The transcription factor T-bet resolves memory B cell subsets with distinct tissue distributions and antibody specificities in mice and humans

Summary

B cell subsets expressing the transcription factor T-bet are associated with humoral immune responses and autoimmunity. Here we examined the anatomic distribution, clonal relationships, and functional properties of T-bet⁺ and T-bet⁻ memory B cells (MBCs) in the context of the influenza-specific immune response. In mice, both T-bet⁻ and T-bet⁺ hemagglutinin-specific B cells arose in germinal centers, acquired memory B cell markers, and persisted indefinitely. Lineage tracing and IgH repertoire analyses revealed minimal interconversion between T-bet⁻ and T-bet⁺ MBCs, and parabionts showed differential tissue residency and recirculation properties. T-bet⁺ MBCs could be subdivided into recirculating T-bet^lo MBCs and spleen-resident T-bet^hi MBCs. Human MBCs displayed similar features. Conditional gene deletion studies revealed that T-bet expression in B cells was required for nearly all HA stalk-specific IgG2c antibodies and for durable neutralizing titers to influenza. Thus, T-bet expression distinguishes MBC subsets that have profoundly different homing, residency, and functional properties, and mediate distinct aspects of humoral immune memory.

eTOC

Johnson, Rosenthal, Knox, Myles et al. find that differential T-bet expression marks subsets of memory B cells (MBCs) with distinct homing and tissue residency patterns and functional properties. Distinguishing features of T-bet⁻, T-bet^hi and T-bet^lo MBCs are seen also in humans.

Graphical Abstract

Introduction

Immune memory relies on the persistence of specialized lymphocytes formed during primary immune responses, collectively termed memory cells. Antigen-specific T and B cell clones expand in representation by 10- to 100-fold during primary responses, and a fraction of these cells persist indefinitely, sustaining ongoing effector functions and participating in responses to subsequent pathogen challenges.

Accumulating evidence shows that memory cells are not monolithic populations, but instead consist of functionally distinct subsets that play different roles in protective immunity. Thus, several subsets of memory T cells are defined based on differences in phenotype, function, and migration patterns (Mueller et al., 2013, Sallusto et al., 1999). Memory B cell (MBC) subsets are currently defined in mice based on differential expression of the surface proteins CD73, CD80 and PD-L2 (Tomayko et al., 2010); MBCs expressing both CD80 and PD-L2 form plasma cells upon re-challenge, whereas the double-negative cells join germinal centers (Zuccarino-Catania et al., 2014).

Different memory fates can be determined by cytokine milieu, metabolic cues and transcriptional programs. For example, reciprocal patterns of T-bet and Eomesodermin expression underlie differentiation of T cells to effector versus memory subsets (Best et al., 2013, Hu and Chen, 2013). While the demarcation of T cell memory subsets by transcription factor expression is well established, analogous relationships have not been extensively explored in MBCs.

The discovery of a T-bet⁺ B cell subset in both mice and humans has piqued interest in the origin and role of these cells in primary and secondary humoral immune responses. T-bet⁺ B cells are observed in the context of murine aging and are thus termed “Age-associated B Cells,” or ABCs (Hao et al., 2011, Rubtsov et al., 2011). Subsequent analyses revealed roles for cognate T cell help and antigen presentation in their development. This, as well as a high frequency of somatically mutated immunoglobulin (Ig) genes in these cells, suggests that T-bet⁺ ABCs are MBCs formed during T-dependent B cell responses (Russell Knode et al., 2017). T-bet⁺ B cells appear and persist following influenza immunization or infection in mice (Naradikian et al., 2016, Russell Knode et al., 2017), providing a means to track T-bet⁺ and T-bet⁻ MBCs in a defined antigen system. Moreover, most humans have been exposed to influenza through immunization and infection and thus have standing influenza hemagglutinin (HA)-specific MBCs, enabling direct comparative analyses between human and murine MBC subsets.

Here we examined whether T-bet⁺ versus T-bet⁻ MBCs differ in their origins, kinetics of generation, trafficking patterns, and functional roles. We found that multiple MBC subsets can be distinguished by T-bet expression. T-bet expression divided influenza-specific MBCs into T-bet⁻, T-bet^lo, and T-bet^hi populations with differing anatomic localization, residency patterns, and antigenic specificity. T-bet⁻, T-bet^lo, and T-bet^hi cells localized to draining lymph nodes, spleen, and infected tissues upon infection; however, T-bet^hi MBCs were selectively maintained in the spleen where they remained resident, being excluded from the lymphatics. B cell receptor (BCR) sequencing analyses revealed infrequent sharing of clones between HA-specific T-bet⁺ and T-bet⁻ MBCs. Divergence within clonal lineages, in conjunction with genetic fate-mapping, demonstrated that T-bet expression in T-bet⁺ MBCs is stable. In mice, T-bet expression in the B lineage was required for the development of HA-specific IgG2c and nearly all HA stalk-specific antibody. Of note, the phenotypic and functional attributes of these subsets are largely shared between mice and humans. Together, these results establish T-bet expression as a distinguishing feature of MBC subsets that have profoundly different homing and functional properties and mediate distinct aspects of humoral immune memory.

Results

T-bet expression distinguishes influenza-specific memory B cell populations

To determine the functional differences between T-bet⁺ and T-bet⁻ MBC subsets, we set out to define T-bet⁻ and T-bet⁺ B cell generation and persistence using influenza infection in Tg(Tbx21-ZsGreen)E3ZJfz mice that are transgenic for a bacterial artificial chromosome (BAC) that reports Tbx21 expression using the fluorescent protein ZsGreen (referred to as Tbx21-ZsGreen) (Zhu et al., 2012). We infected Tbx21-ZsGreen reporter mice with 30 TCID50 of influenza A/Puerto Rico/8/1934 (PR8) and observed weight loss and recovery over a period of 4 weeks post infection, with the nadir at 9 days post infection (dpi; Figure S1A). We harvested mediastinal, mesenteric, and pooled peripheral (superficial cervical, axillary, brachial, and inguinal) lymph nodes (LN), spleen, lungs, and blood from infected mice at multiple time points, and identified influenza hemagglutinin (HA)-specific B cells using biotinylated PR8 HA probes modified to prevent sialic acid binding (Whittle et al., 2014). The HA probes were separately conjugated to two streptavidin-fluorophore conjugates to exclude fluorophore-specific B cells during flow cytometric analysis (Figure S1B).

We identified low numbers of HA-specific B cells in lymphoid organs of naïve mice, in agreement with previous estimates of the primary HA-responsive repertoire (Cancro et al., 1978); these were uniformly T-bet⁻ (Figure S1B). To exclude this primary pool, we focused subsequent analyses on IgD⁻ B cells (Figure S1C). IgD⁻ HA-specific B cells were detected in spleen, mediastinal LN, and lungs of all mice at both acute infection and memory time points (Figure 1A). Examination of Tbx21-ZsGreen and CD11c expression in HA-specific B cells indicated that T-bet⁺ B cells can be phenotypically subdivided into T-bet^hi and T-bet^lo subsets, with different tissues being variably comprised of T-bet⁻, T-bet^lo, and T-bet^hi subsets across infection (Figure 1A). The T-bet^hi subset contained both CD11c⁺ and CD11c⁻ cells with a phenotype and level of T-bet expression matching Age-associated B Cells (Figure S1C). Furthermore, we confirmed that T-bet^lo B cells expressed increased T-bet mRNA transcripts versus T-bet⁻ B cells (Figure S1D).

Figure 1. T-bet expression identifies memory B cell populations with unique tissue distribution.

Tbx21-ZsGreen reporters were intranasally infected with 30 TCID₅₀ influenza A/Puerto Rico/8/1934 (PR8). (A) Fluorescently-conjugated PR8 hemagglutinin (HA) detects HA-specific (HA⁺) B cells, and Tbx21-ZsGreen expression in HA⁺ B cells resolves T-bet⁻, T-bet^lo, and T-bet^hi subsets across tissues at acute (day 15) and memory (day 100) timepoints. (B) Number of HA⁺ B cells in spleen, pooled mediastinal lymph nodes (medLN), lungs, and blood at different time points after infection (left column), and proportions of HA⁺ B cells that are T-bet⁻, T-bet^lo, and T-bet^hi in each tissue (right column). The number of HA⁺ B cells in blood was estimated by calculating their frequency per 100,000 B cells, and proportions of T-bet-defined subsets in blood were calculated after concatenation due to low cell number. (C) Gating scheme identifies splenic HA⁺ GCB cell (GL7⁺CD38⁻), MBC (GL7⁻CD38⁺), and pre-GC cell (CD38⁺GL7⁺) subsets; concatenated flow plots (bottom) depict CD38 and GL7 expression of T-bet⁺ (pooled T-bet^lo and T-bet^hi; green) and T-bet⁻ (purple) HA⁺ B cells at each time point (bottom). Line plots (top) depict number of HA⁺ GCB cells and MBCs separated by T-bet expression phenotype over time. (D) Expression of memory markers (CD80, PD-L2, CD73) in T-bet⁺ (green) and T-bet⁻ (purple) splenic HA⁺ MBCs (GL7⁻ CD38⁺) and naive follicular B cells (IgD⁺; grey). Data in (B) and (C) are compiled from 2 independent experiments with at least 3 mice per experiment. Data in (A) and (D) are representative of 2 independent experiments with at least 3 mice per experiment. Data in (B) and (C) are plotted as mean ± SEM. HA⁺ B cells were identified as live, singlet, DUMP⁻, B220⁺, CD19⁺, IgD⁻ cells, HA-BV421⁺, HA-AF647⁺ cells. DUMP gate includes CD4, CD8, Gr-1, and F4/80.

In agreement with prior studies (Frank et al., 2015, Boyden et al., 2012), HA-specific B cells were readily identified in spleen and mediastinal LN by 7 days post infection (dpi), peaked in number and frequency at 15 dpi in spleen and 22 dpi in mediastinal LNs, and then declined to steady state numbers in both organs by 40 dpi (Figure 1B). HA-specific B cells were occasionally detected in the lungs of some mice as early as day 7, but cell numbers peaked in lungs of all mice by 15 dpi and displayed a gradual decline continuing at least through 100 dpi (Figure 1B). Small numbers of HA⁺ B cells were also detected in mesenteric and peripheral LNs, but these were dwarfed by spleen, mediastinal LN, and lung responses (Figure S1E).

Our longitudinal analysis of the HA-specific B cell pool identified differential induction and maintenance properties for the T-bet⁻, T-bet^lo, and T-bet^hi B cell subsets across tissues. The lung HA-specific response was entirely comprised of T-bet^hi cells at 7 dpi; however, HA-specific cells were not detected in lungs of all mice at this time (Figure 1B). The lung HA-specific population remained T-bet-dominated throughout acute infection but was primarily T-bet⁻ by 100 dpi (Figures 1A and 1B). The majority of mediastinal LN HA⁺ B cells also expressed T-bet at 7 dpi, but T-bet^hi cells rapidly declined by 15 dpi, suggesting rapid tissue exit or differentiation. T-bet^lo B cells similarly declined by 15 dpi, albeit more slowly, and were nearly undetectable by 100 dpi in this tissue. In accord with possible tissue egress, blood T-bet^hi cells peaked in frequency by 22 dpi and rapidly declined to undetectable levels by 40 dpi. In contrast, T-bet^lo and T-bet^hi subsets were consistently maintained in the spleen from 7 to 100 dpi, comprising 27% to 52% (with an average of 37%) of the splenic HA-specific B cell response (Figures 1A and 1B). These findings identified early but transient T-bet⁺ B cell responses in lungs and mediastinal LN and suggest T-bet⁺ HA-specific B cell memory is primarily sustained in the spleen.

We also assessed germinal center B (GCB) cells and MBC marker expression in splenic HA-specific cells at each time-point. T-bet^lo and T-bet^hi cells were pooled for these analyses since they displayed similar memory marker expression throughout infection (data not shown). We delineated GCB cells as CD38⁻GL7⁺ and presumptive MBCs as CD38⁺GL7⁻ (Weisel et al., 2016) (Figure 1C). Notably, GL7 expression closely correlated with other established GCB markers including CD95 and Peanut Agglutinin (PNA) (Figure S1F; data not shown). At 7 dpi, GL7 was present primarily on T-bet⁺ cells without concomitant decreased expression of CD38 (Figure 1C), suggesting a pre-GC phenotype (Taylor et al., 2012, Shinall et al., 2000, Sheikh et al., 2019). Nearly all HA-specific cells exhibited a GCB phenotype by 15 dpi and maintained this through 22 dpi, at which time an MBC population began to emerge (Figure 1C). At 40 dpi, the majority of T-bet⁺ cells had a memory phenotype, whereas nearly half of T-bet⁻ cells still maintained GC markers (Figure 1C). Regardless of T-bet expression, nearly all splenic HA-specific B cells acquired an MBC phenotype by 100 dpi. In contrast, GCB cells persisted in the mediastinal LN, and to a lesser extent in the lung, out to 100 dpi; however, these were T-bet⁻. (Figure S1G). Further analyses of splenic MBC-phenotype cells at each time point identified increased expression of the MBC markers CD80, PD-L2, and CD73 (Zuccarino-Catania et al., 2014) which began by 15–22 dpi and increased further by 100 dpi, suggesting formation of stable T-bet⁻ and T-bet⁺ memory pools by the latter time point (Figure 1D). We also observed early expression of CD80 in CD38⁺GL7⁻ cells as early as 7 dpi (Figure 1D); these may represent other non-GC cells such as extrafollicular plasmablasts, since they were not omitted by our gating strategy. Together, these findings identified similar GC and MBC differentiation between T-bet⁻ and T-bet⁺ subsets during the influenza response, except the T-bet⁺ subset lost GC characteristics and transitioned to a memory phenotype earlier than the T-bet⁻ population. Moreover, T-bet^hi HA-specific MBCs appeared to be spleen-compartmentalized upon resolution of infection.

Human T-bet^hi B cells are an anatomically compartmentalized component of influenza-specific memory

Having identified discrete influenza-specific MBC subsets with differential tissue localization properties in mice, we questioned whether analogous human T-bet-expressing MBCs show a similar anatomical distribution. In humans, T-bet-expressing B cells can be identified alongside T-bet⁻ MBCs in peripheral blood during active viral infections and vaccinations, malaria infection, and autoimmune disease (Obeng-Adjei et al., 2017, Wang et al., 2018, Chang et al., 2017, Lau et al., 2017, Knox et al., 2017). Since T-bet^hi B cells (CD21⁻T-bet^high) display a unique trafficking receptor profile (CD11c⁺CXCR3^+/−CXCR5^lowCD62L^low) compared to classical MBCs (Lau et al., 2017, Knox et al., 2017), we hypothesized that, as in mice, human T-bet^hi B cells might have a distinct tissue distribution pattern. To test this, we obtained peripheral blood, tonsil, iliac and mesenteric lymph nodes, spleen, and bone marrow as donated or discarded surgical tissue and examined B cell phenotypes in these tissues (gating in Figure S2A). In agreement with our previous study (Knox et al., 2017), we observed T-bet^hi B cells in the peripheral blood of all subjects (Figures 2A and 2B). We also readily identified T-bet^hi B cells within spleen and bone marrow compartments (Figures 2A and 2B), and confirmed their antigen-experienced phenotype in spleen (Figure S2B). Conversely, T-bet^hi B cells were largely absent from tonsils and both iliac and mesenteric lymph nodes (Figures 2A and 2B).

Figure 2. Human T-bet^hi B cells do not recirculate via the lymphatics and maintain influenza-specific memory in the spleen.

(A) Identification of human CD21⁻T-bet^hi B cells within total CD19⁺ B cells from peripheral blood (PB), tonsil, iliac lymph node (iLN), mesenteric lymph node (mesLN), spleen, and bone marrow (BM) of representative donors. Different tissue types in (A) or (B) are not matched. (B) Frequency of T-bet^hi B cells in various tissues (n=6 per tissue group). Statistics represent comparisons between PB, spleen, or BM with tonsil, iLN, and mLN; frequencies within PB, spleen, and BM are not statistically different from one another. (C) Identification of T-bet^hi B cells in matched peripheral blood (PB) and thoracic duct fluid (TD) samples from a representative donor. (D) Frequency of T-bet^hi B cells in matched PB and TD samples (n=8). (E) Identification of CD21⁺CXCR3⁺T-bet^lo (blue) and CD21⁺CXCR3⁻T-bet⁻ (black) subsets of memory (IgD⁻/IgD⁺CD27⁺) B cells in matched PB and TD from a representative donor, and mesLN from another donor; T-bet expression by these populations is shown as a histogram. Blood T-bet^hi B cells are included for comparison in grey. (F) Frequency of the CD21⁺CXCR3⁺ population within PB and TD CD19⁺ B cell pools from an 8-donor cohort. (G) Identification of HA-specific, IgD⁻IgM⁻ B cells within CD19⁺CD38^low splenic B cells using two fluorescently-labelled A/California/07/2009 HA probes (H1 strain) or a single fluorescently-labelled A/Wisconsin/67/2005 HA probe (H3 strain). (H) CD21 and T-bet expression in IgD⁻IgM⁻HA⁺ B cells in spleens and mesLNs from representative donors using H1 or H3 probes. (I) Frequency of T-bet^hi phenotype within IgD⁻IgM⁻H1⁺ or H3⁺ B cells in spleens from two 10-donor cohorts and mLN from a 6-donor cohort. (J) T-bet MFI of splenic naïve (IgD⁺CD27⁻) B cells and switched (IgD⁻ IgM⁻) H1-HA-specific CD21⁺ and CD21⁻T-bet^hi B cells from a representative donor. (K) Frequency of isotype expression within human splenic IgD⁻IgM⁻HA⁺ B cells (n=6). Statistical comparisons performed using one-way ANOVA with Tukey post-test (B), paired t-test (D and F), unpaired t-test (I), and repeated measures ANOVA with Tukey post-test (K). Lines depict mean ± SEM. N.S. = not significant, *p<0.05; **p<0.01; ***p<0.001.

These findings suggested restricted trafficking properties of the T-bet^hi B cell subset. To determine whether human peripheral blood T-bet^hi B cells recirculate through tissues, we obtained matched peripheral blood and thoracic duct fluid (chyle) samples from individuals undergoing thoracic duct cannulation. The thoracic duct is the body’s largest lymphatic vessel that collects draining lymph from both lymphoid and non-lymphoid tissues for return to the blood; thus, its contents represent cells undergoing lymphatic recirculation. Despite consistent identification of T-bet^hi B cells in the peripheral blood of these individuals, these cells were essentially absent in matched thoracic duct fluid (Figures 2C and 2D). In contrast, naïve B cells and various CD21⁺ MBC subsets (IgM⁺CD27⁺, IgM⁻CD27⁺, and IgM⁻CD27⁻) were detected at similar frequencies in both peripheral blood and thoracic duct fluid of all subjects, suggesting this anatomical compartmentalization is a feature specific to T-bet^hi B cells, analogous to what we had observed in mice (Figures S2C–E).

We next asked whether T-bet expression per se is associated with tissue restriction. We previously identified two distinct populations of T-bet-expressing B cells in blood of healthy individuals: T-bet^hi B cells (CD21⁻T-bet^high) and T-bet^lo B cells (CD21⁺T-bet^low; (Knox et al., 2017)), which likely correspond to the observed T-bet^hi and T-bet^lo MBC pools in mice (Figure 1). Here, we extended these findings to show that CXCR3 expression enriched for T-bet^lo cells within the greater CD21⁺ memory population (Figure 2E). Using the CD21⁺CXCR3⁺ phenotype, we found that T-bet^lo B cells were present in human blood, thoracic duct fluid, and lymph nodes (Figure 2E), and at similar frequencies between matched blood and thoracic duct fluid samples (Figure 2F). These observations suggest human T-bet^lo B cells recirculate through all lymphoid tissues, while T-bet^hi B cells are restricted to the spleen, blood, and bone marrow in healthy individuals. Further analyses will be necessary to determine the relationship between T-bet^lo and T-bet^hi B cells; however, we propose that these CD21⁺T-bet^lo cells represent the human equivalent of the T-bet^lo B cells observed in Tbx21-ZsGreen reporter mice (Figure 1). Taken together, these findings identify human T-bet^hi B cells as a unique, tissue-restricted subset that does not recirculate via the lymphatic system.

Since murine T-bet^hi HA-specific B cells preferentially populated the spleen at memory time points (Figure 1), we next asked whether the human spleen harbors an HA-specific T-bet^hi MBC population. Using fluorophore-conjugated HA probes from two distantly-related influenza strains, A/California/07/2009 (H1) and A/Wisconsin/67/2005 (H3) (Whittle et al., 2014, Joyce et al., 2016), we assessed T-bet expression by HA-specific class-switched (IgD⁻IgM⁻) B cells in the spleen (Figure 2G) and mesenteric lymph nodes (Figure S2F). Despite our efforts, we were unable to obtain human mediastinal lymph node samples without significant blood contamination for analysis of lung-draining lymphoid tissue. HA-specific T-bet^hi B cells recognizing H1 or H3 strains were identified in the spleens of all donors but were rarely detected in mesenteric lymph nodes, whereas T-bet^loCD21⁺ and T-bet⁻CD21⁺ HA-specific memory B cells were present in all assessed tissues (Figures 2H–J; data not shown). The relative representation of T-bet^hi B cells within the splenic HA-specific population varied considerably (~3–80% of H1⁺ B cells and ~3–53% of H3⁺ B cells; Figure 2I) and positively correlated with age (Figures S2G and S2H). Lastly, we assessed the isotype distribution of the human splenic HA-specific MBC compartment and found that human IgG1, the analog of murine IgG2a/c, dominated the class-switched memory response to influenza (Figure 2K; isotype gating in Figure S2I). IgG3⁺ and IgA⁺ HA-specific cells could be detected at low levels in some donors; however, IgG2⁺ HA-specific B cells were rarely identified (Figure 2K, Figure S2I). Together, these findings suggest T-bet expressing B cells are a critical component of human influenza HA-specific B cell memory and, as in mice, identified the human T-bet^hi HA-specific MBC pool as spleen-localized and absent from lymphatics.

T-bet^hi HA-specific memory B cells are resident in the spleen

The apparent splenic localization of T-bet^hi HA-specific MBCs in both mice and humans led us to rigorously assess tissue residency using a parabiosis-based approach. Thus, Tbx21-ZsGreen reporter mice infected ≥ 40 days prior with PR8 were surgically joined to naïve congenic B6.SJL partners. We reasoned that all HA-specific MBC will originate in the Tbx21-ZsGreen partner, so their presence in the B6.SJL partner would indicate that they are a recirculating population. Conjoined mice were monitored by serial bleeds to assess the portion of circulating CD45.2⁺ (Tbx21-ZsGreen reporter origin) versus CD45.1⁺ (B6.SJL origin) B cells in each partner. Both partners demonstrated mixing of CD45.1⁺ and CD45.2⁺ B cells in the blood as early as day 7, achieving stable proportions between the partners by day 14 (Figure 3A). Accordingly, parabiosed pairs were euthanized ≥ 17 days post-surgery. We observed similar frequencies of CD45.2⁺IgD⁺ B cells – a pool anticipated to circulate freely – in spleen, lungs, and mediastinal lymph nodes of each partner (Figure 3B), suggesting equilibration of recirculating B cells by day 17.

Figure 3. T-bet expression resolves spleen resident versus recirculating MBC pools.

(A). Tbx21-ZsGreen reporters (CD45.2⁺; ≥ 40 dpi) and naïve B6.SJL (CD45.1⁺) were surgically conjoined and showed evidence of blood sharing by day 7, with equilibrium reached by day 14. Parabionts were euthanized at ≥ 17 days post-surgery for analysis. (B) Frequencies of naïve follicular (IgD⁺) B cells expressing CD45.2 in lymphoid and non-lymphoid tissues from each parabiosis pair. (C) Identification of HA⁺IgD⁻ B cells expressing either CD45.1 or CD45.2 in parabiosis partners. (D) Identification of Tbx21-ZsGreen reporter-derived (CD45.2⁺) T-bet⁻, T-bet^lo, and T-bet^hi HA⁺ MBCs in spleens of Tbx21-ZsGreen and B6.SJL partners; data concatenated from 7 pairs. (E) Numbers of T-bet⁻, T-bet^lo, and T-bet^hi HA⁺ splenic MBCs in Tbx21-ZsGreen (red) and B6.SJL (black) partners. (F) Percentage of splenic HA⁺ MBCs that are T-bet⁻, T-bet^lo, or T-bet^hi in each partner. (G and H) Number of HA⁺ MBCs in medLN (G) and lungs (H) of parabiosis partners. (I and J) Tbx21-ZsGreen expression in HA⁺ MBCs from medLN (I) and lungs (J) of Tbx21-ZsGreen partner. HA⁺ MBCs were not detected in the medLN or lung of the B6.SJL partner. Data displayed are from 8 pairs across three independent experiments for spleen and 4 pairs across two-independent experiments for medLN and lungs. HA⁺ B cells were identified as live, singlet, DUMP⁻, B220⁺, CD19⁺, CD45.2⁺, IgD⁻, HA-BV421⁺, HA-AF647⁺ cells. Data in (E), (F), (G), and (H) show individual points with the mean ± SEM indicated. Statistical comparisons performed using paired two-tailed t-test. ns = not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.001

HA-specific MBCs were observed in the spleens of both partners, and virtually all of these were CD45.2⁺ (Figure 3C), consistent with their origin in the previously infected Tbx21-ZsGreen partner. T-bet⁻ and T-bet^lo HA-specific B cells were identified in the spleens of both partners, suggesting these subsets recirculate (Figures 3D–F, Figure S3A). In contrast, T-bet^hi HA-specific MBCs were absent from the naïve B6.SJL partner spleens but remained in spleens of previously infected Tbx21-ZsGreen mice (Figures 3E, 3F, S3A), even when data were concatenated from 7 parabiotic pairs (Figure 3D). To confirm that ZsGreen-expressing cells were not being rejected in the B6.SJL mice, we measured frequencies of donor Tbx21-ZsGreen⁺CXCR3⁺CD8⁺ lymphocytes, which highly express the ZsGreen protein, and found similar frequencies of these cells in both partners (Figure S3B). Moreover, T-bet^lo HA-specific B cells were also present in spleens of both mice (Figures 3D and 3E). Thus, broad rejection of ZsGreen⁺ cells did not occur, consistent with previous studies (Bell et al., 2007). Taken together, these findings identify splenic T-bet^hi HA⁺ MBCs as a tissue-resident memory pool.

We also investigated whether HA-specific B cells showed evidence of residency in mediastinal LNs and lung, the other primary locations of influenza memory cells, as others recently demonstrated (Allie et al., 2019). We identified significant HA-specific B cell numbers in mediastinal LNs of the previously infected partner that were absent in the naïve partner (Figure 3G). We suspect this reflects the extended maintenance of HA-specific GCs in mediastinal LNs, as the local HA-specific B cell population retained a GC phenotype at least through 100 dpi (Figure S1G). This phenomenon appeared to be mediastinal LN-specific, as HA-specific GC B cells were not identified at this late time point in any other lymphoid tissues examined. We also identified HA-specific MBCs in the lungs of the previously infected partner (Figure 3H), but these cells were absent in the naïve partner. Notably, nearly all mediastinal LN- and lung-localized cells were T-bet⁻ (Figures 3I–J). Thus, HA-specific memory was anatomically compartmentalized, encompassing tissue-resident T-bet^hi B cells in the spleen and circulating T-bet⁻ and T-bet^lo B cell populations.

Established T-bet⁺ and T-bet⁻ memory B cells undergo minimal interconversion

The different residency and recirculation properties of T-bet⁻ and T-bet⁺ MBCs raised the question of how T-bet⁻ and T-bet⁺ MBCs arise. We considered four possibilities (Figure S4A): 1) T-bet⁻ and T-bet⁺ MBCs arise independently and are stable, separate subsets; 2) T-bet⁺ cells give rise to T-bet⁻ cells (or vice versa); 3) T-bet⁺ and T-bet⁻ cells undergo shared selection followed by stable commitment to either a T-bet⁺ or T-bet⁻ long-lived MBC population; 4) T-bet⁺ and T-bet⁻ MBCs interconvert by modifying T-bet expression as needed to change localization or functional properties. Immune repertoire profiling of antibody heavy chain variable region gene (VH) rearrangements can be used to distinguish between these four models, as each model makes distinct predictions regarding differences in VH usage (model 1), somatic hypermutation (model 2) and clonal overlap (models 3 and 4; Figure S4A).

We therefore sequenced VH rearrangements of HA-specific MBCs separated by T-bet expression (Figure S4B) in Tbx21-ZsGreen reporter mice immunized intradermally with 30 μg of lipid nano particles (LNP) loaded with HA mRNA (Pardi et al., 2018). The LNP platform generated both T-bet⁻ and T-bet⁺ MBCs (Figure S4B), and we confirmed an HA-antibody response by hemagglutination inhibition (Figure S4C). At 90 days post immunization, all splenic IgD⁻ HA-specific B cells were sorted based upon Tbx21-ZsGreen expression into T-bet⁻ and T-bet⁺ memory subsets, and antibody VH rearrangements of both subsets and IgD⁺ naïve follicular control B cells were sequenced. Similarly, we used the CD21⁻CD85j^hi surface phenotype, which specifically identifies human T-bet^hi B cells (Knox et al., 2017), to sort human splenic HA-specific IgD⁻IgM⁻ MBCs into CD21⁻CD85j^hi and CD21⁺ subsets (Figure S4D) and sequenced the VH rearrangements of these populations along with control bulk splenocytes.

VH analysis revealed comparable VH usage (Figures S4E and S4F) and CDR3 length distributions (Figures 4A and 4B) in T-bet⁺ and T-bet⁻ populations, and there was some clonal overlap between the two populations (Figures 4C and 4D), ruling out a strict separate lineage model (model 1). T-bet⁺ and T-bet⁻ MBCs harbored similar levels of somatic hypermutation, suggesting that one population was not a precursor to the other (model 2), and both populations showed significantly more mutations than naïve B cells or unsorted splenocytes (Figures 4E and 4F).

Figure 4. T-bet⁺ and T-bet⁻ MBCs are selected from a shared pre-immune lineage but do not interconvert.

HA-specific splenic MBCs from Tbx21-ZsGreen reporters (day 100 post immunization) were sorted into T-bet⁻ and T-bet⁺ subsets, with naïve follicular (IgD⁺) B cell controls, for immunoglobulin heavy chain genomic sequencing. Human HA-specific splenic MBCs were similarly sorted into CD21⁺ and CD21⁻CD85j^hi subsets; CD21⁻CD85j^hi phenotype identifies human T-bet^hi B cells (Knox et al., 2017) subsets. (A) CDR3 lengths (in nucleotides) of in-frame sequences from murine T-bet⁻ and T-bet⁺ HA⁺ MBCs and naïve follicular (IgD⁺) B cell controls after all replicates were pooled. (B) CDR3 lengths of in-frame sequences from CD21⁺ and CD21⁻CD85j^hi HA⁺ MBC subsets were quantified (in nucleotides). Bulk splenocytes (largely naive follicular B cells) served as a control. (C) The number of clones that overlap between T-bet⁻ (blue) and T-bet⁺ (red) HA⁺ MBCs in mouse (M. mus, MM). (D) The number of clones that overlap between CD21⁺ (blue) and CD21⁻CD85j^hi (red) HA⁺ MBCs in humans (H. sap; HS). (E) Percentages of clones binned by the level of somatic mutation (expressed as the percent difference in nucleotide sequence to the nearest germline VH gene) in mouse T-bet⁻ and T-bet⁺ HA⁺ MBCs and naïve follicular B cells. (F) Percent of the heavy chain V-gene that is mutated from germline in CD21⁺ and CD21⁻CD85j^hi HA⁺ MBCs and bulk splenocytes in humans. (G) Representative lineage trees of shared clones between T-bet⁻ and T-bet⁺ HA⁺ murine MBCs, with inferred nodes (black), T-bet⁻ nodes (blue), and T-bet⁺ nodes (red). Trees were generated in ImmuneDB and visualized with ETE3 (see Methods). Lineages had to contain at least 10 copies of T-bet⁺ and T-bet⁻ and have at least 4 trunk mutations (shared SHMs) to be included. Numbers indicate the number of mutations compared to the preceding vertical node. The inferred node at the top of the tree indicates the nearest germline sequence. (H) Tbx21-ZsGreen^creERT2-Rosa26^LSL/tdTomato mice (Yu et al., 2015) were treated with tamoxifen to mark T-bet expressing cells with permanent, Rosa21-driven, tdTomato expression and the status of T-bet expression of marked B cells in the blood was tracked over 40 days. For panels (A), (C), (E), and (G), two independent experiments were carried out with at least 4 mice per group. Each gave similar results, and the results for the more recent experiment are shown. For panels (B), (D), and (F), the splenocytes from 4 adult subjects were sorted and sequenced. For genetic fate mapping (H), two independent experiments were carried out with at least 4 mice per group; one experiment is shown here.

Next, we scrutinized the lineage trees of clones that contained T-bet⁺ and T-bet⁻ members. In mice, overlapping clones between T-bet⁺ and T-bet⁻ populations were not as frequent as they were within replicate sequencing libraries from the same subset (Figure S4G), suggesting that many T-bet⁺ and T-bet⁻ clones arise independently, rather than being fully intermingled. The same clonal analysis in humans did not reach statistical significance, likely due to our restricted sample size – we sampled a small portion of the spleen and therefore missed many clonal members (Figure S4H). In further support of this separation, analyses of mouse clonal lineages containing both T-bet⁺ and T-bet⁻ cells revealed that nearly all exhibited segregation of T-bet⁺ and T-bet⁻ sequences onto separate branches (Figure 4G). Taken together, these findings in mice following immunization and in established HA⁺ MBCs in human spleen favor model 3, in which T-bet⁺ and T-bet⁻ MBC precursors undergo shared selection, subsequently commit to a T-bet⁺ or T-bet⁻ MBC population, and thereafter remain stable with respect to T-bet expression status, with minimal, if any, interconversion between established T-bet⁺ and T-bet⁻ MBCs.

To further verify the stability of T-bet⁺ B cells, we used the Tg(Tbx21-ZsGreen,-cre/ERT2)H3Jfz reporter/inducible cre transgenic mouse line (Yu et al., 2015) bred to the Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze} conditional reporter allele background (Madisen et al., 2010) for a combined T-bet reporter/fate mapper mouse (referred to as Tbx21-ZsGreen^creERT2-Rosa26^LSL/tdTomato). These T-bet-sufficient mice contain a BAC with the ZsGreen reporter, T2A peptide, and creER^T2 under control of the T-bet promoter, such that treatment with tamoxifen during active T-bet transcription causes irreversible tdTomato expression. Using these mice, one can delineate cells that expressed T-bet during the tamoxifen treatment period and have subsequently lost expression (tdTomato⁺ZsGreen⁻) from those that retained it (tdTomato⁺ZsGreen⁺). We treated ≥ 20-week-old Tbx21-ZsGreen^creERT2-Rosa26^LSL/tdTomato mice with tamoxifen on days 0, 2, and 4 and performed serial bleeds to assess stability of tdTomato⁺ZsGreen⁺ B cells. All mice demonstrated tdTomato labeling at day 10, with T-bet⁺ cells outnumbering T-bet⁻ cells 10:1 within the tdTomato⁺ B cell population (Figure 4H). The ratio of T-bet⁺ to T-bet⁻ cells was maintained steadily in all mice (Figure 4H), suggesting most B cells expressing T-bet at day 0 maintained expression for 40 days, interconverting rarely if at all during this period. In combination with our clonal overlap and lineage tree analyses, these data show that established T-bet⁺ MBCs represent a separate, stable population.

HA stalk-specific antibody is derived primarily from the T-bet-expressing B cell compartment

The distinct localization and phenotypic stability of T-bet⁺ HA-specific MBCs led us to assess the contribution of the T-bet-expressing B cell compartment to the influenza humoral response. In mice, T-bet promotes antibody class-switching to IgG2a/c (Gerth et al., 2003, Liu et al., 2003, Peng et al., 2002, Wang et al., 2012, Coutelier et al., 1987, Hocart et al., 1989, Markine-Goriaynoff and Coutelier, 2002, Rubtsova et al., 2013), the dominant isotype in influenza and other anti-viral responses (Coutelier et al., 1987, Fazekas et al., 1994). In accordance with this, we observed a greater increase in PR8-specific and HA-specific IgG2c (Figures 5A and 5B) compared to IgG1, evident by day 12 and 15, respectively. This isotype bias confirmed previous studies (Boyden et al., 2012) and suggests a key role for T-bet in regulating influenza antibody production (Piovesan et al., 2017). Therefore, we tested the contribution of T-bet in the B lineage to HA-specific humoral responses by infecting B cell-specific T-bet deficient mice obtained by crossing Cd19^tm1(cre)Cgn mice with Tbx21^F/F mice (referred to as Cd19^cre/+Tbx21^flox/flox), heterozygous Cd19^tm1(cre)Cgn cre controls (referred to as Cd19^cre/+), and wild type mice with PR8 and examined antibody levels and characteristics. All three groups showed similar weight loss kinetics (Figure 5C), total HA-specific B cell numbers (Figure 5D), and HA-specific GCB cell numbers (Figure 5E), although Cd19^cre/+ controls recovered weight more quickly (Figure 5C). As such, CD19 heterozygosity did not appear to significantly impair the influenza response, and initiation of the humoral response did not require T-bet expression in B cells.

Figure 5. T-bet⁺ B cells are required for optimal influenza antibody responses and HA stalk-specific antibody in mice.

(A and B) Total betapropiolactone (BPL)-inactvated PR8-specific IgG1 and IgG2c (A) and PR8 hemagglutinin (HA)-specific IgG1 and IgG2c (B) in sera from infected Tbx21-ZsGreen mice over time. (C). Weight loss and recovery from influenza infection in wild type C57Bl/6, Cd19^cre/+Tbx21^flox/flox, and Cd19^cre/+ mice compared to PBS-treated controls. (D) Number of HA-specific splenic B cells at day 15 and 40 dpi. (E) Number of HA-specific splenic GCB cells at 15 dpi. (F) Hemagglutination inhibition (HAI) titers at 15 and 40 dpi. (G-I) Antibody titers to BPL-inactivated PR8 (G), full-length PR8-HA (H), or chimeric construct comprised of H1 stalk and H6 head (I). Wild type C57Bl/6 were used for naïve controls in (F-I). Data are represented as mean ± SEM from 3 independent experiments with at least 3–5 mice in each group. Statistical comparisons performed using two-sided t-test (G-I) and Wilcoxon rank-sum test (F). *p<0.05, **p<0.01, ***p<0.001. Cells in (D, E) were identified as DUMP⁻, CD19⁺, B220⁺, CD138⁻, IgD⁻, HA-PE⁺, with the additional definition of GC cells in (E) as PNA⁺CD95⁺.

To assess the functionality of antibodies generated in the absence of B lineage T-bet expression, we performed hemagglutination inhibition (HAI) assays from serum. At 15 dpi, the majority of mice displayed HAI titers greater than 40, a level associated in human studies with protection (de Jong et al., 2003, Hobson et al., 1972), although one wild type and several Cd19^cre/+Tbx21^flox/flox mice had titers ranging from 20 to undetectable (Figure 5F). HAI titers declined in all groups by 40 dpi (Figure 5F), likely due to the loss of acute infection-generated IgM titers (Fazekas et al., 1994). However, the Cd19^cre/+Tbx21^flox/flox group displayed significantly reduced HAI titers versus the wild type and Cd19^cre/+ groups at 40 dpi, with 70% of mice showing titers below 40 (Figure 5F). These findings suggest T-bet expression in B cells may be necessary for the development of sustained protective influenza-specific titers.

We hypothesized that decreased HAI titers in Cd19^cre/+Tbx21^flox/flox mice may reflect a loss of specific components of the antibody response. We next assessed antibody titers and found a significant reduction in total PR8-specific IgG2c in Cd19^cre/+Tbx21^flox/flox mice, as expected (Figure 5G). Low IgG2c titers remained in Cd19^cre/+Tbx21^flox/flox mice at 15 dpi but were nearly absent by 40 dpi, suggesting T-bet-independent mechanisms can initiate a degree of IgG2c switching during acute infection (Figure 5G). We focused subsequent analyses on the HA protein, the antigenic target relevant for protective humoral immunity to influenza, and identified significantly reduced IgG2c titers to full-length HA in Cd19^cre/+Tbx21^flox/flox mice at both 15 and 40 dpi compared to wild type and Cd19^cre/+ control groups (Figure 5H). PR8- and HA-specific titers of IgG1, a T-bet-independent isotype, were unaffected in Cd19^cre/+Tbx21^flox/flox mice and did not increase to compensate for IgG2c loss (Figures 5G and 5H). These findings confirmed that the majority of HA-specific IgG2c antibody was derived from T-bet-expressing B cells.

Lastly, we questioned whether T-bet expressing B cells are important for influenza antibody responses to certain specificities. Recent studies highlight a critical role for HA-specific IgG2a/c antibodies for in vivo influenza protection, which primarily skew toward stalk recognition (DiLillo et al., 2014, DiLillo et al., 2016). Thus, we assessed stalk reactivity of IgG1 and IgG2c using a chimeric construct comprised of the PR8-related H1 stalk and unrelated H6 head (2012, Pica et al., 2012, Krammer et al., 2012). This chimera is bound primarily by stalk-specific antibodies since most PR8- generated HA head-binding antibodies are strain-specific and do not recognize H6 head. We found that the stalk response was dominated by IgG2c in wild type mice at both 15 and 40 dpi, while IgG1 stalk titers were negligible (Figure 5I). Moreover, Cd19^cre/+Tbx21^flox/flox mice largely lost IgG2c stalk-reactive titers (Figure 5I), indicating that the bulk of the influenza stalk-specific antibody response arose from T-bet-expressing B cells.

DISCUSSION

Our study reveals multiple MBC subsets delineated by T-bet expression, whose distinct phenotypic and functional attributes are shared by mice and humans. T-bet expression status divided MBCs by anatomic localization and residency, as well as effector function and epitope specificity. Thus, T-bet⁻ and T-bet^lo MBCs originated in all secondary lymphoid tissues and freely recirculated, whereas T-bet^hi MBCs resided in the spleen and were excluded from the lymphatics. Further, clonal and in vivo lineage tracing analyses showed that while HA-specific T-bet⁺ and T-bet⁻ MBCs likely arose from common pre-immune pools, they diverged after antigen encounter and thereafter remained as separate and stable pools. Finally, we showed that the development of mouse IgG2c HA- and HA stalk-specific antibodies, as well as durable neutralizing titers, required T-bet expression in the B lineage. Taken together, these findings show that T-bet expression is a conserved feature of an MBC subset with differential circulatory properties, tissue-residency, and epitope specificity.

Pathogen-driven responses generate both isotype-switched and unswitched T-bet expressing B cells (Kenderes et al., 2018, Barnett et al., 2016, Rubtsova et al., 2013), but detailed analyses of the generation, fate and anatomic characteristics of T-bet⁺ B cells have not been conducted. Our results formally demonstrated antigen-mediated and antigen-specific generation of T-bet⁺ GC B cells during viral infection, followed by the establishment of somatically mutated, antigen-specific T-bet⁺ and T-bet⁻ MBC pools whose numbers were maintained indefinitely. Consistent with memory character, both T-bet⁺ and T-bet⁻ HA-specific B cells expressed the MBC markers CD73, CD80, and PD-L2 with kinetics similar to those in hapten-carrier responses (Weisel et al., 2016). Despite these surface phenotypic similarities, our clonal analyses and genetic fate-mapping experiments suggest it is unlikely that a progenitor-successor relationship exists, or that frequent interconversion occurs, between T-bet⁺ and T-bet⁻ MBCs. Thus, while T-bet⁺ and T-bet⁻ MBCs both resulted from antigen-driven naïve B cell activation, they most often arose independently and remained distinct, rather than representing different stages in a common differentiation pathway. In addition, the role for these HA-specific MBC subsets in recall responses remains an open question. Re-challenge studies in multiple mouse models have found that both CD80⁺PD-L2⁺ MBCs and HA-specific MBCs preferentially differentiate into early antibody-secreting cells (ASCs) as opposed to re-entering germinal centers following antigen encounter (Zuccarino-Catania et al., 2014). These studies suggest that T-bet⁺ MBCs are primed for ASC differentiation, but what influence T-bet has on this fate decision compared to T-bet independent factors such as receptor affinity remains to be determined.

Our tissue distribution analyses indicate that memory B cells are anatomically compartmentalized: T-bet⁻ and T-bet^lo MBCs were found in all secondary lymphoid tissues, whereas T-bet^hi MBCs were primarily in the spleen, blood, and bone marrow. Parabiosis experiments further confirmed that established influenza-specific T-bet^hi MBCs neither exited the spleen to populate the lymphatic system, nor homed to the spleen from blood or other anatomical locations. However, T-bet^hi B cells were identified transiently in mediastinal LN and lungs early after infection, suggesting T-bet^hi B cell generation can occur outside the spleen. We have previously reported the critical role of innate sensors, such as nucleic acid sensing Toll-like receptors (TLRs), and common gamma chain cytokines in regulating T-bet⁺ B cell fate (Naradikian et al., 2016). Thus, the generative signals for T-bet expressing B cells are not spleen-specific per se, and the differential anatomic distribution of established T-bet^hi MBCs is not an immediate consequence of early antigen encounter specifically within the spleen.

Chemokine receptors and integrins regulate the anatomic distribution of immune cells and may contribute to T-bet^hi B cells’ characteristic localization properties. Studies examining human peripheral blood samples found that T-bet⁺ B cells express the integrin CD11c, the chemokine receptor CXCR3, and low levels of CXCR4, CXCR5, and CCR7, chemokine receptors associated with homing to lymphoid organs (Wang et al., 2018, Isnardi et al., 2010, Lau et al., 2017). Thus, the specific combination of these and other surface receptors may impede lymphatic entry and help recruit T-bet^hi B cells to the spleen. Via mechanisms that are unclear, T-bet^hi B cells also appear to enter the blood following activation or recent tissue egress. Consistent with this idea, we observed early loss of HA-specific T-bet^hi B cells in the mediastinal LN and lungs in infected mice, coupled with a temporary wave of HA-specific T-bet^hi B cells in blood, and we previously described an increase in peripheral blood HA-specific T-bet^hi B cells following influenza vaccination in humans (Andrews et al., 2019). While they were absent from the lymphatics in our experiments, recent evidence suggests consistent viremia and/or chronic immune activation may be able to override T-bet^hi MBC compartmentalization: lymphoid tissue infections with pathogens such as HIV and Toxoplasma gondii are associated with a local enrichment of T-bet-expressing B cells in lymph nodes (Johrens et al., 2006, Austin et al., 2019).

The splenic residency associated with T-bet^hi MBCs leads to some intriguing possibilities regarding the role of T-bet⁺ MBCs in immune surveillance. While tissue residency may be critical to protect from local reinfection, this role seems unlikely for spleen-resident HA-specific T-bet^hi MBCs, inasmuch as influenza is a respiratory infection and the virus is not known to replicate in the spleen (Turner et al., 2013). Instead, T-bet^hi splenic resident MBCs may be uniquely positioned to support broad immune surveillance and rapidly produce antibody for systemic dissemination upon reinfection. In support of this notion, T-bet⁺ B cells express elevated quantities of BLIMP-1 (Lau et al., 2017) and, when isolated from the blood of SLE patients, quickly differentiate into plasma cells upon TLR7 stimulation without obligate division (Jenks et al., 2018). It is tempting to speculate that circulating T-bet^lo B cells are short-lived cells derived from a stationary, self-renewing T-bet^hi population; indeed, the possibility of self-renewal and multipotency of T-bet⁺ MBCs has been reported by others (Kenderes et al., 2018), and our findings confirm that T-bet⁺ MBCs are a persistent population. Alternatively, T-bet^lo B cells might be a stable and persistent population with separate maintenance requirements from the T-bet^hi subset. These possibilities are not mutually exclusive and resolving their relative merits and contributions will require examination of the turnover rates and clonal composition of these MBC subsets.

Given the striking parallels between human and mouse T-bet⁺ MBCs, we propose that T-bet expression is a conserved divisor for memory B cell subsets, and that the relative contributions of T-bet⁺ vs T-bet⁻ memory B cells to various aspects of humoral immunity merits detailed investigation. Importantly, in addition to their differences in anatomic localization, T-bet⁺ and T-bet⁻ B cells differed in the quality and specificity of antibodies they generate. Our studies with B cell-specific T-bet deficient mice showed that T-bet drives influenza-specific IgG2c production to HA and the HA stalk. It is tempting to speculate that these subsets could differentially contribute to immunodominance, cross-reactivity, or original antigenic sin, and may thus play distinct roles in immune responses to heterologous challenges. The observed loss of the IgG2c component of anti-influenza responses in Cd19^cre/+Tbx21^flox/flox mice suggests a link between T-bet⁺ MBCs and influenza-specific antibody production; however, the direct contribution of these established memory cells to antibody titer maintenance is unclear, as are the implications of tissue-restriction. T-bet⁺ B cells arise post influenza vaccination in humans (Lau et al., 2017, Koutsakos et al., 2018, Andrews et al., 2019); therefore, based on recent interest in developing HA-stalk-reactive vaccines for broad protection against influenza, we posit that focusing vaccine design efforts on driving T-bet expression in HA-specific B cells and maintaining this population long-term might lead to the development of more effective prophylactic agents and vaccination regimens for influenza.

STAR methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Michael P. Cancro (cancro@pennmedicine.upenn.edu)

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

Raw sequencing data is available on SRA under BioProject number PRJNA610227 for human data and PRJNA610229 for mouse data.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

C57BL/6 and B6.SJL (10–12 weeks old, females, purchased from The Jackson Laboratory); Tbx21^F/F possessing loxP sites flanking exons 2–6 of Tbx21 (Intlekofer et al., 2008) and Cd19^tm1(cre)Cgn containing a targeted cre cassette into exon 2 of Cd19 (Rickert et al., 1997) (obtained from the laboratory of E. John Wherry, University of Pennsylvania); Tbx21-ZsGreen with a ZsGreen reporter gene inserted at the Tbx21 start site in BAC RP23–237M14 (Zhu et al., 2012); Tbx21-ZsGreen^creERT2 possessing the BAC RP23–237M14 with the ZsGreen reporter gene, T2A self-cleaving peptide, and inducible creERT2 recombinase gene (Yu et al., 2015); and Rosa26^LSL/tdTomato with a Rosa26 targeted construct containing CAG-LSL-tdTomato-WPRE (Madisen et al., 2010) were maintained and used in accordance with the University of Pennsylvania Institutional Animal Care and Use Committee guidelines.

Infections

Mice were infected by intranasal infection with 30 tissue culture infectious dose50 (TCID50) of influenza strain A/Puerto Rico/8/1934 (PR8; American Type Culture Collection).

Human samples

All study participants provided written informed consent. Tissue samples were collected with IRB approval at the University of Pennsylvania (809316; 815056; 822686) and Case Western Reserve University (10–09-12). Human peripheral blood mononuclear cell samples were obtained from the University of Pennsylvania Human Immunology core. Human bone marrow mononuclear cell samples were obtained from the University of Pennsylvania Stem Cell and Xenograft core. Paired blood and thoracic duct fluid samples were obtained from individuals with idiopathic or trauma-based chylopericardium or chylothorax requiring intervention at the Hospital of the University of Pennsylvania. Lymphoid tissue samples (mesenteric lymph node, iliac lymph node, tonsil, and spleen) were obtained at the Hospital of the University of Pennsylvania and Case Western Reserve University: mesenteric and iliac lymph nodes were obtained during abdominal surgery and kidney transplant surgery, respectively. Non-enlarged tonsils were obtained from sleep apnea patients. Spleens were removed and obtained due to trauma or surgical intervention. Additional spleen samples were obtained from the Human Pancreas Analysis Program (HPAP) at the University of Pennsylvania. Mononuclear cells were mechanically separated from solid tissues and enriched using a ficoll gradient.

METHOD DETAILS

Parabiotic surgery

Age-matched Tbx21-ZsGreen reporters and B6.SJL adult female mice were conjoined as described previously (Kamran et al., 2013). Briefly, a skin incision was made from the olecranon to the knee of each of the mice to be joined. The elbows and knees of the two paired mice were then tied together with surgical suture, followed by connecting of the skin with surgical sutures and staples. For pain control, mice were given buprenorphine (0.1 mg/kg every 6 hours for 36 hours) and meloxicam (5 mg/kg every 12 hours for 72 hours) and provided with sulfamexathole (400mg/L) and trimethoprim (800mg/L) antibiotics in their drinking water to prevent infection. Mice were monitored for signs of pain, infection, or damage to sutures. Blood was periodically drawn from the tail to check for anastomoses, which appeared complete by d14, therefore, mice were euthanized at day 17. The spleen was harvested from both partners for all pairs, and the lungs and mediastinal lymph nodes were also collected from some pairs.

Flow cytometry

Flow cytometry reagents were purchased from BioLegend (BL), BD Biosciences (BD), eBioscience (eBio), Southern Biotech (SB), or Invitrogen (Inv). The following antibodies were used for mouse studies: T-bet (4B10; BL), CD11c (N418; BL), IgM (R6–60.2; BD), CD38 (90; eBio), CD73 (TY/11.8; BL), CD80 (16–10A1; BD or BL), PD-L2 (TY25; BL), CD138 (281–2; BL), IgD (11–26c.2a; BL), B220 (RA3–6B2; BL or eBio), CD19 (1D3; BD or eBio), CD19 (6D5; BL) peanut agglutinin–FITC (Sigma), CD45.1 (A20; BL), CD45.2 (104; BL), CD183/CXCR3 (CXCR3–173; BL) and CD3 (17A2; BL). DUMP gate comprised CD8 (53–6.7; eBio), CD4 (H129.19; BL), F4/80 (BM8; eBio), Ly-6G/GR1 (RB6–8C5; eBio). The following antibodies were used for human studies: CD38 (HIT2; BL), CD85j (GHI/75; BD; HP-F1; eBio), T-bet (4B10; eBio and BL), IgM (MHM-88; BL), IgD (IA6–2; BD), CD10 (CB-CALLA; eBio), CD27 (O323; BL), CXCR3 (G025H7; BL), IgG (G18–145, BD), CD21 (Bu32, BL; B-ly4, BD), CD19 (HIB19, BL), CD3 (UCHT1, BL), CD14 (M⊘P9, BD), CD16 (3G8, BD), CD11c (3.9, eBio), Bcl-6 (K112–91, BD), Ki67 (56, BD), IgG1 (HP6069, Inv), IgG2 (HP6002, SB), IgG3 (HP6050, SB), and IgA (polyclonal, Inv). For detection of murine influenza-binding B cells, recombinant HA PR8 (Whittle et al., 2014) was obtained from the laboratory of Dr. Barney Graham, National Institute of Allergy and Infectious Disease, biotinylated, and conjugated to streptavidin-fluorophores as previously described (Whittle et al., 2014), or was directly conjugated using the R-phycoerythrin conjugation kit from Abcam (catalog ab102918) as per manufacturer’s instruction. Human HA-specific B cell staining was performed using A/California/07/2009 and A/Wisconsin/67/2005 HA probes prepared as previously described (Whittle et al., 2014, Joyce et al., 2016). Mouse samples were prepared for flow cytometry as follows: Mouse Fc fragment (Jackson ImmunoResearch; 015–000-008) was added to all staining cocktails at a final concentration of 1:200. Mouse spleens were homogenized, on ice, in staining buffer (PBS ⁺ 0.5%BSA ⁺ 2mM EDTA) and passed through nylon mesh (50μM) to obtain single cell suspension. Red blood cells were lysed using ACK lysing buffer (Lonza, cat 10–548E) as per manufacturer’s instructions. Cells were washed with PBS and stained as described previously(Hao et al., 2011, Naradikian et al., 2016). Live/dead discrimination was done using Zombie Aqua fixable viability kit (BL). Prior to T-bet staining, cells were fixed and permeabilized using eBioscience™ Foxp3 / Transcription Factor Staining Buffer Set, at 4°C for 45min-1hr. Human samples were prepared for flow cytometry as previously described (Knox et al., 2017). Data were acquired on BD LSR II flow cytometer and FACS analyses were performed using FlowJo v9 and v10 (Becton Dickinson Co., Ashland, OR).

Tamoxifen Treatment

Tamoxifen (Sigma; cat. T5648–1G) was dissolved in 100% ethanol to a concentration of 80 mg/ml and vortexed and heated at 42°C until completely dissolved. An equal volume of Kolliphor (Sigma, cat. C5135–500G) was added to bring the solution to a concentration of 40 mg/ml. Aliquots were then stored at −20°C. Immediately prior to administration, tamoxifen aliquots were thawed at 42°C and diluted with PBS to 8 mg/ml. Mice were administered 3 doses of 0.8 mg of tamoxifen via intraperitoneal injection every other day.

Serum antibody titers

Serum was harvested by spinning whole blood at 13000g for 10 minutes and stored at −20°C until use. Antibody titers were assessed using ELISA as previously described(Hao et al., 2011, Naradikian et al., 2016) with the following modifications: 96-well medium-binding plates were coated with either 20HAU/well of BPL-inactivated PR8, 2 μg/mL of PR8 HA, or 2 μg/mL of H6/H1 chimeric constructs (expressed in baculovirus system as previously described (Margine et al., 2013)). HA-specific monoclonal antibodies (from Dr. Jonathan Yewdell, National Institute of Allergy and Infectious Diseases) were used as standards to determine concentration of IgG1 and IgG2a/c. Standards were used at a starting concentration of 100 ng/mL for IgG2a and 10 ng/mL for IgG1 and diluted 2-fold across.

HAU (hemagglutination unit) and HAI assays

Viral HAU titers were determined before every HAI assay. All dilutions were prepared in PBS. 50 μL diluted virus, 50 μL heat-inactivated sera and 12.5 μL of 2% turkey erythrocytes were used per well for all assays, which were performed in round-bottom plates.

Starting with a 1:100 dilution of live virus, 2-fold dilutions were mixed with 2% turkey erythrocytes (Lampire biologicals) and incubated for 1 hour at room temperature. Agglutination dose (AD) was determined at the end of the incubation period, and confirmed by repeating the process with a 2-fold dilution series of virus, ranging from 4AD to 0.25 AD. This dose was subsequently used for the HAI assay.

Sera were heat-treated at 55°C for 30 minutes, diluted 2-fold in PBS (staring dilution 1:20), mixed with 4AD and 2% turkey erythrocytes, and incubated as for HAU assay. HAI titers are expressed as inverse of the highest dilution that inhibited agglutination.

mRNA production

The sequence of the Puerto Rico/8/1934 influenza virus hemagglutinin (pTEV-PR8 HA-A101) was codon-optimized, synthetized and cloned to the mRNA production plasmid. The mRNA was produced using T7 RNA polymerase (Megascript, Ambion) on linearized plasmids. The mRNA was transcribed to contain 101 nucleotide-long poly(A) tails. One-methylpseudouridine (m1Ψ)-5’-triphosphate (TriLink) instead of UTP was used to generate modified nucleoside-containing mRNA. Capping of the in vitro transcribed mRNAs was performed co-transcriptionally using the trinucleotide cap1 analog, CleanCap (TriLink). mRNA was purified by cellulose purification, as described (Baiersdorfer et al., 2019). All mRNAs were analyzed by denaturing or native agarose gel electrophoresis and were stored frozen at −20°C.

LNP formulation of the mRNA

Cellulose-purified m1Ψ-containing RNAs were encapsulated in LNPs using a self-assembly process as previously described wherein an ethanolic lipid mixture of ionizable cationic lipid, phosphatidylcholine, cholesterol and polyethylene glycol-lipid was rapidly mixed with an aqueous solution containing mRNA at acidic pH (Maier et al., 2013). The RNA-loaded particles were characterized and subsequently stored at −80°C at a concentration of 1 μg μl^-1. The mean hydrodynamic diameter of these mRNA-LNP was ~80 nm with a polydispersity index of 0.02–0.06 and an encapsulation efficiency of ~95%.

Mouse B cell receptor sequencing

Genomic DNA was extracted from sorted cells using the Qiagen Gentra DNA purification kit (Qiagen, Germantown, MD, Cat. No.158689). Primers used were adapted from Wang et al. (Wang et al., 2000) at the beginning of the FW1 region of VH and were modified to include adaptor sequences for the Illumina NexteraXT kit (sequences are provided in Supplementary Table 1). Samples were amplified in duplicate (2 biological replicates per sample). The sequencing adapters used were:

VHmix (MH1): 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGSARGTNMAGCTGSAGSAGTC-3′

JH1, JH4mix: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGCTANTGAGGAGACGGTGAC-3′

JH2: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGTGAGGAGACTGTGAGAGTGG-3′

The mouse IgH library was generated with one VH primer and a cocktail of JH1,2,4 primers. The VH and JH primer mixes were used at 0.6 μM in a reaction volume of 25 μL using a Multiplex PCR kit (Qiagen, Valencia, CA, Cat. No. 158388). Amplification conditions for the PCR were: primary denaturation at 95^oC for 10 minutes, cycling at 95°C 45s, 60°C 45s, and 72°C for 90s for 35 cycles, and a final extension step at 72°C for 10 minutes.

Amplicons were purified using the Agencourt AMPure XP beads system (Beckman Coulter, Inc., Indianapolis, IN), and second-round PCRs were performed as described (Meng et al., 2017) to add Illumina NexteraXT adaptors to the IgH library. Final sequencing libraries were quantified by Qubit Fluorometric Quantitation (Thermo Fisher Scientific, Grand Island, NY) and loaded onto an Illumina MiSeq instrument in the Human Immunology Core facility at the University of Pennsylvania and sequenced using 2×300 bp paired end kits (Illumina MiSeq Reagent Kit v3, 600 cycle, Illumina Inc., San Diego, Cat. No. MS-102–3003).

Human B cell receptor sequencing

Genomic DNA was extracted from sorted cells using the Qiagen Gentra DNA purification kit. Sequences were generated from genomic DNA using primers that were situated at FR1 and JH (BIOMED2) for IgH V region sequencing. Samples were amplified in duplicate (2 biological replicates per sample). Second-round amplification to generate sequencing libraries used Illumina Nextera XT kit as previously described (Meng et al., 2017, Rosenfeld et al., 2018a). Sequencing were performed on an Illumina MiSeq instrument in the Human Immunology Core facility at the University of Pennsylvania using a 2×300 bp paired end kit.

Sequencing data analysis

Raw sequence data (FASTQ files) were processed through pRESTO version 0.5.10 (Vander Heiden et al., 2014). First, paired reads (R1 & R2) were aligned. Then sequences with an average Phred quality score of less than 30 (an error rate of 1 in 1000 bases) were removed. Of the remaining sequences, the 5’ and 3’ ends were trimmed until a window of 20 nucleotides had an average quality score of at least 30. Short reads of less than 100 bases were discarded after the trimming. Finally, nucleotides with a quality score of less than 30 were masked with an “N,” and any sequence with more than 10 such N’s were discarded. Code 1 shows a script performing these filtering steps.

Code 1

PairSeq.py −1 *R1* *R2*
AssemblePairs.py align −1 *R1_pair-pass* −2 *R2_pair-pass* --coord illumina --rc tail
FilterSeq.py quality -s *assemble-pass* -q 30
FilterSeq.py trimqual -s *quality-pass* -q 30 --win 20
FilterSeq.py length -s *trimqual-pass* -n 100
FilterSeq.py maskqual -s *length-pass* -q 30
FilterSeq.py missing -s *maskqual-pass* -n 10

ImmuneDB (Rosenfeld et al., 2018b) was used for gene identification and clonal inference of heavy chain sequencing data in both humans (using v0.26.0) and mice (using v0.28.0). Sequences were trimmed to IMGT position 20 in mice and 80 in humans to remove 5’ primer sequences. Clones were assembled by grouping sequences with the same V-gene, J-gene, and 85% CDR3 amino-acid similarity as described in (Rosenfeld et al., 2018a).

In the murine dataset, all mice contained a common CDR3 amino-acid string, CARGNRYWYFDVW (or a truncated variant of CARGNRYWYFDV or CARGNRYWYFD), possibly due to contamination, and were excluded from further analysis. Further excluded were two clones that had over 20% mutation in the V-region, due to incorrect V-gene assignment.

For all further analysis of both human and murine data, clones in each subject/subset combination were only included if they contained more than half the mean frequency of copies in that subject/subset.

QUANTIFICATION AND STATISTICAL ANALYSIS

All p values were determined using one of the following as mentioned in figure legends: unpaired non-parametric t-test or one-way ANOVA with Tukey post hoc test, paired t-test or repeated measures ANOVA with Tukey post hoc test, or Spearman correlation, using GraphPad Prism version 7 or version 8 (GraphPad Software, La Jolla, CA 92037 USA). *p < 0.05, **p < 0.01, ***p < 0.001. Data are represented as mean ± SEM. The number of mice and human subjects used in each experiment, as well as the exact number of times an experiment was repeated, is mentioned in the figure legends.

Highlights.

T-bet⁺ B cells are a separate and durable memory subset in mice and humans

T-bet^hi memory B cells are absent from the lymphatic circulation

Influenza-specific T-bet^hi memory B cells are spleen-resident in mice.

B cell-intrinsic T-bet is required for >90% of flu- and HA stalk-specific antibodies.