FOXO1 re-expression with a dual-recombinase allele rescues class switching in germinal center B cells

Site-specific recombination systems, such as Cre/loxP, are key technologies to study gene function and (patho)physiological processes in vivo. Dual-recombinase approaches represent a further technical advance that enables more refined disease modeling, incorporating multistep genetic manipulations and lineage tracing (Li et al., 2023; Sket et al., 2023). However, one limitation of the currently available systems is that the recombinases used are expressed from different loci, thereby requiring the generation of complex compound mutant mice—containing separate transgenes encoding the recombinases together with the respective target alleles—and thus complicating the analysis in terms of both breeding time frame and fidelity of sequential mutagenesis at the level of individual cells. Here, we describe a dual-recombinase mouse strain in the context of germinal center (GC) B cells (GCBCs), where Cre and the tamoxifen (TAM)-dependent DreERT2 recombinases are co-expressed from the immunoglobulin heavy constant gamma 1 (Ighg1/Cγ1) locus and thus at the initiation of the GC response.

Upon antigen encounter, mature B cells undergo clonal expansion in distinct histological structures called GCs. In the course of the GC reaction, GCBCs diversify their immunoglobulin repertoire via class switch recombination (CSR) and somatic hypermutation (SHM) of antibody V region genes (Victora and Nussenzweig, 2022). Positively selected GCBCs differentiate into memory B cells (MBCs) or antibody-secreting plasma cells (PCs), thus shaping the humoral immune response (Victora and Nussenzweig, 2022). Both CSR and SHM require the introduction and efficient repair of DNA strand breaks in the immunoglobulin loci (Chi et al., 2020). Errors during these processes can lead to oncogenic mutations and/or translocations, rendering (post)-GCBCs the origin of most B cell malignancies (Küppers et al., 1999). In order to study the physiology and malignant transformation of GCBCs, a number of Cre mouse lines have been generated (Casola et al., 2006; Crouch et al., 2007; Dogan et al., 2009; Shinnakasu et al., 2016; Weber et al., 2019; Callahan et al., 2024). However, none of the currently available lines allows for the genetic manipulation of GCBC differentiation in a stepwise manner. This technical limitation prevents not only the in vivo investigation of the role and cooperation of genes in different phases of the GC reaction, but also the study of the sequential acquisition of oncogenic events in this microenvironment, underlying malignant transformation. In this study, by sequentially targeting the endogenous Foxo1 locus during the GCBC reaction as a proof-of-concept experiment using a novel dual-recombinase mouse strain, we demonstrate highly efficient Cre-mediated FOXO1 knockout (KO) with perturbation of GC physiology, followed by time-controlled Dre-mediated FOXO1 re-expression and phenotypic rescue, demonstrating that this system can be used in vivo for sequential targeted mutagenesis that is precisely controlled with respect to cellular compartment and time.

We have generated a novel dual-recombinase GCBC-specific mouse strain, in which Cre and the TAM-inducible DreERT2 recombinases are concomitantly expressed from the Ighg1/Cγ1 locus, hereafter called the Cγ1-CDE strain (Fig. 1 A). Cγ1 encodes the constant region of IgG1, and the transgene cassette is inserted in the 3′ untranslated region (UTR), downstream of the last membrane exon. Upon T cell–dependent immunization, most activated B cells that enter the GC produce germline ɣ1 transcripts in response to IL-4 (Casola et al., 2006), allowing the expression of both recombinases—and thus Cre activity—from the Cγ1 locus early on. Cells that have completed switch recombination to IgG1 continue to express the recombinases (from the nonproductive allele) during the course of the GC reaction (Casola et al., 2006). Since DreERT2 activation is dependent on TAM administration, the second (Dre-mediated) recombination event can be induced in a time-controlled manner, selectively in cells having undergone Cre induction. To functionally validate the newly generated strain, Cγ1-CDE mice were crossed to the fluorescent reporter strains R26-BFP^stopF (Sommermann et al., 2020) and R26-ZsGreen^stopRox (Biglari et al., 2021) for detection of Cre- and Dre-mediated recombination, respectively. Cγ1-CDE, R26-BFP^stopF, R26-ZsGreen^stopRox compound mutant mice were immunized with 4-hydroxy-3-nitrophenylacetyl-conjugated chicken gamma globulin (NP-CGG), and TAM was administered at days 2–5, 9–12, or 15–18 after immunization with subsequent flow cytometry analysis at days 7, 14, or 21, respectively (Fig. 1, B–J). Consistent with the previous characterization of the Cγ1-Cre allele (Casola et al., 2006), up to 96% of splenic GCBCs were labeled through Cre-mediated recombination (BFP⁺ GCBCs, Fig. 1, B–D; and Fig. S1, A and B). Further gating on these BFP⁺ GCBCs showed robust and highly efficient Dre-mediated recombination at all time points analyzed (Fig. 1, B–D), with the highest dual-labeling efficiency on day 7 of analysis (up to 60% ZsGreen⁺ cells within the BFP⁺ GCBC population, Fig. 1 B). Importantly, the number of ZsGreen⁺ cells within the BFP⁻ population was negligible at all time points analyzed (Fig. S1 H), demonstrating that Dre-mediated recombination only occurs in cells which have undergone Cre-mediated recombination. GCs are organized into two phenotypically and functionally distinct compartments, the dark zone (DZ) and the light zone (LZ) (Victora and Nussenzweig, 2022). GCBCs in both compartments were successfully labeled (Fig. 1, H–J and Fig. S1 B). In the absence of TAM, no Dre recombination was detected, demonstrating that Dre-mediated recombination is tightly controlled in vivo (Fig. 1, H–J). Of note, GCBCs in Peyer’s patches (PPs) could also be labeled, but at lower levels than splenic GCBCs (around 30% BFP⁺ cells; and 10% ZsGreen⁺ within the BFP⁺ population), similar to the lower Cre-mediated recombination levels previously described in PPs for the original Cγ1-Cre allele (Casola et al., 2006) (Fig. S1 G).

Germline ɣ1 transcripts (ɣ1 GLTs) are detected in the activated B cell pool early upon immunization, irrespective of whether the cells eventually switch to IgG1 expression (Casola et al., 2006; Roco et al., 2019) Accordingly, Cre-mediated followed by Dre-mediated sequential mutagenesis was achieved not only in IgG1⁺ (up to 70% ZsGreen⁺ within BFP⁺ cells) but also in IgM⁺ (up to 42% ZsGreen⁺ within BFP⁺ cells) GCBCs in our mouse model at all three time points analyzed (Fig. S1, E and F).

As expected, labeling also extends to post-GC cells. Within the MBC compartment, class-switched IgG1⁺ MBCs exhibited solid Cre- and Dre-mediated sequential mutagenesis (up to 84% BFP⁺ cells and 60% ZsGreen⁺ within BFP⁺ cells, Fig. 1, B–D and Fig. S1 C). Analysis of the splenic plasmablast/PC compartment (Fig. 1, B–D and Fig. S1 D) exhibited the highest labeling efficiency for this compartment at day 7 of analysis (up to 75% BFP⁺ cells and 50% ZsGreen⁺ within BFP⁺ cells, Fig. 1 B), indicating the feasibility to target and thus study extrafollicular PCs with our Cγ1-CDE strain. However, the fraction of labeled PCs was reduced when analyzing the immunized mice at later time points (days 14 and 21, Fig. 1, C and D), possibly due to the short lifespan of extrafollicular PCs.

The FOXO1 transcription factor, although dispensable for GC development/maintenance, positively regulates both GC compartmentalization (DZ compartment) and CSR (Dominguez-Sola et al., 2015; Sander et al., 2015). In view of this complex role in GCBC physiology, we aimed at genetically manipulating the endogenous Foxo1 locus in a sequential manner, as a proof-of-concept experiment for our dual-recombinase strategy. In addition to the already published Cre-inducible Foxo1 null allele (Foxo1^fl) (Paik et al., 2007), we generated a rox-controlled, Dre-inducible Foxo1 allele (Foxo1^stopRox). The Foxo1^stopRox allele carries a STOP cassette flanked by rox sites in front of the Kozak sequence of Foxo1 (labeled by a silent mutation in exon 1), leading to the inactivation of the allele in the absence of Dre activity (Fig. S2 A). Of note, while hemizygous Foxo1 mice are viable and fertile, homozygous germline deletion of Foxo1 results in embryonic lethality (Furuyama et al., 2004; Hosaka et al., 2004). In order to verify that the Foxo1^stopRox allele is functional upon deletion of the STOP cassette, splenic B cells from Cγ1-CDE, Foxo1^stopRox/wt compound mutant mice were activated in vitro and cultured in the presence or absence of 4-hydroxytamoxifen (4-OHT), followed by sequencing analysis of Foxo1 transcripts (Fig. 2 A). Indeed, the expression of the targeted allele was observed solely in the presence of 4-OHT, demonstrating not only that the Foxo1^stopRox allele is functional, but also the tightness of the system. In vivo, crossing of Cγ1-CDE, Foxo1^stopRox mice with Foxo1^fl transgenic mice (Cγ1-CDE, Foxo1^fl/stopRox) would generate a scenario in which upon NP-CGG immunization, and therefore Cre activation, the Foxo1^fl allele would be deleted, resulting in a Foxo1-KO condition in GCBCs (Fig. 2 B). GCBC-specific Foxo1 deletion is expected to lead to loss of the DZ compartment and to defects in CSR, while maintaining the GCBC phenotype (Dominguez-Sola et al., 2015; Sander et al., 2015). In a second step upon TAM treatment—and hence Dre nuclear translocation and activation—the re-expression of FOXO1 would be achieved from the Foxo1^stopRox allele, thus leading to a rescue of the previously induced FOXO1 deficiency (Fig. 2 B). To address whether the sequential deletion and re-activation of Foxo1 from the endogenous locus could restore GC compartmentalization and CSR, experimental Cγ1-CDE, Foxo1^fl/stopRox, R26-ZsGreen^stopRox (in short Foxo1^fl/stopRox) and control Cγ1-CDE, Foxo1^fl/wt, R26-ZsGreen^stopRox (in short Foxo1^fl/wt) mice were immunized with NP-CGG, subsequently treated with TAM (days 10-12 after immunization), and analyzed on day 14. The R26-ZsGreen^stopRox allele was included as an indirect reporter to identify Dre-recombined cells, i.e., FOXO1 re-expression in Foxo1^stopRox mice (Fig. 2 C). Strikingly, TAM administration with subsequent FOXO1 re-activation in the Foxo1^fl/stopRox group completely restored the DZ compartment in ZsGreen⁺ GCBCs, up to the same levels as in Foxo1^fl/wt control mice (Fig. 2 D and Fig. S2 B). This observation also suggests a high efficiency of Dre-mediated recombination in this setting, since all rox sites available (here the rox sequences in Foxo1^stopRox and R26-ZsGreen^stopRox alleles) are targeted in the same cell. In contrast, ZsGreen⁻ GCBCs in the experimental Foxo1^fl/stopRox group, of which the vast majority had failed to recombine the Foxo1^stopRox allele by Dre and had remained a full FOXO1-KO, lost the physiological levels of the DZ/LZ ratio, at the expense of DZ cells (Fig. 2 D) and in agreement with previous studies (Dominguez-Sola et al., 2015; Sander et al., 2015). Of note, the fraction of ZsGreen⁺ GCBCs in Foxo1^fl/stopRox mice was lower compared with that in controls, possibly reflecting the induction of FOXO1-associated apoptotic effects due to the sudden re-expression of FOXO1 (Dansen and Burgering, 2008; Srinivasan et al., 2009) (Fig. S2 C). Additionally, CSR to IgG1 was robustly restored in ZsGreen⁺ GCBCs in the Foxo1^fl/stopRox group, albeit at a lower frequency compared with controls (Fig. 2 E). We attribute this lower frequency to the fact that reporter ZsGreen⁺ cells from the Foxo1^fl/stopRox mice have expressed FOXO1 only for a maximum of 4 days (after TAM administration), in contrast to the control cells, in which FOXO1 is expressed from the beginning of the GC reaction. The higher frequency of CSR to IgG1 in control ZsGreen⁺ compared with control ZsGreen⁻ cells is due to the design of our Cre/DreERT2 line, since both IgG1 CSR and Dre-mediated ZsGreen recombination are dependent on activation of the Cγ1 locus. Of note, not all control ZsGreen⁺ cells switched to IgG1 (Fig. 2 E), demonstrating that γ1 GLTs can be expressed independently of CSR (Casola et al., 2006; Roco et al., 2019).

Under physiological conditions, CSR mostly occurs in pre-GCBCs, preceding SHM (Roco et al., 2019). Does the rescue of CSR observed following re-expression of FOXO1 at the peak of the GC reaction (Fig. 2, C and E) mainly occur in newly recruited GCBCs that populate the GC as late B cell incomers, rather than matured GCBCs? To address this issue, we performed a SHM analysis of the JH4 intronic region in GCBCs at day 14 after immunization. If CSR rescue predominantly happens in early GCBCs, the presence of mutations in Foxo1^fl/stopRox ZsGreen⁺ rescued cells should be scarce and lower than in control FOXO1-proficient (Foxo1^fl/wt) and FOXO1-KO (Foxo1^fl/stopRox mice, which had not received TAM) cells. As expected, our SHM data showed a similar frequency of mutations between FOXO1-proficient and FOXO1-KO controls (Fig. S2 D), with only a slight increase in the fraction of highly mutated cells (<2% with ≥10 mutations) in the FOXO1-expressing control group (Sander et al., 2015). Similarly, the mutational load in ZsGreen⁺ (rescued) Foxo1^fl/stopRox cells was comparable to that in FOXO1-KO control cells (Fig. S2 D), suggesting that the rescue of CSR upon FOXO1 re-expression is predominantly reflecting CSR in already established and not newly recruited GCBCs.

In a second experimental approach, we asked whether CSR defects could also be rescued in vitro following the sequential KO and re-expression of FOXO1. To this end, we cocultured naïve B cells from control and experimental mice with 40LB feeder cells, a cell line that expresses CD40L and BAFF to mimic T cell–dependent immune responses and can robustly induce GC-like B cells (iGCBCs) in the presence of IL-4 (Nojima et al., 2011). Coculture with 40LB cells efficiently activates the Cγ1 locus in B cells, inducing the Cre-mediated Foxo1-KO from the beginning of the experiment. Dre-mediated FOXO1 re-expression was stimulated by adding 4-OHT either at day 0—early FOXO1 induction upon B cell activation—or day 4—late FOXO1 induction in iGCBCs—of coculture, and CSR events were analyzed at day 4 or 8, respectively (Fig. S2 E). As expected, upon early rescue of FOXO1 at day 0, both FOXO1 levels and the frequency of class-switched iGCBCs were equivalent between control and rescued cells (Fig. S2, H and J) at day 4 of analysis. (FOXO1 was also detected in ZsGreen-negative cells, reducing the fidelity of the reporter as a readout in the in vitro setting.) Of note, control cells with one (heterozygous) or two (homozygous) copies of Foxo1 were comparable in terms of class switching (Fig. S2, F and G). Strikingly, full rescue of CSR events was also detected at the late time point, i.e., when re-expression of FOXO1 was induced in cells already presenting a GC-like phenotype (Fig. S2 I). Together, these findings suggest that FOXO1-deficient GCBCs preserve the ability to undergo class switching if FOXO1 expression is enforced, both in vitro and in vivo, irrespective of the predominance of CSR at the pre-GCBC stage under physiological conditions (Roco et al., 2019).

In order to investigate whether the re-expression of FOXO1 fully restores the FOXO1-dependent transcriptional program, reporter-positive and reporter-negative GCBCs of control Foxo1^wt/wt (Foxo1 WT), Foxo1^fl/wt (Foxo1 het), and experimental Foxo1^fl/stopRox mice were sorted at day 14 after immunization (TAM day 10–12) and subjected to RNA sequencing (Fig. 3 A; mRNA data available in NCBI Sequence Read Archive [SRA] database #PRJNA1365124). In addition, total GCBCs from Foxo1^fl/stopRox mice that had not received TAM were included, representing the FOXO1-KO situation. While the presence or absence of the ZsGreen reporter in control Foxo1-proficient (WT or het) GCBCs should show no differences, the expression status of the ZsGreen reporter in Foxo1^fl/stopRox experimental mice can be used to distinguish Foxo1-rescued (ZsGreen⁺) from cells, which failed to re-express FOXO1 (ZsGreen⁻, Foxo1 not-rescued). Principal component analysis (PCA) based on the top 500 genes showing the highest variability in expression across all samples clustered all Foxo1 control (WT and het) samples together—with no differences between reporter-positive and reporter-negative cells—while Foxo1-KO cells grouped separately (Fig. 3 B), indicating that the Foxo1 expression level was the main determinant for differences between the samples (PC1) (Table S1). Importantly, ZsGreen⁻ cells from Foxo1^fl/stopRox experimental mice (no Foxo1 re-expression) clustered together with the Foxo1-KO controls, whereas the ZsGreen⁺ counterparts (Foxo1 re-expression by Dre^ERT2-mediated recombination) grouped with the Foxo1 WT and het controls (Fig. 3 B), demonstrating that Foxo1-rescued cells present a similar transcriptional program compared with Foxo1 control cells. Similarly, a heatmap of Pearson’s correlation coefficients between log₂ fold changes in gene expression profile showed positive correlations for multiple comparisons between Foxo1-KO vs. control Foxo1-proficient (WT or het) cells, including the independent dataset from Dominguez-Sola et al. (2015), comparing Foxo1 WT and KO cells (Fig. 3 C), corroborating a high degree of similarity between Foxo1 WT and Foxo1 het controls. Remarkably, the comparison between Foxo1-KO and ZsGreen⁺ rescued cells also correlated positively with the multiple comparisons between Foxo1-KO vs. Foxo1-proficient cells (Fig. 3 C), further indicating a similar transcriptome between rescued and WT/het control cells. In order to confirm that FOXO1 re-expression in ZsGreen⁺Foxo1^fl/stopRox cells results in restoration of the FOXO1-associated transcriptional profile, we performed gene set enrichment analysis (GSEA) (Subramanian et al., 2005) using Foxo1 transcriptional signatures (Dominguez-Sola et al., 2015) (Table S2). Since no differences were observed between all Foxo1-proficient control samples (Fig. 3, B and C), the ZsGreen⁺ Foxo1 het was chosen as Foxo1 control condition for the GSEA. Strikingly, Foxo1-rescued cells showed a highly significant enrichment of FOXO1-upregulated genes compared with Foxo1-KO cells, at levels comparable to the enrichment observed for Foxo1 control cells (Fig. 3 D, left panels), while FOXO1-downregulated genes were highly enriched in the Foxo1-KO cells (Fig. 3 D, right panels). Additionally, the DZ signature (Victora et al., 2012)—which is closely associated with Foxo1—was also significantly enriched among genes differentially expressed in rescued cells compared with Foxo1-KO cells, similar to the extent observed in control cells (Fig. 3 E, Table S2). Conversely, ZsGreen⁻ (not-rescued) cells displayed a pattern similar to the Foxo1-KO cells (Fig. S3, A and B). Taken together, the re-induction of FOXO1 at the peak of the GC reaction not only rescued the cells phenotypically, but also fully restored the transcriptional program associated with FOXO1, back to the levels of control cells.

We present proof-of-concept experiments that demonstrate the efficiency of sequential targeted mutagenesis through a novel dual-recombinase allele. This allele mediates cell type–specific expression of Cre recombinase, together with a TAM-regulated form of Dre recombinase (DreERT2) that can be subsequently activated in the same cells in a time-controlled manner. If successful, this approach allows for simplified breeding schemes for compound mutant mice, and promises high fidelity of sequential mutagenesis at the cellular level, in that Dre-mediated recombination is targeted and limited to the very cells that have previously been exposed to Cre activity.

In the Cγ1-CDE strain which we have constructed, Cre and DreERT2 are expressed from the 3′ UTR of the Cγ1 locus that encodes the constant region of γ1 immunoglobulin heavy chain (Fig. 1 A). Cγ1 GLTs are detected in the majority of B cells at the initiation of the GC reaction upon immunization with hapten-carrier conjugates, regardless of whether subsequent CSR to IgG1 takes place (Casola et al., 2006; Roco et al., 2019). Hence, highly efficient Cre- followed by Dre-mediated recombination was observed at the three time points analyzed, with the highest dual-labeling efficiency achieved at the earliest time point (day 7 after immunization), with up to 96% GCBCs expressing the Cre-dependent BFP reporter and more than half co-expressing Dre-dependent ZsGreen (Fig. 1 B). The lower frequency of Dre- as compared to Cre-mediated recombination may be due to the limited time window of TAM application. The reason for the lower labeling efficiencies at the later time points remains to be explored, but may simply reflect higher levels of Cγ1 GLTs in GCBCs early after their recruitment. Dual labeling was also observed in a fraction of post-GC cells, such as MBCs and PCs (Fig. 1). In these compartments, the fractions of labeled cells varied depending on location and time of labeling after immunization, and while all BFP⁺ cells had presumably been labeled at the initiation of the GC response, ZsGreen labeling of some or most of these cells may have occurred already at the GC stage of differentiation. Unlabeled MBCs and PCs in these mice have likely arisen through pathways not involving an IL-4–rich environment and Cγ1 transcription.

We used the Cγ1-CDE strain to analyze the role of the FOXO1 transcription factor at different stages of the GC reaction. In the absence of FOXO1, GCs exhibit a reduction of their fraction of proliferating cells, the so-called DZ, and lose their ability of undergoing CSR (Dominguez-Sola et al., 2015; Sander et al., 2015). Recent data suggest that the latter process is largely restricted to the initiation of the GC response (Roco et al., 2019). Accordingly, knocking out FOXO1 in Cγ1-CDE mice at the initiation of the GC response, through Cre-mediated recombination, resulted in GCs with a reduced DZ and drastically reduced CSR in GCBCs (Fig. 2). These phenotypes were fully rescued by Dre-mediated FOXO1 re-expression, with gene expression and SHM data indicating that this rescue occurred in matured GCBCs (Figs. 2, 3, and S2). These data indicate that GCBCs retain their ability to undergo CSR and respond to the corresponding signals, which apparently remain available at later stages of the GC reaction.

In summary, the Cγ1-CDE allele allows for highly efficient Cre- and subsequent Dre-mediated recombination in GCBCs. This tool can be used in vivo to study the interplay of distinct genetic events or the function of a specific gene at different time points during the GC reaction, and also to model the sequential acquisition of genetic alterations acquired during B cell lymphomagenesis, that often originates from GCBCs (Küppers et al., 1999). Similar dual-recombinase alleles could be profitably constructed for targeted sequential mutagenesis in other cellular contexts.

Cγ1-cre, R26-BFP^stopF, and Foxo1^fl alleles have been described previously (Casola et al., 2006; Paik et al., 2007; Sommermann et al., 2020). R26-ZsGreen^stopRox is a derivative of the R26-CAGS-lox-STOP-lox-rox-STOP-rox-ZsGreen mice crossed to a Deleter-Cre line (Biglari et al., 2021). Cγ1-CDE and Foxo1^stopRox strains were generated by CRISPR/Cas9-mediated homologous recombination in C57BL/6 mouse zygotes according to previously published protocols (Weber et al., 2019). Cγ1-CDE and Foxo1^stopRox strains are available at the Jackson Laboratory Repository with the JAX Stock No. 040961 and 040962, respectively.

Mice were bred and maintained under specific pathogen–free conditions. 8–15-week-old male or female mice were immunized with 100 µg alum-precipitated NP-CGG (ratio 10-19; Cat#N-5055B; LGC Biosearch Technologies) intraperitoneally followed by TAM administration at the indicated time points (4 mg/day oral gavage, TAM; #T5648-5G, 99%; Sigma-Aldrich). Experimental animal procedures were approved by the Landesamt für Gesundheit und Soziales Berlin (G0308/19, G0062/21).

Cells from spleen and bone marrow were collected in B cell medium (DMEM supplemented with 10% FCS, 1× non-essential amino acids (NEAA), 1 mM sodium pyruvate, 2 mM L-glutamine, 1 mM HEPES, 1× penicillin–streptomycin, 50 µM β-ME). Erythrocytes were lysed with Gey’s solution, and single-cell suspensions were stained with antibody conjugates in PBS, pH 7.2, supplemented with 0.5% BSA and 2 mM EDTA. For intracellular staining, single-cell suspensions were stained with Foxp3/Transcription Factor Staining Buffer Set (#00-5523-00; eBioscience), according to the manufacturer’s instructions. Zombie Aqua Fixable Viability Kit (#423102; BioLegend) was used to assess live/dead status. The following antibodies were used: B220-BV785 (clone RA3-6B2; #103246; RRID #AB_2563256; BioLegend), CD19-BV650 (clone 6D5; #115541; RRID #AB_11204087; BioLegend), CD38-AF700 (clone 90; #56-0381-82; RRID #AB_657740; eBioscience), CD86-BV421 (clone GL-1; #105032; RRID #AB_2650895; BioLegend), CD86-PE-Cy7 (clone GL-1; #105014; RRID #439783; BioLegend), CD95/FAS-PE (clone Jo2; #554258; RRID #AB_395330; BD Biosciences), CD95/FAS-PE-Cy7 (clone Jo2; #557653; RRID #AB_396768; BD Biosciences), CXCR4-bio (clone 2B11; #13-9991-82; RRID #AB_10609202; Invitrogen), FOXO1-AF647 (clone C29H4; #72874; RRID #AB_2799829; Cell Signaling), GL7-PE (clone GL7; #144608; RRID #AB_2562926; BioLegend), IgE-BV605 (clone R35-72; #744281; RRID #AB_2742118; BD Biosciences), IgG1-APC (clone A85-1; #560089, RRID #AB_1645625; BD Biosciences), IgG1-BV510 (clone A85-1; #740121; RRID #AB_2739879; BD Biosciences), IgG1-PE (clone A85-1; #550083; RRID #AB_393553; BD Biosciences), IgM-bio (clone Il/41; #13-5790-81; RRID #AB_466675; eBioscience), and Streptavidin-BV605 (#405229; BioLegend). Samples were analyzed on an LSRFortessa (BD Biosciences). Plots were generated using FlowJo software (BD FlowJo) and graphs by Prism software (GraphPad Prism).

Transgenic splenic B cells were isolated by CD43 depletion with magnetic anti-mouse CD43 microbeads (Cat#130-049-081; Miltenyi Biotec) according to the manufacturer’s instructions. To remove the rox-flanked STOP cassette, B cells were activated with IL-4 (25 ng/ml; Cat#404-ML; R&D) and αCD40 (1–2 µg/ml; HM40-3; Cat#102908; BioLegend) and cultured in the presence of 4-OHT (1 µM; Cat#H6278; Sigma-Aldrich). 3 days later, total RNA was extracted using AllPrep DNA/RNA Mini Kit (#80204; Qiagen) according to the manufacturer’s instructions. For cDNA synthesis, 500 ng of RNA was used per reaction and reverse transcription was performed with SuperScript II Reverse Transcriptase (#18064014; Invitrogen). A mixture of random and oligo(dT) primers was used following the manufacturer’s protocol. PCR amplification of Foxo1 cDNA was performed using the following primers: forward 5′-ATG​GCC​GAA​GCG​CCC​CAG​GTG​GTG​GAG​AC-3′ and reverse 5′-CCT​ACT​TCA​AGG​ATA​AGG​GCG​ACA​GCA​AC-3′. PCR products were gel-purified using the NucleoSpin Gel and PCR Clean-Up kit (#740609.250; Macherey-Nagel) and sequenced by Sanger sequencing.

40LB feeder cells (Nojima et al., 2011) were maintained in DMEM supplemented with 10% FCS, 1× penicillin–streptomycin, 1 mM sodium pyruvate, and 1× GlutaMAX. 40LB cells were irradiated with 20 Gγ and cocultured with isolated splenic naïve B cells (supplemented with 1 ng/ml IL-4 and in the presence or absence of 1 µM 4-OHT) to induce GC-like B cells.

GCBCs of the indicated genotypes were sorted (FACSAria, BD Biosciences) and stored at −70°C. DNA was extracted using the AllPrep DNA/RNA micro kit (#80284; Qiagen), and JH4 intronic region was PCR-amplified using the KOD polymerase (#71086-3; Sigma-Aldrich) using the following primers: VHA/VH1 forward primer 5′-ARGCCTGGGRCTTCAGTGAAG-3′, VHE/VH5 forward primer 5′-GTG​GAG​TCT​GGG​GGA​GGC​TTA-3′, and JH4 intronic reverse primer 5′-CTC​CAC​CAG​ACC​TCT​CTA​GAC​AGC-3′. JH4 PCR products (800 bp) were purified by gel electrophoresis and extracted using the NucleoSpin Gel and PCR Clean-Up kit (#740609.250; Macherey-Nagel). Purified products were cloned into the pCR4Blunt-TOPO vector according to manufacturer’s instructions (Zero Blunt TOPO PCR Cloning Kit, #450031; Invitrogen), and transformation was carried out using Top10 competent cells. 30-40 colonies were cultured overnight, and DNA plasmids were isolated (NucleoSpin Plasmid DNA Purification Kit; #740588.250; Macherey-Nagel) and sent for Sanger sequencing (T7 primer, LGC Genomics). Sequences were aligned using Basic Local Alignment Search Tool to quantify the number of somatic mutations.

5,000 GCBCs of the indicated genotypes were sorted (FACSAria, BD Biosciences) into 75 μl RLT plus (RNeasy lysis) buffer (with β-mercaptoethanol) and stored at −70°C. RNA was extracted using the AllPrep DNA/RNA micro kit (#80284; Qiagen), and a low-input protocol for library preparation was conducted as previously described (Rosales et al., 2018). In short, cDNA was synthesized using SuperScript II Reverse Transcriptase (#18064022; Invitrogen), together with Template Switching Oligo and Oligo(dT)-30VN, followed by full-length cDNA amplification using KAPA HiFi HotStart Ready Mix (#KK2601; Roche) and an IS-PCR primer. cDNA libraries were prepared using the Nextera XT DNA library preparation kit (#FC-131-1024; Illumina) and Illumina DNA/RNA UD Indexes Set A with Tagmentation kit (#20091654; Illumina), following manufacturer’s instructions. Libraries were sequenced on a NovaSeq X Plus 25B instrument as paired-end 150-bp reads with a depth of 100 million reads per sample.

Reads were trimmed using bbduk v39.08 for adapters, poly-G tails, and 5′ end nucleotide composition bias due to shearing. RNA quantification was performed with Salmon v1.10.3 using GRCm39 as a mouse reference genome and using GENCODE version M36 for basic feature annotations. Gene-based expression levels were computed using the Bioconductor package tximport v1.34, and the normalization and differential expression using the Bioconductor package DESeq2 v1.46, on all genes with at least 5 reads in at least 3 samples. The log2 fold changes were moderated using the R package ashr v2.2.63. The expression values used for the PCA plot were normalized with the variance-stabilized transformation implemented in DESeq2. The transcriptional profile of one Foxo1^fl/stopRox mouse revealed an underlying infection unrelated to the experiment, which precluded its inclusion in the final analysis. The gene set enrichment running scores were computed with the Bioconductor package DOSE version 4.2.0, and their statistical significance was assessed using the CERNO test (Yamaguchi et al., 2008). All differential expression analysis, tables, and figures were generated using R version 4.4.3 (differential expression) and 4.5.0 (tables and figures).

Fig. S1 shows additional characterization of the Cg1-CDE strain. Fig. S2 shows additional characterization of the sequential mutagenesis at the Foxo1 locus (Cre-mediated FOXO1 KO followed by Dre-mediated re-expression of FOXO1). Fig. S3 shows the GSEA using Foxo1 and DZ signatures comparing FOXO1-proficient vs. ZsGreen⁻ (not-rescued) Foxo1^fl/stopRox conditions. Table S1 shows differentially expressed genes between multiple comparisons. Table S2 shows GSEA results.

The mRNA sequencing data are publicly available in the NCBI SRA database (accession number PRJNA1365124).

We thank all members of the K. Rajewsky lab for discussion, R. Kühn (Max Delbrück Center [MDC] transgenic facility), H.P. Rahn (MDC FACS facility), and G. Natale (K. Rajewsky lab, MDC) for excellent technical support, M. di Virgilio for valuable feedback and discussion, and the MDC animal caretakers for their outstanding technical help. Figure schematics were created using images from BioRender. We apologize to colleagues whose work could not be cited owing to space limitation.

This work was supported by the European Research Council (ERC Advanced Grant #268921) to K. Rajewsky, the Deutsche Krebshilfe (grant #70112800) to K. Rajewsky and M. Janz, and the Berlin School of Integrative Oncology to C. Farré Díaz. Open Access funding was provided by Max-Delbrück-Centrum für Molekulare Medizin in der Helmholtz-Gemeinschaft.

Author contributions: Carlota Farré Díaz: conceptualization, formal analysis, investigation, methodology, resources, visualization, and writing—original draft, review, and editing. Eleni Kabrani: conceptualization, investigation, methodology, visualization, and writing—original draft. Wiebke Winkler: investigation and methodology. Eric Blanc: formal analysis and visualization. Brigitte Wollert-Wulf: investigation. Claudia Salomon: project administration and validation. F. Thomas Wunderlich: methodology. Dieter Beule: data curation, software, and writing—review and editing. Martin Janz: conceptualization, funding acquisition, methodology, project administration, resources, supervision, validation, and writing—review and editing. Klaus Rajewsky: conceptualization, funding acquisition, supervision, and writing—review and editing.