1. Immunology and Inflammation
Download icon

The quantity of CD40 signaling determines the differentiation of B cells into functionally distinct memory cell subsets

  1. Takuya Koike
  2. Koshi Harada
  3. Shu Horiuchi
  4. Daisuke Kitamura  Is a corresponding author
  1. Tokyo University of Science, Japan
Research Article
  • Cited 0
  • Views 1,248
  • Annotations
Cite this article as: eLife 2019;8:e44245 doi: 10.7554/eLife.44245

Abstract

In mice, memory B (Bmem) cells can be divided into two subpopulations: CD80hi Bmem cells, which preferentially differentiate into plasma cells; and CD80lo Bmem cells, which become germinal center (GC) B cells during a recall response. We demonstrate that these distinct responses can be B-cell-intrinsic and essentially independent of B-cell receptor (BCR) isotypes. Furthermore, we find that the development of CD80hi Bmem cells in the primary immune response requires follicular helper T cells, a relatively strong CD40 signal and a high-affinity BCR on B cells, whereas the development of CD80lo Bmem cells does not. Quantitative differences in CD40 stimulation were enough to recapitulate the distinct B cell fate decisions in an in vitro culture system. The quantity of CD40 signaling appears to be translated into NF-κB activation, followed by BATF upregulation that promotes Bmem cell differentiation from GC B cells.

https://doi.org/10.7554/eLife.44245.001

Introduction

Memory B (Bmem) cells are crucial for humoral immunity, preventing the spread of re-infecting viruses and bacteria by rapidly producing large amounts of class-switched antibodies (Abs) against these pathogens. Bmem cells also regenerate themselves with improved affinity for their cognate antigen through the germinal center (GC) reaction upon each iterative infection. The rapid and exaggerated response of Bmem cells has been attributed to their B-cell receptor (BCR) isotype, namely membrane-bound IgG (mIgG). The relatively long cytoplasmic tail of mIgG, as compared to that of membrane-bound IgM (mIgM) which consists of only three amino acids, contains specific motifs that recruit signaling molecules such as Grb2 and SAP97 after BCR cross-linking; these motifs and molecules are required for enhanced BCR signaling and plasma cell (PC) formation upon rechallenge (Engels et al., 2009; Kaisho et al., 1997; Liu et al., 2012; Lutz et al., 2015). At odds with this model is the finding that naïve B cells expressing NP-specific mIgG1, which were derived from cloned mice generated from a single IgG1+ Bmem cell, expanded to a similar extent as NP-specific (IgH-knock-in) IgM+ naïve B cells, and both B cell types predominantly differentiated into GC B cells rather than PCs upon primary immunization (Kometani et al., 2013). These observations suggest that the heightened capacity of Bmem cells to differentiate into PCs cannot be attributed solely to the expression of mIgG1, but also depends on some cell-intrinsic status, such as a reduced expression of Bach2, a transcription factor that suppresses PC differentiation (He et al., 2017; Kometani et al., 2013). In support of this notion, human Bmem cells differentiate into PCs better than do naïve B cells in antigen-free in vitro cultures (Arpin et al., 1997).

As a further refinement in our understanding of the functional properties of Bmem cells, it has recently been proposed that the Bmem cell pool can be divided into distinct subsets on the basis of their potential to generate PCs or GC B cells upon antigen encounter. Earlier reports indicated that Bmem cells consist of mIgG+ cells and mIgM+ cells, with the former prone to becoming PCs and the latter GC B cells (Dogan et al., 2009; Pape et al., 2011), although a recent report suggests a more complex scenario (McHeyzer-Williams et al., 2015). Another study proposed that functionally different subsets can be phenotypically defined using surface markers, CD80 and PD-L2: CD80+ PD-L2+ Bmem cells are prone to differentiate into PCs, whereas CD80 PD-L2 Bmem cells preferentially form GC upon secondary immunization, and CD80 PD-L2+ Bmem cells are intermediate between the two but functionally closer to the CD80 PD-L2 Bmem cells (Zuccarino-Catania et al., 2014). Consistent with the BCR-isotype-based classification, the majority of IgG1+ Bmem cells had the CD80+ PD-L2+ phenotype, whereas IgM+ Bmem cells were dominated by CD80 PD-L2 cells. In addition, CD73 has been used to further dissect the PC-prone subpopulation as CD80+ PD-L2+ CD73+ Bmem cells (He et al., 2017).

Circulating memory T cells have also been functionally divided into two subsets termed effector memory T (TEM) and central memory T (TCM) cells. TEM cells, which lack the lymph node homing receptors CD62L and CCR7, produce abundant effector cytokines and cytotoxic granules, whereas TCM cells expressing both CD62L and CCR7 have a greater potential for proliferation and self-renewal (Mueller et al., 2013). The PC-prone Bmem cells may resemble the TEM cells in terms of their effector function, whereas the GC-B cell-prone Bmem cells are similar to the self-renewable TCM cells.

At present, it is not clear how these two subsets diverge during the primary immune response. It has been reported that CD80+ PD-L2+ Bmem cells have the highest affinity for antigen, whereas the double negative cells have the lowest affinity (Tomayko et al., 2010; Zuccarino-Catania et al., 2014). Moreover, it is generally accepted that B cells expressing BCR of higher affinity are prone to differentiate into PCs and those of lower affinity into GC B cells (Ochiai et al., 2013; Paus et al., 2006; Sciammas et al., 2011). Together with the analysis of the frequency of BCR somatic mutations and the timing of Bmem cell generation (Weisel et al., 2016), it has been suggested that PC-prone Bmem cells, including IgG1+ and CD80+ PD-L2+ Bmem cells, are mainly derived from the GC, whereas most of GC-B cell-prone Bmem cells, including IgM+ and CD80 PD-L2 Bmem cells, are generated before GC formation. It remains unclear, however, which signals determine the distinct Bmem cell fates. Thus, we sought to define the signaling mechanism for the bifurcated generation of Bmem cells.

Results

Characterization of CD80hi and CD80lo memory B cells

On the basis of previous reports showing that IgG1+ Bmem cells are mainly composed of CD80+ PD-L2+ and CD80 PD-L2+ Bmem cells and that CD80 PD-L2+ and CD80 PD-L2 Bmem cells are functionally similar (He et al., 2017; Zuccarino-Catania et al., 2014), we hypothesized that the proposed Bmem cell subsets could be distinguished simply by the expression of CD80, as CD80hi or CD80lo Bmem cells. Consistent with these reports, most of the CD80hi Bmem cells expressed PD-L2 and CD73, thus constituting the reported affinity-matured subset (He et al., 2017; Tomayko et al., 2010; Zuccarino-Catania et al., 2014), whereas the CD80lo Bmem cells consisted of PD-L2+ and PD-L2, as well as CD73+ and CD73 subpopulations (Figure 1a). Although CD62L was reported to be expressed on Bmem cells (Anderson et al., 2007), the majority of CD80hi Bmem cells express a lower level of CD62L (a phenotype that somewhat resembles TEM cells), whereas most CD80lo Bmem cells express a higher level of CD62L (a phenotype that somewhat resembles TCM cells). Both CD80hi and CD80lo Bmem cells expressed FAS (also called CD95 or APO1) at a higher level than naïve B cells, as previously reported (Anderson et al., 2007), but at a lower level than GC B cells, and, as expected, they were GL7.

Figure 1 with 1 supplement see all
Characterization of CD80hi and CD80lo memory B cells.

(a) Splenocytes from B6 mice immunized with NP-CGG in alum 4 weeks earlier were analyzed by flow cytometry (FCM). The gating strategy for naïve B cells (CD19+ IgM+), GC B cells (CD19+ IgM IgG1+ CD38), and CD80lo and CD80hi Bmem cells (CD19+ IgM IgG1+ CD38+) is shown in the top panels, and the expression of the indicated cell-surface proteins in each population is shown in the bottom panels. Data are representative of two independent experiments with similar results. (b) Outline of the experimental protocol (top). Splenic B cells from CD45.1 B1-8 ki mice were transferred into B6 mice (CD45.2), which were then immunized with NP-CGG in alum. Four weeks later, four subsets of donor-derived Bmem cells (CD45.1+ CD19+ CD38+), defined by the expression of IgG1 and CD80, were sorted from recipient spleens, cultured on 40LB feeder layers with IL-21 for 2 days, and analyzed by FCM. The frequency of CD138+ GL7 plasmablasts or PCs and CD138 GL7+ GC B cells in each subset is represented by a dot (bottom; combined data from two triplicate experiments). (c) Splenocytes from Cd4-Cre, Bcl6+/+ or Bcl6f/f mice immunized with NP-CGG in alum 6 weeks earlier were analyzed by FCM. The representative data indicate the gating strategy with percentages of the gated population. (d) The frequency (%) and absolute number (#) of CD80hi and CD80lo cells among IgG1+ Bmem cells in each spleen from individual recipient mice, as analyzed in (c) (n = 8). The mean of the values in each group is indicated by a horizontal bar (b, d). n.s., not significant (p>0.05); *, p<0.05; ***, p<0.001; ****, p<0.0001; as determined by one-way ANOVA followed by Tukey’s multiple comparisons test (b) and unpaired Student’s t test (d). All data are representative of two independent experiments, except (b and d), where data from two independent experiments are combined.

https://doi.org/10.7554/eLife.44245.002

In order to examine in vitro whether the CD80hi and CD80lo Bmem cells are intrinsically biased in their differentiation fate toward PCs or GC B cells, we transferred into B6 mice allotypically marked (CD45.1+) B cells of B1-8 knock-in (ki) mice, whose knock-in IgH chain, when combined with the λL chain, forms an NP-specific BCR, and immunized these mice with NP-CGG. From these mice, we sorted CD80hi and CD80lo Bmem cells, either IgG1+ or IgG1, and cultured them with IL-21 on feeder cells that express exogenous CD40L and BAFF (40LB) (Nojima et al., 2011; Takatsuka et al., 2018). Under these conditions, CD80hi Bmem cells differentiated more preferentially into CD138+ plasmablasts or PCs and less into GL7+ GC-like B cells, as compared with CD80lo Bmem cells, regardless of their BCR isotype (Figure 1b and Figure 1—figure supplement 1a,b). These in vitro data were consistent with the previous in vivo data (Zuccarino-Catania et al., 2014), and further revealed that the biased differentiation of the CD80hi or CD80lo Bmem cells is determined in a cell-intrinsic manner, and is essentially independent of BCR isotype and BCR affinity for antigen.

Strong CD40 signaling induced by TFH cells is required for the development of CD80hi Bmem cells

We next sought to clarify a need for GC in the development of CD80hi and CD80lo Bmem cells. A previous report indicated that CD80 and PD-L2 were expressed at normal levels on Bmem cells in B-cell-specific BCL6-deficient mice that lack GCs (Kaji et al., 2012). To examine a role for the GC environment in Bmem cell development from normal B cells, we used CD4+ T-cell-specific BCL6-deficient mice, which lack TFH cells and GCs (Kaji et al., 2012). Six weeks after immunization, the number of CD80hi Bmem cells decreased by approximately ten-fold in Cd4-Cre Bcl6f/f mice as compared to the control Cd4-Cre Bcl6+/+ mice, while the number of CD80lo Bmem cells was essentially unchanged (Figure 1c,d). These data suggested that the GC environment, or more specifically TFH cells, facilitate the development of CD80hi Bmem cells.

TFH cells differ from naïve or effecter CD4+ T cells in that they express a much higher level of CD40L (Breitfeld et al., 2000), as we confirmed using purified T-cell subsets (Figure 2a and Figure 2—figure supplement 1a). As previously indicated (Lenschow et al., 1994; Ranheim and Kipps, 1993), stimulation of B cells through CD40, but not BCR, induced CD80 expression in a time- and dose-dependent manner (Figure 2—figure supplement 1b,c). Supposing that CD80 expression on B cells that is induced during the primary response is maintained until the Bmem cell stage, we hypothesized that TFH cell stimulation through CD40 promotes CD80hi Bmem cell development. To test this, we first used CD40L-deficient mice as recipients of antigen-specific (BCR-knock-in) naïve B cells. After immunization of such mice, however, Bmem cell development from the donor B cells was abrogated altogether in the absence of CD40L, indicating that CD40L-mediated stimulation is indispensable for Bmem cell development (Figure 2b).

Figure 2 with 1 supplement see all
Strong CD40 signaling favors CD80hi Bmem cell development.

(a) Naïve T (CD4+ CD62L+ CXCR5 PD-1), effector T (CD4+ CD62L CXCR5 PD-1), and TFH (CD4+ CD62L CXCR5+ PD-1+) cells were sorted from spleens of mice immunized with NP-CGG in alum 7 days earlier and then stimulated with phorbol myristate acetate (PMA) and ionomycin for 2 hr. CD40L expression on each cell subset was analyzed by FCM (left) and represented as geometric mean fluorescence intensity (gMFI, right) (mean + s.d. of triplicates). (b) Cd40lg+/+ or Cd40lg−/− mice were transferred with splenic B cells from B1-8 ki mice, and immunized with NP-CGG in alum. Six weeks later, the frequency of donor-derived (CD45.1+) NP+ CD19+ B cells (representative data on the left) and the number of the donor-derived Bmem cells (CD19+ CD45.1+ NP+ CD38+; plotted on the right; n = 7) in each spleen were analyzed by FCM. (c, d) B6 mice were transferred with splenic B cells from B1-8 ki mice, immunized with NP-CGG in alum, and injected subcutaneously (s.c.) with an inhibitory CD40L (MR-1: 1 mg/kg) mAb (αCD40L) or an isotype-matched control (ctrl IgG) Ab every day from day −1 to day 5 after immunization. Ten days or 6 weeks after immunization, splenocytes of the recipient mice were analyzed by FCM. (c) Representative data (day 10) of the analysis showing the gating strategy. (d) The frequency (%) of CD80hi and CD80lo cells in the donor-derived, class-switched Bmem cells (CD45.1+ NP+ CD19+ CD38+ IgM), and their absolute numbers (#) at 10 days (top, n = 9) and 6 weeks (bottom, n = 8) after immunization are plotted. (e, f) B6 mice transferred with B1-8 ki B cells and immunized as in (c, d) were injected intraperitoneally (i.p.) with PBS or a stimulatory CD40 mAb (αCD40) (FGK4.5: 250 μg) at 8 days after immunization. Ten days after immunization, splenocytes from the recipient mice were analyzed by FCM. (e) Representative data of the analysis showing the gating strategy. (f) The frequency (%) and absolute numbers (#) of CD80hi and CD80lo cells in the donor-derived, IgG1+ Bmem cells (CD45.1+ NP+ CD138 CD19+ IgG1+ CD38+), and the numbers of the donor-derived GC B cells (CD45.1+ NP+ CD19+ IgG1+ CD38) or of plasmablasts (CD45.1+ NP+ CD138+) in splenocytes are plotted (n = 5). (g–i) B6 mice, co-transferred on day −1 with B1-8 ki B cells (1 × 105) and OT-II T cells (1 × 105) that had been transduced with control (shCtrl) or shCd40lg retroviral vectors on the previous day, were immunized with NP-OVA in alum. Ten days after immunization, spleen cells from the recipient mice were analyzed by FCM. (g) Outline of the experimental protocol. (h) Representative data showing the gating strategy. (i) The frequencies of CD80hi and CD80lo cells among the donor-derived, class-switched Bmem cells, defined as in (d) (n = 8). The mean of the values in each group is indicated by a horizontal bar (b, d, f, i). n.s., not significant (p>0.05); *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001; unpaired Student’s t test (b, d, f, i). All data are representative of two independent experiments except (b) and (i), where data from two independent experiments are combined.

https://doi.org/10.7554/eLife.44245.005

Next, we treated immunized mice with anti-CD40L blocking antibody (Ab) in a dose that we had determined only partially inhibited Ab production and GC B-cell formation (Figure 2—figure supplement 1d,e). This treatment preferentially affected CD80hi Bmem cell development. At 10 days after immunization, the frequency of CD80hi Bmem cells was significantly reduced among class-switched Bmem cells, of which ~90% were IgG1+ and ~10% were IgM IgG1, both normally containing the CD80hi cells at a similar frequency (Figure 2c,d and Figure 2—figure supplement 1f–h). The absolute number of the CD80hi Bmem cells was also severely reduced while there was only a moderate reduction in the absolute number of CD80lo Bmem cells, in a condition where the number of GC B cells was reduced by about ten-fold (Figure 2d and Figure 2—figure supplement 1e). The frequency and the number of the CD80hi Bmem cells were somewhat recovered by 6 weeks after immunization, presumably because of generation of such cells in the late GC after the lapse of the injected anti-CD40L Ab (Figure 2d). In order to focus on Bmem cell generation, avoiding possible effects of alteration in GC formation or Bmem cell maintenance, we hereafter mainly analyzed Bmem cells by 10 days after immunization, referring to previous reports (Suan et al., 2017; Wang et al., 2017; Weisel et al., 2016). As an opposite experiment, administration of agonistic anti-CD40 Ab to immunized mice markedly increased the frequency and the number of CD80hi Bmem cells, while the numbers of CD80lo Bmem cells, GC B cells and plasmablasts remained unchanged (Figure 2e,f).

Finally, antigen (NP)-specific B cells and carrier [ovalbumin (OVA)]-specific (OT-II) CD4+ T cells, which had been transduced with short hairpin (sh) RNA targeting CD40L (shCd40lg) or unrelated control (shCtrl), were co-transferred into B6 mice, which were then immunized with NP-OVA (Figure 2g). Among the generated donor-derived class-switched Bmem cells, the frequency of CD80hi Bmem cells was significantly lower in mice that had received the CD40L-knockdown T cells as compared to those that had received the control T cells (Figure 2g–i and Figure 2—figure supplement 1i). These data together suggest that, although CD40L-mediated stimulation is required for development of both Bmem cell types, stronger CD40 stimulation by TFH cells selectively facilitates the development of CD80hi Bmem cells.

CD40 signal strength in vitro affects differentiation into CD80hi or CD80lo Bmem cells in vivo

The data discussed above strongly suggested that quantitative differences in CD40 signaling in B cells during the primary response determine their developmental fate into either CD80hi or CD80lo Bmem cells. However, these in vivo experiments cannot exclude the possibility that some other factors that are affected by the CD40/CD40L manipulation might contribute to the fate decision. In order to demonstrate a direct contribution of CD40 signaling quantity in B cells, we utilized our in vitro induced GC B (iGB) cell culture system, in which naïve B cells massively proliferate, efficiently switch to IgG1, and differentiate into GC-like B (iGB) cells after being cultured with IL-4 on a feeder layer of 40LB cells (Nojima et al., 2011). In addition, these iGB cells differentiate into memory-like B cells [termed induced memory B (iMB) cells] in vivo when transferred into irradiated mice (Nojima et al., 2011). To stimulate B cells through CD40 at different levels in this in vitro system, we derived 40LB sublines that express CD40L at low, intermediate and high levels, termed 40LB-lo, 40LB-mid, and 40LB-hi, respectively (Figure 3a). As expected, B cells cultured on 40LB-hi (iGB-hi cells) expressed the highest level of CD80, whereas those on 40LB-mid (iGB-mid cells) or 40LB-lo (iGB-lo cells) exhibited intermediate or the lowest levels of CD80 expression, respectively (Figure 3b). The iGB-hi, iGB-mid and iGB-lo cells underwent class switching to IgG1 or IgE to similar extents and were mostly CD138 (Figure 3—figure supplement 1a). It is of note that the iGB-lo cells expressed Fas at a lower level than iGB-hi and iGB-mid cells, whereas they expressed CD38 and CD62L at higher levels than iGB-hi and iGB-mid cells, perhaps reflecting their relatively lower activation status (Figure 3—figure supplement 1b).

Figure 3 with 1 supplement see all
The quantity of CD40 signaling determines the differentiation of B cells into distinct Bmem cell subsets.

(a) Expression of CD40L on 3T3-BAFF cells and 40LB sublines (40LB-lo, 40LB-mid, and 40LB-hi) was analyzed by FCM. (b) Splenic B cells were cultured with IL-4 for the indicated number of days on feeder layers of each 40LB subline. The expression of CD80 on the expanded B (iGB) cells was analyzed by FCM and presented as gMFI (mean of triplicates). (c) A schematic representation of a method used to generate the induced memory B (iMB) cells. Splenic B cells from CD45.1+ congenic B6 mice were cultured for 4 days, as in (b). The resultant iGB-lo, iGB-mid, or iGB-hi cells were transferred intravenously (i.v.) into γ-irradiated mice (CD45.2+), and the donor-derived Bmem-like cells (CD19+ CD45.1+ CD38+) detected in the recipient spleens 2 weeks after the transfer were designated iMB-lo, iMB-mid or iMB-hi cells, respectively. (d–f) Expression of CD80 on IgG1+ or IgG1 iMB cells (iMB-lo, iMB-mid or iMB-hi) generated as in (c) was analyzed by FCM. (d) Representative data showing the gating strategy. (e) The frequencies of CD80hi and CD80lo cells among the IgG1+ or IgG1 iMB cells (%) and their absolute number per spleen (#) are plotted (n = 8). (f) Expression of the indicated surface markers on the recipient total B cells (CD45.1 CD19+), spontaneous GC B cells (CD45.1 CD19+ CD38 GL7+), the iMB-lo and the iMB-hi cells (CD19+ CD45.1+ CD38+). The mean of the values in each group is indicated by a horizontal bar (e). n.s., not significant (p>0.05); *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001; as determined by one-way ANOVA followed by Tukey’s multiple comparisons test (e). All data are representative of two independent experiments except (e), where data from two independent experiments are combined.

https://doi.org/10.7554/eLife.44245.008

These iGB cells were then transferred into irradiated mice and, 2 weeks later, B cells in the spleen were analyzed by FCM (Figure 3c). The B cells derived from all of these iGB cells were mostly CD38+ (iMB) cells and contained similar percentages of IgG1+ cells. However, the iMB cells derived from iGB-hi cells (iMB-hi cells) contained almost exclusively CD80hi cells, whereas iMB cells from iGB-lo cells (iMB-lo cells) were dominated by CD80lo cells, irrespective of their BCR isotypes (IgG1+ or IgG1). iMB cells derived from iGB-mid cells exhibited an intermediate phenotype (Figure 3d,e). In addition, iMB-hi cells expressed PD-L2, CD73 and FAS at higher levels than iMB-lo cells, whereas both expressed equally low levels of GL7 (Figure 3f). Thus, iMB-hi and iMB-lo cells phenotypically resembled CD80hi and CD80lo Bmem cells, respectively, that are generated in a physiological immune response (Figure 1a).

iGB-hi cells grow far more extensively than iGB-lo cells beyond two days of culture (Figure 3—figure supplement 1c). As cell cycling may cause epigenetic and transcriptional changes, it is possible that different cell cycles caused by different CD40 signaling strength affected the bifurcated Bmem cell development. To address this issue, iMB cells were generated by transferring iGB-lo and iGB-hi cells on day 2 of culturing, when these cells had expanded to similar levels. As a result, about 80% of iMB cells derived from the day 2 iGB-hi cells were CD80hi, whereas about 20% from iGB-lo cells were CD80hi. These results are reminiscent of those from iMB cells generated from the day 4 iGB cells, except that the iMB cells from day 2 iGB cells contained fewer IgG1+ and fewer PD-L2+ cells (Figure 3—figure supplement 1e,f). iGB-hi cells, rather than iGB-lo cells, tended to dominate the generation of CD80hi iMB cells, even when we used day 1 iGB cells (Figure 3—figure supplement 1g). These data indicate that the bifurcated Bmem cell fate was not determined by different levels of cell proliferation. However, they also demonstrated that the frequency of the CD80hi population in iMB-hi cells increased as the culture period of iGB-hi cells became longer (i.e. from 1 to 2 to 4 days), thus indicating that the duration of CD40 signaling may also affect the development of CD80hi Bmem cells. Although IL-21 is a hallmark TFH cytokine, which is known to support GC B cell proliferation, the addition of IL-21 to the iGB-hi and iGB-lo cell culture did not affect the frequencies of CD80hi iMB cells derived from each type of iGB cells (Figure 3—figure supplement 1h,i), suggesting that the contribution of IL-21 to the differential Bmem cell fate decision in vivo is less likely.

In order to examine whether the iMB-hi and iMB-lo cells recapitulate the functional differences seen in CD80hi and CD80lo Bmem cells, we analyzed their differentiation in culture on the 40LB feeder cells with IL-21. Regardless of their BCR isotype, iMB-lo cells preferentially differentiated toward GC B cells, as clearly seen on day 2, whereas iMB-hi cells preferentially differentiated into plasmablasts or PCs, becoming evident on day 3. The behavior of the iMB-mid cells was intermediate (Figure 4a,b and Figure 4—figure supplement 1). Next, we investigated the in vivo fate of these iMB cells in response to immunization with a cognate antigen. We sorted allotypically marked NP-binding iMB-lo and iMB-hi cells, derived from iGB-lo and iGB-hi cells of B1-8ki Igκ−/− mice, respectively, and co-transferred the iMB-lo and the iMB-hi cells at an equal ratio into B6 mice together with carrier (CGG)-primed T cells. The recipient mice were immunized with NP-CGG, and their spleen cells were analyzed 4 or 10 days later by FCM (Figure 4c,d). Four days after immunization, most donor-derived, NP-binding plasmablasts were derived from iMB-hi cells. By contrast, the vast majority of donor-derived, NP-binding GC B and Bmem cells at day 10 were found to originate from iMB-lo cells (Figure 4e,f). These in vitro and in vivo data together indicate that iMB-hi and iMB-lo cells functionally represent CD80hi and CD80lo Bmem cells, respectively. Taken together, these data indicate that the quantity of CD40 signaling in B cells determines their differentiation fate toward phenotypically and functionally distinct Bmem cell subsets.

Figure 4 with 1 supplement see all
Primary CD40 signaling strength affects secondary Bmem cell differentiation, either to PCs or to GC B cells.

(a, b) Splenic B cells from each recipient mouse (containing iMB cells), generated as in Figure 3 (d), were cultured on 40LB feeder layers with IL-21 for 2 (a) or 3 (b) days. The expression of GL7 (a, b) and CD138 (b) on gated IgM+ or IgG1+ CD45.1+ (iMB cell-derived) cells was analyzed by FCM. (c–f) iMB-lo and iMB-hi cells were generated from B1-8 ki Igκ−/− CD45.1/CD45.2 or B1-8 ki Igκ−/− CD45.1 iGB cells, respectively, as in Figure 3 (d). 2.5 × 104 (for ‘day 4’) or 1 × 104 (for ‘day 10’) of NP+ iMB-lo and iMB-hi cells were mixed and co-transferred into WT B6 recipient mice with 1 × 107 CGG-primed splenocytes. The recipient mice were then immunized with NP-CGG in alum and analyzed 4 or 10 days after immunization. (c) A schematic of the experimental procedure. (d) A representative FCM profile of the mixture of iMB-lo (CD45.2+) and iMB-hi (CD45.2) cells, gated on CD45.1+ NP+ cells, before the transfer. (e) Representative FCM data at day 4 and day 10 after immunization showing the gating strategy. (f) The frequencies of iMB-lo- and iMB-hi-derived cells among CD45.1+ NP+ CD138 or CD138+ cells at day 4, and among CD45.1+ NP+ Bmem cells (CD138 GL7 CD38+) or GC B cells (CD138 GL7+ CD38) at day 10. The mean of the values in each group is indicated by a horizontal bar (f). n.s., not significant (p>0.05); **, p<0.01; ****, p<0.0001; as determined by paired Student’s t tests (f). All data are representative of two independent experiments.

https://doi.org/10.7554/eLife.44245.011

Higher BCR affinity for antigen favors the development of CD80hi Bmem cells

On the basis of the data described above, it is likely that the CD40 signaling quantity is primarily determined by the expression level of CD40L on cognate T cells. As CD40L on T cells was shown to be induced in an antigen-dose-dependent manner (Jaiswal and Croft, 1997), it seemed plausible that the quantity of antigen presented on B cells would determine the expression level of CD40L on cognate T cells, which in turn determines the differentiation fate toward each Bmem subset. To confirm that antigen presentation by B cells induces CD40L on cognate T cells in a dose-dependent manner, OT-II-mouse-derived activated T cells were co-cultured with B cells and various concentrations of OVA peptide (Figure 5a and Figure 5—figure supplement 1). CD40L expression on the T cells was rapidly induced and its levels positively correlated with antigen dose (Figure 5b), as was also the case for CD80 expression on B cells on day 2 (Figure 5c). This CD80 induction was suppressed by blocking with anti-CD40L mAb, confirming that the CD40L–CD40 interaction leads to CD80 induction on B cells (Figure 5d).

Figure 5 with 1 supplement see all
High-affinity B cells preferentially differentiate into CD80hi Bmem cells, possibly through stronger induction of CD40L on cognate T cells.

(a–d) OT-II-derived Th0 cells and splenic B cells were co-cultured with the indicated concentration of OVA peptide for 6 hr or 2 days and analyzed by FCM. (a) An outline of the procedure for the T-B co-culture (see Materials and methods). (b) Expression levels of CD40L on CD4+ T cells after 6 hr co-culture are presented as gMFI (mean + s.d. of triplicates). (c) Expression levels of CD80 on CD19+ B cells after 2 days co-culture are presented as gMFI (mean + s.d. of triplicates). (d) Expression levels of CD80 on B cells after 2 days co-culture with 5 μM OVA peptide in the presence of the indicated concentration of anti-CD40L blocking Ab. Data are shown as in (c). (e, f) 1 × 105 NP+ splenic B cells from B1-8hi ki (CD45.1/45.2) or B1-8 ki (CD45.1) mice were co-transferred into the recipient B6 mice, which were immunized with NP-CGG in alum on the next day and analyzed by FCM 7 days later. (e) Representative FCM data showing the gating strategy. (f) The frequency (%) of CD80hi among IgG1+ Bmem cells (CD19+ CD38+) derived from either B1-8hi ki or B1-8 ki cells is plotted (n = 7). (g–j) B6 mice transferred with B1-8 ki B cells and immunized as in (e, f) were analyzed by FCM at 10 days after immunization. (g, i) Representative FCM data showing the gating strategy. (h, j) The frequencies (%) of CD80hi cells among NPmedhi or NPmedlo cells in Igλ+ (h) or IgG1+ (j) Bmem cells (CD19+ CD45.1+ CD38+), gating of each as shown in (g) and (i), respectively (n = 8). The mean of the values in each group is indicated by a horizontal bar (f, h, j). ****, p<0.0001; as determined by paired Student’s t tests (f, h, j). All data are representative of two independent experiments except (f, h, j), where data from two independent experiments are combined.

https://doi.org/10.7554/eLife.44245.014

During an immune response, B cells expressing a high-affinity BCR would take up more of the cognate antigen, and thus would present a larger amount of antigenic peptide–MHC complex to cognate T cells (Schwickert et al., 2011). This would lead to greater CD40L induction than would occur on T cells interacting with B cells with a lower affinity BCR. To investigate the correlation between BCR affinity and B cell differentiation fate, we used B1-8hi ki mice whose λ+ B cells express BCR with ten-fold higher NP affinity than BCR expressed on λ+ B1-8 ki B cells (Allen et al., 1988). B cells from B1-8hi ki and B1-8 ki mice, expressing discriminating allotypic markers, were co-transferred into B6 mice, and immunized with NP-CGG in alum. FCM analysis on day 7 after the immunization revealed that IgG1+ Bmem cells developed from B1-8hi ki B cells contained a higher frequency of CD80hi cells than those from B1-8 ki B cells (Figure 5e,f).

In the next experiment, we stained Bmem cells with NPmed-APC, allophycocyanin (APC) conjugated with NP at a relatively lower valency, which only binds to high affinity anti-NP BCRs (Nishimura et al., 2011). Mice were transferred with B1-8 ki B cells, immunized with NP-CGG in alum, and analyzed by FCM 10 days later. Among donor-derived Igλ+ or IgG1+ Bmem cells, those stained more brightly with NPmed (NPmedhi) contained a higher frequency of CD80hi Bmem cells than those stained less brightly (NPmedlo) (Figure 5g–j). These data indicated that B cells with higher antigen affinity preferentially differentiate into CD80hi Bmem cells rather than CD80lo Bmem cells, probably through more extensive antigen presentation to cognate T cells, which results in greater induction of CD40L. Alternatively, B cells with lower affinity may be excluded from the GC and therefore fail to access to CD40L on TFH cells. In any case, BCR affinity appears to be a primary determinant for the differential Bmem subset development that is dependent on CD40 signaling quantity.

Possible mechanisms through which CD40 signaling facilitates GC B-cell differentiation into CD80hi Bmem cells

We next investigated the CD40 signaling mechanisms that are responsible for the development of the CD80hi Bmem cell subset. NF-κB is a typical transcription factor that are induced by CD40 stimulation (Berberich et al., 1994), and the p50/p65 heterodimer was reported to bind to the Cd80 gene locus and to induce CD80 expression in a B cell line after stimulation (George et al., 2006). In accord with these data, among the constitutively active (CA) forms of protein kinases that are known to be activated by CD40 stimulation, CA-IKKβ, an activator of the canonical NF-κB pathway (Mercurio et al., 1997), but not CA-Akt or CA-MKK4, upregulated CD80 expression on iGB-lo cells (Figure 6a). Stimulation of splenic B cells with a higher concentration (10 μg/ml), but not a lower concentration (1 μg/ml), of anti-CD40 Ab induced the nuclear translocation of the NF-κB subunits c-Rel and RelA (Figure 6b). Furthermore, knockdown of Rel and Rela gene expression in iGB-hi cells resulted in downregulation of surface CD80 expression (Figure 6c,d and Figure 6—figure supplement 1a). These data indicated that CD80 expression on B cells induced by strong CD40 signaling is mediated by NF-κB, c-Rel and RelA. We next examined whether a similar signaling pathway is involved in the generation of Bmem cells. By using the iGB cell system, we showed that c-Rel- and Rela-knockdown iGB-hi cells generated fewer CD80hi iMB cells than mock-treated iGB-hi cells in vivo (Figure 6c–f and Figure 6—figure supplement 1b). Furthermore, we transferred NP-specific B cells transduced with the Rela-knockdown vector or a mock vector into mice, which were then immunized with NP-CGG. The knockdown of Rela selectively suppressed the development of CD80hi Bmem cells among the B cells that had responded to an immunized antigen (Figure 6g,h and Figure 6—figure supplement 1c). These data indicated that canonical NF-κB signaling plays a role in CD80hi Bmem cell development that is facilitated by stronger CD40 stimulation.

Figure 6 with 1 supplement see all
NF-κB signaling is involved in CD80hi Bmem cell development.

(a) Constitutively active (CA) variants of Akt, IKKβ or MKK4 were retrovirally transduced into B cells cultured on 40LB-lo feeder cells (iGB-lo cells) on day 2 of the culture. The expression of CD80 on the gated IgG1+ CD138 infection-marker-positive iGB-lo cells was then analyzed by FCM on day 5. The number on each histogram indicates gMFI. (b) Cytoplasmic and nuclear lysates from B cells stimulated with CD40 mAb (1 or 10 mg/ml) for the indicated time periods were analyzed by immunoblotting using Abs against c-Rel and p65/RelA. Tubulin and Lamin B were used as loading controls for cytoplasmic or nuclear proteins, respectively. (c) B cells cultured on 40LB-hi feeder cells (iGB-hi cells) were transduced with shCtrl, shRel, or shRela retroviral vectors, each carrying a GFP gene as an infection marker. Three days after the transduction, the expression of Rel and Rela mRNA in the sorted GFP+ cells was analyzed by qRT-PCR (mean + S.D. of triplicates). (d) Expression of CD80 on the GFP+ IgG1+ CD138 iGB-hi cells analyzed by FCM at 3 days after gene transduction, as in (c). (e, f) The iGB-hi cells transduced with the knock-down constructs as shown in (c,d) were transferred into γ-irradiated mice, spleens of which were analyzed by FCM 2 weeks later. (e) Outline of the experimental procedure. (f) Representative FCM data showing the the expression of CD80 (above) and the frequency (%) of CD80hi cells (bottom; n = 3) in the gene-transduced iMB cells (CD19+ CD45.1+ CD38+ GFP+) formed in the recipients’ spleens. (g, h) In vivo activated B cells derived from B1-8hi ki mice were transduced with shCtrl or shRela vectors, as described in the Materials and methods, and the resultant B cells (1 × 106) were transferred into WT B6 mice. The recipient mice were immunized with NP-CGG in alum on the next day. Splenocytes from these mice were analyzed by FCM at 10 days after immunization. (g) Outline of the experimental procedure. (h) Representative FCM data showing the gating strategy (left). The frequencies (%) of CD80hi cells among donor-derived, vector-transduced, and class-switched Bmem cells (CD45.1+ NP+ GFP+ IgM CD19+ CD38+) at 10 days after immunization (right; n = 4). The mean of the values in each group is indicated by a horizontal bar (f, h). *, p<0.05; **, p<0.01; ***, p<0.001; as determined by unpaired Student’s t tests. All data are representative of two independent experiments.

https://doi.org/10.7554/eLife.44245.017

It has been observed that CD40 stimulation induces IRF4 in B cells and that bone marrow-derived dendritic cells from IRF4-deficient mice express a reduced level of CD80 upon LPS stimulation (Saito et al., 2007; Suzuki et al., 2004). It was also reported that transient or intermediate expression of IRF4 induced GC-related genes through the formation of heterodimers with BATF or PU.1, whereas its sustained or high expression induced PC-related genes through an IRF4 homodimer (Ochiai et al., 2013). Thus, we examined whether these transcription factors are involved in CD40 signaling in GC B cells. When ex-vivo GC B cells were cultured with anti-CD40 or anti-BCR Abs, of either high or low doses, or with various cytokines, IRF4 expression was found to be upregulated by a high dose of anti-CD40 or anti-BCR Abs, whereas BATF expression was selectively upregulated by a high dose of anti-CD40 Ab (Figure 7a,b and Figure 7—figure supplement 1a). Then, we tested whether BATF and IRF4 are involved in the induction of CD80 by using tamoxifen-inducible ERT2-BATF or ERT2-IRF4 constructs. Induced activation of BATF selectively upregulated CD80 expression in iGB-lo cells, although IRF4 alone did not, and co-activation of BATF and IRF4 slightly enhanced CD80 expression (Figure 7c and Figure 7—figure supplement 1b,c). In addition, a BATF mutant (BATF- HKE), which is defective in IRF4 binding (Tussiwand et al., 2012), failed to upregulate CD80 expression regardless of exogenous IRF4 (Figure 7d), suggesting that the exogenous BATF formed a heterodimer with endogenous IRF4 for CD80 upregulation. These data together indicate that the BATF–IRF4 heterodimer that is induced by strong CD40 signaling enhances CD80 expression in activated B cells. As an IKKβ inhibitor suppressed CD40-induced expression of CD80 as well as of BATF and IRF4, the canonical NF-κB pathway appears to upregulate the expression of BATF and IRF4 (Figure 7e).

Figure 7 with 1 supplement see all
CD40-induced BATF may be involved in CD80hi Bmem cell development.

(a, b) GC B cells (CD19+ CD138 CD38 GL7+) were sorted from splenocytes of mice at 7 days after immunization with NP-CGG in alum and cultured without (−) or with the addition of the following reagents for 6 hr: anti-IgM plus anti-IgG Abs (αBCR, 10 μg/ml or 1 μg/ml), anti-CD40 Ab (αCD40, 10 μg/ml or 1 μg/ml), IL-4 (10 ng/ml), IL-21 (10 ng/ml) or BAFF (10 ng/ml). Expression of the indicated proteins in these B cells (CD19+ CD138) was analyzed by intracellular staining followed by FCM. (a) Representative FCM data. (b) gMFIs of the histograms shown in (a). Data are mean + s.d. of triplicates. (c) Splenic B cells cultured on 40LB-lo feeder cells (iGB-lo) were transduced with a mock vector or with the indicated vectors expressing each factor fused with ERT2 (generated as described in the Materials and methods) on day 2 of the culture, and then treated with vehicle (EtOH) alone or with 4-OHT from day 3 to day 5. The expression of CD80 on these cells was analyzed on day 5, and shown as in (a). (d) iGB-lo cells were transduced with the indicated combination of the ERT2–fusion vectors and treated with 4-OHT as in (c). The CD80 expression on these cells is shown as in (c). (e) Splenic B cells were cultured with anti-CD40 Ab (20 μg/ml) for 2 days without (−) or with an IKKβ inhibitor (BAY11-7082). Expression of the indicated proteins in these cells was analyzed by FCM. The shadowed histograms represent the cells cultured with medium alone. (f, g) Splenic B cells from B1-8 ki CD45.1 mice were transferred into WT B6 mice, which were immunized with NP-CGG in alum on the next day. At 10 days after immunization, splenocytes from the recipients were analyzed by FCM. (f) A representative data showing the gating strategy. (g) gMFI of the indicated proteins in the donor-derived GC B cells (CD19+ CD45.1+ IgM GL7+ Ephrin B1+ CD38), pre-Bmem cells (CD19+ CD45.1+ IgM GL7+ Ephrin B1+ CD38+), and Bmem cells (CD19+ CD45.1+ IgM GL7 CD38+) (n = 5). (h–j) iGB-lo cells were transduced with the retroviral vectors expressing ERT2-BATF-ires-GFP and ERT2-IRF4-ires-CFP (BATF-GFP +IRF4 CFP), or with empty vectors expressing GFP and CFP (mock) on day 2 of culture treated with 4-OHT from day 3 to day 5, and then transferred into γ-irradiated mice. Two weeks after the transfer, spleen cells of the recipient mice were analyzed by FCM. (h) Outline of the experimental procedure. (i) A representative data showing the gating strategy. (j) The frequency (%) of CD80hi cells in the ERT2-BATF gene-transduced iMB cells formed in the recipients’ spleens (CD19+ CD45.1+ CD38+ GFP+) (n = 6). The mean of the values in each group is indicated by a horizontal bar (g, j). *, p<0.05; **, p<0.01; as determined paired Student’s t tests. All data are representative of two independent experiments except (j), in which data from two independent experiments are combined.

https://doi.org/10.7554/eLife.44245.020

Considering that the NF-κB pathway upregulates CD80 expression on iGB cells and facilitates CD80hi Bmem cell development in vivo (Figure 6), it is possible that the BATF–IRF4 heterodimer plays a role in the strong CD40 signal that drives GC B cell differentiation into CD80hi Bmem cells. This idea was supported by our finding that GL7+ Efnb1+ CD38+ GC-derived memory precursors (pre-Bmem) cells (Laidlaw et al., 2017) expressed BATF and CD80 at higher levels than GL7+ Efnb1+ CD38 GC B cells at 10 days after immunization (Figure 7f,g and Figure 7—figure supplement 1d).

Finally, to investigate whether the BATF–IRF4 heterodimer is involved in the development of CD80hi Bmem cells, iMB cells were generated from iGB-lo cells transduced with both ERT2-BATF and ERT2-IRF4, which were then activated in the culture with 4-OHT (Figure 7h). In the iMB cells, ERT2-IRF4-expressing (CFP+) cells could not be detected, possibly because cells that expressed an excess amount of IRF4 had differentiated into PCs. On the other hand, ERT2-BATF-expressing (GFP+) iMB cells were present and contained a significantly higher proportion of CD80hi cells than mock-transduced iMB cells (Figure 7i,j and Figure 7—figure supplement 1e). Taken together, our data indicate that strong CD40 signaling is converted into the activation of NF-κB and the following upregulation of BATF, resulting in the generation of CD80hi Bmem cells.

Discussion

The regulation of the bidirectional response of Bmem cells, either to PCs or to GC B cells upon secondary antigen challenge, has recently been explained by defining the functionally different Bmem cell subsets. According to the report by Shlomchik and colleagues, CD80+ PD-L2+ Bmem cells preferentially differentiate into PCs, whereas CD80 PD-L2+ and CD80 PD-L2 Bmem cells differentiate to GC B cells (Zuccarino-Catania et al., 2014). As the class-switched Bmem cell population that we mainly focused on is mostly composed of CD80+ PD-L2+ and CD80 PD-L2+ cells, and also because CD80 PD-L2+ and CD80 PD-L2 Bmem cells are functionally similar to each other (Zuccarino-Catania et al., 2014), we reasoned that the proposed subsets can simply be distinguished by CD80 expression as CD80hi and CD80lo Bmem cells, a distinction that we applied in this study to make a multi-color FCM analyses easier. The CD80hi Bmem cells were mostly PD-L2+, CD73+, CD62Llo, whereas CD80lo Bmem cells included PD-L2+ and PD-L2 cells, CD73+ and CD73 cells, and mostly CD62Lhi, which was largely consistent with previous reports (Dogan et al., 2009; He et al., 2017; Pape et al., 2011; Zuccarino-Catania et al., 2014).

Our data showing preferential differentiation of the CD80hi and CD80lo Bmem cells into plasmablasts/PCs and GC B cells in vitro, respectively, was also consistent with the previous in vivo data showing the preferential differentiation of each Bmem cell subset during the recall response. Reports showing that Bmem cells corresponding to the CD80hi Bmem cells have BCRs with higher antigen affinity than those corresponding to the CD80lo Bmem cells, and that high affinity B cells are more prone to become PCs, seemed to suggest that BCR-signal strength determines the fates of CD80hi and CD80lo Bmem cells upon the secondary challenge (Phan et al., 2006; Zuccarino-Catania et al., 2014). However, our in vitro culture system without specific antigens clearly showed that the distinct fates of these Bmem subsets upon re-stimulation are not determined by the BCR affinity or isotype, although they might be affected by slightly different levels of CD40 expression on CD80hi and CD80lo Bmem cells (Figure 2—figure supplement 1j). Thus, cell status represented by, for example, transcriptomic or epigenetic profiles may largely define the function of each Bmem subset (He et al., 2017; Kometani et al., 2013; Zuccarino-Catania et al., 2014).

Previously it was reported that GC depletion by anti-CD40L mAb treatment reduced the frequency of CD80hi Bmem cells, and that the transcriptomic signature of CD80hi Bmem cells more closely correlated with CD40-stimulated B cells than did that of CD80lo Bmem cells (He et al., 2017; Weisel et al., 2016). We demonstrated that the development of CD80hi Bmem cells was largely dependent on the presence of TFH cells, which express CD40L at a markedly higher level than naïve or effector T cells, and that partial blocking by anti-CD40L Ab or knock-down of CD40L on CD4 T cells during the primary response dominantly affected the generation of CD80hi Bmem cells rather than CD80lo Bmem cells. Conversely, in vivo stimulation with anti-CD40 Ab during the primary response increased the number of CD80hi Bmem cells but not the numbers of CD80lo Bmem cells nor GC B cells. As CD40L in the recipient mice was essential for the generation of both types of Bmem cells from transferred B cells, we hypothesized that the generation of either CD80hi or CD80lo Bmem cells is determined by a difference in the quantity of CD40 signaling.

This hypothesis was strongly supported by our simplified experimental system, which enables in vivo generation of memory-like B (iMB) cells without immunization from naive B cells cultured on feeder cells (40LB) and transferred into mice. Using 40LB cells expressing different levels of CD40L, we clearly showed that stronger in vitro stimulation via CD40 promoted the generation of CD80hi iMB cells, whereas weaker stimulation facilitated the generation of CD80lo iMB cells from B cells with essentially the same BCR repertoire of specificity and isotype. The CD80hi and CD80lo iMB cells phenocopied the CD80hi and CD80lo Bmem cells, respectively, in that CD80hi iMB cells preferentially differentiate into plasmablasts/PCs and CD80lo iMB cells into GC B cells (and Bmem cells in vivo) in ex-vivo culture and after antigen challenge in vivo. The commitment to differentiate into either CD80hi or CD80lo iMB cells, as determined by the different quantities of CD40 signaling in B cells, was made within two days of culture when the proliferation did not differ among the conditions, and partially made in just one day when CD80 expression was hardly detectable on B cells. Therefore, it seems that the quantity of CD40 signaling in B cells directs cellular programming, which determines the differentiation into distinct Bmem cell subsets.

Although the B cells cultured on 40LB feeder cells (iGB cells) mimic some aspect of GC B cells, naturally they differ from genuine GC B cells in that iGB cells are uniformly proliferating, are CD80+ (to distinct levels depending on the strength of CD40 stimulation), and do not mutate Ig genes. We consider that the iGB cells that were primarily cultured with IL-4 may represent a certain state of T-cell-activated B cells that are destined to become Bmem cells, such as naïve B cells that are activated in the initial phase of primary response or GC B cells that have just undergone selection as what we call pre-Bmem cells. Therefore, our data describing the in vivo differentiation of iGB cells into Bmem-like iMB cells probably explain the mechanism for the induction of bidirectional Bmem cell differentiation upon T-B interactions, during both the early phase and the GC phase of the primary immune response. Thus, in either phase, B cells receiving relatively stronger CD40 signaling are committed to CD80hi Bmem cells, whereas those receiving weaker CD40 signaling are committed to CD80lo Bmem cells. Our additional data suggest that, in the early phase, B cells expressing BCR with higher affinity to antigen present more antigenic peptide to cognate T cells, thus inducing CD40L on these T cells more strongly so that they acquire stronger CD40 signaling than lower affinity B cells. Similarly, in the GC phase, higher-affinity GC B cells dominantly present antigenic peptide, and will acquire more frequent and more durable interactions with TFH cells that express a high level of CD40L (Figure 8).

Proposed model for the generation of CD80hi and CD80lo Bmem cells.

(Top) In the pre-GC phase of the primary response, the BCR affinity to antigen or the amount of available antigen determine the quantity of antigen presentation to T cells, and the extent of the induction of CD40L on T cells. Thus, the strength of CD40 signaling in B cells is determined by the interacting T cells, which then directs the differentiation fate to distinct Bmem subsets: relatively stronger CD40 signal commits B cells towards CD80hi Bmem cells, whereas weaker CD40 signal commits B cells towards CD80lo Bmem cells. After GC formation, TFH cells, being able to express a high level of CD40L after TCR stimulation, strongly stimulate CD40 on relatively high-affinity B cells and facilitate their differentiation to CD80hi Bmem cells. (Bottom) Activation of NF-κB, and the downstream BATF–IRF4 heterodimer, may transmit the strong CD40 signaling into a mechanism that facilitates the differentiation towards CD80hi Bmem cells.

https://doi.org/10.7554/eLife.44245.023

Thus, stronger CD40 signaling in the GC may direct the development of CD80hi Bmem cells, as supported by our data. Despite this supposition, it was reported that strong T cell help and CD40 signaling in vivo induce the differentiation of GC B cells into PCs (Ise et al., 2018; Schwickert et al., 2011), raising a question as to the mechanism for the differentiation of GC B cells into either CD80hi Bmem cells or PCs. The fact that CD40 stimulation suppresses PC generation in vitro (Hawkins et al., 2013; Randall et al., 1998; Satpathy et al., 2010) suggests that the strong CD40 signaling alone does not directly promote PC differentiation in the GC. It is possible that additive BCR signaling affects the differentiation of GC B cells into PCs (Kräutler et al., 2017), although it has been reported that BCR signaling is inactive in most GC B cells (Khalil et al., 2012). Supposing that TFH cells are heterogeneous in terms of cytokine production (Weinstein et al., 2016), a cytokine produced from a particular TFH cell subset that interacts with the GC B cells may play a key role. Thus, a combination and integration of signaling pathways, one from the strong CD40 stimulation, the other from a particular cytokine, and maybe more, may ultimately determine the fate of GC B cells. IL-21 is known to induce B cell differentiation into PCs, while IL-21R-defficiency attenuated PC development and accelerated Bmem cell development (Zotos et al., 2010). In addition, we previously reported that iGB cells that were secondarily cultured with IL-21 preferentially develop into bone marrow PCs but not Bmem cells in vivo after adoptive transfer (Nojima et al., 2011). Thus, when GC B cells interact with IL-21-producing TFH cells, and receive a strong CD40 signal, they will differentiate into PCs. A TFH cell subset that induces the differentiation of GC B cells into CD80hi Bmem cells has yet to be defined. IL-4-producing TFH cells may be this subset, because iGB cells that are cultured with IL-4 on the CD40Lhigh feeder preferentially differentiated into CD80hi iMB cells in vivo. Considering a report showing that Bmem cells develop from B cells in the earlier GC, whereas long-lived PCs are generated during the later GC (Weisel et al., 2016), it is possible that distinct TFH subsets may work dominantly in B cell selection along the time course of the GC reaction.

Our data demonstrating that CD80lo Bmem cell generation was little affected by the absence of GC resulting from TFH cell-deficiency is consistent with a report indicating that the majority of CD80lo PD-L2 Bmem cell cells were generated prior to GC formation (Weisel et al., 2016). Given that low-affinity B cells do enter into the GC reaction and could maintain their low affinity even after mutation, why are only few CD80lo Bmem cells generated during the GC phase? It has been shown that half of GC B cells undergo apoptosis every 6 hr (Mayer et al., 2017), and that this response can be avoided by CD40 signaling (Luo et al., 2018; Mayer et al., 2017). These data suggest that weaker CD40 signaling in the GC phase may not be enough to prevent the apoptosis of B cells and therefore could fail to induce CD80lo Bmem cell development. Alternatively, lower-affinity B cells may be excluded from the GC in a competitive situation (Schwickert et al., 2011), and therefore fail to receive a strong CD40 stimulation from TFH cells.

As mentioned earlier, CD80hi and CD80lo Bmem cells phenotypically and functionally resemble effector memory T (TEM) and central memory T (TCM) cells, respectively, in that TEM cells are CD62L and produce abundant effector cytokines upon an antigen re-challenge, whereas TCM cells are CD62L+ and have a greater potential for proliferation (Mueller et al., 2013). Our finding that CD40 signal strength directs the generation of CD80hi or CD80lo Bmem cells also resembles mechanistic aspects of the current model for the generation of TEM and TCM cells: stronger TCR signaling favors TEM cells, whereas weaker TCR signaling favors the generation of TCM cells (Daniels and Teixeiro, 2015). It has been reported that the commitment to CD4+ TEM or TCM cells is determined by the expression of T-bet and BCL6 transcription factors, respectively (Pepper et al., 2011), and that low-affinity TCR signaled greater induction of BCL6 expression but less expression of T-bet compared to high-affinity TCR (Knudson et al., 2013). Therefore, it is likely that TCR affinity/signal strength determines the direction of differentiation to distinct Tmem subsets, through the induction of distinctive transcription factors (Daniels and Teixeiro, 2015).

Previous studies and our observations imply transcriptomic predisposition of CD80hi and CD80lo Bmem cells that may account for their preferential differentiation upon re-stimulation, and suggest that the transcriptomic statuses may be established during the primary response as proposed for the T cell memory. It has been demonstrated that the CD40-NF-κB-IRF4 pathway represses the transcription of Bcl6 (Saito et al., 2007), and that Bcl6 mRNA is more abundantly expressed in CD80 PD-L2 cells than in CD80+ PD-L2+ cells (Zuccarino-Catania et al., 2014). Combined with our data suggesting that NF-κB and the downstream IRF4–BATF heterodimer play a role in generation of CD80hi Bmem cells, the development to CD80hi and CD80lo Bmem cells is determined by the balance of the expression levels of transcription factors such as IRF4, BATF and BCL6, which are regulated by CD40 signaling quantity.

Canonical NF-κB mainly consists of heterodimer p50/c-Rel or p50/RelA (Jost and Ruland, 2017). It has recently been proposed that c-Rel and RelA have different roles in late B cell development: c-Rel promotes proliferation, whereas RelA upregulates Blimp1, a master regulator of PCs (Heise et al., 2014; Roy et al., 2019). Our data indicate that RelA is contributes more to the generation of CD80hi iMB cells than does c-Rel, which may be involved in determining the nature of CD80hi Bmem cells that are predisposed to PC development. Although RelA and c-Rel may function redundantly to some extent to generate CD80hi Bmem cells, our data showing reduced generation of CD80hi iMB cells by c-Rel knockdown alone also suggests a unique role for c-Rel.

What would be the survival advantage for individuals of generating bifurcated Bmem cell subsets upon infection? CD80hi Bmem cells express high-affinity BCR (Zuccarino-Catania et al., 2014) and therefore produce high-affinity Abs. On the other hand, CD80lo Bmem cells mainly express low-affinity BCRs, but have a greater proliferative potential and preferentially generate secondary GC upon reencounter with antigen, when they can further diversify their BCR repertoire. Therefore, CD80lo Bmem cells, unlike high-affinity CD80hi Bmem cells, can cope with a broader array of epitopes that may be generated by pathogens through mutations. Taken together, robust and rapid high-affinity Ab production by CD80hi Bmem cells eliminates the majority of reinfecting pathogens, whereas pathogens that escape destruction because of epitope changes resulting from mutations are eliminated by Abs derived from CD80lo Bmem cells that evolved in GCs. Thus, if we could dissect the CD40 signaling pathways that direct differentiation into CD80hi Bmem cells and those promote cell proliferation or survival, selective suppression of the former pathways during vaccination might convert the generation of CD80hi Bmem cells into CD80lo Bmem cells that would eventually produce broadly reactive Abs. Alternatively, additive stimulation of CD40 with proper timing would facilitate the generation of CD80hi Bmem cells that rapidly produce highly specific Abs upon actual infection.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Genetic
reagent (M. musculus)
B1-8kiLam et al., 1997Dr. Rajewsky (Max Delbrück Center for Molecular Medicine)
Genetic reagent (M. musculus)B1-8hikiShih et al., 2002IMSR Cat# JAX:007775; RRID:IMSR_JAX:007775Dr. Nussenzweig (The Rockefeller University)
Genetic reagent (M. musculus)Bcl6floxKaji et al., 2012IMSR Cat# RBRC05663; RRID:IMSR_RBRC05663Dr. Takemori (RIKEN)
Genetic reagent (M. musculus)Cd4-CreLee et al., 2001IMSR Cat# JAX:017336; RRID:IMSR_JAX:017336Dr. Kubo
(Tokyo University of Science)
Genetic reagent (M. musculus)Cd40lg−/−Xu et al., 1994RRID:MGI:2449454Dr. Flavell (Yale School of Medicine)
Genetic reagent (M. musculus)Igκ−/−Chen et al., 1993Dr. Tsubata (Tokyo Medical and Dental University)
Genetic reagent (M. musculus)OT-IIBarnden et al., 1998IMSR Cat# JAX:004194; RRID:IMSR_JAX:004194Dr. Kubo (Tokyo University of Science)
Cell line (M. musculus)40LBNojima et al., 2011Dr. Kitamura (Tokyo University of Science)
Cell line (M. musculus)40LB-hiTakatsuka et al., 2018Dr. Kitamura (Tokyo University of Science)
Cell line (M. musculus)40LB-lothis paperDr. Kitamura (Tokyo University of Science)
Cell line (M. musculus)40LB-midthis paperDr. Kitamura (Tokyo University of Science)
Antibodyrat monoclonal anti-mouse CD40; FGK4.5Bio X CellBio X Cell Cat# BE0016-2; RRID:AB_1107647250 μg
Antibodyarmenian hamster monoclonal anti-mouse CD40L; MR-1Noelle et al., 1992ATCC Cat# HB-11048; RRID:CVCL_89641 mg/kg; Dr. Abe (Tokyo University of Science)
Sequence-based reagentshRNAthis paperSee Supplementary file 1
Software, algorithmFlowJohttps://www.flowjo.com/solutions/flowjoRRID:SCR_008520
Software, algorithmGraphPad Prismhttps://graphpad.comRRID:SCR_002798

Mice and immunization

Request a detailed protocol

C57BL/6 NCrSlc (B6) mice were purchased from Sankyo Labo Service. All of the following mice were backcrossed to B6 or congenic B6 CD45.1+ mouse strains: B1-8 ki (Lam et al., 1997), B1-8hi ki (Shih et al., 2002), Bcl6f/f (Kaji et al., 2012), Cd4-Cre (Lee et al., 2001), Cd40lg−/− (Xu et al., 1994), Igκ−/− (Chen et al., 1993), and OT-II (Barnden et al., 1998). Mice were immunized i.p. with 100 μg of NP32-CGG, or NP14-OVA where indicated, in alum. Sex-matched, 7-week-old or older mice were used for all experiments. All mice were bred and maintained under specific pathogen-free conditions, and all animal experiments were performed under protocols approved by the Animal Care and Use Committee of the Tokyo University of Science (Approval No.: S15021, S16019, S17004, S18018).

Flow cytometry

Request a detailed protocol

For all the flow cytometry (FCM) analyses, single-cell suspensions were depleted of red blood cells (RBC) by ammonium chloride lysis, blocked with anti-CD16/32 (FcγRII/III) Ab (2.4G2), and then stained with the appropriate mAbs listed in Supplementary file 1, in PBS supplemented with 0.5% BSA, 2 mM EDTA, and 0.05% sodium azide. Stained cells were analyzed using FACSCalibur or FACSCantoII (BD Biosciences) instruments. The data were analyzed using Flowjo (Tree Star). Dead cells, detected by using propidium iodide or Fixable Viability Dye (eBioscience), were gated out in all FCM experiments. For intracellular staining, cells were fixed and permeabilized using a Foxp3 staining kit (eBioscience) before staining.

Cell purification and culture

Request a detailed protocol

Naïve B cells were purified as described previously (Nojima et al., 2011). Naïve T cells were purified from OT-II mice as follows: RBC-depleted splenocytes were stained with fluorochrome-conjugated mAbs to CD4, CD25, CD44, and CD62L, and then naïve T cells (CD4+ CD25 CD44 CD62L+) were sorted using FACSAriaII or FACSAriaIII (BD Biosciences) instruments. GC B cells were purified from the mice immunized with NP-CGG in alum 7 days previously as follows. Cells from pooled spleens were stained with FITC-conjugated anti-GL7 and anti-FITC microbeads (Miltenyi Biotec), and GL7+ cells were enriched using a MACS system (Miltenyi Biotec). The enriched cells were stained with anti-CD19, anti-CD38 and anti-CD138, and then GC B cells (CD19+ CD38 CD138 GL7+) were sorted using FACSAriaII or FACSAriaIII.

B and T cells were cultured in 37°C/5% CO2 conditions in complete medium: RPMI-1640 medium (Wako) supplemented with 10% heat-inactivated fetal bovine serum, 1 mM sodium pyruvate, 50 μM 2-mercaptoethanol, 10 mM HEPES pH7.5, 100 U/ml penicillin and 100 μg/ml streptomycin (GIBCO). Naïve B cells (5 × 106/ml) were cultured with anti-CD40 (1C10; Southern Biotech) or IKKβ inhibitor (BAY11-7082; Merck). Sorted T cell subsets (2.5 × 105/ml) were cultured with PMA (20 ng/ml; Sigma) and ionomycin (1 μg/ml; Sigma) for 2 hr. To generate Th0 cells, naïve OT-II T cells (1 × 106/ml) were cultured in six-well plates (Corning) coated with anti-CD3ε (8 μg/ml; 145–2 C11; Biolegend) and anti-CD28 (8 μg/ml; 37.51; Biolegend) for 3 days, and then cultured without Abs for 1 day. The resultant Th0 cells (1 × 106/ml) and naïve B cells (1 × 106/ml) were co-cultured with OVA peptide (Figure 4a). GC B cells (1 × 106/ml) were cultured with anti-IgM (10 or 1 μg/ml; Jackson ImmunoResearch), anti-IgG (10 or 1 μg/ml; Jackson ImmunoResearch), anti-CD40 (1C10; 10 or 1 μg/ml), IL-4 (10 ng/ml; PeproTech), IL-21 (10 ng/ml; PeproTech) or BAFF (10 ng/ml) for 6 hr.

Adoptive transfer and memory B cell purification

Request a detailed protocol

Naïve B cells were purified from B1-8ki CD45.1 mice and the frequency of NP+ cells was determined by FCM. The naïve B cells containing 1 × 104 NP+ B cells per mouse were transferred into B6 mice, which were then immunized i.p. with NP-CGG in alum on the next day. Four weeks after the immunization, Bmem cells were purified from pooled spleens through two-step negative sorting and final positive sorting: cells stained with biotinylated antibodies against CD4, CD8a, CD11b, CD43, CD45.2, CD49b, and Ter119, followed by streptavidin particle DM (BD Biosciences), were negatively sorted sequentially with the iMag (BD) and MACS systems. The resultant cells were stained with fluorochrome-conjugated CD19, CD38, CD45.1 mAbs, and NP-BSA, and then Bmem cells (all positive) were sorted using FACSAriaII or FACSAriaIII.

In vivo administration of antibodies

Request a detailed protocol

To inhibit CD40 signaling, mice were injected s.c. with the antagonistic CD40L mAb MR-1, (in house; 30 μg per mouse) or with control IgG (IR-AHT-GF, Innovative Research), every day from day −1 to day 5 after immunization. To activate CD40 signaling, mice were injected i.p. with agonistic CD40 mAb (FGK4.5, Bio X Cell; 250 μg per mouse) or PBS at day 8 after immunization.

Cell lines, iGB cell culture and iMB cell generation

Request a detailed protocol

Production of cell lines, iGB cell culture and iMB cell generation were performed as previously described (Nojima et al., 2011). 40LB-hi cells were generated by repeated transduction of a CD40L expression vector (pMXs-CD40L-IRES-GFP) into 40LB cells (Takatsuka et al., 2018). 40LB-mid and 40LB-lo cells were subclones of 40LB cells made by cell sorting followed by limiting dilution. The CD40L expression level in 40LB-mid cells was equivalent to that of the parental 40LB cells. A parental cell line for 40LB, BALB/c 3T3 fibroblast (clone A31), was provided by RIKEN BRC, Japan. All the BALB/c 3T3-derived cell lines were checked routinely using the PCR Mycoplasma Detection Set (Takara) and proved to be mycoplasma-free.

Adoptive transfer of iMB cells for immunization

Request a detailed protocol

Splenocytes from mice that had been transfected with iGB cells 2 week earlier were analyzed by FCM to estimate the numbers of donor-derived (CD45.1+) iMB cells. To examine the response of the iMB cells to a NP antigen in vivo, spleen B cells, including a fixed number of iMB cells derived from B1-8 ki B cells, were co-transferred with CGG-primed spleen cells into WT B6 mice, which were immunized i.p. with NP-CGG in alum on the next day.

Plasmid constructions

Request a detailed protocol

BATF and IRF4 cDNAs were cloned by PCR using iGB cell mRNA. The ERT2 segment was fused to the 5′-terminus of the BATF or IRF4 cDNAs by ligations using PCR-generated de novo restriction enzyme sites. The BATF-HKE mutant (H55Q, K63D, and E77K) was generated by PCR-based mutagenesis (Iwata et al., 2017; Tussiwand et al., 2012). Constructs encoding BATF or BATF-HKE, each fused with ERT2, were cloned into a pMXs-IRES-GFP vector. A construct encoding IRF4 fused with ERT2 was cloned into a pMXs-IRES-CFP vector, derived from the pMXs-IRES-GFP, in which the GFP sequence was replaced with CFP. CA-IKKβ (S177E and S188E) (Mercurio et al., 1997), CA-Akt (E40K) (Arimura et al., 2004), and CA-MKK4 (S257E, T261D) constructs were cloned into the pMXs-IRES-GFP. For RNAi, the target sequences of shRNAs, as listed in Supplementary file 1, were inserted into a pSIREN-GFP vector, which was made by replacing a puromycin resistance gene in a pSIREN-RetroQ vector (Clontech) with an EGFP sequence.

Retroviral transduction

Request a detailed protocol

Retroviral transduction of iGB cells was performed as previously described (Haniuda et al., 2016). For T cells, naïve T cells were stimulated with plate-coated anti-CD3ε (8 μg/mL) and anti-CD28 (8 μg/mL) for 36 hr, and then transduced with retrovirus vectors by spin-infection (Haniuda et al., 2016). Retroviral transduction of in-vivo-activated primary B cells and their transfer into mice were performed as previously described (Inoue et al., 2017). In brief, B1-8hi ki mice were injected i.p. with NP-Ficoll (50 μg), and then B cells were purified from the spleens of these mice 6 hr later and stimulated in vitro with anti-CD40 Ab (2 μg/ml) for 18 hr. Cultured B cells were spin-infected with retroviral vectors and further cultured for 3 hr. The resultant viable B cells (1 × 106) were transferred into WT mice for immunization with NP-CGG.

Immunoblot analysis

Request a detailed protocol

Cells were lysed in cytoplasmic extraction (CE) buffer (10 mM HEPES pH 7.9, 10 mM KCl, 0.1 mM EDTA pH 8.0, 0.1 mM EGTA, and 1 mM DTT) for 10 min at 4°C and then NP-40 were added to the final concentration of 0.5%. The cell lysates were centrifuged and supernatants were collected as the cytosolic fraction. Precipitates were washed twice with CE buffer and the final precipitates were lysed in a nuclear extraction buffer (20 mM HEPES pH7.9, 400 mM NaCl, 1 mM EDTA pH 8.0, 1 mM EGTA, 25% glycerol, and 1 mM DTT) for 40 min at 4°C with aggressive mixing every 10 min. The lysates were centrifuged and the supernatants were used as the nuclear fractions. The cytosolic and nuclear fractions were mixed with SDS sample buffer, boiled, and used for SDS-PAGE, followed by immunoblotting using Abs listed in the Supplementary file 1.

Quantitative RT-PCR

Request a detailed protocol

The procedures for RNA extraction and reverse transcription to cDNA have been described previously (Nojima et al., 2011). Quantitative real-time PCR was performed with a 7500 fast Real-time PCR system or with QuantStudio 3 (Applied Biosystems). Gene expression levels were determined by the relative standard curve method and normalized to that of Gapdh.

ELISA

Request a detailed protocol

NP-specific IgG1 was detected by ELISA using NP-BSA as a plate-coated antigen as described previously (Nojima et al., 2011).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
    Cross-linking CD40 on B cells rapidly activates nuclear factor-kappa B
    1. I Berberich
    2. GL Shu
    3. EA Clark
    (1994)
    Journal of Immunology 153:4357–4366.
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    CD40 ligand induction on T cell subsets by peptide-presenting B cells: implications for development of the primary T and B cell response
    1. AI Jaiswal
    2. M Croft
    (1997)
    Journal of Immunology 159:2282–2291.
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
    Differential up-regulation of the B7-1 and B7-2 costimulatory molecules after ig receptor engagement by antigen
    1. DJ Lenschow
    2. AI Sperling
    3. MP Cooke
    4. G Freeman
    5. L Rhee
    6. DC Decker
    7. G Gray
    8. LM Nadler
    9. CC Goodnow
    10. JA Bluestone
    (1994)
    Journal of Immunology 153:1990–1997.
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65

Decision letter

  1. Facundo D Batista
    Reviewing Editor; Ragon Institute of MGH, MIT and Harvard, United States
  2. Tadatsugu Taniguchi
    Senior Editor; Institute of Industrial Science, The University of Tokyo, Japan

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "The quantity of CD40 signaling determines the differentiation of B cells into functionally distinct memory cell subsets" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Tadatsugu Taniguchi as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Essential revisions:

As you might see from their reports, all three reviewers have found the manuscript interesting and worthy of inviting for a re-submission. Moreover, they have raised several points that if addressed will certainly increase the general impact. In particular, there was a general concern among reviewers about a strict definition of memory B cells as B central memory (BCM) and B effector memory (BEM) cells. The consensus was that there needs to be an effort to better characterize this population not only in terms of markers (such as surface markers or Bcl6 and Blimp) but also in terms of their time of generation in relation B cell division or after transfer. It will also be beneficial, if within the reach of the authors, to validate their observations in an infectious model and not only with model antigens. In parallel, the text also needs to be softened in several parts. The general view was that the data, as it stands, does not warrant a rigorous definition of BCM and BEM definition as in T cells. In line with this, the manuscript will also benefit from a robust discussion on how the in vitro system does not completely recapitulate the results observed in vivo. A final important point raised by reviewers is how the levels of CD40 stimulation might influence the outcome of BCM and BEM. This has to be addressed and discussed extensively. As the manuscript stands, the authors views on this point and the role of CD40 in B cell differentiation appears to contrast with some of the current literature, and this needs to be discussed thoroughly.

Please find below the full reviewers' comments so that you get an idea of their different views as well as experimental work that you might carry out to add to accommodate their concerns.

Reviewer #1:

In this manuscript, Koike et al. investigate the signals involved in the differentiation of distinct populations of memory B cells upon immunization: a CD80-high population that can rapidly differentiate into plasma cells and a CD80-low population that can re-enter germinal centers upon re-challenge. By performing in vivo experiments and taking advantage of an ex vivo germinal center system, authors found that the development of distinct populations of memory B cells is independent of BCR isotypes, but rather associated with the affinity of their BCR and the strength of CD40 signal received from T cells. Authors found that memory B cells with high affinity to antigen (CD80-high), receive strong CD40 signals and preferentially differentiate into effector plasma cells, while low affinity memory B cells (CD80-low) receive less CD40 signal and mostly differentiate into germinal centers. In addition, authors add some mechanistic clues to how CD40 might regulate B cell fate decision through NF-κB. Some of the results present in this study were already shown by the group of Mark Shlomchik (Zuccarino-Catania et al., 2014). However, the present manuscript digs into the signals required for the generation of distinct memory B cell populations. Overall, experiments are well performed, they follow a logic and add sufficiently to our knowledge of memory B cell biology. I think it could be of potential interest for its publication in eLife after several changes and additional experiments.

- My main concern relies on the fact that authors try to define these two populations of memory B cells as B central memory (BCM) and B effector memory (BEM) cells, making a parallel to the memory T cell convention. However, in T cells, these two populations are clearly defined by specific surface markers (for instance, CD44 and CD62L in mice) and are also transcriptionally different. I am not convinced that we are in the presence of two functionally different populations of memory B cells. For instance, even though authors state that BCM are CD80- and BEM are CD80+, Figure 1A clearly shows that memory B cells display a gradient of CD80 expression rather than being separated into two distinct populations. In fact, authors arbitrary place a gate and decide which ones will be called CD80- and which ones CD80+. If authors wish to create this new nomenclature for memory B cell subsets, they should do a better characterization of these 2 populations, either by flow cytometry or transcriptionally. Authors may combine CD80 with other stainings, like CD62L or PDL2, on the same FACS plot to better separate/define these two populations.

- Authors show that interfering with CD40 signalling affects the balance between CD80-high and CD80-low memory B cells (Figure 2). Do these changes translate into different B cell effector programs (plasma vs germinal center) upon re-challenge?

- All experiments are performed with protein antigens, which raises the question of whether these "two populations" would appear in a real infection model. If that is the case, would they also have distinct CD40 signalling?

Reviewer #2:

The work of Koike et al. starts by reprising work of Shlomchik in describing subsets of memory B cells using CD80 and PDL2. They do add that CD80hi memory B cells are predisposed on restimulation in vitro with CD40L fibroblasts to produce antibody secreting cells (ASC) while CD80lo Bmem are predisposed to produce germinal center (GC) cells. They show that GC and TFH are required to bias towards CD80+ Bmem and that there is a relationship between the amount of CD40L stimulation and proportional representation of the Bmem subsets. Diminished amounts of CD40L are associated with reduced CD80+ Bmem. Curiously, when one looks at the different time points when this was assessed, 10 days and 6 weeks, the effect of reduced CD40L becomes less the longer out from the immunisation one goes. This part of the work highlights CD80 as a marker for memory subsets, introduces the potentially confusing (but seductive) nomenclature of central and effector memory B cells and promotes the idea of amount of CD40L exposure as influencing memory subset representation.

The second part is a detailed analysis of the memory B cells that are generated in mice after transfer of B cells activated in vitro with CD40L, IL4, BAFF and fibroblasts, a system called in vitro GC B cells (iGC). This process has been described by this group and involves transferring 20x10^6 activated B cells into an irradiated recipient and waiting two weeks in this case when approximately 10^6 memory B cells are recovered. The authors make the interesting observation that these Bmem can be partitioned according to Shlomchik's system and that there is a relationship between the proportion of CD80hi and in vitro exposure to high amounts of CD40L. They also show that these Bmem populations show differentiation predispositions that mimic that of the in vivo generated subsets (CD80+ to ASC and away from GC; CD80- to GC and away from ASC), although the in vivo test of this differentiation bias (Figure 3I-K) failed to reveal ASC from either subset.

The work then addresses the idea that amount of CD40L a B cell 'sees' in a GC will determine its memory subset outcome by showing, somewhat curiously, that high affinity precursor B cells produce proportionately more CD80+ memory B cells at day 7 than low affinity precursor B cells when in competition and that among day 10 memory B cells, high affinity cells are more likely to be in the CD80+ subset. This is somewhat complicated by the fact that others using this system (Nussenzweig) have shown that in competition, low affinity cells are excluded from the GC, so one wonders if this is what is being measured here (Figure 4E). Even so, these data are also more or less in agreement with Shlomchik in terms of mutational load and therefore possible affinity improvements. This part of the work also shows that there is a relationship between amount of CD80 upregulation on naïve B cells after their exposure to activated CD4 T cells that is dependent on the amount of peptide the B cells display, providing a link between degree of T cell help and CD80 expression. A major caveat of this conclusion is that this is an in vitro proliferating B cells and not a memory B cell subset.

Finally, the authors look to mechanism and things become a little murky in so far as the confusion between in vitro activated and proliferating B cells and memory subsets remains. The conclusions that NFkB is involved and that its effects are mediated by IRF4 and BATF are to me highly problematic in themselves and also emblematic of my major issue with the work itself. These biochemistry studies, which also ignore a body of work on the consequences of CD40 signalling on GC B cells, are on actively proliferating B cells. One has to seriously question what the relationship is between a B cell cycling in optimal and unchanging conditions in vitro, to that in GC where presumably conditions and stimuli are under constant flux. Indeed, the authors show quite clearly that GC B cells are in fact CD80 negative while all their cells are CD80+. It would therefore seem that in vivo, Bmem formation is associated with acquisition of CD80 while in the in vitro transfer system, Bmem formation is the retention of CD80. Even in this context it is hard to understand how if NFkB is so crucial that knockdown of Rela has such a modest effect (Figure 5F), although the degree of knockdown is not shown. The final figure showing increased expression of CD80 on what are labelled memory precursors, together with increased IRF4 and BATF, is somewhat unconvincing as it is not clear if this amount of CD80 is the baseline for memory B cells, let alone if the slightly increased amounts of IRF4 and BATF are biologically relevant.

My major concern is that this study represents two potentially independent and unrelated observations that are joined by the coincidence of CD80 expression. That is, why should I think that the processes occurring in the in vitro cultures reflect the processes in GC that give rise to memory? In vitro, the B cells are exposed to a constant and unlimited amounts of stimulus, whereas in GC they cycle between exposure to an unknown amount of CD40L, cytokines and cell-cell contact. Equally, IL21 is a crucial component of normal GC function and a defining characteristic of TFH, yet it is left out of these cultures. Previously this group reported that the Bmem produced by in vivo transfer showed different differentiation potential based on IL21 presence in the first culture. That confuses the CD40L as major factor quite a bit. Similarly, what is the timing of CD80 upregulation in the cultures? Is it immediate and irrelevant to time or is it a timing event? What is the extent of proliferation in the different cultures? Is this phenomenon related to extent of division and would different results be achieved by transferring earlier? Last why are all the mice examined at such early time points? What kind of memory is present at days 7 or 10, and doesn't that change dramatically over the weeks that follow? As shown in this reports Figure 2?

Reviewer #3:

In this manuscript Koike and colleagues investigate the role of CD40-CD40L interaction for the development of functional different memory B cell (MBC) subsets i.e. MBCs that upon re-activation are mostly prone to differentiate into plasma cells (CD80+) and those (CD80-) mostly prone to re-enter the germinal center (GC) reaction. In light of these functional properties, and in parallel with terms used for T cells, the authors respectively call these subsets effector memory B (BEM) and central memory B (BCM). Notably, a productive GC reaction is largely required for the formation of the so-called BEM, while a GC is largely dispensable for the formation of BCM. Overall the study demonstrates that CD40 is required for the formation of MBC cell subsets characterized by CD80+ or CD80-. The authors also show that BEM formation requires increased CD40 signaling compared to BCM and this correlates to their GC and pre-GC formation, respectively. BCR affinity contributes to the ability of cells to present antigen and therefore receive increased help via CD40, as previously shown by others. The work also shows that once BEM and BCM are formed their functional properties are per se largely independent on BCR receptor signal strength, suggesting a different transcriptomic or epigenetic profile. MBCs displaying IgG1 positivity can also be formed in the absence of a GC reaction, and the authors do not perform experiments where somatic mutation is determined. Does it remain to be shown whether in the GC context the choice of a B cell to differentiate into BEM or BCM exists and if that choice is dependent on CD40 signaling strength. There are additional conceptual and technical aspects of the work that need to be addressed.

1) Figure 1B (Figure 1—figure supplement 1) – The Flow Cytometry gating strategy in Figure 1—figure supplement 1B that pertains the results summarized in Figure 1B needs to be revisited. What is the justification for inclusion of GL7 within the plasma cell gating? Also, GL7 expression within the CD138- subset displays a broad range of intensities and the gating in this case includes GL7 low and likely also GL7 negative cells. How would the results in Figure 1B appear if the gating strategy is more conservative? Staining for Bcl6, as a GC marker, and for Blimp1, as a plasma cell marker, would clarify this issue.

2) Figure 2B – Transfer of B1-8 B cells into CD40lg-/- yields very few cells with an MBC phenotype demonstrating that CD40 engagement is largely required for MBC formation in general. How much or how little do the few formed MBC cells in CD40lg-/- express CD80?

3) Figure 2C, D – What is the impact in GC B cell frequency and numbers and how does that confound the results presented and interpretation? What is the authors interpretation for the cell number recovery of both BEM and BCM by 6 weeks?

4) Figure 3 – How do the varied levels of CD40L expression (40LB-lo, mid, hi) in the 40LB culture system impact on the generation of iGB cells in vitro? What is the phenotype, beyond CD80 expression, of the cells to be transferred into mice? As mentioned previously the study would have increased credibility and interpretations could be made with higher confidence if throughout the study Bcl6, as a GC marker, and for Blimp1, as a plasma cell marker are used.

5) Figure 5E – Also in this experiment it is important to describe the phenotype of the transferred cells. Upon transfer what is the impact of RelA knockdown on BCM, the GC reaction, plasma cell formation?

6) Figure 6. What would be the outcome in the formation of either BEM and BCM upon transfer of iGC-lo cells constitutively or conditionally expressing BATF, IRF4 or both?

7) The authors propose, in the fourth paragraph of the Discussion, that the reason why the so-called BCM cells are hardly generated during the GC reaction is because a) these cells fail to induce a high level of CD40L in the cognate T cells and hence would die by apoptosis. b) The authors also write that apoptosis can be avoided by c-Myc expression downstream of CD40 signaling. While the first point (a) is likely to be true given the survival properties of CD40 signaling, the suggestion that c-Myc curtails apoptosis downstream of CD40 (b) is an experimentally unsubstantiated assumption. Current data suggests that c-Myc positivity in the GC reaction represent positive selection downstream of BCR and CD40 signaling and that is most likely the reason why these cells display reduced apoptosis compared to c-Myc negative LZ B cells.

8) The authors go through great lengths to discredit a fundamental role for CD40 in plasma cell differentiation, suggesting that CD40 signaling merely enhances the effect of cytokine from particular TFH cells, likely IL21 expression ones. Clearly this is a gross misinterpretation, previous work has demonstrated that ablation of RelA is critical for GC derived plasma cell formation. Having said this, CD40 signaling may not be per se sufficient; as the authors themselves write IL21 has also been shown to be required for plasma cell formation to ensue. It is not contradictory that CD40 is involved in both MBC and plasma cell formation given that the cell fate is likely to be ultimately determined by a combination and integration of signaling pathways.

9) The authors propose that modulation of CD40 signaling could facilitate the generation of BCM in vaccination possibly beneficial against evolved viruses with epitope mutations. This proposal is however negated by the authors data in Figure 2: if CD40 signaling is impaired (Figure 2C) the formation of both BEM or BCM is impaired (cell numbers); if CD40 signaling is enhanced (Figure 2E) there is no impact on the production of BCM (cell numbers), but a boost occurs in the production of BEM (cell numbers). An avenue with respect to skewing the BEM GC fate to a BCM one may be possible, however as mentioned above, the authors do not perform experiments addressing whether in the GC context a B cell fate choice to BEM or BCM exists and if that choice is dependent on CD40 signaling strength.

https://doi.org/10.7554/eLife.44245.027

Author response

Reviewer #1:

In this manuscript, Koike et al. investigate the signals involved in the differentiation of distinct populations of memory B cells upon immunization: a CD80-high population that can rapidly differentiate into plasma cells and a CD80-low population that can re-enter germinal centers upon re-challenge. By performing in vivo experiments and taking advantage of an ex vivo germinal center system, authors found that the development of distinct populations of memory B cells is independent of BCR isotypes, but rather associated with the affinity of their BCR and the strength of CD40 signal received from T cells. Authors found that memory B cells with high affinity to antigen (CD80-high), receive strong CD40 signals and preferentially differentiate into effector plasma cells, while low affinity memory B cells (CD80-low) receive less CD40 signal and mostly differentiate into germinal centers. In addition, authors add some mechanistic clues to how CD40 might regulate B cell fate decision through NF-κB. Some of the results present in this study were already shown by the group of Mark Shlomchik (Zuccarino-Catania et al., 2014). However, the present manuscript digs into the signals required for the generation of distinct memory B cell populations. Overall, experiments are well performed, they follow a logic and add sufficiently to our knowledge of memory B cell biology. I think it could be of potential interest for its publication in eLife after several changes and additional experiments.

- My main concern relies on the fact that authors try to define these two populations of memory B cells as B central memory (BCM) and B effector memory (BEM) cells, making a parallel to the memory T cell convention. However, in T cells, these two populations are clearly defined by specific surface markers (for instance, CD44 and CD62L in mice) and are also transcriptionally different. I am not convinced that we are in the presence of two functionally different populations of memory B cells. For instance, even though authors state that BCM are CD80- and BEM are CD80+, Figure 1A clearly shows that memory B cells display a gradient of CD80 expression rather than being separated into two distinct populations. In fact, authors arbitrary place a gate and decide which ones will be called CD80- and which ones CD80+. If authors wish to create this new nomenclature for memory B cell subsets, they should do a better characterization of these 2 populations, either by flow cytometry or transcriptionally. Authors may combine CD80 with other stainings, like CD62L or PDL2, on the same FACS plot to better separate/define these two populations.

In our original manuscript, we intended to simplify the Bmem cell subsets defined and extensively characterized by Shlomchik and colleagues (Zuccarino-Catania et al., 2014), namely, CD80+ PD-L2+, CD80 PD-L2+, and CD80 PD-L2, and divided into two (CD80+ and CD80), since they demonstrated that class-switched Bmem cells were mostly PD-L2+, and that the CD80 PD-L2+ and CD80 PD-L2 Bmem cells were functionally similar. We did not intend to define a new Bmem cell subsets. Therefore, agreeing to the reviewer’s comment, we have got rid of the nomenclature, BEM and BCM. As the expression of CD80 on Bmem cells was not so discrete, as the reviewer pointed out, we decided to call them as CD80hi and CD80lo Bmem cells throughout our revised manuscript. We have described above explanation in the first paragraph of the Discussion section of our revised manuscript.

- Authors show that interfering with CD40 signalling affects the balance between CD80-high and CD80-low memory B cells (Figure 2). Do these changes translate into different B cell effector programs (plasma vs germinal center) upon re-challenge?

In the experiment shown in Figure 2C, D, the number of Bmem cells in the mice injected with anti-CD40L antibody was too small to perform the secondary transfer for the recall response experiment or ex vivo re-stimulation experiment (as in Figure 1B). Even if the primed mice were directly re-challenged, the markedly lower number of total Bmem cells in the antibody-injected mice as compared to the control mice (Figure 2C) would make it difficult to properly compare the numbers of PCs or GC B cells after the secondary challenge. In any case, the in vivo CD40-blocking would inhibit proliferation and therefore affinity maturation of B cells in the GC, and thus reduce the affinity of Bmem cells to an antigen as compared to the control, which would affect the outcome of the in vivo recall response. Another experiment using the memory-like iMBhi and iMBlo cells with uniform, unmutated BCRs to examine their recall (-like) response in the same individuals (the former Figure 3I-K; the new Figure 4C-F) clearly demonstrated that the primary CD40 signaling quantity determines the different effector programs of Bmem cells that are independent of BCR affinity.

- All experiments are performed with protein antigens, which raises the question of whether these "two populations" would appear in a real infection model. If that is the case, would they also have distinct CD40 signalling?

We agree that the infection model experiments would be informative, but we could not perform such experiments since it will take several months to have an application for such animal experiments approved by our committee and then to perform such experiments.

Reviewer #2:

The work of Koike et al. starts by reprising work of Shlomchik in describing subsets of memory B cells using CD80 and PDL2. They do add that CD80hi memory B cells are predisposed on restimulation in vitro with CD40L fibroblasts to produce antibody secreting cells (ASC) while CD80lo Bmem are predisposed to produce germinal center (GC) cells. They show that GC and TFH are required to bias towards CD80+ Bmem and that there is a relationship between the amount of CD40L stimulation and proportional representation of the Bmem subsets. Diminished amounts of CD40L are associated with reduced CD80+ Bmem. Curiously, when one looks at the different time points when this was assessed, 10 days and 6 weeks, the effect of reduced CD40L becomes less the longer out from the immunisation one goes. This part of the work highlights CD80 as a marker for memory subsets, introduces the potentially confusing (but seductive) nomenclature of central and effector memory B cells and promotes the idea of amount of CD40L exposure as influencing memory subset representation.

Before responding to each comment, we note here that we have got rid of the nomenclature, BEM and BCM, and called them as CD80hi and CD80lo Bmem cells throughout our revised manuscript, as mentioned in the response to the first reviewer.

In response to the comment “Curiously, when one looks at the different time points when this was assessed, 10 days and 6 weeks, the effect of reduced CD40L becomes less the longer out from the immunisation one goes.”:

As the reviewer commented, the effect of injection of anti-CD40L antibody on the Bmem cell development was less impressive at 6 weeks after immunization (Figure 2D). We think this is because CD80hi Bmem cells may have been generated in later GC after the injected antibody had lapsed, as we have described in the Results section of the revised manuscript (subsection “Strong CD40 signaling induced by TFH cells is required for development of CD80hi Bmem cells”, third paragraph).

The second part is a detailed analysis of the memory B cells that are generated in mice after transfer of B cells activated in vitro with CD40L, IL4, BAFF and fibroblasts, a system called in vitro GC B cells (iGC). This process has been described by this group and involves transferring 20x10^6 activated B cells into an irradiated recipient and waiting two weeks in this case when approximately 10^6 memory B cells are recovered. The authors make the interesting observation that these Bmem can be partitioned according to Shlomchik's system and that there is a relationship between the proportion of CD80hi and in vitro exposure to high amounts of CD40L. They also show that these Bmem populations show differentiation predispositions that mimic that of the in vivo generated subsets (CD80+ to ASC and away from GC; CD80- to GC and away from ASC), although the in vivo test of this differentiation bias (Figure 3I-K) failed to reveal ASC from either subset.

In the experiment shown in the former Figure 3I-K, NP-specific plasmablasts could hardly be detected on day 10 or later in the spleen, but were readily detectable on day 4 after immunization. Thus, we have included the data of the plasmablast analysis on day 4, which shows that vast majority of the donor-derived plasmablasts was derived of iMB-hi cells, as shown in the new Figure 4E, F. We have described about this data in the Results section of the revised manuscript (subsection “CD40 signal strength in vitro affects differentiation into CD80hi or CD80lo Bmem cells in vivo”, last paragraph).

The work then addresses the idea that amount of CD40L a B cell 'sees' in a GC will determine its memory subset outcome by showing, somewhat curiously, that high affinity precursor B cells produce proportionately more CD80+ memory B cells at day 7 than low affinity precursor B cells when in competition and that among day 10 memory B cells, high affinity cells are more likely to be in the CD80+ subset. This is somewhat complicated by the fact that others using this system (Nussenzweig) have shown that in competition, low affinity cells are excluded from the GC, so one wonders if this is what is being measured here (Figure 4E). Even so, these data are also more or less in agreement with Shlomchik in terms of mutational load and therefore possible affinity improvements.

We agree to the possibility that low affinity cells are excluded from the GC and therefore fail to access to CD40L on TFH cells. This possibility does not contradict to our original idea that higher affinity B cells preferentially differentiate into CD80hi Bmem cells through getting stronger stimulation with CD40L on TFH cells. Thus, we have included the new possibility and described in the Results section (subsection “Higher BCR affinity for antigen favors development of CD80hi Bmem cells”, last paragraph) and the Discussion section (seventh paragraph) of the revised manuscript.

This part of the work also shows that there is a relationship between amount of CD80 upregulation on naïve B cells after their exposure to activated CD4 T cells that is dependent on the amount of peptide the B cells display, providing a link between degree of T cell help and CD80 expression. A major caveat of this conclusion is that this is an in vitro proliferating B cells and not a memory B cell subset.

With the data in the former Figure 4A-D, we intended to show that CD40L expression was induced on activated T cells to the levels in proportion to the extent of antigen presentation on B cells, and that the induced levels of CD40L quantitatively correlated to its function to stimulate cognate B cells through CD40, as assessed by CD80 induction. These data indicate quantitative relationship among B-cell antigen presentation, CD40L expression on T cells and B cell activation, and would serve as an introduction for the following experiments. We did not mean that the CD80 expression levels on naïve B cells directly correlated to the development of Bmem cells with distinct expression levels of CD80. However, it is possible that the levels of CD80 expression induced by the T-B interaction during the early primary response or during GC reaction may be maintained through the development to Bmem cells. We have included a sentence mentioning this possibility for the logical sequence of the experiments shown in Figure 2, in the Results section of the revised manuscript (subsection “Strong CD40 signaling induced by TFH cells is required for development of CD80hi Bmem cells”, second paragraph).

Finally, the authors look to mechanism and things become a little murky in so far as the confusion between in vitro activated and proliferating B cells and memory subsets remains. The conclusions that NFkB is involved and that its effects are mediated by IRF4 and BATF are to me highly problematic in themselves and also emblematic of my major issue with the work itself. These biochemistry studies, which also ignore a body of work on the consequences of CD40 signalling on GC B cells, are on actively proliferating B cells. One has to seriously question what the relationship is between a B cell cycling in optimal and unchanging conditions in vitro, to that in GC where presumably conditions and stimuli are under constant flux. Indeed, the authors show quite clearly that GC B cells are in fact CD80 negative while all their cells are CD80+. It would therefore seem that in vivo, Bmem formation is associated with acquisition of CD80 while in the in vitro transfer system, Bmem formation is the retention of CD80. Even in this context it is hard to understand how if NFkB is so crucial that knockdown of Rela has such a modest effect (Figure 5F), although the degree of knockdown is not shown.

We understand the reviewer’s concern, but we were aware that the mechanisms of CD40 signaling to induce CD80 expression in in vitro cultured B cells are primarily irrelevant to those to promote development of CD80hi or CD80lo Bmem cells in vivo. However, we thought that the biochemical analyses in the former could provide a clue to elucidate the latter mechanisms that appeared to be difficult to directly tackle. Thus, after having found that RelA was involved in the strong CD40 signaling in cultured B cells (the former Figure 5A, B), we tested whether RelA was involved in the development of CD80hi Bmem cells (the former Figure 5E, F). The result showed that the RelA-knockdown in the in vivo activated B cells resulted in significant reduction of development of CD80hi Bmem cells, although the difference was modest as the reviewer pointed out. However, the knockdown efficiency was unknown and could not be measured in this experiment because the ex vivo B cells transduced with the knockdown vector containing GFP marker had to be transferred back into mice 3 hours later, before the GFP expression became apparent (please see the new Figure 6G). Thus, we only tested the knockdown vector efficacy using iGB cells (the former Figure 5C). To convince the reviewer as well as the readers, in the revised manuscript, we have included the data of an experiment in which iGB cells were transduced with the knockdown vectors, now including that of c-Rel, and then transferred into mice to let them differentiate into memory-like iMB cells. As shown in the new Figure 6C-F and described in the Results section subsection “Possible mechanisms by which CD40 signaling facilitates GC B-cell differentiation into CD80hi Bmem cells”, first paragraph) of the revised manuscript, the knockdown of either RelA or c-Rel, which was now evaluated (Figure 6C), significantly attenuated in vivo generation of CD80hi iMB cells.

Although the reviewer seems to consider that a body of our work is on the consequences of CD40 signaling on GC B cells, we considered that the CD40 signaling affect B cells not only in the GC phase but also during the initial activation phase before the GC phase, for their differentiation into Bmem cells. Indeed, stimulation of naïve B cells for only two day, or even one, on the feeder cells differentially expressing CD40L affected their development into CD80hi or CD80lo iMB cells (the new Figure 3—figure supplement 1E-G). Thus, we consider that the iGB culture system represents the both phases, although being quite artificial, and is useful to elucidate signals for B cells to differentiate into Bmem cells (iMB cell in the experiments). Although majority of GC B cells are CD80, unlike in vitro activated B cells, a minor fraction that has been selected for Bmem cells (pre-Bmem cells) expressed CD80 (the former Figure 6F, G; the revised Figure 7F, G). Therefore, the mechanism of CD40 signaling for the Bmem cell development may be similar between naïve B cells in the early phase and GC B cells in the GC phase of the primary immune response, and they both induce CD80 expression upon the strong and durable CD40 signaling and may retain the CD80 expression until and after they become Bmem cells. We have stated this view in the Discussion section of the new manuscript (fifth paragraph).

The final figure showing increased expression of CD80 on what are labelled memory precursors, together with increased IRF4 and BATF, is somewhat unconvincing as it is not clear if this amount of CD80 is the baseline for memory B cells, let alone if the slightly increased amounts of IRF4 and BATF are biologically relevant.

To answer this comment, we have repeated the same experiment but now including Bmem cells in the analysis. The expression levels of CD80 and BATF in pre-Bmem cells were intermediate between those of GC B cells and Bmem cells, although the increase of IRF4 expression in pre-Bmem cells was not significant in this experiment (the new Figure 7F, G). The data were evaluated by control staining with isotype-matched antibodies (the new Figure 7—figure supplement 1D). Thus, we revised the Results section in the revised manuscript (subsection “Possible mechanisms by which CD40 signaling facilitates GC B-cell differentiation into CD80hi Bmem cells”, third paragraph).

My major concern is that this study represents two potentially independent and unrelated observations that are joined by the coincidence of CD80 expression. That is, why should I think that the processes occurring in the in vitro cultures reflect the processes in GC that give rise to memory? In vitro, the B cells are exposed to a constant and unlimited amounts of stimulus, whereas in GC they cycle between exposure to an unknown amount of CD40L, cytokines and cell-cell contact. Equally, IL21 is a crucial component of normal GC function and a defining characteristic of TFH, yet it is left out of these cultures. Previously this group reported that the Bmem produced by in vivo transfer showed different differentiation potential based on IL21 presence in the first culture. That confuses the CD40L as major factor quite a bit.

As answered in the previous response, we think the iGB culture system, from which the cultured B cells differentiate into Bmem-like iMB cells in vivo, represents a common aspect of T-cell-mediated stimulation, but not the whole process, of naïve B cells and GC B cells during the initial phase of the immune response and the GC phase, respectively. The resultant CD80hi and CD80lo iMB cells and genuine CD80hi and CD80lo Bmem cells do not just coincide with each other on the CD80 expression, but were similar in their secondary responses in vitro as well as in vivo (Figure 1B, the new Figure 4, and Zuccarino-Catania et al., 2014). As the reviewer pointed out, the condition of the iGB culture system is far simpler than that of genuine GC, but it is well known that a numerous number of such simplified experimental systems have greatly contributed to our understanding of the complex immune system.

As for the IL-21, we previously reported that the addition of IL-21 secondary to the IL-4 in the iGB culture suppresses the generation of the iMB cells but instead promotes the generation of bone-marrow PCs (Nojima et al., 2011). Although we used only the first culture with IL-4 that allows generation of iMB cells in the original manuscript, we have now tested the effect of additional IL-21 in the first culture. As shown in the new Figure 3—figure supplement 1H, I (subsection “CD40 signal strength in vitro affects differentiation into CD80hi or CD80lo Bmem cells in vivo”, third paragraph) in the revised manuscript, the addition of IL-21 did not affect the frequency of CD80hi iMB cells generated from iGB cells cultured on either 40LBhi or 40LBlo cells.

Similarly, what is the timing of CD80 upregulation in the cultures? Is it immediate and irrelevant to time or is it a timing event? What is the extent of proliferation in the different cultures? Is this phenomenon related to extent of division and would different results be achieved by transferring earlier?

We showed that CD80 was upregulated later than one day after the start of the iGB culture, and increased by day 3, to different extents depending on CD40L expression levels on feeder cells (Figure 3B). In our revised manuscript, we have included the data showing that the proliferation of iGB cells was equally minimum on day 2 but markedly differed depending on the feeder cells, as evident on day 4 after the start of the culture (the new Figure 3—figure supplement 1C). Transferring with the day 2 iGB cells, or even day 1, resulted in the similar result as with day 4 iGB cells, although the frequency of CD80hi iMB cells corresponded to the length of the culture period, as shown in the new Figure 3—figure supplement 1E-G. This result indicates that the differential Bmem cell development is not related to the extent of division of B cells in culture, but related to strength and duration of CD40 stimulation. These results and view are stated in the Results section of the revised manuscript (subsection “CD40 signal strength in vitro affects differentiation into CD80hi or CD80lo Bmem cells in vivo”, third paragraph).

Last why are all the mice examined at such early time points? What kind of memory is present at days 7 or 10, and doesn't that change dramatically over the weeks that follow? As shown in this reports Figure 2?

Bmem cells were examined at 6 weeks after immunization in the experiment shown in Figure 1C and Figure 2B, but mainly at 10 days (or 7 days in some) in other experiments. This is because we focused on the generation of Bmem cells and tried to avoid outcomes of possible alterations in the maintenance of Bmem cells or in their late development from GC, which might be caused by the experimental interventions we performed. The previous reports indicated that class-switched Bmem cells are readily detectable by day 7~9 in experiments where antigen-specific Ig-knockin B cells are transferred into mice, and most Bmem cells are formed by day 11 after immunization (Wang et al., 2017; Suan et al., 2017; Weisel et al., 2016). We stated the above view in the Results section of the revised manuscript (subsection “Strong CD40 signaling induced by TFH cells is required for development of CD80hi Bmem cells”, third paragraph).

Reviewer #3:

In this manuscript Koike and colleagues investigate the role of CD40-CD40L interaction for the development of functional different memory B cell (MBC) subsets i.e. MBCs that upon re-activation are mostly prone to differentiate into plasma cells (CD80+) and those (CD80-) mostly prone to re-enter the germinal center (GC) reaction. In light of these functional properties, and in parallel with terms used for T cells, the authors respectively call these subsets effector memory B (BEM) and central memory B (BCM). Notably, a productive GC reaction is largely required for the formation of the so-called BEM, while a GC is largely dispensable for the formation of BCM. Overall the study demonstrates that CD40 is required for the formation of MBC cell subsets characterized by CD80+ or CD80-. The authors also show that BEM formation requires increased CD40 signaling compared to BCM and this correlates to their GC and pre-GC formation, respectively. BCR affinity contributes to the ability of cells to present antigen and therefore receive increased help via CD40, as previously shown by others. The work also shows that once BEM and BCM are formed their functional properties are per se largely independent on BCR receptor signal strength, suggesting a different transcriptomic or epigenetic profile. MBCs displaying IgG1 positivity can also be formed in the absence of a GC reaction, and the authors do not perform experiments where somatic mutation is determined. Does it remain to be shown whether in the GC context the choice of a B cell to differentiate into BEM or BCM exists and if that choice is dependent on CD40 signaling strength. There are additional conceptual and technical aspects of the work that need to be addressed.

Before responding to each comment, we note here that we have got rid of the nomenclature, BEM and BCM, and instead called them as CD80hi and CD80lo Bmem cells throughout our revised manuscript, in response to other reviewers’ and editors’ comments.

Our response to the sentence “Does it remain to be shown whether in the GC context the choice of a B cell to differentiate into BEM or BCM exists and if that choice is dependent on CD40 signaling strength.”:

Although we have not done such experiments, Shlomchik and colleagues showed in their paper that injection of blocking anti-CD40L antibody at the peak of GC reaction (day 12-14 after immunization) resulted in reduction of the frequency of CD80+ Bmem cells and increase of CD80 Bmem cells, suggesting that the choice of a B cell to differentiate into either Bmem cell type, depending on CD40 signaling strength, exists in the GC context (Weisel et al., 2016).

1) Figure 1B (Figure 1—figure supplement 1) – The Flow Cytometry gating strategy in Figure 1—figure supplement 1B that pertains the results summarized in Figure 1B needs to be revisited. What is the justification for inclusion of GL7 within the plasma cell gating? Also, GL7 expression within the CD138- subset displays a broad range of intensities and the gating in this case includes GL7 low and likely also GL7 negative cells. How would the results in Figure 1B appear if the gating strategy is more conservative? Staining for Bcl6, as a GC marker, and for Blimp1, as a plasma cell marker, would clarify this issue.

We have re-gated the FCM data by quadrants (the new Figure 1—figure supplement 1B) and re-assessed the frequencies of CD138+ GL7− plasmablasts and CD138− GL7+ GC B cells (the new Figure 1B), which resulted in the similar data as the former Figure 1B.

2) Figure 2B – Transfer of B1-8 B cells into CD40lg-/- yields very few cells with an MBC phenotype demonstrating that CD40 engagement is largely required for MBC formation in general. How much or how little do the few formed MBC cells in CD40lg-/- express CD80?

The number of the Bmem cells in CD40lg–/– mice was so few in some individuals that we could not make statistically reliable data. If we took data from a few CD40lg–/– mice with relatively higher number of Bmem cells, the frequency of CD80hi Bmem cells was decreased by about half as compared to those in CD40lg+/+ mice (data not shown).

3) Figure 2C, D – What is the impact in GC B cell frequency and numbers and how does that confound the results presented and interpretation?

In the mice injected with anti-CD40L antibody, the frequency and the numbers of GC B cells declined by about one-tenth as deduced by the data in Figure 2C, and as shown in the new Figure 2—figure supplement 1E, respectively. Therefore, it is unclear whether the reduction of CD80hi Bmem cells was due to the decrease of GC B cells or to the reduction of CD40 signaling in GC B cells, or both. That is why we used the iGB culture system with differential in vitro CD40 stimulation in Figure 3.

What is the authors interpretation for the cell number recovery of both BEM and BCM by 6 weeks?

We think this is because CD80hi Bmem cells may have been generated in later GC after the injected anti-CD40L antibody had lapsed, as we have described in the Results section of the revised manuscript (subsection “Strong CD40 signaling induced by TFH cells is required for development of CD80hi Bmem cells”, third paragraph).

4) Figure 3 – How do the varied levels of CD40L expression (40LB-lo, mid, hi) in the 40LB culture system impact on the generation of iGB cells in vitro? What is the phenotype, beyond CD80 expression, of the cells to be transferred into mice? As mentioned previously the study would have increased credibility and interpretations could be made with higher confidence if throughout the study Bcl6, as a GC marker, and for Blimp1, as a plasma cell marker are used.

In response to this comment, we analyzed surface phenotype, proliferation, and expression of Bcl6 and Prdm1 (for Blimp1) mRNA (the new Figure 3—figure supplement 1A-D). The extent of class switching and expression levels of GL7, PD-L2, CD73 were almost the same among the B cells cultured on 40LB-lo, 40LB-mid, or 40LB-hi. Expression levels of CD38 and CD62L were higher and those of Fas were lower in B cells cultured on 40LB-lo feeder cells, probably reflecting less activated state of the cells. Proliferation was correlated to the expression levels of CD40L on these feeder cells. Curiously, the mRNA levels of both Bcl6 and Prdm1 exhibited inverse correlation to the CD40L levels on these feeder cells. We described about these data in the Results section of the revised manuscript (subsection “CD40 signal strength in vitro affects differentiation into CD80hi or CD80lo Bmem cells in vivo”, first paragraph).

5) Figure 5E – Also in this experiment it is important to describe the phenotype of the transferred cells.

In this experiment (the new Figure 6G, H), the B cells activated in vivo, and transduced with the knockdown or control vectors containing GFP marker ex vivo, had to be transferred back into mice 3 hours after the gene transduction, before the GFP expression became apparent (the new Figure 6G). Since we could not distinguish the vector-transduced (GFP+) cells before transfer, we could not assess the phenotypic change by the gene knockdown.

To compensate for the weakness of this experiment, we performed an experiment in which iGB cells were transduced with the knockdown vectors, now including that of c-Rel, and then transferred into mice to let them differentiate into memory-like iMB cells. The knockdown of RelA or c-Rel did not affect class-switching and did not induce PC differentiation of iGB cells before the transfer (the new Figure 6—figure supplement 1A), but significantly attenuated in vivo generation of CD80hi iMB cells, as shown in the new Figure 6C-F and described in the Results section (subsection “Possible mechanisms by which CD40 signaling facilitates GC B-cell differentiation into CD80hi Bmem cells”, first paragraph) of the revised manuscript.

Upon transfer what is the impact of RelA knockdown on BCM, the GC reaction, plasma cell formation?

We have now included data showing the frequency of GC B cells among the gene-transduced, antigen-specific, class-switched B cells, and CD80lo cells among the gene-transduced Bmem cells, as shown in the new Figure 6—figure supplement 1C in the revised manuscript. PCs could not be seen in the spleen at this time point. The RelA knockdown resulted in a little increase of the frequency of GC B cells and increase of CD80lo Bmem cells complementary to the decrease of CD80hi Bmem cells.

6) Figure 6. What would be the outcome in the formation of either BEM and BCM upon transfer of iGC-lo cells constitutively or conditionally expressing BATF, IRF4 or both?

We have performed this experiment, as shown in the new Figure 7H-J in the revised manuscript. The result demonstrated that BATF-expressing (GFP+) iMB cells contained a significantly higher proportion of CD80hi cells than mock-transduced iMB cells. IRF4-expressing (CFP+) cells could not be detected in iMB cells, possibly because those over-expressing IRF4 had differentiated into PCs. We described about this data in the Results section of the revised manuscript (subsection “Possible mechanisms by which CD40 signaling facilitates GC B-cell differentiation into CD80hi Bmem cells”, last paragraph).

7) The authors propose, in the fourth paragraph of the Discussion, that the reason why the so-called BCM cells are hardly generated during the GC reaction is because a) these cells fail to induce a high level of CD40L in the cognate T cells and hence would die by apoptosis. b) The authors also write that apoptosis can be avoided by c-Myc expression downstream of CD40 signaling. While the first point (a) is likely to be true given the survival properties of CD40 signaling, the suggestion that c-Myc curtails apoptosis downstream of CD40 (b) is an experimentally unsubstantiated assumption. Current data suggests that c-Myc positivity in the GC reaction represent positive selection downstream of BCR and CD40 signaling and that is most likely the reason why these cells display reduced apoptosis compared to c-Myc negative LZ B cells.

In accordance with the reviewer’s comment, we have removed the statement of our assumption about c-Myc involvement from the Discussion section.

8) The authors go through great lengths to discredit a fundamental role for CD40 in plasma cell differentiation, suggesting that CD40 signaling merely enhances the effect of cytokine from particular TFH cells, likely IL21 expression ones. Clearly this is a gross misinterpretation, previous work has demonstrated that ablation of RelA is critical for GC derived plasma cell formation. Having said this, CD40 signaling may not be per se sufficient; as the authors themselves write IL21 has also been shown to be required for plasma cell formation to ensue. It is not contradictory that CD40 is involved in both MBC and plasma cell formation given that the cell fate is likely to be ultimately determined by a combination and integration of signaling pathways.

We agree to this comment, and therefore we have removed the statement that the strong CD40 signaling merely enhance the interaction between GC B cells and TFH cells from our Discussion, and extensively revised this paragraph of the Discussion in the revised manuscript, adapting the reviewer’s view, namely, a combination and integration of CD40 signaling and cytokine signaling may determine the fate of GC B cells (Discussion, sixth paragraph).

9) The authors propose that modulation of CD40 signaling could facilitate the generation of BCM in vaccination possibly beneficial against evolved viruses with epitope mutations. This proposal is however negated by the authors data in Figure 2: if CD40 signaling is impaired (Figure 2C) the formation of both BEM or BCM is impaired (cell numbers); if CD40 signaling is enhanced (Figure 2E) there is no impact on the production of BCM (cell numbers), but a boost occurs in the production of BEM (cell numbers).

We completely agree to this comment and have revised our Discussion in our revised manuscript as following: “Thus, if we could dissect the CD40 signaling pathways that direct differentiation into CD80hi Bmem cells and those promote cell proliferation/survival, selective suppression of the former pathways during vaccination might convert the generation of CD80hi Bmem cells into CD80lo Bmem cells that would eventually produce broadly reactive Abs.”.

An avenue with respect to skewing the BEM GC fate to a BCM one may be possible, however as mentioned above, the authors do not perform experiments addressing whether in the GC context a B cell fate choice to BEM or BCM exists and if that choice is dependent on CD40 signaling strength.

As for this part of the comment, we have described our response following the “General assessment and major comments”.

https://doi.org/10.7554/eLife.44245.028

Article and author information

Author details

  1. Takuya Koike

    Division of Molecular Biology, Research Institute for Biomedical Sciences (RIBS), Tokyo University of Science, Noda, Japan
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
  2. Koshi Harada

    Division of Molecular Biology, Research Institute for Biomedical Sciences (RIBS), Tokyo University of Science, Noda, Japan
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  3. Shu Horiuchi

    Division of Molecular Biology, Research Institute for Biomedical Sciences (RIBS), Tokyo University of Science, Noda, Japan
    Contribution
    Supervision, Investigation, Methodology
    Competing interests
    No competing interests declared
  4. Daisuke Kitamura

    Division of Molecular Biology, Research Institute for Biomedical Sciences (RIBS), Tokyo University of Science, Noda, Japan
    Contribution
    Conceptualization, Supervision, Funding acquisition, Validation, Writing—original draft, Writing—review and editing
    For correspondence
    kitamura@rs.noda.tus.ac.jp
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5195-0474

Funding

Japan Society for the Promotion of Science (16H05206)

  • Daisuke Kitamura

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thanks K Rajewsky (Max Delbrück Center for Molecular Medicine) for B1-8ki mice, M Nussenzweig (The Rockefeller University) for B1-8hiki mice, T Takemori (RIKEN) for Bcl6-flox mice, R Fravell (Yale School of Medicine) for Cd40lg−/− mice, T Tsubata (Tokyo Medical and Dental University) for Igκ−/− mice, T Azuma and Y Tashiro (Research Institute for Biomedical Sciences, RIBS) for NPmed-APC, J Yagi (Tokyo Women’s Medical University School of Medicine) for CA-Akt, M Hibi (Nagoya University) for CA-MKK4, K Haniuda, S Fukao, and S Konishi for plasmid constructs and regents, M Funatsu and M Nomoto for technical assistance, M Kubo, R Goitsuka, and other members of RIBS for technical advice and comments, and P Burrows for critical reading. This works was supported by Grant-in-Aid for Scientific Research (B) (to DK). T Koike is a Research Fellow of Scholarship for Doctoral Students in Immunology of the Japanese Society for Immunology.

Ethics

Animal experimentation: All animal experiments were performed under protocols approved by the Animal Care and Use Committee of the Tokyo University of Science (Approval No.: S15021, S16019, S17004, S18018). All surgery was performed under Isoflurane anesthesia, and every effort was made to minimize suffering.

Senior Editor

  1. Tadatsugu Taniguchi, Institute of Industrial Science, The University of Tokyo, Japan

Reviewing Editor

  1. Facundo D Batista, Ragon Institute of MGH, MIT and Harvard, United States

Publication history

  1. Received: December 9, 2018
  2. Accepted: June 14, 2019
  3. Accepted Manuscript published: June 21, 2019 (version 1)
  4. Version of Record published: July 17, 2019 (version 2)

Copyright

© 2019, Koike et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,248
    Page views
  • 323
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Immunology and Inflammation
    Ksenya Cohen Katsenelson et al.
    Research Article
    1. Immunology and Inflammation
    Eric E Irons et al.
    Research Article