation of naive B cells, including BCR and CD40 coengagement, induced MIZ1. Whereas MYC expression was already increased at 4 h after stimuli (Luo et al., 2018), MIZ1 expression was only increased after 16 h (Fig. S2,D and E). The absence of MYC-MIZ1 complexes did not significantly impact cell proliferation of in vitro-activated naive cells (Fig. S2 F). We next determined the expression of MYC and MIZ1 at the very early stages of the (pre-)GC reaction. Contrary to positively selected B cells in mature GCs, coexpression of MYC and MIZ1 was seldom observed in (pre-)GC B cells (Fig. S2, G-I). Overall these data indicated that the requirements for MIZ1 induction varied according to B cell stage and that MYC-MIZ1 complexes played a function in GC B cell expansion rather than formation. Specifically, in LZ MYC + cells the absence of MYC-MIZ1 complexes dissociated MYC expression from cell cycle engagement.

MIZ1 target genes are enriched in MBCs
The impaired cell cycle engagement of LZ MYC + cells of Myc VD and the knowledge that MBC precursors are found within quiescent LZ B cells (Laidlaw et al., 2017;Suan et al., 2017;Wang et al., 2017) led us to investigated if the LZ Myc VD GEP displayed a shift toward that of MBCs. We first generated an MBC GEP by performing RNA-seq of FACS-purified MBCs from day 10 SRBCimmunized wild-type mice. Using these data, we generated signatures of genes up-regulated (MBC vs. LZ UP) or downregulated (MBC vs. LZ DOWN) in MBCs compared with LZ B cells. We found that the LZ Myc VD GEP was significantly enriched for the "MBC vs. LZ UP" gene signature, whereas the LZ Myc WT GEP was significantly enriched for the "MBC vs. LZ  DOWN" gene signature (Fig. 5 A). To investigate whether this enrichment was the consequence of decreased cell cycle activity in the Myc VD LZ we performed RNA-seq of FACS-purified FO B cells, known to be quiescent, and generated signatures of genes up-regulated (MBC vs. FO UP) or down-regulated (MBC vs. FO DOWN) in MBCs compared FO B cells. Similar to the previous analysis, we found that the Myc VD LZ GEP was significantly enriched for the "MBC vs. FO UP" gene signature, whereas the LZ Myc WT GEP was significantly enriched for the "MBC vs. FO DOWN" gene signature (Fig. 5 A). Identical results were observed when analyzing "MBC vs. FO" signatures using GEP datasets generated by others ( Fig. S3 A), in agreement with the significant overlap between the MBC GEP generated in this work and that of others (Fig. S3 B; Kaji et al., 2012;Weisel et al., 2016). We also found a significant enrichment in the Myc VD LZ GEP for genes up-regulated in MBC vs. GC B cells of human origin (human MBC vs. GC_UP), whereas the Myc WT LZ GEP was enriched for genes down-regulated in MBC vs. GC B cells (human MBC vs. GC_DOWN;Fig. 5 B;Luckey et al., 2006). To further validate our analysis, we used a publicly available RNA-seq dataset comparing LZ B cells expressing high Bach2 levels (LZ Bach2 hi ) that are enriched for MBC precursors and LZ B cells expressing low Bach2 levels (LZ Bach2 low ) that are enriched for cells fated for GC expansion and PC differentiation (Shinnakasu et al., 2016). We found that the Myc WT LZ GEP was significantly enriched for the gene signature "LZ Bach2 hi vs. LZ Bach2 low DOWN" that is associated with GC expansion and PC fates (Fig. 5 A). In contrast, Myc VD LZ GEP was significantly enriched for the gene signature "LZ Bach2 hi vs. LZ Bach2 low UP" that is associated with MBC precursors (Fig. 5 A). In agreement with Bach2 levels being down-regulated by T cell help (Shinnakasu et al., 2016), we found that the expression of Bach2 was significantly reduced in LZ MYC + compared with LZ MYC neg cell subsets of wild-type mice (Fig. S3 C). However, Bach2 levels were not different between LZ B cells of Myc VD and Myc WT (Fig. S3 D), further suggesting that CD40 signaling is not impaired in Myc VD compared with Myc WT . Overall, these analyses indicate that the absence of MYC-MIZ1 complexes in LZ MYC + cells altered the GEP of LZ B cells in part through the enrichment of genes expressed in MBCs.
Next, we investigated whether the identified 60 MIZ1 target genes bound by MYC and significantly up-regulated in the LZ of Myc VD compared with that of Myc WT (Fig. 2 G and Fig. S1 E) were enriched in the MBC GEP. 41 out of the 60 MIZ1 target genes (68%) had increased expression in MBCs vs. LZ B cells (Fig. 5,C and D), and the majority of these 41 genes (30/41, 73%) had reduced expression in LZ MYC + compared with LZ MYC neg cells (Fig. S3 D). These data reflected the repressive activity of MYC-MIZ1 complexes in LZ MYC + cells and that these complexes directly repressed genes enriched in MBCs.

MYC-MIZ1 complexes restrict MBC differentiation
To assess whether MBC formation was altered in the absence of MYC-MIZ1 complexes, we initially immunized Myc VD and Myc WT with SRBC ( Fig. 6 A). The fraction of IgM + CD273 + MBCs within B cells was significantly increased in Myc VD compared with Myc WT at day 10 after SRBC immunization, albeit the absolute cell number not being statistically different between genotypes (Fig. 6 B). Nevertheless, considering that Myc VD had reduced GC B cell numbers compared with Myc WT , a significantly increased IgM + CD273 + MBC to IgM + GC B cell ratio was observed, indicating that Myc VD produced more IgM + CD273 + MBCs than Myc WT (Fig. 6 B). The fraction of IgG1 + CD273 + MBCs within B cells was significantly increased at day 5 after immunization in Myc VD compared with Myc WT , but only a trend in terms of absolute cell numbers was observed (Fig. 6 C). However, at day 10 after SRBC immunization, the difference was clear, as both the fraction within B cells and the absolute number of IgG1 + CD273 + MBCs were significantly increased in Myc VD compared with Myc WT (Fig. 6 C). We concluded that the absence of MYC-MIZ1 complexes in LZ MYC + cells increased MBC differentiation.
It is acknowledged that MBCs can be generated by both GCdependent and GC-independent paths (Blink et al., 2005; Chan  Inamine et al., 2005;Takemori et al., 2014;Toyama et al., 2002). We therefore investigated the kinetics of MBC differentiation in Myc VD and Myc WT by performing EdU pulse-chase experiments in which "3-d labeling windows" are generated through EdU injection at specific time points after immunization (Weisel et al., 2016). The labeling windows chosen with respect to the immunization time point were: (1) days 0-2, which encompass the earlier stage of the GC reaction; (2) days 5-7, which encompasses critical time points of GC expansion; and (3) days 12-14 for a later GC stage (Calado et al., 2012). In these experiments, analysis of mice at day 15 after SRBC immunization revealed a significantly increased number of EdU + IgM + CD273 + MBCs and to a greater degree for EdU + IgG1 + CD273 + MBCs in Myc VD compared with Myc WT at the day 5-7 labeling window (Fig. 6, D and E). These data showed that the absence of MYC-MIZ1 complexes in LZ MYC + cells originated a marked increase of MBC formation at critical time points of GC expansion. We also found that the fraction of IgG1 + LZ B cells of Myc VD was significantly enriched for cells expressing the MBC precursor marker CCR6 (Suan et al., 2017) compared Myc WT (Fig. 6, F and G), whereas the fraction of CCR6 + cells within IgG1 + CD273 + MBCs was identical between genotypes (Fig. 6, H and I). These data indicated that MBC precursors were significantly increased within IgG1 + LZ B cells of Myc VD compared Myc WT . Previous work in which GC B cell formation was abrogated demonstrated that the expression of CD73 marks GC-derived MBCs, whereas MBCs lacking CD73 can be of either GC or non-GC origin (Anderson et al., 2007;Kaji et al., 2012;Taylor et al., 2012). We found that ∼45% of the IgM + CD273 + MBCs generated in Myc VD were CD73 + compared with ∼50% in Myc WT (Fig. 6 J), whereas ∼40% of IgG1 + CD273 + MBCs in Myc VD were CD73 + compared with ∼35% in Myc WT (Fig. 6 K). These data indicated a similar origin of MBCs between genotypes, and we concluded that the increase in MBC differentiation observed in the absence of MYC-MIZ1 complexes occurred primarily through a GC-dependent path. We also investigated the distribution of known MBC surface marker combinations (Anderson et al., 2007;Tomayko et al., 2010) within IgG1 + MBCs in Myc VD compared with Myc WT (Fig. S3, E-H). We found significantly increased fractions of CD73 + CD80 neg and CD80 neg CD35 + MBCs in Myc VD , whereas a decreased fraction of CD80 + CD35 neg was observed in these mice (Fig. S3, F-H). Expression of CD80 was previously associated with a higher number of somatic mutations in the IgH variable region, whereas that of CD35 with reduced mutational load (Anderson et al., 2007). These data therefore indicated that the enlarged IgG1 + MBC pool of Myc VD is composed of cells with lower affinity toward the antigen, consistent with increased MBC differentiation at critical time points of GC expansion (Fig. 6 E), when GC antigen affinity is likely to be low.
Reduced affinity of the MBC pool in the absence of MYC-MIZ1 complexes Following the analysis using SRBC immunization, we wanted to determine whether the absence of MYC-MIZ1 complexes impacted antigen affinity maturation. To study BCR antigen affinity, we generated compound mutant Myc VD and Myc WT carrying the SWHEL allelic system, in which B cells express a transgenic BCR recognizing hen egg lysozyme (HEL; Phan et al., 2003). In this system, the BCR affinity to HEL is very high (2 × 10 10 M −1 ), but affinity maturation can be studied by immunizing mice with a mutant version of HEL called HEL 3X , for which SWHEL BCR has much lower affinity (1.5 × 10 6 M −1 ; Paus et al., 2006). According to established protocols , we purified B cells from Myc VD SWHEL and control Myc WT SWHEL and transferred them into congenic CD45.1 mice, after which recipient mice were immunized with HEL 3X (Fig. 7 A). We found at all time points of analysis (days 8, 10, and 15) a reduction in GC B cells and an increase of CD273 + MBCs within B cells derived from Myc VD SWHEL compared with Myc WT SWHEL (Fig. 7, B-D). These data reproduced the phenotypes observed in Myc VD (Fig. 3 and Fig. 6) and demonstrated that the reduced GC expansion and  (D) and IgG1 + CD273 + (E) MBC production determined by EdU pulse-chase experiments at the indicated labeling windows. Cumulative FACS analyses of splenic MBCs analyzed as in A at day 15 after SRBC immunization (SRBCi). (F) Representative flow cytometry of CD23 and CCR6 expression within splenic IgG1 + LZ B cells of Myc WT and Myc VD at day 10 after SRBC immunization. (G) Cumulative data for the fraction of CD23 + CCR6 + within splenic IgG1 + LZ B cells from Myc WT and Myc VD , analyzed as in F. (H) Representative flow cytometry of CD23 and CCR6 expression within splenic IgG1 + CD273 + MBCs of Myc WT and Myc VD at day 10 after SRBC immunization. (I) Cumulative data for the fraction of CD23 + CCR6 + within splenic IgG1 + CD273 + MBCs from Myc WT and Myc VD , analyzed as in H. (J) Cumulative data for the fraction of CD73 + cells within splenic IgM + CD273 + MBCs analyzed as in A at day 10 after SRBC immunization. (K) Cumulative data for the fraction of CD73 + cells within splenic IgG1 + CD273 + MBCs analyzed as in A at day 10 after SRBC immunization. Each symbol (B and C: day 0 Myc WT n = 5, Myc VD n = 4; day 5 Myc WT n = 4, Myc VD n = 4; day 10 Myc WT n = 22, Myc VD n = 19; G and I: Myc WT n = 15, Myc VD n = 16; J and K: Myc WT n = 5, Myc VD n = 5) represents an individual mouse; small horizontal lines show median, minimum, and maximum values or mean and SEM (D and E). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001 (unpaired two-tailed Student's t test). Data are representative of two (B and C at days 0 and 5; D and E) or three (B and C at day 10; G-K) independent experiments. ns, not significant.
increased MBC differentiation was due to the specific absence of MYC-MIZ1 complexes in B cells.
To determine the BCR somatic mutation pattern and affinity maturation to HEL 3X , we FACS-purified GC cells and MBCs at day 8 and 15 after HEL 3X immunization, followed by cloning and sequencing of the IgH variable region. The replacement of a tyrosine at position 53 to aspartic acid in SWHEL IgH (Y53D) leads to ∼100-fold increased affinity for HEL 3X (Phan et al., 2003). At day 8 after HEL 3X immunization, ∼9% of Myc VD S-WHEL GC B cells had the Y53D amino acid change compared with ∼24% of Myc WT SWHEL cells (Fig. 7 E and Fig. S4 A). The average number of amino acid substitutions was also reduced in Myc VD SWHEL GC B cells (average, 1.6/sequence) compared with Myc WT SWHEL (average, 2.1/sequence; Fig. S4, A-C). However, the number of nucleotide mutations in the SWHEL IgH of GC B cells was similar between genotypes ( Fig. S4 D), suggesting that somatic hypermutation per se was not impaired. The observed differences in GC B cells at day 8 were nevertheless transient given that at day 15 after HEL 3X immunization no significant differences were found between genotypes (Fig. 7, E-G; and Fig. S4, A-C). MBCs at day 8 after HEL 3X immunization of either genotype did not display Y53D mutations (Fig. 7 H  and Fig. S4 E). However, the average number of amino acid substitutions was lower in Myc VD SWHEL MBCs (average, 1/sequence) compared with Myc WT SWHEL (average, 1.4/sequence; Fig. S4, E-G). Still, only small differences were found with respect to nucleotide mutations, with ∼80% of Myc VD SWHEL MBCs carrying a somatically mutated BCR compared with ∼85% of Myc WT SWHEL MBCs (Fig. S4 H). At day 15 of analysis, the absence of MYC-MIZ1 complexes significantly impacted BCR antigen affinity of MBCs, with ∼60% of Myc VD SWHEL MBCs cells carrying Y53D mutations compared with ∼87% of Myc WT SWHEL cells (Fig. 7 H and Fig. S4 E). This was accompanied by a reduced number of amino acid substitutions in  E-G). Nevertheless, the vast majority of Myc VD SWHEL MBCs (∼90%) displayed a somatically mutated BCR (Fig. 7 J), further demonstrating that the observed increase in MBC differentiation when MYC-MIZ1 complexes are absent occurs primarily through a GC-dependent path. To investigate if the absence of MYC-MIZ1 complexes altered antigen affinity of the PC pool, we evaluated the binding of IgG1 antibodies in the serum of mice to HEL 3X at days 8 and 15 after immunization (Fig. S4 I). No significant differences were found between genotypes, indicating that the absence of MYC-MIZ1 complexes primarily impacts the antigen affinity of the MBC pool. In summary, the absence of MYC-MIZ1 complexes had a lasting impact in the MBC pool by increasing its size and decreasing its affinity for the antigen.
MYC-MIZ1 complexes restrict MBC differentiation when affinity-based selection is absent LZ B cells bearing BCRs with lower antigen affinity are favored to differentiate into MBCs (Shinnakasu et al., 2016). Therefore, it was possible that increased MBC differentiation in Myc VD was primarily the consequence of the transitory delay in GC B cell affinity maturation in these mice. To test this hypothesis, we used the SWHEL system and immunized recipient mice with HEL, for which BCR affinity is very high (2 × 10 10 M −1 ; Paus et al., 2006), effectively curtailing antigen affinity-based selection (Fig. 8 A). Similar to the results using HEL 3X immunization (Fig. 7, A-D), the fraction of GC B cells was reduced within Myc VD SWHEL B cells, whereas the fraction of CD273 + MBCs was increased, compared with Myc WT SWHEL at all time points of analysis (days 8 and 15; Fig. 8, B-D). We also performed an experiment using mice deficient for the enzyme Aicda that is critically required for somatic hypermutation and class-switch recombination (Muramatsu et al., 2000). Aicda-deficient Myc VD , similarly to Myc VD (Fig. 3), displayed a significant reduction in the fraction of GC B cells compared with mice with Aicda deficiency alone (Fig. S5, A and B). We also observed that LZ B cells of Aicda-deficient Myc VD were impaired in cell cycle engagement compared with Aicda-deficient mice, whereas the fraction of DZ B cells engaged in cell cycle was not different between genotypes (Fig. S5 C). Lastly, Aicda-deficient Myc VD displayed a significantly increased fraction of MBCs compared with mice with Aicda deficiency alone (Fig. S5, D and E). We concluded that MYC-MIZ1 complexes in LZ MYC + cells restricted MBC differentiation even when affinity-based selection was absent.

Discussion
MBCs are key for long-term protection from reinfection. However, the mechanisms underlying MBC differentiation are unclear. Here, we found that LZ MYC + cells mostly coexpressed the transcription activators MYC and MIZ1. Complexes formed by MYC and MIZ1 in these cells were required for repression of MIZ1 target genes, effective cell cycle engagement of LZ B cells, GC expansion, and PC formation. Notably, most of the MIZ1 target genes repressed by MYC-MIZ1 complexes were enriched in MBCs, and the absence of MYC-MIZ1 complexes increased MBC differentiation. Whether these events are interdependent or independent, the first suggesting that restriction of MBC differentiation in LZ MYC + cells is required for effective GC expansion and PC formation, remains to be determined. The signals driving MIZ1 expression in LZ MYC + cells also need further investigation, because in contrast to MYC (Luo et al., 2018), ex vivo BCR and CD40 coengagement was not sufficient to induce MIZ1 in GC B cells. Further, the discrepancy between Miz1 gene and protein expression argues for a tight regulation at the posttranscriptional level. HUWE1 is an E3 ubiquitin ligase that ubiquitinates MIZ1, triggering its proteasomal degradation (Yang et al., 2010). However, previous analysis of HUWE1deficient mice suggest a limited function for this E3 ubiquitin ligase in the GC reaction (Hao et al., 2012).
MYC-MIZ1 complexes were not equally required in all LZ MYC + cells, given that apparently a fraction entered cell cycle normally. These data and of others (Ise et al., 2018) highlight the need to investigate populational diversity within LZ MYC + cells. We also observed that cellular proliferation of activated B cells in vitro was mostly unimpaired. Thus it is possible that MYC-MIZ1 complexes are primarily required at specific B cell stages and/or to counteract autocrine and/or paracrine (i.e., microenvironmental) antiproliferative signals (van Riggelen et al., 2010).
Most of the identified MIZ1 target genes up-regulated in the absence of MYC-MIZ1 complexes were enriched in MBCs. Although for many a function in B cells is unknown, for others (namely Arhgef1, Gadd45g, Sh2b3, and Map2k7), tumorsuppressive and/or antiproliferative activity was described (Lu et al., 2001;Lv et al., 2017;Muppidi et al., 2014;Perez-Garcia et al., 2013;Thalheimer et al., 2014). This knowledge is in agreement with work demonstrating that quiescent B cells in the LZ are enriched for MBC precursors (Laidlaw et al., 2017;Suan et al., 2017;Wang et al., 2017). Thus, MIZ1 transcriptional activity may promote a quiescent state, which in turn permits MBC differentiation. Such activity is restrained by MYC-MIZ1 complexes in LZ MYC + cells. Whether MIZ1 transcriptional activity is itself required for MBC differentiation needs further investigation.
The finding that MIZ1 targets genes with B cell tumor suppressor and/or antiproliferative function may be relevant for GC B cell lymphomagenesis. It is tempting to speculate that interference with MBC differentiation could be oncogenic, similar to what we and others demonstrated for PC differentiation (Calado et al., 2010;Mandelbaum et al., 2010;Zhang et al., 2015).
In cancer, Cdkn1a is a well-characterized target of MYC-MIZ1 complexes (Herold et al., 2002;Walz et al., 2014). However, Cdkn1a expression was unchanged in LZ B cells in the absence of MYC-MIZ1 complexes. Previous work showed that BCL6 can form a transcriptional repressor complex with MIZ1 that represses CDKN1A expression in GC-derived lymphoma cell lines (Basso and Dalla-Favera, 2010;Phan et al., 2005). Also, others have also shown that EZH2 is required in GC B cells for repression of Cdkn1a expression (Béguelin et al., 2017). Thus, it is possible that redundancy in the regulation of Cdkn1a expression exists in GC B cells. In contrast, we found increased expression of Cdkn1b in the absence of MYC-MIZ1 complexes. However, and although our analysis and that of others have not identified Cdkn1b as a direct MIZ1 target (Walz et al., 2014), this has been suggested by other studies (Basu et al., 2009;Yang et al., 2001), urging further investigation.
LZ B cells with high Bach2 expression are favored to differentiate into MBCs (Shinnakasu et al., 2016). However, Bach2 expression was not altered in the absence of MYC-MIZ1 complexes. Bach2 expression is inversely correlated to the strength of T cell help (Shinnakasu et al., 2016). On the contrary, Myc is induced upon positive selection downstream of T cell help (Luo et al., 2018), and MYC levels were not altered by the absence of MYC-MIZ1 complexes, suggesting that T cell help was not affected. As a consequence, changes in Bach2 expression levels would not be expected in the absence of MYC-MIZ1 complexes. This does not contradict the role of Bach2 in the regulation of MBC differentiation (Shinnakasu et al., 2016). First, Bach2 may be required for MBC differentiation of a different LZ B cell subset; second, we found that MYC-MIZ1 complexes repressed the expression of the gene encoding the MAFK transcription factor, with which BACH2 is bound at most target genes (Huang et al., 2014;Oyake et al., 1996). Thus, given that strong T cell help reduces Bach2 expression but does not extinguish it (Shinnakasu et al., 2016), increased Mafk expression in the absence of MYC-MIZ1 complexes could favor MBC differentiation.
Increased MBC differentiation, together with impaired GC B cell expansion, was also found when LZ B cells cannot sense IL21 and when LZ to DZ migration is abrogated by impaired CXCR4 signaling (Bannard et al., 2013;Barinov et al., 2017;Linterman et al., 2010;Zotos et al., 2010). In both scenarios, antigen affinity was reduced, possibly favoring MBC differentiation (Shinnakasu et al., 2016;Weisel et al., 2016). The absence of MYC-MIZ1 complexes led only to a transient delay in GC B cell affinity. Thus, and although we cannot exclude a contribution of this phenomenon toward increased MBC differentiation, we found that MYC-MIZ1 complexes restricted MBC differentiation even when affinity-based selection was absent. The MBC phenotype in IL21/IL21R deficiency was attributed to a GCindependent path given that most MBCs had unmutated BCRs, whereas for CXCR4 deficiency, the involvement of GCs was suggested (Bannard et al., 2013;Linterman et al., 2010;Zotos et al., 2010). The latter conclusion was based on the detection of CD73 expression and MBC BrdU incorporation in pulse-chase experiments (Bannard et al., 2013). We found that increased MBC differentiation in the absence of MYC-MIZ1 complexes occurred primarily through a GC-dependent path: (1) MYC-MIZ1 complexes were required for GC expansion rather than formation; (2) an increased number of EdU + MBCs was formed during critical time points of GC expansion; (3) the fraction of IgG1 + CD73 + MBCs and LZ MBC precursors (CCR6 + LZ B cells) was increased in Myc VD compared with Myc WT ; and (4) the vast majority (∼90%) of MBCs in Myc VD display somatically mutated BCRs.
Our work uncovered that MYC-MIZ1 complexes in LZ MYC + cells are required for effective GC expansion and PC formation and to restrict MBC differentiation. Until now, a physiological function for MYC-MIZ1 complexes was unknown (Wiese et al., 2013). MIZ1, similarly to MYC, is required for early B and T cell development; however, these functions were shown to be independent of MYC-MIZ1 complexes (Douglas et al., 2001;Kosan et al., 2010;Saba et al., 2011;Vallespinós et al., 2011). MIZ1 homologues and the conservation of a valine in position 394 in MYC, critical for the interaction with MIZ1, are found only in vertebrates (Conacci-Sorrell et al., 2014). Thus, compared with other MYC network members, like MAX and MXD proteins, MYC-MIZ1 complexes are a late addition in evolution (Conacci-Sorrell et al., 2014). Interestingly, such evolutionary timeframe is similar to that of AICDA (Conticello et al., 2005).
MYC-MIZ1 complexes per se are not essential for life (Wiese et al., 2013). As a consequence, this protein complex is a viable candidate for intervention in cancer and vaccination. With respect to the latter, an increased MBC pool size with lower affinity for the primary immunizing antigen, as observed in the absence of MYC-MIZ1 complexes, may permit the recognition of similar but different antigens and further affinity maturation (Bannard and Cyster, 2017;Inoue et al., 2018;McHeyzer-Williams et al., 2015;Mesin et al., 2016;Takahashi and Kelsoe, 2017;Tas et al., 2016). This could be important in vaccination for protection against pathogenic substrains and evolved mutants (Inoue et al., 2018;Victora and Wilson, 2015). Thus, interventions that modulate the activity of MYC-MIZ1 complexes may tailor the GC response to meet individual humoral memory requirements for infection control and prevention.

Mice
Myc VD , Cdkn1a KO, AID-Cre-ERT2, and the transgenic SWHEL allelic BCR system mouse strains have been previously described (Brugarolas et al., 1995;Calado et al., 2012;Casola et al., 2006;de Alboran et al., 2001;Dogan et al., 2009;Phan et al., 2003;Saba et al., 2011). Mice were maintained on the C57BL/6 background and bred at the Francis Crick Institute biological resources facility under specific pathogen-free conditions. Animal experiments were performed in accordance with national and institutional guidelines for animal care and approved by The Francis Crick Institute biological resources facility strategic oversight committee (incorporating the Animal Welfare and Ethical Review Body) and by the Home Office, UK.
Histology, immunohistochemistry, and immunofluorescence Spleens were fixed with 10% neutral buffered formalin (Thermo Fisher) and embedded in paraffin. Sections were stained with hematoxylin (Sigma) and biotinylated peanut agglutinin (Vector). Images were acquired with Zeiss Axio Scan.Z1 Slide Scanner and visualized with Photoshop (v12.1; Adobe). ImageJ (v2.0.0) was used to quantify GC B cell area and number of GCs per spleen section. For immunofluorescence, spleens were embedded in optimum cutting temperature compound (Sakura) and flash frozen in liquid nitrogen. 8-to 30-µM tissue sections were cut on an OTF5000 cryostat (Bright Instruments), fixed in 4% paraformaldehyde (Thermo Fisher), permeabilized and blocked in PBS with 0.3% Triton X-100 (Sigma) and 10% normal goat serum (Sigma). Samples were stained with antibodies in PBS containing 0.3% Triton X-100 with 1% bovine serum albumin (Thermo Fisher). Every incubation was followed by three washes with PBS. To prevent cross-reactivity, samples were blocked with PBS containing 10% normal rabbit serum (Sigma). Images were acquired with a Leica Sp5 confocal microscope, using sequential acquisition between frames with 405-, 488-, 555-, and 647-nm laser excitations. Images were analyzed with Imaris software (Bitplane); cells were automatically identified by the software based on nuclear Hoechst 33342 staining.

In vitro cell stimulation and transduction
Splenic naive B cells were purified by CD43 depletion (Miltenyi). GC B cells were purified as previously described (Luo et al., 2018). Naive and GC B cells were cultured in vitro in the presence of 10 µg/ml (anti-IgM + anti-IgG; Jackson ImmunoResearch) and/or 20 µg/ml anti-CD40 antibody (FGK45; Bio X Cell) with or without 25 ng/ml IL-4. For cell proliferation studies, naive B cells were stained with CellTrace Violet (Thermo Fisher) according to the manufacturer's protocol. HEK293T cells were transiently transduced with a MIZ1-IRES-GFP expressing vector or GFP control (VectorBuilder) using Lipofectamine (Invitrogen). Cells were analyzed 48 h after transduction.

Flow cytometry and ELISA
Single-cell suspensions of spleen were prepared in FACS buffer (2% FBS and 2 mM EDTA, in PBS; Gibco) and treated with Gey's solution for erythrocyte lysis. Single-cell suspensions were stained with antibodies. The use of biotinylated antibodies was followed by incubation with fluorochrome labeled streptavidin (1/200 dilution). For the analyses of SWHEL mice, HEL-binding cells were stained with 50 ng/ml HEL (Sigma) followed by HyHEL9-A647 (R. Brink). Dead cells were excluded using Zombie NIR Fixable Viability Kit (BioLegend). For detection of EdU incorporation, cells were fixed for 15 min at room temperature in 4% paraformaldehyde (Thermo Fisher) after surface marker and viability dye staining. Fixation was followed by Click-iT EdU A647 flow cytometry assay kit or Click-iT Plus EdU Alexa Fluor 647 flow cytometry assay kit (Life Technologies) as indicated by supplier. For the detection of cleaved caspase-3, samples were fixed for 15 min at room temperature in 4% paraformaldehyde (Thermo Fisher) after surface marker and viability dye staining, followed by intracellular staining with BD Cytofix/Cytoperm staining kit (BD Biosciences), as per manufacturer's specifications. The detection of the transcription factor was performed after surface marker and viability dye staining with True-Nuclear Transcription Factor Buffer Set (BioLegend), following supplier's instructions. To prevent cross-reactivity, samples were blocked with 10% normal rabbit serum. Samples were acquired on an LSR-Fortessa (BD Biosciences) with FACS-Diva software (BD Biosciences), and data were analyzed with FlowJo software (v10.3; Tree Star). MBCs were FACS purified using the following antibody panel: CD273 + CD138 neg B220 + CD19 + CD38 high IgD low . Cell populations were defined by the following surface markers: LZ GC B cells (B220 + CD19 + CD38 low CD95 high CXCR4 low CD86 high ), DZ GC B cells (B220 + CD19 + CD38 low CD95 high CXCR4 high CD86 low ), PCs (CD19 low CD138 + ), and IgM + and IgG1 + MBCs (B220 + CD19 + CD38 high IgD low CD273 + ); FO B cells (AA4.1 neg B220 + CD19 + CD38 high IgD + CD21 high CD23 high ). Anti-HEL 3X IgG1 antibodies in sera were analyzed by ELISA as previously described (Phan et al., 2003.

IgH somatic mutation analysis
IgG1 + CD273 + HEL + MBCs and IgG1 + HEL + GC B cells from SWHEL mice were sorted using a FACSAria III or FACSAria Fusion (BD Biosciences). Cells were sorted into proteinase K (Qiagen) containing buffer, as previously described (Brink et al., 2015;Paus et al., 2006). The SWHEL Ig heavy-chain variable region exon was PCR amplified using Taq DNA polymerase (Thermo Fisher) with the primers 59-GTTGTAGCCTAAAAGATG ATGGTG-39 and 59-GATAATCTGTCCTAAAGGCTCTGAG-39. The primary PCR product was further amplified with the nested primers 59-TTGTAGCCTAAAAGATGATGGTGTTAAGTC-39 and 59-CAACTTCTCTCAGCCGGCTC-39. Nested PCR product was isolated using QIAquick Kit (Qiagen) and cloned into pGEM-T Easy vector (Promega), as per the manufacturer's instructions. DH5α-competent bacteria (Life Technologies) were transformed and single colonies were grown in 96 deep-well culture blocks (Macherey-Nagel). DNA was extracted automatically with Biorobot Universal System (Qiagen) using Nucleospin 96 Plasmid kits (Macherey-Nagel) and Sanger-sequenced with T7, SP6, M13 forward, and/or M13 reverse primers. Sequenced DNA was run against VBASE2 database (Retter et al., 2005). Using HyHEL10 sequence as a reference, mature VDJ DNA sequences were analyzed for the presence of somatic mutation events, and the translated protein sequences were examined for amino acid substitutions.

Gene expression analysis and ChIP-seq
For the gene expression profiling of Myc WT and Myc VD LZ GC B cells (CD138 neg , B220 + , CD19 + CD38 low , CD95 high , CXCR4 low CD86 high ), Myc WT MBCs (CD138 neg , B220 + , CD19 + , AA4.1 neg , CD38 + , IgD neg , CD273 + ), and Myc WT FO B cells (AA4.1 neg , B220 + , CD19 + , CD38 high , IgD + , CD21 high , CD23 high ) were sorted by flow cytometry at day 10 after immunization with SRBCs using a FACSAria III or a FACSAria Fusion (BD Biosciences). Cells were sorted in RTL Buffer Plus (Qiagen) containing 1% β-mercaptoethanol (Sigma) and RNA purified using AllPrep DNA/RNA Mini and Micro Kits (Qiagen) following the manufacturer's instructions. RNA-seq was performed at the Francis Crick Institute advanced sequencing facility. The RSEM package (v1.3.0; Li and Dewey, 2011) and STAR (v2.5.2a;Dobin et al., 2013) were used to align reads to the mouse mm10 transcriptome, taken from Ensembl (vGRCm38) and available at the University of California, Santa Cruz (https://genome.ucsc.edu). For RSEM, all parameters were run as default using the "-forward-prob 0" option for a strand-specific protocol. Differential expression analysis was performed with DESeq2 package (v.1.12.4;Love et al., 2014) within R (v3.3.1; R Core Team, 2008; http://www.R-project.org). Genes were considered to be differential expressed if P < 0.05. GSEA was performed using GSEA software (v2.2.3) from the Broad Institute (Subramanian et al., 2005). All analyses for RNA-seq-generated expression profiles were done with ranked gene lists using Wald statistics. For ChIP, splenic B cells from C57BL/6 mice were purified using antimouse CD43 (Ly-48) MicroBeads (Miltenyi Biotec) depletion and treated with Gey's solution for erythrocyte lysis. Cells (2 × 10 6 cells/ml) were cultured at 37°C (5% CO 2 ) in DMEM supplemented GlutaMAX, nonessential amino acids, penicillinstreptomycin, Hepes (Gibco), β-mercaptoethanol (Sigma), and 10% fetal bovine serum (Thermo Fisher) containing anti-IgM (5 µg/ml), CD40 ligand (0.5 µg/ml), and IL-4 (10 ng/ml) and harvested after 16 h. ChIP protocol was adapted from Seitz et al. (2011). Briefly, cells in B cell media were fixed with formaldehyde (Sigma) at final concentration of 1% for 10 min at room temperature. Glycine (Thermo Fisher) was added at a final concentration of 0.125 M for 5 min at room temperature. Cells were washed, pelleted, and stored at −80°C. After cell lysis, the DNA of 35 × 10 6 cells was fragmented using BIORUPTOR 200 immersion sonicator under the following conditions: high power, 30 s on, 30 s off (40 cycles). 100 μl of Pierce Protein A/G Magnetic Beads were coupled to 10 µg of antibody or control rabbit IgG antibody (Sigma). DNA was incubated with antibodycoupled beads, and after uncoupling, DNA was purified with NucleoSpin Gel and PCR Clean-up (Macheray-Nagel) following the manufacturer's protocol for SDS-rich samples using Buffer NTB (Macheray-Nagel). Samples were sequenced on an Illumina HiSeq2500 generating 100bp single ended reads. ChIP-seq reads were aligned to the mouse mm10 genome assembly using BWA version 0.7.15 (Li and Durbin, 2010) with a maximum mismatch of two bases. Picard tools version 2.1.1 (http://www. broadinstitute.github.io/picard) was used to sort, mark duplicates, and index the resulting alignment bam files. Files were normalized, and tdf files (for visualization purposes) were created using IGVtools version 2.3.75 software (http://www. broadinstitute.org/igv) by extending reads by 50 bp and normalizing to 10 million mapped reads per sample. Peaks were called using the standard parameters by comparing immunoprecipitated samples to their respective input and/or IgG controls using MACS version 2.1.1 (Zhang et al., 2008). Peaks called by MACS were annotated using the "annotatepeaks" function in the Homer version 4.8 software package (Heinz et al., 2010). Common and unique peaks across experiments were determined using a custom script. Datasets are available at the National Center for Biotechnology Information Gene Expression Omnibus under the following accession numbers: GSE129262, GSE80669, GSE77319, GSE4142, GSE76502, GSE11961, and GSE98419.
Quantification and statistical analysis Data were analyzed with unpaired two-tailed Student's t test; a P value of ≤0.05 was considered significant. Prism (v7 and v8, GraphPad) was used for statistical analysis. A single asterisk (*) in the graphs of figures represents a P value ≤0.05, double asterisks (**) a P value ≤0.01, triple asterisks (***) a P value ≤0.001, and quadruple asterisks (****) a P value ≤0.0001; "ns" stands for not statistically significant (i.e., a P value >0.05). Data in text and figures are presented as median or mean ± SEM; each case is indicated in the figure legends.
Online supplemental material Fig. S1 shows MYC and MIZ1 expression in GC B cell populations, the identified MYC-MIZ1 complex target genes, and PC and CD4 + CXCR5 + PD1 + T cell populations in the absence of MYC-MIZ1 complexes. Fig. S2 shows the gene expression of cell cycle-related genes in the absence of MYC-MIZ1 complexes, the characterization of MYC, MIZ1 in activated non-GC B cells, their cell proliferation in the absence of MYC-MIZ1 complexes, and MYC and MIZ1 expression in (pre-)GC B cells. Fig. S3 shows the enrichment for genes up-regulated in MBCs in the absence of MYC-MIZ1 complexes and MBC populations. Fig. S4 shows GC and MBC affinity in the absence of MYC-MIZ1 complexes. Fig.  S5 shows that MYC-MIZ1 complexes restrict MBC differentiation in the absence of Aicda.