Whole-Virion Influenza Vaccine Recalls an Early Burst of High-Affinity Memory B Cell Response through TLR Signaling Free

This article is featured in In This Issue, p.3973

Many licensed vaccines aim to elicit humoral memory responses that lead to a rapid and robust increase in serum neutralizing Abs after antigenic re-exposure (1, 2). Among the best examples are inactivated influenza vaccines, which are repeatedly administered to naive infants to prime and boost humoral memory and to adults to boost pre-existing memory (3). Ab recall responses are mediated by memory B cells that rapidly proliferate and differentiate into plasma cells in response to booster vaccination (4). Therefore, the rapid and robust reactivation of memory B cells contributes significantly to the effectiveness of influenza vaccination.

Several influenza vaccines with different formulations are licensed worldwide, including live attenuated and inactivated influenza vaccines. Among them, the inactivated influenza vaccine has long been used in many countries and is composed of whole- and split-virion (subvirion) vaccines (3, 5, 6). Formalin-inactivated whole-virion vaccines retain particulate structures along with internal ssRNA (5, 6). Conversely, split-virion vaccines are ether-treated viruses composed of disrupted viral proteins lacking particulate structure and internal ssRNA (5, 6). Most inactivated vaccines manufactured since the 1970s have been of the split-virion type because it has reduced reactogenicity but seemingly retains the comparable immunogenic properties that boost pre-existing memory response in humans (5–7). However, a detailed analysis of the characteristic of the recall responses elicited by both vaccine types in animal models is still lacking. Therefore, the effect of the structural differences of the vaccines on the quality and magnitude of secondary Ab responses and whether structure influences the effectiveness of booster vaccination are unknown.

ssRNAs within inactivated whole-virion vaccines are the ligands for endosomal TLR7, which is expressed on several types of cells including dendritic cells (DCs) and B cells (8–11). The role of TLRs in B cell responses in vivo has been controversial (12–14). However, recent evidence confirms that both B cell–intrinsic and –extrinsic TLRs shape virus-specific B cell responses in vivo, although the extent of that response depends on the experimental system. For example, Koyama et al. (15) found that TLR7 on plasmacytoid DCs (pDCs) potentiates primary Ab response through the secretion of type I IFN, whereas secondary Ab response is largely independent of pDCs. The roles of B cell–intrinsic TLRs are also elucidated in B cell–selective MyD88-deficient mice, in which primary IgG response and germinal center (GC) formation decrease profoundly after immunization with virus-like particles or nanoparticles plus TLR agonists (16, 17). The role of B cell–intrinsic TLR7/MyD88 signaling has also been addressed in several virus infection models, which have revealed the requirement of this signaling pathway for primary Ab and GC response in some viral infections (18, 19). Thus, B cell–intrinsic TLR expression on naive B cells indeed contributes to the primary B cell responses against particulate Ags and viruses, including TLR agonists, but its role on memory B cells during secondary Ab responses against viruses and vaccines is currently unknown. This issue is especially crucial to the efficacy of influenza vaccines, which in many instances are administered to reactivate memory B cells.

By tracking the in vivo dynamics of virus-binding B cells in the murine system, we analyzed the recall Ab response by memory B cells after booster vaccination with whole-virion vaccines. We show in this article that, compared with split-virion vaccines lacking TLR7 agonist (ssRNA), whole-virion vaccines reactivate memory B cells in a T cell–independent (TI) manner with the help of B cell–intrinsic TLR signals, which contribute to the faster supply of higher-affinity Abs. Notably, we found that this TLR-mediated pathway was unique to memory B cells, because naive B cells underwent TI activation without TLR signaling. Thus, our data indicate a novel role of B cell–intrinsic TLR signaling in humoral memory responses to booster influenza vaccination.

The NIBRG-14 (H5N1) virus was provided by the National Institute for Biological Standards and Controls. The NIBRG-14 viruses were propagated on embryonated chicken eggs and purified through a 10–50% sucrose gradient as previously described (20). For inactivated whole-virion vaccines, purified viruses were treated with 0.1% formalin at 4°C for a week. For split-virion vaccines, viruses were suspended in 0.1% Tween 80 and mixed with an equal volume of ether. After a 30-min stirring, the solution was centrifuged to separate each phase and the ether phase was discarded. The remaining aqueous phase was collected and evaporated. Both whole- and split-virion vaccines were prepared via mixing with 9% aluminum potassium sulfate solution. Then, after adjusting to pH 7 by adding 1N KOH solution, the mixtures were left to precipitate for over 3 h at 4°C before use.

C57BL/6J mice were purchased from Japan SLC, and C57BL/6-scid mice were purchased from Jackson Laboratories. MyD88^−/−TRIF^−/− and TLR7^−/− mice were kindly provided by S. Akira (Osaka University). IFNAR2^−/− mice were purchased from B&K. mb-1 Cre^KI/WT mice were kindly provided by M. Reth (University Freiburg) (21) and crossed with MyD88-floxed mice (Jackson Laboratories) to generate MyD88^f/f-mb-1 Cre^KI/WT mice. Anti-hemagglutinin (anti-HA) V_H knock-in mice were generated by targeting the recombined V_H gene to the H chain locus. The targeting vector was constructed as follows: a 0.9-kb fragment of recombined V_H gene was isolated from anti-HA IgG2a-producing mouse hybridoma (OMb) (22) by using PCR with the primers (5′-CATGCTGCAGATGAACTTCGGGCTCAGCTTGA-3′, 5′-CATTGGATCCTGA GGAGACGGTGACTGAGGTTCCTTGACCCAGTAGTCC-3′) and inserted into a pBSK vector with diphtheria toxin A and FRT-flanked PGK-neo^r genes. The long arm and short arm of homology, which consisted of a 6.1-kb fragment 5′ upstream of DQ52 and a 1.7-kb fragment 3′ downstream of J_H4, respectively, were inserted into the vector. The targeting vector was linearized and transfected into Bruce4 embryonic stem cells. G-418–resistant clones were screened for homologous recombination by PCR, and positive clones were used to generate chimeric mice. The resulting mice were bred with flippase recombinase transgenic mice to delete the neo^r gene and then backcrossed to the C57BL/6-Ly5.1 genetic background. For the experiments, these mice were crossed with TLR7^−/− mice to generate TLR7-deficient V_H knock-in mice. For analysis of the genotypes, tail DNA was subjected to PCR with the primer pair: 5′-CAGTATTCTCTGTTTGCAGGTGTCCAC-3′ and 5′-GTAGTCCATAGCATAT CCGTAGTATTC-3′, and PBMCs were subjected to flow cytometry analysis with a fluorescent HA probe to confirm the expression of BCRs bearing the V_H knock-in gene in their B cells. For priming, mice were s.c. immunized with alum-precipitated whole- or split-virion vaccines (20 μg/head) and boosted with the same vaccines at 3-wk intervals. Similarly, for generating V_H knock-in–derived memory B cells, naive C57BL/6 (Ly5.2) mice receiving 10⁶ cells of V_H knock-in naive B cells (Ly5.1) were immunized as described earlier.

Anti-FcγRII/III (2.4G2) and anti-mouse CD38 (CS2) mAb were purified in our laboratory. Recombinant HA (rHA) was prepared and conjugated with PE as previously described (23). The CD38 (CS2) mAb was conjugated with AlexaFluor647 and Pacific orange (Molecular Probes) in our laboratory. Anti–IgM (II/41)-biotin, anti–IgD (11-26c)-biotin, anti–B220/CD45R (RA3-6B2)-biotin, anti–CD93 (AA4.1)-biotin, anti–Thy1.2/CD90.2 (53-2.1)-biotin, anti–CD3 (145-2C11)-biotin, anti–Gr-1 (RB6-8C5)-biotin, anti–F4/80 (BM8)-biotin, anti–CD11b (M1/70)-biotin, anti–CD11c (N418)-biotin, anti–CD117 (2B8)-biotin, anti–CD8α (53-6.7)-biotin, anti–TER-119 (TER-119)-biotin, anti–CD5 (53-7.3)-biotin, anti–CD4 (GK1.5)-biotin, anti–CD24 (M1/69)-PE, and anti–CD93 (AA4.1)-allophycocyanin were purchased from eBioscience. Anti-CD43 (S7)-biotin/FITC, anti–CD138 (281-2)-biotin/allophycocyanin, anti–CD19 (1D3)-biotin, anti–IgG1 (A85-1)-FITC, anti–IgM (II/41)-FITC, and streptavidin (SA)-PE-Texas Red were purchased from BD Biosciences. Anti–CD38 (90)-FITC/allophycocyanin, anti–CD80 (16-10A1)-allophycocyanin, anti–CD45.1(A20)-Pacific blue, anti–CD21 (7E9)-Pacific blue, anti–CD23 (B3B4)-PE Cy7, and anti–B220/CD45R (RA3-6B2)-PE/Pacific blue/AlexaFluor700 were purchased from BioLegend. SA-Tricolor was purchased from Invitrogen. Goat anti-mouse IgG2c-alkaline phosphatase (AP), goat anti-mouse IgG2b-AP, goat anti-mouse IgM-AP, goat anti-mouse IgG1-HRP, goat anti-mouse IgG3-HRP, and goat anti-mouse IgG-HRP were purchased from Southern Biotech.

Single-cell suspensions were prepared from spleens in DMEM containing 2% FCS, 2 mM l-glutamine, 100 IU/ml penicillin, 100 mg/ml streptomycin, and 5 × 10⁻⁵ M 2-ME as previously described (23). Cells were pretreated with the anti-FcγRII/III mAb and then incubated with biotinylated mAbs against IgM, IgD, CD43, CD138, CD5, CD93, CD90, CD3, CD4, CD8α, F4/80, CD11b, CD117, CD11c, Gr-1, and TER-119 for memory B cells and IgM, IgD, CD90, CD3, CD4, CD8α, F4/80, Ter-119, and Gr-1 for plasma cells. This incubation was followed by staining with anti–B220-Pacific Blue/Alexa 700, anti–CD38-allophycocyanin/Pacific orange, rHA-PE, SA-PE-Texas red, anti–IgG1-FITC, and propidium iodide (PI) for memory B cells, and anti–B220-Pacific Blue, rHA-PE, SA-PE-Texas Red, PI, and anti–CD138-allophycocyanin for plasma cells. In addition, anti–IgM-FITC, anti–Ly5.1-Pacific blue, and anti–CD80-allophycocyanin Abs were also used in some experiments. For the sorting of memory B cells, the cells stained with biotinylated Abs were depleted by using a MACS system (Miltenyi Biotec), followed by staining with fluorescence-conjugated reagents. Stained cells were analyzed or purified by using a FACS Canto II or FACSAria (BD Biosciences). A total of >2,000,000 events were collected, and data were analyzed with FlowJo software (TreeStar).

For the detection of anti-HA plasma cells with ELISPOT assay, nitrocellulose membranes were coated with 20 μg/ml rHA, and cells were incubated on the membranes for 2 h at 37°C. After the cells were washed off, the membranes were incubated with anti-mouse IgG2c-AP, anti-mouse IgG2b-AP/anti-mouse IgG1-HRP, or anti-mouse IgM-AP/anti-mouse IgG3-HRP. AP and HRP activities were visualized as previously described (23). For the detection of anti-HA Ab titers, ELISA was performed with anti-mouse IgG-HRP, anti-mouse IgG1-HRP, or anti-mouse IgG2c-HRP as previously described (23). The avidity index was calculated by dividing anti-HA IgG Ab titers resistant to 7 M urea treatment (15 min) by total anti-HA IgG Ab titers. Virus-neutralization Ab titers were determined with a microneutralization assay using MDCK cells (20). In brief, RDE-treated serum was incubated with NIBRG-14 virus (100 median tissue culture infectious dose) for 30 min at 37°C and then added to MDCK cells together with acetyltrypsin (Sigma). After 3 d of incubation, anti–HA-IgG Ab titers, at which 50% of the cultures were protected from infection, were determined and the reciprocal numbers were plotted as virus-neutralization Ab titers.

IgM/D⁻HA-binding CD38⁺ memory B cells were plated in 96-well plates at 10³ cells/well with irradiated naive B cells (1 × 10⁵ cells/well) and cultured in complete IMDM medium (Sigma-Aldrich) supplemented with murine IL-2 (20 U/ml; Sigma-Aldrich), murine IL-4 (400 U/ml; Peprotech), murine IL-5 (50 U/ml; Peprotech), murine IL-21 (50 ng/ml; R&D Systems), and anti-CD40 (10 μg/ml; FGK45) for 6 d. For determining the avidities of the memory B cells, the supernatants were subjected to a modified ELISA, and the avidity index was calculated as described earlier.

Spleens were embedded in Tissue-Tek OCT Compound (Sakura Finetek, Tokyo, Japan). After freezing, 7-μm sections were deposited on slides and fixed in ice-cold acetone. For immunohistochemistry, sections were pretreated with anti-FcγRII/III mAb and stained with fluorescein-labeled peanut agglutinin (PNA) (VECTOR), anti–B220-PE, and anti–CD38-Alexa 647. Stained sections were scanned under a confocal laser-scanning microscope LSM 510 (Carl Zeiss).

B cells and CD4⁺ T cells were purified from pooled spleens by using a MACS system with biotinylated mAbs against CD3, CD90, CD4, CD8α, F4/80, Gr-1, CD11b, CD117, and CD138 for B cells and CD19, B220, IgM, IgD, F4/80, Gr-1, CD11b, and CD8α for CD4⁺ T cells, followed by SA-microbeads (Miltenyi Biotec). B cells (2 × 10⁶/head) from naive mice and CD4⁺ T cells (1 × 10⁶/head) from vaccinated wild-type mice were i.v. injected into C57BL/6-scid mice with or without sorted memory B cells. On the next day, the recipient mice were i.v. boosted with 20 μg alum-free vaccines.

Single HA-binding memory B cells were directly sorted into 10 μl H₂O containing 50 ng carrier RNA (Qiagen). RT reaction was performed with a SuperScript III CellsDirect cDNA Synthesis Kit (Invitrogen) and random hexamers, according to the manufacturer’s instruction. V_H genes were amplified with two rounds of PCR from 1 μl cDNAs by using Taq DNA polymerase (Qiagen). The PCR primers were 5′ MsVHE (5′-GGGAATTCGAGGTGCAGCTGCAGGAGTCTGG-3′) plus 3′ Cγ1 outer (5′-GGAAGGTGTGCACACCGCTGGAC-3′) for the first round, and 5′ MsVHE plus 3′ Cγ1 inner (5′-GCTCAGGGAAATAGCCCTTGAC-3′) for the second round. PCR was performed with 1 cycle of 15 min at 94°C, 40 cycles of 30 s at 94°C, 20 s at 50°C, and 60 s at 72°C. The amplified PCR products were purified and directly subjected to DNA sequencing analysis with Cγ1 inner primer. Germline V_H-D_H-J_H genes and the location of somatic hypermutations were identified with the IgBLAST software from the National Center for Biotechnology Information.

Statistical significance was determined using an unpaired, two-tailed Student t test. The p values <0.05 were considered significant and are indicated by asterisks: *p < 0.05, **p < 0.01, ***p < 0.001.

Complete sequence data are available from GenBank (http://www.ncbi.nlm.nih.gov/genbank) under the following accession numbers: whole-virion–induced memory B cells (AB734714-AB734734); split-virion–induced memory B cells (AB734735-AB734765).

Compared with split-virion vaccines, whole-virion vaccines possess superior immunogenicity in naive hosts (5, 6). Although whole-virion vaccines enhance the primary Ab response through TLR-mediated stimulation of both pDCs and B cells (15, 16), the possible differential impacts of both types of vaccines on the generation and maintenance of virus-specific memory B cells have not been previously clarified. Therefore, by using fluorescent HA probes, we first sought to trace virus-specific memory B cell responses after s.c. vaccination as previously reported (Fig. 1) (23, 24). We selected whole-virion vaccines of the H5N1 subtype for this study, because this is the only whole-virion vaccine currently stockpiled in Japan for humans. To follow the same vaccination protocols with those for humans, we primed mice with alum-adjuvanted H5N1 vaccines twice at 3-wk intervals for the priming and, after resting >60 d, boosted the animals with nonadjuvanted vaccines to assess the booster effects.

Immunization by both whole- and split-virion vaccines increased the numbers of HA-binding isotype-switched B cells with CD38⁺ (memory) and CD38^dull (GC) phenotypes (25, 26) at >20-fold higher levels than those in naive mice (Fig. 1). Compared with spit-virion vaccines, whole-virion vaccines generated higher amounts of CD38^dull GC B cells, which were more evident in the IgG1⁻ fraction. The dominance of the IgG1⁻ fraction was also observed in CD38⁺ memory B cells, whereas the numbers of IgG1⁺ memory B cells generated by both types of vaccine were comparable (Fig. 1B). For comparison of V_H gene mutations, HA-binding IgG1⁺ memory B cells were sorted at one cell/well, and their V_H genes, switched to IgG1 isotype, were amplified with a mixture of a universal V_H and Cγ1 primers (27). Mutational analysis of the amplified V_H genes revealed that both types of vaccine generated HA-binding memory B cells with comparable mutation numbers, suggesting that affinity maturation during the course of memory B cell generation was similar, regardless of vaccine formulation (Fig. 1C). These data indicate that there were no remarkable differences in the induction of HA-binding IgG1⁺ memory B cells among the two types of vaccines, although whole-virion vaccine evoked larger amounts of GC B cells and IgG1⁻ memory B cells.

We next examined the booster effects of whole- and split-virion vaccines on HA-binding isotype-switched memory B cells. To analyze the reactivation of memory B cells in vivo, we performed adoptive transfer experiments in which purified HA-binding isotype-switched memory B cells were restimulated with either whole- or split-virion vaccines in recipient mice (Fig. 2A, 2B) (23). Flow cytometry analysis revealed that compared with split-virion vaccines, whole-virion vaccines generated higher numbers of HA-binding CD138⁺ cells in the early phase of recall responses (days 4–6), although the numbers eventually reached a comparable level with both vaccines (Fig. 2C, 2D). ELISPOT analysis showed a similar result in the number of HA-binding IgG⁺ plasma cells (Fig. 2E). These data indicate that compared with split-virion boosters, whole-virion booster vaccinations recall an accelerated secondary response.

The affinity maturation of virus-binding Abs is a key determinant of virus neutralization activity (28–30). Therefore, we next examined the average affinity of the HA-binding Abs by using a modified ELISA to dissociate plate-bound, low-affinity Abs selectively via treatment with 7 M urea (31). Remarkably, compared with IgG Abs at day 7, those at day 4 were more resistant to urea detachment after boosting with whole-virion vaccines (Fig. 2F). Moreover, the avidity indices of whole-virion–boosted Abs at day 4 were higher than those recalled by split-virion–boosted Abs at day 7. These results support the conclusion that compared with split-virion vaccines, whole-virion vaccines stimulate high-affinity memory B cells for Ab production with faster kinetics.

The earlier emergence of plasma cells observed with whole-virion booster vaccination was reminiscent of TI Ab responses that have been well characterized in primary responses from innate-like B cells (B1 and marginal zone B cells) (32). Although memory B cells share the phenotypic features with follicular B cells (Supplemental Fig. 1), those with follicular phenotypes may be reactivated in a TI manner by whole-virion vaccines because there are several examples for TI reactivation of memory B cells in other virus systems (e.g., vesicular stomatitis virus and CMV) (33, 34). To test this possibility, we assessed the T cell dependency of memory B cell reactivation with an adoptive transfer experiment in which HA-binding memory B cells were stimulated with or without CD4⁺ T cells. In the presence of CD4⁺ T cells, both whole- and split-virion vaccines induced comparable numbers of plasma cells at day 7 after boosting (Fig. 2G). On the contrary and similar to many protein Ags, split-virion vaccines failed to induce plasma cells in the absence of CD4⁺ T cells (34–36). Notably, however, whole-virion vaccines recalled unaltered Ab responses even in the absence of CD4⁺ T cells. Thus, whole-virion vaccines reactivate memory B cells in a TI manner, which potentially accounts for the prompt reactivation.

The TLR agonists included in virus particles modulate primary Ab responses through both B cell–intrinsic and –extrinsic pathways at variable levels depending on the types and forms of viral Ags (15, 16). To explore the role of TLR signaling in the memory response elicited by whole-virion vaccines, we immunized MyD88^−/−TRIF^−/− mice, in which all TLR signaling is absent, with whole-virion vaccines. As previously shown in MyD88^−/− mice, serum IgG2c Ab titers were significantly decreased in MyD88^−/−TRIF^−/− mice, whereas IgG1 Ab titers remained fairly intact (data not shown) (37, 38).

Flow cytometry analysis revealed that the number of HA-binding isotype-switched IgG1⁻ GC B cells decreased in MyD88^−/−TRIF^−/− mice at days 21 and 60 after immunization with whole-virion vaccines. Conversely, the number of HA-binding IgG1⁺ GC B cells modestly increased (Fig. 3A, 3B). Histological analysis confirmed the presence of the GC structure (PNA⁺CD38^dullB220⁺) on spleen sections at day 21 after immunization, supporting the conclusion that MyD88/TRIF deficiency does not impair the development of GC pathways after immunization with whole-virion vaccines with alum adjuvant (Fig. 3C). The discrepancy between our results and previous findings of reduced GC formation in MyD88-deficient mice (16) could be caused by the use of alum and/or repeated vaccination in our experiments because alum may have the complementary capability to enhance GC response owing to inflammasome activation, mobilization of other cell types, or both (39–45).

The formation of IgG1⁺ GCs allowed the generation and maintenance of HA-binding IgG1⁺ memory B cells in MyD88^−/−TRIF^−/− mice at slightly increased levels (Fig. 3A, 3B). We purified MyD88^−/−TRIF^−/−IgG1⁺ memory B cells to assess their capacity to respond against booster vaccination in adoptive hosts (Fig. 3D). Consistent with the lack of TI response by split-virion vaccines, HA-binding IgG1⁺ memory B cells without MyD88/TRIF expression could not respond to the booster vaccination in the absence of CD4⁺ T cells, although they exhibited normal levels of T cell–dependent (TD) response (Fig. 3E). These data indicate that the TI activation of memory B cells depends on MyD88/TRIF signaling.

In addition to TLR7, MyD88/TRIF adaptor molecules mediate the signaling pathway through several receptors (i.e., other TLRs, transmembrane activator and cyclophilin ligand interactor, and IL-1βR) (46, 47). To explore the contribution of TLR7 on TI reactivation of memory B cells by whole-virion vaccines, we immunized several gene-targeted mice with whole-virion vaccines. The development of memory B cells and the TD memory response were unaffected in all genotypes of mice (Fig. 4A), which demonstrated that there was no intrinsic defects of memory B cells for terminal differentiation into plasma cells under the provision of T cell help (Fig. 4B). However, we found severely impaired TI response in TLR7^−/− and MyD88^−/− memory B cells, whereas TRIF^−/− memory B cells retained the response at almost normal levels, which indicated that the TLR7-mediated MyD88 pathway is responsible for the TI reactivation of memory B cells. Furthermore, the dependence of TI memory response on TLR7 and MyD88 expression demonstrates the exclusive use of the TLR7/MyD88 cascade for the response.

Type I IFN upregulates TLR7 expression on naive B cells, thereby enhancing B cell sensitivity to TLR7 agonists in an autocrine manner (48). To explore the possible contribution of this signaling circuit to TLR7-dependent memory B cell responses, we analyzed TI memory response by using IFNAR2^−/− memory B cells, in which IFN-mediated signaling is impaired. Whereas memory B cells lacking IFNAR2 normally elicited the recall response in the presence of T cells, the response was partially reduced in the absence of T cells (Fig. 4B). Collectively, these results support the model that type I IFN signaling sensitizes memory B cells to the TLR7/MyD88-dependent reactivation of memory B cells by whole-virion vaccines.

The TLR7/MyD88 pathway is active in several cell types besides B cells. To clarify the B cell–intrinsic role of this pathway, we generated V_H knock-in mice that express the targeted V_H gene of anti-HA mAb [clone OMb (22); Supplemental Fig. 2]. In this V_H knock-in mouse strain, B cell development was slightly disturbed, as reflected in reduced numbers of bone marrow B cell populations from pre–B cell stages (Supplemental Fig. 3). Splenic B cell populations, T1, T2, T3, and follicular B cells, were also reduced, but conversely, the marginal zone B cell population was increased (Supplemental Fig. 3). Despite such altered B cell development, we observed that >5% of peripheral B cells were bound to HA in these mice under the same conditions in which <0.1% of wildtype naive B cells were bound to HA (Supplemental Fig. 3).

By crossing TLR7^−/− and Ly5.1 mice, we established a mouse strain with TLR7-deficient HA-binding Ly5.1⁺ B cells. We immunized C57BL/6 mice receiving 1 × 10⁶ V_H knock-in B cells with whole-virion vaccines (Fig. 5A) and quantified IgG1⁺Ly5.1⁺ memory B cells derived from transferred V_H knock-in B cells at day 60 after vaccination (Fig. 5B). The numbers of memory B cells were comparable between both genotypes of B cells regardless of TLR7 deficiency, further supporting the conclusion that a B cell–intrinsic TLR7 response is not required for the generation and maintenance of memory B cells.

We then adoptively transferred purified IgG1⁺Ly5.1⁺ memory B cells into recipient mice to analyze TD and TI memory responses to booster vaccination (Fig. 5C, 5D). Both TD and TI responses were elicited from V_H knock-in memory B cells, although TI responses were somehow attenuated in V_H knock-in memory B cells. On the contrary, TLR7^−/− V_H knock-in memory B cells generated no HA-binding plasma cells at all in the absence of CD4⁺ T cells, whereas TD response was elicited normally. These results clearly indicate that B cell–intrinsic TLR7 signaling is not required for the TD reactivation of memory B cells but is crucial for TI reactivation to whole-virion vaccination.

We next sought to determine whether the requirement of TLR7/MyD88 signaling in TI responses was a specific feature of memory B cells. To do so, we set up the adoptive transfer experiment with V_H knock-in mice as described in Fig. 5 but injected purified HA-binding IgM⁺CD38⁺ B cells into SCID mice (Fig. 6A). The IgM⁺CD38⁺ B cell population included both naive and IgM⁺ memory B cells, and we quantified CD80 expression to distinguish CD80⁺ memory B cells from CD80⁻ naive B cells (49). HA-binding IgM⁺CD38⁺CD80⁻ B cells were purified from naive V_H knock-in mice as naive B cells, and HA-binding IgM⁺CD38⁺CD80⁺ B cells from the same Ly5.2⁺ donor with IgG1⁺ memory B cells were used as IgM⁺ memory B cells (Fig. 6A). It is important to note that the BCR affinities of IgM⁺ naive and IgM⁺ memory B cells used in this experiment were expected to be equivalent to those of IgG1⁺ memory B cells. After boosting these B cell populations in recipient mice, all three populations generated plasma cells independently of TLR7 in the presence of CD4⁺ T cells (Fig. 6B). Moreover, all of the cell populations exhibited TI Ab responses, although at levels lower than those of TD responses, especially in IgM⁺ memory B cells. Possible reasons for the attenuated TI response are explained in the 22Discussion. Notably, IgM⁺ naive B cells elicited TI responses without TLR7 expression in contrast with TLR7 dependency of IgM⁺ and IgG1⁺ memory B cells (Fig. 6B). Thus, an essential requirement of TLR7/MyD88 signaling in TI response is the unique properties of memory B cells, regardless of IgM and IgG1 isotypes.

Finally, to reveal the distinctive function of Ab responses derived from TI reactivation of memory B cells, we comparatively analyzed the promptness, affinity, and virus-neutralizing activity of Abs from TD and TI memory responses. Whole-virion vaccines elicited TI responses, but the responses were masked by TD responses if we used wild-type memory B cells as the responders (Fig. 2). Therefore, to visualize the TD memory response alone, we boosted memory B cells from B cell–intrinsic MyD88^−/− mice (MyD88^flox/flox mb-1-Cre^KI/WT) (16, 21), which lack the ability to mount TI memory response to a whole-virion vaccine. Conversely, for visualizing TI memory response alone, MyD88^+/+ memory B cells were boosted in the absence of CD4⁺ T cells. First, the promptness of plasma cell development was analyzed daily with flow cytometry and ELISPOT at days 4–7 after the boosting of recipient mice (Fig. 7A–C). Compared with the TD response, the TI response generated CD138⁺ plasma cells at a 5-fold higher frequency at day 4. However, these levels plateaued in the early time point and remained constant until day 7. On the contrary, TD responses continuously increased the numbers of plasma cells by day 7, and they reached levels similar to those of TI responses. The boosting of MyD88^+/+ memory B cells in the presence of CD4⁺ T cells confirmed the result as the combined TI + TD responses. The enumeration of HA-binding IgG⁺ plasma cells by ELISPOT analysis also gave comparable results (Fig. 7C).

Next, the avidities of anti-HA IgG Abs derived from both responses at day 7 after boosting were estimated with modified ELISA in which low-avidity Abs were selectively dissociated by treatment with 7 M urea (Fig. 7D). Compared with those from TD responses, HA-binding IgG Abs from TI responses were more resistant to detachment by urea treatment, which indicated that TI responses evoked anti-HA IgG Abs with higher avidity. Although we did not see dramatic impacts of B cell–selective TLR7 deficiency on IgG1⁺ memory B cell development (Fig. 5), it remains possible that the memory B cell compartment established without MyD88 has impaired affinity maturation leading to the reduced affinity of a recall response. To exclude this possibility, we compared the extent of affinity maturation in both genotypes of memory B cells by stimulating HA-binding memory B cells in vitro in the presence of a cytokine mixture and anti-CD40 mAbs (Fig. 7E). This stimulation led to the differentiation of memory B cells into plasma cells polyclonally in an Ag-independent manner, thereby allowing us to compare the levels of affinity maturation between both genotypes of memory B cells. Detachment by urea treatment resulted in comparable resistance in the Abs secreted from both genotypes of memory B cells, which indicated that the extent of affinity maturation was equivalent in MyD88⁺ and MyD88⁻ memory B cells. Therefore, the production of higher-affinity Abs in the TI memory response appeared to reflect the affinity-based selection of an established memory B cell repertoire through competitive reactivation by boosted Ags.

Given the differential affinities in the secreted Abs, we next sought to determine whether the affinity difference affects the virus-neutralizing activities of the Abs produced in both memory responses. In parallel to affinity difference, anti-HA Abs produced by the TI response exhibited neutralization activity even at 10-fold lower concentrations than those produced by the TD response (Fig. 7F). Therefore, B cell–intrinsic TLR signaling improved the promptness, affinity, and neutralization activity of the recall Ab responses to whole-virion vaccines by eliciting TI memory Ab responses upon booster vaccination. A proposed model for memory B cell responses against influenza vaccines is illustrated in Fig. 8.

Improving influenza vaccine strategy requires a better understanding of the cellular and molecular basis for neutralizing Ab response after the vaccination. Although memory B cells are the main players in supplying neutralization Abs (5, 50–53), little is known about how memory B cells are reactivated by influenza vaccines. This gap has occurred largely because previous analyses of memory B cell responses in mice have used purified protein Ags that incompletely mimic the immunogenic properties of influenza vaccines. In this study, we traced virus-binding memory B cell responses after influenza vaccination in mouse models and demonstrated that influenza vaccines recalled two waves of secondary Ab responses: the first wave promptly emerged in a manner dependent on B cell–intrinsic TLR7 signaling, and the second conventional wave followed later in a TLR7-independent but TD manner (Fig. 8).

Notably, compared with the TLR7-independent conventional response, the TLR7-mediated recall response provided high-affinity Abs with increased neutralizing activity and faster kinetics. Given the requirement of a costimulatory signal for Ag-driven B cell activation, it is conceivable that such accelerated recall response is driven by immediate delivery of TLR7-mediated costimulatory signals soon after BCR-mediated uptake of vaccines including TLR7 agonists. Indeed, many previous reports (54–56) have shown that TI Ab responses arise a few days sooner than TD responses, supporting the conclusion that the delivery of costimulatory signals from cognate Th cells requires extra days. However, the possibility remains that such prompt memory response driven by whole-virion vaccines is a specific feature of the H5N1 virus strain (NIBRG-14) used in this study. Indeed, compared with other strains, NIBRG-14 contains HA Ags at a lower density, which could substantially affect the magnitude of TI B cell responses because the density of B cell epitope is a key factor for the extent of BCR cross-linking and subsequent B cell response (57, 58). Thus, it is important to determine whether the phenomenon can be generalized to other vaccine strains.

B cell–intrinsic TLR signaling has great impacts on GC formation, IgG2 class switching, and enhancement of Ab responses during primary TD responses against other virus particles or viral infection (16, 18, 59). Therefore, there is no doubt that B cell–intrinsic TLR signaling plays key roles in the activation of naive B cells under some conditions. However, we found no dramatic effects of B cell–intrinsic TLR7 signaling on memory B cell development or maintenance in our study. We think that this is, at least partly, due to the use of alum adjuvants and repeated vaccinations, which would mask the effects of B cell–intrinsic TLR7 signaling. On the contrary, we did find novel roles for B cell–intrinsic TLR7 signaling in the prompt TI recall response from memory B cells; the effect was unique to the memory stage, not the naive stage, because no defect was found in the TI IgM response from naive B cells even after the simulation without alum adjuvant (Fig. 6). The differential impacts on the TLR7 requirement for the memory stage versus the naive stage may represent either BCR-intrinsic or -extrinsic effects, or a combination of both; however, we believe that BCR-extrinsic effects are more significant for several reasons. First, we used HA-binding IgM⁺ naive B cells from V_H knock-in mice as the reference to HA-binding IgG1⁺ memory B cells; therefore, BCR affinity was largely comparable between IgM⁺ naive and IgG1⁺ memory B cells in this experiment. Second, HA-binding IgM⁺ memory B cells exhibited TLR7 dependency closer to that of IgG1⁺ memory B cells than to that of IgM⁺ naive B cells. Whereas transcriptome analysis of murine memory versus naive B cells showed the relative conservation of gene expression profiles among the two B cell populations, several transcription factors were recently identified to be differentially expressed in memory B cells, accelerating their terminal differentiation into plasma cells upon TD Ag stimulation (60). Moreover, several TLR genes themselves, including TLR7, were upregulated in memory B cells (61, 62). Indeed, compared with naive B cells, both human and murine memory B cells showed greater responses to TLR ligand stimulation in vitro regardless of isotype (61, 63, 64). These observations, together with our findings, suggest that BCR-extrinsic factors render memory B cells more susceptible to TLR-mediated activation.

Increased affinity maturation in TI memory response during the early phase may be because of the antigenic competition of transferred memory B cells against boosted vaccines, and the dominant reactivation of higher-affinity memory B cells. Memory B cells can display a wide variety of affinities for B cell epitopes present in HA (65–67) because they are developed from both GC-dependent and -independent pathways (68, 69). Consistent with this development, memory B cells stimulated in vitro secreted HA-binding IgG1 Abs with a wide avidity spectrum after whole-virion vaccination (Fig. 7E). BCR affinity is a critical determinant of B cell fate decisions and the magnitude of B cell response in the primary response (70, 71). More recently, B cells have been shown to take up and present Ags on their surfaces at densities proportional to BCR affinity, which increases the availability of cognate T cell help in higher-affinity B cells and their preferential selection in the pre-GC and GC phases (72–74). In a similar scenario, high-affinity BCRs may take up greater amounts of Ag, which may be translated into a differential magnitude or quality of TLR7 signaling required for TI response. Notably, OMb Ab binds to HA Ags at an avidity index of 0.28, which is lower than that of TI response (0.44) and closer to that of TD response (0.23; Fig. 7D). As we repeatedly observed, compared with endogenous memory B cells, memory B cells expressing OMb-derived V_H genes showed weaker TI memory response (Figs. 5, 6), which suggest that the affinity maturation of V_H knock-in memory B cells may not reach the levels required for strong TI response as seen in endogenous memory B cells. Given the normal TD response from the same V_H knock-in memory B cells, these results suggest that TI activation requires higher BCR affinity than TD activation. As a result, compared with TD activation, TLR7-dependent activation of memory B cells can produce higher-affinity Abs with neutralization activity.

Our present study also provides new insights for vaccine strategies with improved recall Ab responses and reduced adverse effects. Whereas whole-virion vaccines are more immunogenic than other formulations (e.g., split and subunit vaccines) in humans and mice (7, 75–78), they have been replaced by split-virion types because of the high incidence of adverse effects such as febrile illness and local reactions at the injection site (79–82). It is generally accepted that inflammatory cytokines produced from innate immune cells (i.e., DCs) are the primary cause of adverse effects in subjects vaccinated with whole-virion vaccines (75, 83). Therefore, it is conceivable that a vaccine strategy selectively targeting B cell–intrinsic TLR signaling at the memory stage could reduce adverse effects while keeping the capacity for prompt recall response fairly intact. Notably, another possible advantage of this strategy is related to the unique features of TI response, which produces more sialylated Abs than does TD response (84). Sialylation of IgG Fc glycan suppresses excessive inflammatory responses; therefore, the enhanced production of sialylated IgG Abs may also contribute to reducing the adverse effects of vaccination. Thus, targeting B cell–intrinsic TLR signaling at the memory stage would not only accelerate the recall response with high-affinity neutralizing Abs, but also may reduce adverse effects through multiple mechanisms.

We thank Dr. M. Reth (University of Freiburg) for providing the mb-1 Cre^KI/WT mice and R. Aizawa, E. Izumiyama, and Dr. Y. Aiba for technical assistance.

The authors have no financial conflicts of interest.