Memory B cells contribute to rapid Bcl6 expression by memory follicular helper T cells

Significance

Follicular helper T (T_FH) cells have emerged as the key cell type required for the formation of germinal centers and subsequent long-lasting antibody responses. It has been demonstrated that T_FH cells enter the memory pool. However, it is unclear how the generation, survival, or activation of those T_FH memory cells is regulated. Here we show that B-cell lymphoma 6 (Bcl6), a master regulator of T_FH generation, is required for maintenance of T_FH memory cells and subsequent humoral memory. In these recall responses, antigen-specific memory B cells majorly contribute to the quick induction of Bcl6 in T_FH memory cells. Our results reinforce the importance of cognate interaction between memory T_FH and memory B cells and give important implications for development of better vaccines.

Abstract

In primary humoral responses, B-cell lymphoma 6 (Bcl6) is a master regulator of follicular helper T (T_FH) cell differentiation; however, its activation mechanisms and role in memory responses remain unclear. Here we demonstrate that survival of CXCR5⁺ T_FH memory cells, and thus subsequent recall antibody response, require Bcl6 expression. Furthermore, we show that, upon rechallenge with soluble antigen Bcl6 in memory T_FH cells is rapidly induced in a dendritic cell-independent manner and that peptide:class II complexes (pMHC) on cognate memory B cells significantly contribute to this induction. Given the previous evidence that antigen-specific B cells residing in the follicles acquire antigens within minutes of injection, our results suggest that memory B cells present antigens to the cognate T_FH memory cells, thereby contributing to rapid Bcl6 reexpression and differentiation of the T_FH memory cells during humoral memory responses.

The development of high-affinity B-cell memory is essential in most effective vaccines that are in use today. Because most protein antigens require T-cell help to induce B-cell responses, understanding the mechanisms by which memory T and B cells are generated and maintained, as well as how their swift activation is executed, is of fundamental importance for vaccine development.

In primary immune responses, it is widely accepted that among several differentiated helper T-cell subsets follicular helper CD4 T cells (T_FH cells) are the major subset to deliver help to B cells (1). T_FH cells express CXC-chemokine receptor 5 (CXCR5), the chemokine receptor for the B-cell homing chemokine CXCL13. Surface expression of CXCR5 enables T_FH cells to migrate into B-cell follicles, where they provide help to B cells to form germinal centers. In addition, T_FH cells are needed for the crucial affinity-maturation process of B cells in germinal centers, whereby Ag-specific B cells undergo repeated rounds of somatic hypermutation and positive selection by T_FH cells to rapidly evolve high-affinity somatically mutated B-cell receptors. B-cell lymphoma 6 (Bcl6) has recently been identified as a T_FH lineage regulator (2–4); it is highly expressed by T_FH cells and is required for their development. According to the current view, during a primary response Bcl6 expression by T cells is induced by priming with dendritic cells (5–7) and ICOS is a key coreceptor molecule for induction of Bcl6 (5, 8). The initial Bcl6 induction and subsequent CXCR5 expression allow CD4 T cells to migrate toward the T–B border, where T_FH cells interact with antigen-specific B cells. According to this model, cognate B cells are not required for the induction of Bcl6 but support the expansion of T_FH cells (9).

Although the importance of Bcl6 and its expression kinetics in naïve T-cell differentiation have been well elucidated, its role and activation mechanisms in T_FH memory cells still remain obscure. Hence, in this paper we first focus upon the roles of Bcl6, demonstrating its importance for maintenance of T_FH memory cells. Then, we show that Bcl6 in memory T_FH cells was rapidly induced upon rechallenge with soluble antigen and that this response was mainly mediated through antigen presentation by the cognate memory B cells. Given the good association between Bcl6 with IL-21 expression in differentiated memory T_FH cells, our results suggest that memory B cells are the primary antigen-presenting cells (APCs) to induce the rapid differentiation of memory T_FH cells toward effector cells, further accelerating memory B-cell responses during recall.

Results

CXCR5⁺ Memory T Cells Provide Potent Help to Memory B Cells.

To identify T_FH memory cells we set up adoptive transfer experiments in which naïve T cells purified from TEa T-cell receptor transgenic (TCR Tg) mice were transferred to congenically marked mice and immunized with (4-hydroxy-3-nitrophenyl)acetyl (NP)-Eα-GFP in alum. Six weeks after immunization, expression of several surface molecules on the surviving memory T cells was examined. TEa memory T cells were found to be heterogeneous in their expression of CXCR5 or CD62L (Fig. 1A); 20–30% of memory TEa T cells were CXCR5⁺ with low or high expression of CD62L.

Fig. 1.

CXCR5⁺ memory T cells with potent ability to activate memory B cells. Naïve T cells from TEa TCR transgenic mice were transferred into congenically marked hosts that then were immunized i.p. with NP-Eα-GFP in alum. Six weeks after immunization, spleens were harvested. (A) Expression of CD62L or CXCR5 on CD45.2⁺ TEa memory T cells was analyzed. Representative data of more than five independent experiments are shown. (B) Naïve TEa T cells, CXCR5⁻ memory, or CXCR5⁺ memory T cells (CD45.1⁺) were transferred (10,000 cells) into B6 mice (CD45.2⁺) in conjunction with CD45.1⁺B1-8^hi memory IgG1 B cells. On the next day, the mice were immunized with NP-Eα-GFP in PBS. Seven days later, development of plasma cells (CD138⁺ cells) or germinal center cells (Fas^hiGL7^hi cells) from transferred memory B cells (NIP⁺CD45.1⁺) was examined. The data are representative of two experiments. Bar graphs represent mean ± SD, n = 3, *P < 0.05, **P < 0.01, ***P < 0.005; NS, not significant. (C) CXCR5⁺ or CXCR5⁻ memory TEa T cells (CD45.2⁺) were isolated and transferred (10,000 cells) into CD45.1⁺ congenic hosts. One day after transfer, localization of the transferred T cells in spleen was analyzed. A total of 100 CD45.2⁺ cells (derived from each donor) were detected on sections from three independent experiments. The graph summarizes the localization of those CD45.2⁺ cells in the indicated area of spleen.

To determine the capacity of the above distinct memory populations to help IgG1⁺ memory B cells, purified CXCR5⁻ or CXCR5⁺ memory TEa cells were transferred into congenic WT mice, together with B1-8^hi IgG1⁺ memory B cells, followed by immunization with soluble NP-Eα-GFP. As shown in Fig. 1B, naïve TEa T cells could not induce differentiation of IgG1⁺ memory B cells into plasma cells. By contrast, CXCR5⁺ memory T cells were potent at inducing differentiation of the transferred IgG1⁺ memory B cells. CXCR5⁻ memory T cells also activated memory B cells but the magnitude of the B-cell responses was weaker than the response induced by CXCR5⁺ memory T cells. This differential B-cell helper activity was similarly observed when NP-Eα-GFP in alum was used instead of soluble antigen alone (Fig. S1A). Because CXCR5⁺ memory T cells consist of CD62L⁺ and CD62L⁻ subsets, we next determined whether these two subsets exhibited differential memory B-cell helper activity. As shown in Fig. S1B, both CXCR5⁺CD62L⁻ and CXCR5⁺CD62L⁺ memory cells possessed similar potency.

CXCR5 mAbs do not work well for immunohistochemical analyses to examine the location of CXCR5⁺ memory cells. Therefore, we purified CXCR5⁻ or CXCR5⁺ memory TEa cells from mice immunized 6 wk before and transferred them into congenic mice. As expected, about 40% and 30%, respectively, of the CXCR5⁺ memory T cells were found in the T–B border and B-cell follicle, although some cells were in the T-cell zone (Fig. 1C).

The above experiments depended on the adoptive transfer of relatively large numbers of antigen-specific T cells from TCR-transgenic mice. A limitation of this approach is that T cells may receive less intense antigenic stimulation when present in large numbers, perhaps because of intraclonal competition. To validate the use of the adoptive transfer system in this model, we investigated the status of an endogenous antigen-specific T-cell population. B6 mice were injected intraperitoneally with NP-2W1S-GFP in alum. 2W1S-specific endogenous CD4⁺ T cells could be detected with a 2W1S/I-A^b tetramer (10, 11) (Fig. S2 A–C). We found that the 2W1S-specific memory T-cell population consisted of both CXCR5⁺ and CXCR5⁻ cells with low or high CD62L expression (Fig. S2D). To examine their helper activity, CXCR5⁺ or CXCR5⁻ 2W1S-specific memory T cells were transferred together with B1-8^hi IgG1⁺ memory B cells into congenic mice, followed by immunization with soluble NP-2W1S-OVA. As shown in Fig. S2 E and F, CXCR5⁺ 2W1S-specific memory T cells were superior to their CXCR5⁻ counterparts in inducing IgG1 response by memory B cells. Thus, the similarity between the endogenous and TCR-transgenic T cells validated the use of the adoptive transfer approach for subsequent studies of memory T cells.

CXCR5⁺ Memory T Cells Are Derived from T_FH Effectors.

Several recent studies have provided evidence that T_FH effector T cells can further differentiate into resting memory CD4⁺ T cells (12–16). To confirm that CXCR5⁺ memory T cells in our system are derived from T_FH effectors, we purified CXCR5^hiPD1^hi (T_FH) or CXCR5^lo/-PD1⁻ (non-T_FH) TEa effector cells, transferred them into congenic naïve mice, and then analyzed the phenotype of the surviving cells 4 wk later. As shown in Fig. S3A, the majority of the cells derived from T_FH effectors maintained high levels of CXCR5 expression and were localized in the T-cell area, T–B border, or B-cell follicles, whereas non-T_FH–derived cells gave rise to cells with lower CXCR5 expression and were found exclusively in the T-cell area. More importantly, T_FH-effector-derived memory cells had potent helper activity for antibody responses when B1-8^hi memory B cells were transferred into the same recipients (Fig. S3 B and C). The T_FH-effector-derived memory T cells elicited stronger plasma cell differentiation and IgG1 response than their non-T_FH–derived counterparts. Thus, we concluded that the majority of CXCR5⁺ memory T cells originate from T_FH effector cells.

Bcl6 Is Expressed in CXCR5⁺ Memory T Cells.

Bcl6 is a master regulator of T_FH generation, and T_FH cells express high levels of Bcl6 (1). However, it is not clear whether Bcl6 is also required for maintenance or function of memory T cells derived from T_FH effector cells. Before addressing this question, we determined the expression status of Bcl6 by examining Bcl6 protein levels in Bcl6-YFP reporter mice (17). As shown in Fig. 2A, a fraction of CXCR5^hi effector T cells at 1 wk after immunization expressed high levels of Bcl6-YFP. The Bcl6-YFP^hi cells were not present in the memory T-cell population. CXCR5⁺ memory T cells expressed quite low but slightly higher levels of Bcl6-YFP than naïve or CXCR5⁻ memory T cells. Bcl6 mRNA levels correlated well with its protein levels (Fig. 2B). Twenty-four hours after rechallenge with soluble NP-Eα-Ova, cells with high levels of Bcl6-YFP were induced. CXCR5 was also up-regulated and high levels of Bcl6-YFP were observed in CXCR5^hi cells (Fig. 2C). Bcl6-YFP expression was detected in the T–B border and B-cell follicles in the spleen (Fig. 2D), where CXCR5⁺ memory T cells were also enriched, suggesting the involvement of B cells in the activation of CXCR5⁺ memory T cells.

Fig. 2.

Rapid induction of Bcl6 expression in CXCR5⁺ memory T cells upon rechallenge. T cells from TEa Tg × Bcl6-YFP reporter mice (CD45.1⁺) were transferred into B6 hosts (CD45.2⁺) that were then immunized with NP-Eα-GFP in alum. (A) Bcl6-YFP expression on CD45.1⁺ T cells at day 7 or 56 was examined. The data are representative of three independent experiments. Mean fluorescence intensity (MFI) of Bcl6-YFP in naïve, T_FH (at day 7), CXCR5⁻, or CXCR5⁺ memory TEa T cells (at day 56) is summarized (Right) (n = 3). (B) Bcl6 mRNA levels in TEa T cells at the indicated time points (Left, n = 3) or at 8 wk after transfer (Right, n = 3) were measured. (C) After 8 wk of transfer, Bcl6-YFP expression in CD45.1⁺ T cells 24 h after rechallenge with NP-Eα-OVA was measured. (D) Bcl6-YFP expression in a frozen spleen section 24 h after rechallenge was analyzed. A total of 100 of YFP⁺ cells were counted from three independent experiments. The graph summarizes the localization of those YFP⁺ cells in the indicated area of spleen. The tissue section is representative of three experiments. (E) CXCR5⁺ or CXCR5⁻ memory TEa T cells were isolated and transferred (8,000 cells) into B6 mice, followed by stimulation with NP-Eα-OVA. Four days later, the expression of Bcl6-YFP in CD45.1⁺ cells was measured. Flow cytometric data are representative of two independent experiments. Bar graphs represent mean ± SD, n = 3. (F) Kinetics of Bcl6-YFP expression during primary activation of naïve T cells or recall response by memory T cells primed 6 wk before were compared. Data are from three independent experiments and are shown as mean ± SD *P < 0.05, **P < 0.01, ***P < 0.005; NS, not significant.

To investigate whether CXCR5⁺ memory T cells are the cells that express Bcl6 upon rechallenge, CXCR5⁺ or CXCR5⁻ memory T cells were purified, transferred, and restimulated with soluble antigen. As shown in Fig. 2E, Bcl6-YFP expression was induced by CXCR5⁺ donor cells, but barely by CXCR5⁻ counterpart, clearly suggesting that CXCR5⁺ memory T cells retain a capacity to redifferentiate into Bcl6⁺ T_FH cells.

Bcl6 Expression Is Rapidly Induced in CXCR5⁺ Memory T Cells upon Restimulation.

We investigated the kinetics of Bcl6 expression in memory versus naïve TEa T cells in more detail, revealing the rapid and robust Bcl6 induction in memory TEa cells (Fig. 2F). Bcl6 expression reached its maximum in memory T cells on day 2 after rechallenge, whereas it took 6–7 d for naïve TEa cells to express maximum levels of Bcl6. Moreover, although fewer than 10% of naïve TEa cells expressed Bcl6 during primary activation, 20–30% of memory TEa cells expressed high levels of Bcl6. We confirmed the Bcl6 expression in endogenous memory T cells as well (Fig. S4). As in the case of TEa TCR Tg memory T cells, 2W1S-specific memory T cells rapidly expressed high levels of Bcl6-YFP 2 d after rechallenge. In addition, by monitoring IL-21 reporter mice (18) we found that more than 20% of activated 2W1S-specific memory T cells expressed IL-21 at 24 h after restimulation (Fig. S5). Again, high levels of IL-21 were observed in the CXCR5^hi population. Taken together, these data demonstrate that Bcl6 is induced much earlier in memory T_FH cells than in naïve T cells and that this rapid Bcl6 expression has a good correlation with IL-21 expression.

Bcl6 Is Required for Maintenance of CXCR5⁺ Memory T Cells.

To examine the function of Bcl6 in T_FH memory cells we used mutant mice carrying a loxP-flanked bcl6 exon 7–9 allele (Bcl6 f/f). The mice were crossed with Cre-ERT2 and TEa TCR transgenic mice, which allowed conditional deletion of the bcl6 gene from TEa memory T cells by administration of tamoxifen. TEa CD4⁺ T cells were purified from Cre-ERT2 or Cre-ERT2 × Bcl6 f/f mice and were adoptively transferred into C57BL6 mice. Six weeks after immunization with NP-Eα-GFP/alum, tamoxifen was administered on three consecutive days to delete the bcl6 gene from the transferred T cells (Fig. 3A). We confirmed that 90% of loxP-flanked DNA was deleted in memory TEa T cells in Cre-ERT2 × Bcl6 f/f mice (Fig. 3B). The phenotype of TEa memory T cells was examined 10 d after the last tamoxifen treatment. Deletion of the bcl6 gene by tamoxifen administration did not affect the number of CXCR5⁻ memory T cells (Fig. 3C). However, we observed a significant decrease in the number of CXCR5⁺ memory T cells (Fig. 3C), suggesting that Bcl6 is required for the survival of CXCR5⁺ memory T cells.

Fig. 3.

The requirement of Bcl6 for the maintenance of CXCR5⁺ memory T cells. (A) Schematic illustration of the experiment. Naïve CD4 TEa T cells (CD45.1⁺) from Cre-ERT2 or Cre-ERT2 × Bcl6 f/f mice were transferred into B6 mice. The mice were immunized with NP-Eα-GFP in alum. Six weeks after immunization, tamoxifen (2 mg per mouse) was administered i.p. for three consecutive days. Spleens were isolated 10 d later for analysis. (B) Deletion efficiency of floxed exon 8 of the bcl6 gene in memory TEa CD4 T cells was examined by real-time PCR (n = 5). (C) Expression of CXCR5 and PD1 on memory TEa T cells was examined. The number of CXCR5⁻ and CXCR5⁺ TEa T cells is summarized (Right). Data are expressed as mean ± SD (n = 5). (D and E) Memory TEa T cells were transferred (10,000 cells) into B6 mice that had received B1-8^hi memory B cells. Four days after rechallenge, the expression of CXCR5 or PD1 on transferred T cells (n = 3) (D) and CD138 expression by transferred B1-8^hi memory B cells (E) was analyzed (n = 3). Data are shown as mean ± SD *P < 0.05; NS, not significant.

The requirement of Bcl6 for the survival of CXCR5⁺ memory T cells was further confirmed. CXCR5⁺ memory TEa T cells derived from Cre-ERT2 × Bcl6 f/f mice were purified and transferred to congenic mice, followed by tamoxifen treatment. As shown in Fig. S6, bcl6 deletion by tamoxifen treatment significantly decreased the number of donor-derived cells, suggesting that loss of CXCR5⁺ memory T cells was due to cell death, but not to phenotypic change.

We purified surviving memory T cells 10 d after the last tamoxifen treatment and transferred them into C57BL6 mice that had received B1-8^hi memory B cells. Upon rechallenge with NP-Eα-OVA, generation of CXCR5^hiPD1^hi T cells from transferred memory T cells was strongly inhibited by bcl6 deletion (Fig. 3D). Consequently, generation of CD138⁺ plasma cells from NP-specific memory B cells was also compromised (Fig. 3E). Collectively, these results show that Bcl6 is essential for maintenance of CXCR5⁺ memory T cells and subsequent memory antibody response.

CD11c⁺ Cells Are Dispensable for the Activation of CXCR5⁺ Memory T Cells and Secondary Ab Responses.

The induction of Bcl6 in naïve T cells has been reported to require two steps (5–7, 9, 19); Bcl6 is first induced by interactions between conventional dendritic cells (DCs) and T cells. Then, although cognate B cells are not required for the induction of Bcl6, they support the expansion of Bcl6-expressing T_FH cells. Rapid induction of Bcl6 in memory T cells, mainly in the T–B border region as described above, suggested that Bcl6 in CXCR5⁺ memory T cells might be induced in a manner different from that in naïve T cells. Therefore, we first determined whether interactions between CXCR5⁺ memory T cells and conventional DCs are necessary for up-regulation of Bcl6 and subsequent secondary antibody responses. We transferred naïve TEa × Bcl6-YFP T cells into CD11c-DTR Tg mice, followed by immunization with NP-Eα-GFP/alum. Six weeks later, the mice were administered diphtheria toxin (DT) to ablate CD11c⁺ cells and then were rechallenged on the next day with soluble NP-Eα-OVA to induce memory T-cell activation (Fig. 4A). We confirmed that DT treatment efficiently depleted CD11c^hi cells (Fig. S7A). Rechallenge with NP-Eα-OVA induced up-regulation of Bcl6 in CD11c-DTR mice (Fig. 4A). Notably, depletion of CD11c⁺ cells in CD11c-DTR mice did not compromise Bcl6 up-regulation in CXCR5⁺ memory T cells. Along with these data, the secondary anti-NP IgG1 response was normally induced even after CD11c⁺ cells were depleted (Fig. 4B). When NP-Eα-Ova with alum were used for the rechallenge, Bcl6 up-regulation was again unaffected by depletion of CD11c⁺ cells (Fig. S7B). As has been previously demonstrated (7), ablation of CD11c⁺ cells resulted in impaired Bcl6 induction by naïve T cells (Fig. S7C). Thus, we concluded that upon rechallenge with soluble antigen CD11c⁺ cells are not required for generation of Bcl6^hi T cells and subsequent memory antibody responses.

Fig. 4.

Bcl6 expression by memory T cells and recall antibody response in the absence of CD11c⁺ cells. (A) Naïve TEa × Bcl6-YFP T cells (CD45.1⁺) were transferred into CD11c-DTR Tg mice (CD45.2⁺) that were then immunized with NP-Eα-GFP in alum. Six weeks after immunization, the mice were administered DT and were rechallenged on the next day with NP-Eα-OVA in PBS. Bcl6-YFP expression by memory T cells (CD45.1⁺) in mice with or without DT treatment was examined 1 d after rechallenge with NP-Eα-OVA. Flow cytometric data are representative of three independent experiments. Bar graphs represent mean ± SD (Right, n = 3). (B) CD11c-DTR Tg mice immunized with NP-CGG in alum 6 wk before were left untreated or administered DT. The mice were rechallenged with NP-CGG. On day 6 after rechallenge, serum anti–NP-IgG1 responses were examined. Data are expressed as mean ± SD (Right, n = 4), NS, not significant.

Antigen-Specific Memory B Cells Efficiently Present Antigen and Activate CXCR5⁺ Memory T Cells.

We next attempted to determine which cells could present antigen to activate CXCR5⁺ memory T cells during secondary immune responses. Soluble NP-Eα-GFP antigen was administered to WT mice that were unprimed or previously primed with NP-CGG/alum. In this setting, presentation of the Eα peptide could be monitored with the Y-Ae mAb, which is specific for Eα:I-A^b complexes. We examined antigen presentation by DCs (CD11c^hi MHC class II^hi), total B cells (B220⁺) or NP-specific naïve B cells (B220⁺NIP⁺CD38^hi), and NP-specific memory B cells (B220⁺NIP⁺CD38^hiCD273⁺). As demonstrated in Fig. 5A, if we compared total B220⁺ B cells and total CD11c^hi DCs, 0.8% and 2.4%, respectively, presented Eα peptide 24 h after antigen injection. However, if we focus on NP-specific B cells, 8.9% of NP-specific naïve B cells present the antigen. Notably, NP-specific memory B cells presented the antigen more efficiently; 17% of them were Y-Ae positive.

Fig. 5.

Antigen presentation by memory B cells in vivo. (A) C57BL6J mice were immunized with NP-CGG in alum. Eight weeks later, the mice were left untreated or rechallenged with NP-Eα-GFP in PBS. Spleens were harvested on the next day and antigen presentation by total B cells (B220⁺), NP-specific memory B cells (B220⁺NIP⁺CD38^hiCD273⁺), or dendritic cells (CD11c^hi MHC class II^hi) was detected with the Y-Ae mAb. Antigen presentation by naïve B cells (B220⁺NIP⁺CD38^hi) purified from naïve mice challenged with NP-Eα-GFP in PBS 1 d before was also examined. Flow cytometric data are representative of three independent experiments. Bar graphs represent mean ± SD (Right, n = 3),*P < 0.05, **P < 0.01. (B) Naïve TEa × Bcl6-YFP T cells (CD45.1⁺) were transferred into B6 mice (CD45.2⁺) that were then immunized with Eα-GFP in alum. Six weeks later, NP-specific or NP-nonspecific memory IgG1 B cells isolated from WT B6 mice were transferred (5,000 cells), followed by rechallenge with NP-Eα-OVA. Bcl6-YFP expression by CD45.1⁺ memory T cells was examined 2 d later. Flow cytometric data are representative of two independent experiments. The graph represents mean ± SD, n = 3, *P < 0.05.

To examine whether antigen-specific memory B cells could indeed contribute to the activation of CXCR5⁺ memory T cells, we transferred TEa × Bcl6-YFP T cells into congenic mice, followed by immunization with Eα-GFP/alum. Then, we transferred NP-specific or NP-nonspecific memory B cells into the primed mice, just before the rechallenge with NP-Eα-OVA. As shown in Fig. 5B, transferred NP-specific memory B cells were able to enhance Bcl6-YFP expression by TEa memory T cells, whereas NP-nonspecific memory B cells failed to induce Bcl6-YFP. These results suggest that antigen-specific memory B cells activate cognate memory T cells to express Bcl6. This induction of Bcl6 was blocked by pretreatment of NP-specific memory B cells with anti-class II Ab (Fig. S8). Collectively, these results demonstrate that antigen-specific memory B cells indeed contribute to rapid Bcl6 up-regulation in CXCR5⁺ memory T cells by virtue of APC function.

Discussion

We previously demonstrated that IgG1 memory B cells require CD4 T cells to initiate secondary humoral immune responses (20). In this study, we have addressed two issues: which T-cell subset is mainly responsible for helping IgG1 memory B cells and how these T cells are activated upon secondary challenge, thereby inducing rapid and robust humoral responses. Here, we show that CXCR5⁺ T_FH memory cells localize close to B-cell follicles and are superior in helping memory B-cell activation. Furthermore, we show that the CXCR5⁺ T_FH memory cells promptly reexpress Bcl6 upon secondary challenge, and that this is primarily induced by the APC function of memory B cells. Thus, our results reinforce the importance of cognate interactions between T_FH memory cells and memory B cells in humoral memory responses.

Regarding the commitment or plasticity of CXCR5⁺ T_FH memory cells, several recent studies have come to different conclusions (13, 14, 21). Although our study has not directly addressed this issue, two lines of our evidence from our studies are more consistent with the idea that CXCR5⁺ T_FH memory cells are committed populations that are poised for the lineage-specific reexpression of effector molecules upon recall. First, resting T_FH and non-T_FH memory cells arose from the corresponding effector cells, respectively. Second, CXCR5⁺ T_FH, but not CXCR5⁻ non-T_FH, memory cells up-regulated Bcl6 upon antigen rechallenge. In regard to the mechanistic aspect of the commitment of T_FH memory cells, the low level of Bcl6 expression at the memory phase, as discussed below, might not allow the T_FH memory cells to assume other lineages.

Based on studies using germ-line knockout or transgenic mice, Bcl6 has previously been suggested to be involved in CD8 and CD4 T-cell memory development (22, 23). However, it has not been clear whether Bcl6 is required for maintenance of memory T cells. By taking advantage of a conditional deletion system for the bcl6 gene we could demonstrate that T_FH memory cells rely on Bcl6 for their survival. Inducible deletion of bcl6 from the antigen-specific memory T-cell compartment selectively decreased the number of CXCR5⁺ memory T cells. Consistent with a previous report (24), CXCR5⁺ T_FH memory cells have quite low levels of Bcl6, only slightly higher than those in their CXCR5⁻ counterparts or in naïve T cells. Conceivably, such low levels of Bcl6 are sufficient and required for survival of these cells. The molecular mechanisms by which Bcl6 controls survival of T_FH memory cells are currently speculative. Given that Blimp-1 and Bcl6 are antagonistic transcription factors, repression of Blimp-1 by Bcl6 might be one of the potential survival mechanisms. Indeed, in the case of Blimp-1–deficient CD8 T cells, memory precursor cells survived better (25).

We and others previously proposed that memory B cells are the primary APCs in the memory response and that locally confined T_FH memory cells are the cognate regulators of the memory B-cell response (26, 27). These proposals are well substantiated by the following two lines of evidence presented in this study. First, memory B cells present antigens with high efficiency upon soluble antigen rechallenge compared with naïve B cells. Furthermore, memory B cells are significant contributors to the rapid up-regulation of Bcl6 on CXCR5⁺ T_FH memory cells upon rechallenge. Second, the rapid and robust Bcl6 expression in CXCR5⁺ T_FH memory cells was observed in locally confined regions (at the T–B border or in B-cell follicles), strongly suggesting the occurrence of cognate interactions between memory B cells and locally confined T_FH memory cells. Although our data define memory B cells as the major APCs, it still remains possible that other APCs, such as DCs, can participate at least to some extent. Indeed, a recent report shows that even in a B-cell–deficient condition recall T_FH-like response can occur. In these studies in a lymphocytic choriomeningitis virus infection system in B-cell–deficient μMT mice T_FH memory cells were able to recall a T_FH-like response, although the efficiency was lower compared with WT mice (21). These observations, at first glance, seem to contradict our conclusion. However, in the life-long B-cell–deficient condition there may be some compensation and other APCs probably play a more crucial role in activating T_FH memory cells.

Because the kinetics of IL-21 and Bcl6 up-regulation in CXCR5⁺ T_FH memory cells upon rechallenge are correlated, it is likely that rapid Bcl6 up-regulation is a primary inducer of rapid differentiation of T_FH memory cells toward effector cells. In regard to the rapid Bcl6 up-regulation, three mechanisms can be envisaged. First, memory B cells with relatively high-affinity B-cell antigen receptors are able to rapidly capture low levels of secondary antigen and present this antigen to the cognate T_FH memory cells. In this context, increased levels of CD80 and MHC class II on memory B cells could contribute to efficient activation of T_FH memory cells (28). Second, cognate memory T_FH cells reside in close proximity to memory B cells, which should facilitate their interactions. Finally, T_FH memory T cells might undergo positive epigenetic modification of genes that allow them to swiftly up-regulate Bcl6. For instance, in the case of T_H1 memory cells, Hale et al. recently demonstrated the epigenetic modification of the granzyme B locus (21).

Given the functional heterogeneity of memory B-cell subsets (e.g., IgM⁺ vs. IgG1⁺ memory B cells) (29, 30), it is possible that each subset might differentially contribute to activation of T_FH memory cells. Thus, better understanding of the regulatory mechanisms in the interactions of these memory B-cell subsets and T_FH memory cells should provide important insights for development of better vaccines.

Materials and Methods

Mice.

Mice were maintained under specific pathogen-free conditions in accordance with the guidelines of the Animal Care and Use Committee of Osaka University. C57BL/6 mice were purchased from CLEA Japan. TEa TCR transgenic mice (CD45.2/CD45.2 or CD45.1/CD45.2) (31), Bcl6-YFP reporter mice (17), CD11c-DTR transgenic mice (32), B1-8^hi IgH knock-in mice (33), and IL-21 hCD2 Bac Tg reporter (15) mice were bred on-site. Bcl6 flox mice were independently generated using the same strategy as in a previous study (34) and were bred with Rosa-CreERT2 mice (purchased from Taconic) and TEa TCR transgenic mice.

Antigens.

The Eα-GFP-expression vector (Eα-GFP-pTrcHis2) was kindly provided by M. Jenkins, University of Minnesota Medical School, Minneapolis. The 2W1S-GFP-expression vector (2W1S-GFP-pTrcHis2) was generated by PCR mutagenesis of an Eα cDNA. Eα-OVA- or 2W1S-OVA-expression vectors were generated by replacing the GFP cDNA with an OVA cDNA (provided by T. Aoshi, National Institute of Biomedical Innovation, Osaka). Protein production was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside overnight. Eα-GFP, Eα-OVA, 2W1S-GFP, and 2W1S-OVA fusion proteins were purified from bacterial lysates using a Ni-NTA superflow cartridges (Qiagen). Endotoxin was removed from the purified proteins using EndoTrap red (Hyglos GmbH). NP conjugates were generated by coupling NP-succinimide ester (NP-Osu; Biosearch Technologies) to the purified fusion proteins or chicken γ-globulin (CGG) using NP-Osu (Biosearch Technologies).

Adoptive Transfer and Injections.

To generate antigen-specific memory T cells, 1–2 × 10⁶ TEa naïve T cells were transferred i.v. into WT B6 mice, followed by immunization i.p. with 100 μg of NP-Eα-GFP or Eα-GFP in alum. Six weeks later, spleens were harvested for analysis or sorting of memory T-cell subsets. For generation of memory B cells, 1 × 10⁵ B1-8^hi B cells were transferred into B6 mice, followed by immunization i.p. with 100 μg of NP-CGG in alum. Four to six weeks later, CD45.1⁺B220⁺CD38^hiNP⁺IgG1⁺ cells were sorted as NP-specific IgG1⁺ memory B cells. Sorted TEa memory T cells (1–2 × 10⁴ cells) or B1-8^hi memory B cells (10,000 cells) were transferred back to B6 mice, followed by immunization with 50 μg of NP-Eα-OVA in PBS. For deletion of the bcl6 gene from memory T cells, 2 mg of tamoxifen (Sigma) was administered i.p. for three consecutive days to B6 recipient mice that were previously transferred with TEa × CreERT2⁺ × Bcl6 f/f T cells. For depletion of CD11c⁺ cells, DT (Sigma) was administered i.p. to CD11c-DTR mice at 4 ng/g body weight.