The Transcription Factor T-bet Resolves Memory B Cell Subsets with Distinct Tissue Distributions and Antibody Specificities in Mice and Humans

Abstract

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 (HA)-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.

Keywords: Age-associated B cells; B cell memory; BCR sequencing; Humoral immunity; T-bet(+) B cells; antibody; hemagglutinin stalk; immune repertoire profiling; influenza; tissue-resident.

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.

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.

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

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.

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⁺.