C57BL/6 NCrSlc (B6) mice were purchased from Sankyo Labo Service. All the following mouse strains were backcrossed to B6 or congenic B6 CD45.1⁺ mice: SpiB ^f/f (see below), Blimp1 ^f/f, Blimp1-GFP (18), and Bach2 ^-/- (19). SpiB ^f/f and Blimp1 ^f/f mice were crossed with Cγ1-cre (20) and CD19-cre mice (21), respectively, to generate each conditional knockout strain. As the control for the SpiB ^f/f Cγ1-cre (here called ‘SpiB cKO’) mice, SpiB ^+/+ Cγ1-cre (here called ‘SpiB WT’) mice were used. Sex-matched 7-12 week-old mice were immunized i.p. with 100 μg of 4-hydroxy-3-nitrophenyl acetyl (NP)₃₂-chicken gamma globulin (CGG) in alum or NP-conjugated sheep red blood cell (SRBC) where indicated. Mice were maintained in the Tokyo University of Science mouse facility under specific pathogen-free conditions. Mouse procedures were performed under protocols approved by the Animal Care and Use Committee of Tokyo University of Science (Approval No S13009).

To generate the Spib-floxed (SpiB ^f/f) mice, a targeting vector was designed to delete exon 6 encoding the Ets domain of Spi-B. To construct the targeting vector, C57BL/6N mouse genomic DNA fragments were amplified by PCR using KOD -plus- DNA polymerase (Toyobo, Osaka, Japan). An 8.0 kbp long arm containing exon 6, which has an insertion of a loxP site in the intron downstream of the exon 6, was connected to an MC1 promoter-driven neomycin resistance gene (neo) cassette flanked by two FRT sites and a loxP site. Then, a 1.9 kbp short arm containing exons 3-5 was ligated upstream of the neo cassette. A herpes simplex virus thymidine kinase gene (HSV-TK) was attached at the downstream end of the long arm. The targeting vector was linearized with SalI and electroporated into the Bruce4 ES cells. Clones with resistance to G418 and gancyclovir were screened for homologous recombination by PCR and confirmed by Southern blot analysis. ES cell clones carrying the targeted allele were used to generate Spib mutant mice. To delete the neo cassette flanked by two FRT sites, the Spib-targeted mice were intercrossed with C57BL/6-Tg(CAG-flpe)36Ito/ItoRbrc (RBRC01834) (RIKEN BioResource Research Center, Japan) ( Supplementary Figure 1A ).

Genotyping of Spib mutant mice was performed by PCR using the primers listed below. PCR using primers #1 and #2 generates 301bp and 475 bp bands in the wild-type and floxed alleles, respectively. PCR using primers #1 and #3 generates a 454 bp band in the deleted allele. Although PCR using primers #1 and #3 should also generate a band with larger than 2700 bp in both wild-type and floxed alleles, it was too large to be detected ( Supplementary Figures 1B, C ).

Primer #1: 5’-AACCAGGCCCAACTCCATTGTGAAAAC-3’.

Primer #2: 5’-GCCAGCTTCTGATACGTCATGCGCTTG-3’.

Primer #3: 5’-AGTCCTGTCACACCATTGGTTGCAGTG-3’.

Single-cell suspensions from spleens were depleted of RBCs by ammonium chloride lysis, incubated with mAb to FcγRII/III (2.4G2; BD Pharmingen) to block FcγRs and were stained for 30 min on ice in PBS containing 0.5% BSA, 2 mM EDTA, 0.05% sodium azide with various combinations of antibodies and reagents. Intracellular SpiB staining was performed with FoxP3 Staining Buffer Set according to the manufacturer’s protocol (eBioscience) with anti-SpiB sheep polyclonal antibody (R&D systems, AF7204) followed by Alexa 647-labeled donkey anti-sheep IgG (Invitrogen, A-21448). All samples were analyzed by either FACS Callibur or FACS Canto II (BD Biosciences). The data were analyzed by FlowJo (Tree Star).

Splenic naïve B cells were purified as described previously (7). B cells were cultured in RPMI-1640 medium (Wako) supplemented with 10% FCS, 10 mM HEPES, 1 mM sodium pyruvate, 5.5 × 10⁻⁵ M 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin (GIBCO). For iGB cell culture, naive B cells were cultured on 40LB feeder cells with IL-4 (10 ng/ml) or IL-21 (10 ng/ml) where indicated. To purify iGB cells or to change the culture condition, 40LB feeder cells were removed from the total cells of iGB cultures by anti-H-2Kd (BioLegend) Ab, Streptavidin Particle Plus DM and the IMag system (BD Biosciences).

Early B_mem and GC B cells were purified from spleen of the mice immunized with NP-SRBC, 7 or 14 days previously, as follows: splenocytes were stained with anti-CD19, anti-CD38, anti-GL7, and NP-BSA-biotin, followed by fluorescence-labeled streptavidin. B_mem cells (CD19⁺ NP⁺ CD38⁺ GL7^–) and GC B cells (CD19⁺ NP⁺ CD38^– GL7⁺) were sorted by FACS AriaII. For iGB cells, cultured cells depleted of 40LB feeder cells as described above, were stained with anti-IgG and anti-IgE and sorted by FACS Aria II.

Retroviruses were produced in Plat-E cells transfected with pMXs-based vectors with Fugene HD (Promega) or PEI “Max” (Mw 40,000; Polysciences). The virus-containing supernatant was collected 2 days after transfection and added to iGB cells cultured for 2 days previously. Cells were spin-infected at 2,000 rpm, room temperature for 90 minutes with 10 μg/ml DOTAP Liposomal Transfection Reagent (Roche) and IL-4 (10 ng/ml). One day later, cells were collected and cultured on new 40LB feeder layers with IL-4 for further culture.

iMB cells were generated as previously described (7). Retrovirally transduced iGB cells from CD45.1⁺ B6 mice were collected by removing 40LB feeder cells with anti-H-2Kd Ab (BioLegend), Streptavidin Particle Plus DM, and the IMag system (BD Biosciences). The iGB cells were adoptively transferred to γ-irradiated (6.5Gy) B6 mice (CD45.2⁺). Splenocytes were harvested from the recipient mice after 2 weeks or later for further analysis.

The procedures for RNA extraction and reverse transcription to cDNA have been described previously (7). Quantitative real-time PCR was performed with a 7500 fast Real-time PCR system (Applied Biosystems). Gene expression levels were determined by the relative standard curve method and normalized to that of either Gapdh or 18S rRNA.

Total and high affinity NP-specific IgG1 was detected by the method described previously (7). NP_13.6-BSA or NP_2.7-BSA for total or high affinity antibodies, respectively, was coated on 96 well flat bottom plate. Bound Abs were detected by HRP-conjugated goat anti-mouse IgG1 Ab (Southern Biotechnology) and TMB substrate (Sigma).

To understand the regulation of B_mem cells by transcription factors, we analyzed the gene-expression profiles of induced memory-like B (iMB) cells and compared the result with publicly available gene expression data (ImmGen). We noticed that mRNA for the transcription factor SpiB was expressed selectively high in B_mem cells among B cells, while hardly expressed in plasmablasts and PCs. In contrast, Bcl6 was expressed substantially lower in B_mem cells than in GC B cells ( Supplementary Figure 2A ). To confirm these expression patterns, we analyzed the mRNA levels of SpiB, Bach2 and Bcl6 in early B_mem and GC B cells from mice immunized with NP-CGG, and naïve follicular B cells from unimmunized mice. In slight contrast to the ImmGen data, SpiB expression in the early B_mem cells was lower than that in GC B cells, but appeared to be higher than that in follicular B cells. These expression patterns are similar to those of Bach2 and Bcl6 ( Figure 1A ; Supplementary Figures 2B, C ).

Our previous finding proposed that the quantity of CD40 stimulation on B cells during the primary response is a key factor in regulating B_mem cells (22). Additionally, SpiB is reported to be induced by OBF-1 which is upregulated by CD40 signaling (23, 24). Thus, we assessed whether CD40 stimulation induces the expression of SpiB in B cells. By culturing naïve B cells on feeder cells expressing CD40L and BAFF (40LB) without any cytokines, the expression of SpiB was induced after 3 days ( Figure 1B ). We then tested other stimuli known to activate naïve B cells, namely, anti-CD40 or/and anti-IgM antibodies, or LPS, and observed that anti-CD40 induced the highest levels of SpiB expression among these stimuli, and no synergistic effect was seen between anti-CD40 and anti-IgM antibodies ( Figure 1B ). These results indicate that CD40 stimulation is sufficient to induce SpiB expression in B cells. Immunofluorescence microscopy analysis of spleens from mice immunized with NP-SRBC 7 days earlier showed that a small number of discernible SpiB⁺ cells were scattered in GC, B cell follicles, and T cell zones in the white pulp, while SpiB expression was undetectable in the vast majority of GC and follicular B cells ( Figure 1C , top). At a closer look, nuclear SpiB⁺ cells in the B cell follicle were also B220⁺ ( Figure 1C , bottom), either IgD⁺ or IgD^–, and those in the GC and T cell zones were discernible as IgD^– ( Figure 1 , top). Together with our mRNA expression data and previous reports showing that SpiB mRNA expression is limited to B cells and plasmacytoid dendritic cells (9; ImmGen), these results suggest that a significant level of SpiB is expressed in selected GC B cells, class-switched or unswitched B_mem cells in the B cell follicle, and plasmacytoid dendritic cells in the T cell zone.

It has been shown that SpiB suppresses the expression of Blimp1 in human cells (17) and potentially in mice (16, 25). To further infer the role of SpiB in the late B-cell development, we analyzed the expression of SpiB in B cells cultured in the iGB system (7). The expression of SpiB mRNA was readily detected in IgG1⁺ B cells cultured on 40LB feeder cells with IL-4 (iGB-4 cells), and in IgE⁺ iGB-4 cells albeit at a reduced level, but declined to an almost undetectable level following culture with IL-21 (iGB-21 cells) ( Figure 1D ). SpiB protein expression was also decreased in the iGB-21 cells as compared to iGB-4 cells ( Figure 1E ). Co-staining of SpiB protein with memory/naive B cell marker (CD38) and PC marker (CD138) revealed that SpiB^high cells expressed higher levels of CD38 and little CD138, whereas SpiB^low cells expressed lower CD38 and were mostly CD138⁺ ( Figure 1E ). Among the iGB-4 cells, IgG1⁺ cells were largely SpiB⁺ CD38^high CD138^–, whereas IgE⁺ cells contained a substantial population of SpiB^– CD38^low CD138⁺ cells, which markedly increased in iGB-21 cells, either of IgG1⁺ or IgE⁺ ( Supplementary Figures 2D, E ). This result corresponded to the notion that IgE⁺ B cells rapidly differentiate into PCs (26, 27) and IL-21 facilitates PC differentiation (7). These results suggest that SpiB may have a function to suppress PC differentiation while inducing the phenotype of B_mem cells.

To better understand the function of SpiB in B cell differentiation, we performed retroviral overexpression or knockdown of SpiB in iGB cells ( Figure 2A ). Overexpression of SpiB resulted in the decrease, while knockdown of SpiB resulted in the increase, of the proportion of CD138⁺ cells in both IgE^– (mostly IgG1⁺) and IgE⁺ fractions of iGB-4 and iGB-21 cells ( Figure 2B ; Supplementary Figure 3A ). PC differentiation is known to be induced by a transcription factor Blimp1 which is coded by the gene Prdm1 (28). To further elucidate the mechanism of SpiB function, we analyzed the expression of Prdm1 gene by using iGB cells derived from Blimp1-reporter (Blimp1-GFP) mice. Overexpression of SpiB resulted in the decrease, whereas knockdown of SpiB resulted in the increase, of the proportion of Blimp1 GFP⁺ cells within IgE⁺, IgM⁺ and IgE^– IgM^– (mostly IgG1⁺) B cells in iGB-4 and iGB-21 cells ( Figure 2C ). These results suggest the function of SpiB to suppress Blimp1 expression at a transcription level and thus PC differentiation.

We further analyzed the mRNA expression levels of B-cell-related genes in the iGB-4 cells over-expressing SpiB and confirmed that the expression of Prdm1 was suppressed by SpiB ( Supplementary Figure 3B ). Of note, SpiB overexpression resulted in upregulation of the expression of Bach2, Pax5, and Bcl6, which encode transcription factors known to reciprocally suppress the expression of Blimp1 (6, 29–31) ( Supplementary Figure 3B ). Blimp1 is known to suppress several B cell-related genes to positively regulate PC differentiation (32–34). To clarify whether the SpiB-mediated upregulation of Bach2, Pax5, and Bcl6 is independent of Blimp1, we used B cells from B-cell-specific Blimp1-deficient mice (Blimp1^f/f CD19-cre; here called Blimp1-KO mice) for iGB cell culture. Overexpression of SpiB in the Blimp1-KO iGB-4 cells resulted in an increase of the expression of Bach2, Pax5, and Bcl6, whereas knockdown of SpiB resulted in the opposite outcome ( Figures 2D, E ). These results suggest that SpiB functions to suppress the expression of Blimp1 and independently induce Bach2, Pax5, and Bcl6 in B cells.

The above results indicating that SpiB suppresses the expression of Blimp1 while inducing the expression of Bach2, suggesting that SpiB functions to generate or maintain B_mem cells. To address this possibility, we applied our system in which iGB-4 cells differentiate to memory-like B (iMB) cells in the recipient mice (7). We retrovirally overexpressed or knocked down SpiB in iGB-4 cells and transferred these cells to irradiated mice respectively to generate iMB cells ( Figure 3A ). SpiB overexpression resulted in an increase in the frequency of the iMB cells ( Figure 3B ). On the contrary, SpiB knockdown resulted in the reduced frequency of the iMB cells ( Figure 3C ). These data indicated that SpiB promoted the generation or maintenance of iMB cells in a dose-dependent manner. These results strongly suggest that SpiB is a factor that regulates the generation or maintenance of B_mem cells.

To confirm this, we next analyzed the generation of B_mem cells from SpiB-deficient B cells. Since SpiB null-knockout mice were previously reported to have defective GCs (15), we analyzed SpiB-floxed (SpiB ^f/f) mice (see Materials and Methods 2.2) crossed with Cγ1-cre mice expressing cre recombinase only in B cells class-switched to IgG1 (20) (here called ‘SpiB cKO’ mice). SpiB ^+/+ Cγ1-cre (called ‘SpiB WT’) mice were used as controls. The deletion of SpiB was confirmed in IgG1⁺ B cells in the immunized mice and in iGB cells ( Supplementary Figures 4A–E ). We first examined the iMB cell generation from SpiB cKO iGB-4 cells in vivo. As expected from the result of SpiB-knockdown experiments ( Figure 3C ), very few IgG1⁺ iMB cells were generated from SpiB cKO iGB-4 cells ( Figure 4A ).

Next, we analyzed T cell-dependent (TD) immune response in the SpiB cKO mice by immunizing the mice with a typical TD antigen, NP-CGG, precipitated in alum. On day 7, 14, and 63 after the primary immunization, we analyzed GC B cells (GL7⁺ CD38^–) and B_mem cells (GL7^– CD38⁺) within NP-specific IgG1⁺ and IgG1^– B cells by FCM ( Figure 4B ). The frequency and the absolute number of GC B and B_mem cells did not significantly differ between SpiB WT and SpiB cKO mice in both NP⁺ IgG1⁺ and NP⁺ IgG1^– B cell populations at 7- and 14-days post-immunization ( Figure 4C ). The concentration of blood anti-NP-IgG1 antibody was also similar between the two strains during the primary immune response ( Figure 4D ). On day 63 post-immunization, when GC B cells almost disappeared, the number of B_mem cells in the NP⁺ IgG1⁺ population, but not in the NP⁺ IgG1^– population, was significantly reduced in SpiB cKO mice as compared to SpiB WT mice. We further analyzed the memory antibody response in these mice to determine the effect of the reduction of IgG1⁺ B_mem cells. After re-immunization of the mice with NP-CGG without adjuvants, the generation of serum anti-NP IgG1 antibodies was attenuated in SpiB cKO mice during the secondary immune response, which was highlighted when we detected only high-affinity antibodies that are mostly produced from B_mem cells ( Figure 4D ). These results suggest that SpiB functions to maintain, rather than to induce, the antigen-specific B_mem cells.

It is known that Bach2 functions as a suppressor for Blimp1 and regulates B_mem cell differentiation (6, 35). Our in-vitro analysis demonstrated that SpiB functions to suppress PC differentiation by suppressing Blimp1 expression while inducing the expression of Bach2 ( Figures 2C–E ). To elucidate the mechanism of SpiB function in B cells, we focused on the relationship of SpiB and Bach2. We first analyzed expression of SpiB and CD138 in the iGB-4 cells derived from splenic B cells of normal B6 (WT), Blimp1-KO, Bach2-KO, and Blimp1/Bach2-double KO mice by FCM. As expected, CD138⁺ cells were induced in Bach2 KO iGB cells. However, the induction of CD138⁺ cells was strongly suppressed in Blimp1/Bach2 double KO iGB-4 cells irrespective of their BCR isotype, indicating that Bach2 suppresses PC differentiation through Blimp1 suppression ( Figure 5A ; Supplementary Figure 5A ).

We next analyzed the expression of Bach2 and Prdm1 along with other B cell-related genes, such as Bcl6, Pax5 and IRF4, in SpiB cKO and SpiB WT iGB-4 cells. SpiB-deficiency resulted in decrease of the expression of Bach2 while increase of Prdm1 and IRF4 ( Figure 5B , Supplementary Figure 5B ). To verify the functional meaning of the SpiB-regulated expression of Bach2 and Blimp1, we retrovirally transduced SpiB and Bach2 into Bach2 KO iGB cells. The frequency of CD138⁺ cells was strongly suppressed by SpiB overexpression in Bach-2 KO iGB-4 cells as observed by Bach2 reconstitution, and to a lesser extent in iGB-21 cells, indicating that SpiB can suppress PC differentiation independently of Bach2 ( Figure 5C ). To further analyze the relation between Bach2 and SpiB in vivo, we sorted IgG1⁺ GC and early B_mem cells from SpiB WT and SpiB cKO mice 14 days after immunization with NP-CGG in alum, and analyzed the expression of Bach2, Prdm1 and other B cell-related genes. In GC B cells, the expression of Bach2 was decreased while that of Prdm1 was increased by the SpiB-deficiency, with other B cell-related genes largely unaffected, resembling the result with iGB-4 cells. In the early B_mem cells, however, the expression of Bach2 and Prdm1, as well as of other genes, was not apparently affected by the SpiB-deficiency ( Figure 5D ; Supplementary Figure 5C ). These results indicate that SpiB upregulates Bach2 expression and suppresses Blimp1 expression and PC differentiation independently of Bach2 in GC B cells, which may render the longevity to the descendant B_mem cells.

Since the phenotype of SpiB cKO mice in the TD immune response was apparent only in the late B_mem cells and their recall response, we suspected the function of SpiB to be more related to the maintenance of B_mem cells rather than their generation. Previous studies have shown that autophagy is one of the mechanisms to inhibit apoptosis and to support the survival of memory T and B cells (36–39). Therefore, we analyzed the expression levels of anti-apoptotic genes and autophagy genes in GC B and B_mem cells sorted from SpiB cKO and SpiB WT mice 14 days after immunization with NP-CGG. As a result, SpiB cKO GC B cells express lower levels of anti-apoptotic genes (Bcl-X_L , Bcl2, Mcl1) and some autophagy genes (Atg5, Atg6, Atg7) as compared to SpiB WT GC B cells and the difference was more profound in B_mem cells, except for Bcl2 ( Figures 6A, B ).

Autophagy is known to be necessary for the clearance of dysfunctional mitochondria that produce mitochondrial reactive oxygen species (mROS) (40) which is one of the factors to induce apoptosis (41). Given the almost undetectable levels of the autophagy genes, Atg5, Atg6, Atg7, in SpiB cKO B_mem cells, we assessed mROS production in GC B and B_mem cells sorted from immunized SpiB WT and SpiB cKO mice by staining them with MitoSox and analyzing by FCM. MitosSox positive cell population was small and similar between SpiB WT and SpiB cKO GC B and B_mem cells before culture (0 h), but the frequency of MitoSox-positive cells increased after 9 hours’ culture more abundantly in SpiB cKO GC B and B_mem cells than SpiB WT ones ( Figure 6C ). We next addressed autophagy activity in SpiB cKO GC B and B_mem cells by evaluating the expression of a known autophagy marker LC3B (42–44). When compared with those from SpiB WT mice, GC B and B_mem cells from SpiB cKO mice showed reduced expression of LC3B, indicating lower activity of autophagy in the absence of SpiB in these cells ( Figure 6D ). Finally, we found that IgG1⁺ B_mem cells in the immunized SpiB cKO mice contained more dead cells than those in the immunized SpiB WT mice after 9-hour incubation in the plain medium ( Figure 6E ). Taken together, these results indicate that the lack of SpiB in GC B and B_mem cells results in reduction of gene expression and activity for autophagy, leading to the accumulation of mROS and increased susceptibility to apoptosis.