Temporal and transcriptional programming of effector memory B cells shapes vaccine platform-dependent humoral immunity

Summary

This study investigates how influenza vaccine platform shapes the memory B cell responses that underlie durable antibody protection. Using fluorescent hemagglutinin (HA) probes, the authors compared HA-specific IgD-negative B cells in adults who received either live attenuated influenza vaccine (LAIV) or inactivated influenza vaccine (IIV). Both platforms induced circulating T-bet⁺ HA⁺ B cells with similar phenotypic features, but their relationship to antibody outcomes diverged sharply: while these cells track with durable IgG responses after IIV, only early (day 7–14) T-bet⁺ HA⁺ B cell frequencies predicted durable IgA responses after LAIV. Single-cell transcriptional profiling further revealed that LAIV and IIV elicit distinct effector-memory B cell states, including an LAIV-enriched, terminally differentiated TBX21/ZEB2-high population with limited plasmablast recall potential. Together, these findings show that influenza vaccination drives platform-dependent diversification of HA-specific B cell states, highlighting the need for vaccine-platform-specific correlates of immune protection.

Graphical abstract

Highlights

• •LAIV induces circulating T-bet⁺ HA⁺ B cells phenotypically similar to IIV
• •T-bet⁺ HA⁺ B cells don’t predict HA-IgG after LAIV, unlike after IIV
• •Early T-bet⁺ HA⁺ B cells predict durable HA-IgA responses after LAIV
• •LAIV and IIV elicit transcriptionally distinct HA⁺ B cell states

Immunology; Immunology; Cell biology

Introduction

Influenza is a significant global threat and causes 290,000 to 650,000 annual fatalities world-wide.¹ As an RNA virus, influenza continuously evolves via antigenic shift or drift to result in seasonal or pandemic spread of disease; vaccines are the primary means of immune protection. Influenza vaccines are formulated as intramuscular and inactivated (IIV) or as live attenuated (LAIV) for intranasal mucosal delivery. Vaccines act by inducing antigen-specific humoral immune responses to immunogenic influenza proteins, including hemagglutinin (HA). In the context of influenza vaccination of adults, who are almost certainly flu experienced, this antigen-specific humoral immune response is driven by expansion of pre-existing memory B cell populations. Some of these memory B cells are recalled into germinal centers (GCs) to interact with memory T follicular helper cells (Tfh) and to produce daughter memory and antibody-secreting cells (ASCs), which can mediate long-term protection. Memory B cells are heterogeneous and can be broadly grouped into (1) stem-like populations with the capacity to re-enter GC reactions and generate new memory B cells and ASCs^2,3; (2) effector memory subsets that rapidly differentiate into ASCs upon recall^4,5,6; and (3) activated effector B cell states capable of rapid cytokine production in inflammatory contexts.^7,8,9,10

Mucosally delivered influenza viral antigen has been shown in murine models and among immunized humans to elicit mucosal, tissue-resident memory B cells as critical mediators of first-line defense against challenge infections.^11,12,13 However, the efficacy of LAIV versus IIV in humans is inconsistent. Historically, in pediatric cohorts, LAIV has been shown to elicit superior immunogenicity than IIV.¹⁴ In contrast, in adults, LAIV is not uniformly protective.¹⁵ In fact, during the 2015–2016 influenza season, LAIV was removed from the US market because of a lack of efficacy, even among pediatric patients, against the H1N1 strain.¹⁶ Importantly, no easily assayed circulating correlate for immune protection after LAIV has been defined and this limits our ability to understand the quality of immune protection after LAIV.¹⁷

Until recently, the only antigen-specific outputs that could be assayed after IIV were plasmablasts (PBs) and circulating Tfh (cTfh) as measures of IIV-responsiveness.^18,19,20 Now fluorochrome-labeled influenza HA antigen probes or tetramers (some with modifications of the Y98F residue to attenuate non-specific sialic acid binding^21,22) have been utilized as reagents that permit the direct study of antigen-specific B cells in humans that circulate at various time points after vaccination. Using these tetramer reagents, the circulating IIV-elicited HA-specific (HA⁺) B cell compartment in adults has been defined as mutated, consistent with the properties of memory B cells.^23,24,25 These HA⁺ memory B cells have also been described as functionally heterogeneous. For example, we and others have classified IIV-elicited circulating HA⁺ memory B cells according to expression of the lineage-defining effector transcription factor (TF), T-bet, as well as atypical memory markers CD21, FcRL5, and CD11c.^23,24,25,26 T-bet⁺ HA⁺ memory B cells circulate as early as 7 days after IIV, are identified in the blood as late as 60–120 days after vaccination, and predict durable HA IgG immune responses to IIV.²⁵ T-bet⁺ HA⁺ memory B cells are transcriptionally programmed as effector memory and comprise clones that are preferentially recalled as PBs after antigen re-challenge for immediate immune protection from disease.²⁵ Thus, circulating T-bet⁺ HA⁺ memory B cells are a promising, easily assayed biomarker of vaccine response, but it is unclear whether this applies across platforms such as LAIV.^23,24,25

Here we apply fluorochrome-labeled B cell tetramers to evaluate HA⁺ memory B cells in the blood in a cohort of inoculated healthy adults who seroconvert after LAIV. We specifically compare the functional properties of HA⁺ memory B cells that circulate after LAIV to those that circulate after IIV, in order to assess which, if any, circulating memory B cells function as candidate correlates of durable immune protection. We find that T-bet⁺ HA⁺ memory B cells that circulate after LAIV are phenotypically similar to those in circulation after IIV. However, unlike what we report from subjects after IIV, T-bet⁺ HA⁺ memory B cells do not correlate with durable HA IgG Ab after LAIV; rather, early (day 7–14) T-bet⁺ HA⁺ memory B cells predict durable HA IgA after LAIV. Bulk RNA sequencing (RNA-seq) and single-cell sequencing analyses demonstrate that LAIV-elicited T-bet⁺ HA⁺ memory B cells are transcriptionally unique from IIV-elicited counterparts. We additionally identify a subset of LAIV-elicited HA⁺ memory B cells that preferentially accumulate after LAIV, are detectable early in blood, and are characterized by high expression of TBX21 and ZEB2, consistent with a highly differentiated effector B cell state. These cells show limited detectable recall as circulating PBs following antigen re-challenge, suggesting restricted recall potential. Together, these findings reveal that the phenotype and function of circulating HA⁺ memory B cell subsets, including their ability to predict durable antibody responses, are shaped by vaccine platform, underscoring the need for platform-specific correlates of immune protection.

Results

HA⁺ IgD^neg B cells circulate in adults who seroconvert after LAIV

During the 2015–2016 influenza vaccine season, we enrolled 36 healthy adults who received either LAIV (n = 17) or IIV (n = 19), with serial blood sampling following vaccination. Participants reported no influenza vaccination in the prior season; prior influenza infection history was not captured. The LAIV cohort included a higher proportion of women than the IIV cohort, but groups were otherwise comparable in age and ethnicity (Figure S1A).

To benchmark early antigen-specific immune responses, we quantified established day 7 circulating correlates of vaccine immunogenicity: CD27⁺CD38⁺ PBs^18,19 and CXCR5⁺PD1⁺ cTfh.^20,27,28 Gating strategies are shown in Figures S1A and S1B. Both vaccines induced increases in PB (Figures 1A and 1B) and cTfh frequencies (Figures 1D and 1E); however, these responses were markedly attenuated following LAIV compared to IIV (p < 0.0001 for both) (Figures 1C and 1F), indicating reduced canonical early circulating antigen-specific, cellular responses after LAIV.

Figure 1.

HA B cell tetramers identify circulating HA⁺ IgD^neg B cells after LAIV

(A) Percentage of plasmablasts (PBs) from total live CD19⁺ B cells (see Figure S1A for gating strategy) that circulate at day 0 and day 7 after IIV inoculation.

(B) Percentage of PBs from total live CD19⁺ B cells that circulate at day 0 and day 7 after LAIV inoculation.

(C) Dot plot depicting the percent PBs that circulate after LAIV (square, gray) versus IIV (circle, orange).

(D) Percentage of circulating T follicular helper cells (cTfh) from total live CD4⁺ T cells (see Figure S1B for gating strategy) at day 0 and day 7 after IIV inoculation.

(E) Percentage of cTfh cells from total live CD4⁺ T cells at day 0 and day 7 after LAIV inoculation.

(F) Dot plot depicting the percent cTfh cells that circulate after LAIV (square, gray) versus IIV (circle, orange).

(G and H) Arbitrary units (AU) were calculated by normalizing optical density (OD) values to a reference control serum included on each assay plate. (G) IgG antibody titer (AU) against the Switzerland H3N2 strain in IIV (orange) and LAIV (gray) inoculated subjects. (H) IgA antibody titer (AU) against the Switzerland H3N2 strain in IIV (orange) and LAIV (gray) inoculated subjects.

(I) Percent of Sw15-H3-binding (H3⁺) IgD^neg B cells that circulate after IIV (circle, orange) and LAIV (square, gray) inoculation.

(J) Percent of Sw15-H3-binding (H3⁺) T-bet^neg IgD^neg B cells (see Figure S1A for gating strategy) that circulate after IIV (circle, orange) and LAIV (square, gray) inoculation.

(K) Percent of Sw15-H3-binding (H3⁺) T-bet⁺ IgD^neg B cells (see Figure S1A for gating strategy) that circulate after IIV (circle, orange) and LAIV (square, gray) inoculation.

Nineteen healthy subjects were immunized with the 2015–2016 IIV and seventeen healthy subjects were immunized with the 2015–2016 LAIV (see Table S1A for demographics and Table S1B for influenza A vaccine antigens). Peripheral blood mononuclear cells (PBMCs) were drawn at weekly time points after inoculation. Wilcoxon test used in (A), (B), (D), (E), (G), (H), and (I) and Mann-Whitney test used in (C) and (F).

We next assessed antigen-specific B cell responses using HA B cell tetramers.^23,24,25 Because the A/California/09 (H1N1) component of the 2015–2016 LAIV was poorly immunogenic,¹⁶ LAIV-inoculated recipients failed to mount durable, day 120 H1-specific IgG responses and showed minimal expansion of H1⁺ IgD^neg B cells relative to IIV recipients (Figures S1C and S1D). In contrast, both cohorts generated durable H3-specific IgG and IgA (H3-IgG and H3-IgA) responses to the A/Switzerland/9715293/2013 (H3N2)-like antigen (Figures 1G and 1H). Consistent with this, H3⁺ IgD^neg B cells were readily detected in circulation after LAIV, with frequencies comparable to those observed after IIV across the first month following vaccination (Figure 1I). Thus, despite matched attenuated PB and cTfh responses, LAIV can elicit similarly robust circulating antigen-specific B cell responses to immunogenic strains relative to IIV.

LAIV and IIV elicit transcriptionally and kinetically distinct HA⁺ IgD^neg B cells subsets defined by T-bet expression

Prior studies have shown that antigen-specific IgD^neg B cells elicited by IIV are transcriptionally heterogeneous and stratified by expression of the TF T-bet.^24,25 Consistent with this, H3⁺ IgD^neg B cells detected after LAIV could be partitioned into T-bet⁺ and T-bet^neg subsets that expanded over time in circulation (FigureS1A).

However, the temporal dynamics of these subsets differed by vaccine platform. Following IIV, T-bet^neg cells peaked at day 7 and declined thereafter, whereas after LAIV, they accumulated more gradually, reaching maximal frequencies at day 14 (Figure 1J). T-bet⁺ cells peaked at days 14–21 after IIV but exhibited delayed expansion after LAIV, with maximal frequencies observed at days 21–28 (Figure 1K). These data demonstrate that LAIV elicits a temporally shifted circulating HA⁺ B cell response, characterized by delayed accumulation of both T-bet^neg and T-bet⁺ HA-specific populations relative to IIV.

Transcriptional profiling reveals platform-specific programming for FcRL5⁺ and FcRL5^neg HA⁺ IgD^neg B cell subsets

To determine whether these kinetically distinct populations also differ transcriptionally, we used FcRL5 as a surrogate marker of T-bet expression to segregate circulating H3⁺ IgD^neg B cells into subsets at day 14 following LAIV (N = 3 subjects) for bulk RNA-seq. Comparison of FcRL5⁺ and FcRL5^neg subsets identified 273 differentially expressed genes (DEGs) (Figure 2A). FcRL5⁺ cells upregulated canonical, effector-associated genes, including TBX21, ZEB2, BATF, and ITGAX, while downregulating lymphoid homing markers such as CXCR5.

Figure 2.

Bulk RNA-seq demonstrate unique transcriptional signature of LAIV versus IIV elicited HA⁺ IgD^neg B cells by FcRL5 expression status

(A) Volcano plot depicts differentially expressed genes (DEGs). Genes of interest are annotated in blue. Threshold of q value at 0.05 is annotated by red line.

(B and C) GSEA comparing the ranked gene list from LAIV-elicited H3⁺ FcRL5⁺ IgD^neg B cells over H3⁺ FcRL5^neg IgD^neg B cells that are either downregulated (B) or upregulated (C) in IIV-elicited H1-specific FcRL5⁺ IgD^neg B cells over H1-specific FcRL5^neg IgD^neg B cells.²⁵

(D) Scatterplot depicts the fold change of differentially expressed genes in LAIV versus IIV-elicited FcRL5⁺ over FcRL5^neg subsets. Genes are colorized by assignment to IIV dataset (green), LAIV dataset (orange) or both datasets (black).

(E) Bar plot of Normalized Enrichment Score for each Hallmark gene set is shown. Significant bar plots (q < 0.05) are colorized purple.

(F–I) GSEA comparing ranked gene lists from IIV-elicited FcRL5⁺ over FcRL5^neg subsets (F and H) LAIV-elicited FcRL5⁺ over FcRL5^neg subsets (G and I) to Hallmark gene sets TNF-α signaling via NF-κb (F and G) and Inflammatory response (H and I).

FcRL5⁺ HA⁺ IgD^neg B cells and FcRL5^neg HA⁺ IgD^neg B cells were sort purified from N = 3 LAIV immunized subjects at day 14 for bulk RNA-seq.

We next compared these LAIV-elicited transcriptional programs to previously defined FcRL5⁺ versus FcRL5^neg subsets following IIV.²⁵ Gene set enrichment analysis (GSEA)²⁹ demonstrated concordant enrichment of core gene programs distinguishing FcRL5⁺ and FcRL5^neg cells across both vaccine platforms (Figures 2B and 2C). However, direct comparison of DEGs revealed substantial divergence: only 97 genes were shared, whereas the majority (176 genes) were unique to either LAIV or IIV (Figure 2D). These findings indicate that, despite a conserved core program, FcRL5-defined B cell subsets are differentially programmed depending on vaccine modality.

To further define these differences, we performed GSEA of LAIV-elicited FcRL5⁺ versus FcRL5^neg cells against MSigDB Hallmark pathways.³⁰ Twenty-seven pathways were positively enriched (false discovery rate [FDR] < 0.05), with no negatively enriched pathways detected (Figure 2E). Comparison with IIV-elicited subsets revealed shared enrichment of metabolic and activation-associated programs, including oxidative phosphorylation, mTORC1 signaling, reactive oxygen species, and IL-6-JAK-STAT3 signaling.

Strikingly, TNF-α signaling via NF-κB and the Inflammatory Response Hallmark gene sets showed opposite enrichment patterns across platforms, with positive enrichment in LAIV-elicited FcRL5⁺ over FcRL5^neg cells and negative enrichment in IIV-elicited counterparts (Figures 2F–2I), indicating divergent inflammatory programming. This pattern was preserved when comparing LAIV day 14 responses to previously published day 7 IIV datasets, despite variability in other pathways (Figure S2A), supporting the robustness of this distinction across time points. Collectively, these data demonstrate that HA⁺ IgD^neg B cells elicited by LAIV and IIV segregate into transcriptionally distinct FcRL5-defined states that share a conserved core effector-associated program but differ in broader transcriptional features, with inflammation emerging as a key axis of platform-specific divergence.

Distinct maturation and isotype programs characterize HA⁺ IgD^neg B cells after LAIV and IIV

While bulk RNA-seq identified platform-dependent differences in FcRL5⁺ versus FcRL5^neg HA⁺ IgD^neg B cell subsets, these analyses were constrained by antigen mismatch and reliance on a single surrogate marker of T-bet expression. We therefore sought to define the HA⁺ IgD^neg B cell compartment after LAIV and IIV in a more comprehensive and unbiased manner.

To do this, we performed single-cell RNA-seq with paired BCR sequencing on vaccine antigen-matched HA⁺ IgD^neg B cells collected within one month of vaccination. We analyzed samples from LAIV recipients (n = 6 across 2018 and 2019 seasons) and IIV recipients (N = 4 across 2018 and 2019 seasons), integrating datasets using Seurat³¹ and correcting for batch effects with Harmony.³² After filtering out cells with high mitochondrial DNA (including PB clusters 6 and 11, not shown), we obtained transcriptomes from 13,902 HA⁺ IgD^neg B cells, with paired VDJ sequences recovered from 34% of cells. Unsupervised clustering resolved nine transcriptionally distinct B cell states that were reproducibly identified across donors (Figures 3A, S3A, and S3B). The donor lacking all 9 transcriptionally discrete clusters had the least cells amplified (Figure S3B). All clusters exhibited high transcriptional purity (mean ratio of global unshifted entropy [ROGUE]³³ score > 0.9; Figure S4C), suggestive of robust separation of cell states.

Figure 3.

Distinct maturation, isotype and effector memory programs characterize HA⁺ IgD^neg B cells following LAIV versus IIV

(A) UMAP projection of HA⁺ IgD^neg B cells clustered by transcriptional similarity. (Figure S3A shows total cell number by cluster. Figure S3B shows cellular contributions to clusters by individual. Figure S3C shows ROGUE scoring as a metric of transcriptional similarity by cluster).

(B) Violin plot depicts IGHV mutation rates for clusters from IIV (pink) versus LAIV (blue) inoculated subjects. ∗∗∗∗p < 0.0001, unpaired t test. Figure S3D shows mutation rates across clusters without stratification by vaccine type.

(C) Bar plot depicts isotype frequency by cluster assignment and vaccine platform. ∗denotes clusters with significant differences in isotype distribution by PERMANOVA testing. Table S4 shows mean isotype frequencies by cluster and by vaccine platform.

(D) GSEA comparing differentially expressed genes in isotype switched clusters to the ranked gene list of IIV-elicited H1-specific FcRL5⁺ over FcRL5^neg IgD^neg B cells.²⁵

(E) UMAP projection shows areas of differential abundance (DA-seq³⁴) as overrepresented in LAIV (blue, DA1) versus IIV inoculated (red, DA2) subjects. Figure S3H–I lists genes predicted as overexpressed in these DA regions.

Six individuals received LAIV (3 received 2018–2019, 3 received 2019–2020) and four (2 received 2018–2019, 2 received 2019–2020) individuals received IIV (see Table S1B for vaccine antigens). Vaccine antigen-matched HA⁺ IgD^neg B cells were sort-purified for single cell sequencing and data were integrated using Harmony.³²

To assign biological identity to each cluster, we integrated somatic hypermutation (Figure S3D), isotype usage (Figure S3E), and clonal expansion (Figure S3F). One cluster (cluster 7) exhibited minimal mutation, had a predominantly IgM isotype, and was found as polyclonal, consistent with a naive B cell population. In contrast, clusters 2 and 4 were highly mutated, almost entirely isotype-switched, and composed of expanded clones, indicative of antigen-experienced memory populations. Clusters 0, 1, and 8 were also isotype-switched but less mutated and largely polyclonal, suggesting less differentiated memory states. Cluster 3 comprised isotype-switched cells with clonal expansion but significantly lower mutation rates than clusters 2 and 4. Cluster 5 was notable for a predominant IgM isotype despite clonal expansion, including a large clone derived from a single IIV recipient, suggesting an IgM⁺ memory-like population. A small cluster, cluster 10 (< 100 cells), was excluded from further analysis.

Having defined these transcriptional states, we next asked whether their maturation and isotype characteristics differed by vaccine platform. Clusters 2, 3, and 4 exhibited significantly higher levels of somatic hypermutation in LAIV recipients compared to IIV recipients, suggesting differences in precursor recruitment and/or GC selection (Figure 3B). In parallel, several clusters showed platform-dependent differences in isotype composition (Figure 3C). PERMANOVA analysis identified significant differences in isotype distributions in clusters 0 (p = 0.015, r² 0.59), 1 (p =0.03, r² = 0.53), and 3 (p = 0.015, r² = 0.53). Increased IgM over IgG usage in IIV-inoculated subjects drove differences in clusters 0 and 1, while increased IgA over IgG usage in LAIV-inoculated subjects contributed to differences in cluster 3 (Figure S3).

Together, these data demonstrate that LAIV and IIV differentially shape the maturation and isotype composition of HA⁺ IgD^neg B cell responses. While both vaccines generate a shared set of transcriptional B cell states, LAIV preferentially enriches for a more highly mutated and IgA-biased population (cluster 3), consistent with distinct pathways of B cell differentiation following mucosal attenuated infection versus systemic inactivated immunization.

Differential enrichment of effector memory programs in HA⁺ IgD^neg B cells following LAIV versus IIV

Building on the observed differences in mutation rates and isotype usage, we next examined whether LAIV and IIV also differentially shape the functional states of circulating HA⁺ IgD^neg B cell populations. Prior studies have shown that isotype-switched HA⁺ IgD^neg B cells can be partitioned into stem-like and effector memory states, which represent divergent fate trajectories of proliferation and persistence versus rapid ASC differentiation or cytokine secretion after antigen re-challenge. We therefore mapped our isotype-switched clusters (0, 1, 2, 3, 4, and 8) onto these established gene programs, using GSEA (Figure 3D).

This analysis revealed a structured functional landscape across clusters. Clusters 0,1, and 8 enriched for stem-like memory programs, consistent with a quiescent, self-renewing state, whereas cluster 3 mapped onto an effector memory program. Notably, clusters 2 and 4 did not conform to either canonical state. Instead, cluster 2 exhibited loss of stem-like features without acquisition of effector programs, while cluster 4 displayed only partial induction of effector-associated genes. These populations may therefore represent transitional or alternative differentiation states that fall outside the classical central to effector memory B cell continuum. Collectively, these findings suggest that the circulating HA⁺ IgD^neg B cell compartment comprises a spectrum of functionally distinct B cell states.

We next asked whether these functional states are differentially distributed between vaccine platforms. Visualization of cells by vaccine origin revealed non-random segregation within uniform manifold approximation and projection (UMAP) transcriptional space, suggesting that LAIV-elicited cells and IIV-elicited cells preferentially populate distinct regions (termed LAIV-enriched versus IIV-enriched) of the B cell state landscape (Figure S3G).³⁴ To formally quantify this, we applied DA-seq, a multiscale framework that identifies differentially abundant cellular neighborhoods across conditions. This analysis defined an LAIV-enriched region corresponding to cluster 3 (DA1) and IIV-enriched region corresponding to cluster 4 (DA1) (Figure 3E). At the transcriptional level, the LAIV-associated DA1 region was marked by genes linked to cytotoxicity, effector differentiation, and exhaustion programs, including CST7, SLAMF7, SRGN, ZEB2, and TOX respectively as well as components of the TNF-α signaling pathways. In contrast, the IIV-associated DA2 region was enriched for genes involved in cell cycle control, NF-kB signaling, and endoplasmic reticulum stress responses, including CDKN1A, SIK3, NFKB1, NFKBID, ATF6, and EIF2AK3 respectively (Figures S3H and S3I).

Together these data suggest that LAIV and IIV do not simply vary the magnitude of the antigen-specific B cell response but instead bias the distribution of circulating antigen-specific B cells across a continuum of functional states.

Transcriptional states of LAIV-elicited HA-specific IgD^neg B cells associate with differences in circulating kinetics and clonal recall

To further define functional differences between clusters 3 and 4, which are differentially enriched in circulation after LAIV and IIV, respectively, we first examined the expression of canonical memory B cell TFs and transcriptional regulators (TRs) (Figure 4A).⁶ Both clusters expressed near equivalent levels of the memory B cell TF Hhex, consistent with a shared memory B cell identity.³⁵ In contrast, TBX21 was differentially expressed across clusters, with intermediate levels in cluster 4 and highest expression in cluster 3, suggesting a graded acquisition of a T-bet-associated effector program. Cluster 4 additionally expressed RUNX3 and BHLHE40, consistent with an effector-memory-like state that is maintained without terminal differentiation.^36,37,38,39 By comparison, cluster 3 was distinguished by higher expression of ZEB2, a TF associated with terminal effector CD8 T cell differentiation, consistent with a more differentiated effector state.⁴⁰

Figure 4.

Transcriptional states of LAIV-elicited HA⁺IgD^neg B cells associate with distinct circulating kinetics and clonal recall trajectories

(A) Dot plot of transcription factors (TFs, colorized as red) and transcription regulators (TRs, colorized as blue) for isotype switched B cell clusters depicted in Figure 3A.

(B) UMAP colored by pseudotime from Monocle trajectory inference.⁴¹

(C) Sankey plot depicts the proportion of cells by cluster assignment and day after immunization for one LAIV-inoculated subject. Table S5 shows cell frequencies by cluster and day after immunization.

(D) Heatmap depicts percent clonal sharing across clusters at an individual time point and at weekly time points after LAIV inoculation of N =1 healthy subject. Gray diagonal boxes list total numbers of clones in each cluster by time point.

(E) Schematic of prime-boost immunization and blood draw schedule.

(F) Heatmap depicts clonal lineage relationships between antigen-specific memory B cells following vaccine prime and PBs following vaccine boost, stratified by transcriptional cluster and individual subject grouped by vaccine regimen (LAIV-LAIV, red; LAIV-IIV, blue). Expanded lineages (defined lineages in top 30% of clones) are above the white lines. Asterisks denote samples with a significant enrichment of HA⁺ IgD^neg clones from single cell lineages within expanded over unexpanded ASC lineages (Fisher’s exact test).

(G) Stacked bar plot depicts the proportion of lineages per cluster after vaccine prime that are related to PB lineages after vaccine boost. Values are adjusted for the number of lineages per cluster. Subjects are shown individually (LAIV-LAIV, red; LAIV-IIV, blue).

To place these transcriptional differences within a global organizational framework, we next performed trajectory-based manifold embedding of HA⁺ IgD^neg B cells across subjects (Figure 4B). This analysis revealed a structured transcriptional landscape in which clusters 2, 3, and 4 occupied discrete, minimally overlapping regions, consistent with distinct transcriptional end states rather than a continuum of transitional intermediates.

We next examined the temporal dynamics of each cluster following LAIV (Figure 4C; Table S4). Cluster 3 cells decreased by approximately 50% between days 7 and 14 following LAIV vaccination, indicating a rapid contraction of this transcriptionally distinct end state in circulation. In contrast, cluster 4 cells remained relatively stable over the same period, whereas cluster 2 cells, which exhibited minimal TBX21 expression among isotype-switched clusters (Figure 4A), increased in frequency over time. Notably, although T-bet⁺ HA-specific IgD^neg B cells are first detectable at days 7 and 14 following LAIV immunization, they do not reach maximal abundance in blood until day 21 (Figure 1K), suggesting that early cluster 3 contraction precedes the later accumulation of the broader circulating T-bet⁺ compartment. Following IIV, cluster 2 contracted early, cluster 4 expanded over time, and cluster 3 remained relatively stable, highlighting a distinct, vaccine modality-dependent deployment of these transcriptional end states (Figure S4B).

We next asked whether these transcriptionally defined end states represented stable populations over time or transient intermediates within a shared differentiation trajectory. To address this, we performed longitudinal VDJ and gene expression profiling of HA⁺ IgD^neg B cells from days 7, 14, and 21 following LAIV vaccination in a single deeply sampled individual. Cluster identities were stable across time points (data not shown), with minimal evidence of transition between clusters, supporting the interpretation that these represent stable end states rather than transient intermediates. Importantly, we observed minimal clonal overlap between cluster 3 and other clusters, arguing against appreciable transition of cluster 3 cells into alternative states within the limits of sampling resolution (Figure 4D). In contrast, clusters 2 and 4 showed clonal sharing over time, including among expanded clones (Figures 4D and S4A). Notably, analysis of a deeply sampled IIV-inoculated subject revealed a different pattern, in which clusters 2, 3, and 4 were clonally related over time (Figure S4C), indicating that these transcriptionally defined states can arise from shared antigen-specific lineages depending on vaccination context. Although derived from a single deeply sampled subject, these longitudinal dynamics after LAIV recapitulate the temporal patterns observed in the 2015–2016 cross-sectional cohort (Figures 1J and 1K).

Divergent effector B cell states differentially contribute to recall responses and are associated with durable immunity

Together these data support a model in which transcriptionally distinct HA⁺ B cell clusters 2, 3, and 4 represent stable end states whose clonal relationships are differentially constrained by the nature of vaccine antigen exposure. However, whether these kinetically and transcriptionally distinct populations contribute differentially to downstream humoral responses after LAIV remained unclear. In particular, the transient kinetics of cluster 3 in blood following LAIV raised the possibility that this population may be functionally distinct from clusters 2 and 4 in its contribution to recall responses.

To address this, we next evaluated whether transcriptionally discrete LAIV-elicited HA⁺ IgD^neg B cell populations contributed differently to recall responses following antigen re-challenge. Across five LAIV-inoculated individuals (N = 3 boosted with IIV and N = 2 boosted with LAIV), PBs were sorted at day 7 post-boost for paired VDJ analysis (Figure 4E). Shared clones were defined as ≥ 85% CDR3 nucleotide similarity with identical V gene, J gene, and CDR3 length. Analyses were restricted to clusters 2-4, encompassing the principal isotype-switched populations with effector-associated transcriptional programs (Figure 3D). We reasoned that expanded PB lineages are more likely to reflect antigen-driven expansion of HA-specific precursor clones, particularly when there is clonal sharing between single HA⁺ IgD^neg B cells and expanded PB lineages. To quantify contributions to the expanded PB response, we annotated PB lineages using two complementary criteria: (1) large lineages, defined as those in the top 30% of lineage sizes at day 7 post-boost, and (2) lineages connected to clusters 2-4. We then tested whether lineage expansion was associated with cluster connectivity within each donor and cluster. This analysis revealed that clusters 2 and 4 were consistently enriched among expanded PB lineages across donors, suggesting preferential participation in the recall response (Figure 4F). This association reached statistical significance (Fisher’s exact test) for cluster 2 in two independent individuals (IIV- and LAIV-boosted) and cluster 4 in one individual (LAIV-boosted). In contrast, cluster 3 single-cell lineages showed minimal connectivity to expanded PB lineages (Figure 4F). Consistent with this, cluster 3-derived clones were underrepresented within the total PB repertoire, whereas clusters 2 and 4 were relatively more enriched (Figure 4G). Together, these data indicate that TBX21-expressing HA⁺ IgD^neg B cell states differ in their contribution to recall responses, with cluster 3 exhibiting limited participation in antigen-driven circulating PB expansion following re-challenge.

To determine whether features of the antigen-specific B cell response were associated with durable antibody, we returned to our original 2015–2016 cohort and first considered PB and cTfh cells that expand in blood within 7 days of vaccine (Figures 1B and 1C). We observed no correlation between the magnitude of PBs and cTfh cells and the fold change of H3-IgG or H3-IgA between day 0 and day 120 after LAIV (Figures S5A–S5D).

Given that our transcriptional analyses revealed time-dependent differences in effector B cell programs following LAIV, we next related the frequency of T-bet^neg and T-bet⁺ H3⁺ IgD^neg B cells at early and late time points to longitudinal antibody responses in our original cohort. We found no correlation between the frequency at any time point of T-bet^neg cells and the fold change of H3-IgG or H3-IgA between day 0 and day 120 after LAIV (Figures S5E–S5L). Instead, we found the frequency of T-bet⁺ HA⁺ IgD^neg B cells at day 7 and day 14 was positively and significantly correlated with the fold change in H3-IgA between day 0 and day 120 after LAIV (Figures 5A and 5C). In contrast, no association was observed between early T-bet⁺ B cell frequencies and HA-IgG responses over the same time interval (Figures 5B and 5D). Despite reaching peak frequencies in circulation at day 21 and day 28 (Figure 1K), the magnitude of T-bet⁺ H3⁺ IgD^neg B cells at day 21 or day 28 did not correlate with the fold change between day 0 and day 120 H3-IgA or H3-IgG responses (Figures 5E–5H).

Figure 5.

Circulating T-bet⁺ and T-bet^neg H3⁺ IgD^neg B cells do not predict durable H3 antibody response

As described in Figure 1, seventeen healthy subjects received 2015–2016 LAIV. Correlation between the fold change in H3-IgA Ab (A, C, E, and G) and H3-IgG Ab (B, D, F, and H) at days 0 and 120 and the magnitude of H3⁺T-bet⁺ IgD^neg B cells at day 7 (A and B), day 14 (C and D), day 21 (E and F), or day 28 (G and H) from the seventeen LAIV inoculated subjects. Gating strategy for H3⁺ T-bet⁺ IgD^neg B cells are shown in Figure S1A. Non-parametric testing was used with one tailed t test.

Collectively, these data support a model in which HA⁺ IgD^neg B cells differentiate into transcriptionally distinct effector-like states with vaccine modality-dependent clonal connectivity and kinetics in blood. After LAIV, cluster 3 represents a transient circulating TBX21-high population that is enriched in IgA isotype usage, minimal clonal expansion, and contributes minimally to recalled circulating PB responses, whereas clusters 2 and 4 exhibit greater clonal persistence in blood and preferential participation in circulating recall responses. Importantly, early induction of T-bet⁺ HA⁺ IgD^neg B cells, coincident with the transient circulation of cluster 3, is selectively associated with durable HA-IgA responses after LAIV, linking early effector programming of circulating HA⁺ IgD^neg B cells to long-term systemic HA-IgA immunity after LAIV.

Discussion

Traditionally defined by surface phenotypes, human memory B cells are now understood as a spectrum of antigen-experienced states with distinct transcriptional programs and differentiation potential.⁴ A recurring feature across acute and chronic infection, autoimmunity, and vaccination is the emergence of T-bet-expressing B cells, yet it is not completely clear whether these cells represent a unified functional population or multiple context-dependent states.^42,43 By comparing HA⁺ IgD^neg B cell responses after LAIV and IIV, our findings demonstrate that T-bet associated HA⁺ IgD^neg B cells comprise heterogeneous populations with distinct transcriptional programs, clonal behaviors, and circulation kinetics shaped by vaccine platform.

Unlike IIV, which primarily delivers antigen systemically, LAIV acts as a transient localized infection of the upper airway mucosa. This distinction likely imposes different constraints on antigen persistence, innate immune activation, and tissue localization that shape antigen-specific B cell differentiation. Consistent with this framework, prior studies have shown that LAIV can induce discordant mucosal and systemic antibody responses, with a substantial fraction of individuals failing to exhibit robust systemic seroconversion and, in some cohorts, minimally detectable mucosal antibody responses as well.^44,45 In these settings, circulating antigen-specific B cell responses in blood are correspondingly limited but mucosal antigen-specific B cells can still be detected, suggesting that blood may incompletely capture the magnitude or distribution of LAIV-induced immunity rather than reflecting overall immunogenicity.^13,44 In contrast, other cohorts, including the one studied here, exhibit both circulating HA-specific B cell responses and durable HA-IgA and HA-IgG induction,⁴⁶ indicating a response that is more readily detectable across systemic and mucosal compartments. Together, these observations suggest that variability in the magnitude, kinetics, and compartmental distribution of the response may underlie differences in the detectability of circulating antigen-specific B cells following LAIV.

Within this context, in our cohort, we find that LAIV preferentially enriches for circulating HA⁺ IgD^neg B cell states characterized by high somatic hypermutation, IgA skewing, and inflammatory transcriptional programs. In particular, we identify a TBX21-high, ZEB2-expressing population (cluster 3) that preferentially circulates after LAIV. These cells exhibit transient kinetics, minimal clonal expansion in blood, and limited contribution to circulating PB lineages following antigen re-challenge, consistent with a short-lived effector-like state that is not efficiently recalled into circulation. In contrast, a distinct population with lower TBX21 expression and partial effector programming (cluster 4) displays clonal persistence and preferential contribution to recall PB responses, consistent with a recall-competent effector memory state. Although detectable in circulation after both vaccine platforms, this population is comparatively more represented in blood following IIV. Together, these findings suggest that attenuated mucosal infection preferentially promotes a transient TBX21-high effector-like B cell states in circulation while comparatively limiting representation of circulating recall-competent memory populations.

The transcriptional features of the TBX21-high cluster further parallel principles established in CD8⁺ T cell biology, where graded TBX21 expression coordinates terminal effector versus memory precursor fate decisions in concert with ZEB2.^40,47 Whether a similar cooperative interaction between TBX21 and ZEB2 enforces this differentiation program in human B cells remains an important question for future epigenetic and functional studies. Consistent with this possibility, ZEB2 has been implicated in TLR7/8-driven T-bet⁺ inflammatory B cell differentiation, suggesting that high TBX21/ZEB2 expression may define a shared effector-like fate axis across lymphocyte lineages.⁴⁸ However, the upstream innate sensing pathways that initiate this program in the context of LAIV remain undefined.

At the same time, our data argue against strict separation of mucosal and systemic immunity. Prior studies have shown that intestinal and circulating IgA responses in celiac disease exhibit substantial clonal overlap, supporting continued connectivity between mucosal and systemic B cell compartments.⁴⁹ Similarly, tonsillar Tfh responses after LAIV have been associated with the magnitude of systemic influenza-specific IgG responses, suggesting coordinated interactions between mucosal lymphoid tissues and systemic humoral immunity following intranasal vaccination.⁴⁶ Consistent with this framework, certain HA⁺ IgD^neg B cell populations identified after LAIV, particularly clusters 2 and 4, comprise expanded clonotypes that repeatedly circulate in blood and contribute to PB recall responses following antigen re-challenge. The expression of tissue-associated TFs including RUNX3⁵⁰ and BHLHE40⁵¹ in cluster 4 further raises the possibility that these cells represent mucosal-adapted states capable of transient recirculation after LAIV. These observations support a model in which mucosal vaccination generates a dynamic network of antigen-specific B cell states that can traffic between mucosal and systemic compartments despite localized antigen exposure.

In summary, our findings demonstrate that influenza vaccine platform influences the differentiation and circulation of antigen-specific B cell states with distinct transcriptional programs and inferred fate potential. By framing LAIV as a transient attenuated mucosal infection rather than fixed systemic antigen exposure, these data provide a mechanistic context for how vaccine modality shapes B cell differentiation. More broadly, our results suggest that T-bet-associated B cell responses cannot be interpreted as a single conserved memory population, but instead comprise multiple context-dependent states whose immunologic significance depends on the anatomical and inflammatory environment of antigen encounter.

Limitations of the study

This study has several limitations. The cohort size is modest and not matched for sex, introducing the possibility of sex-associated biases⁵² and necessitating validation in larger and more diverse populations. Our analyses are restricted to circulating cells and do not directly sample mucosal tissues, limiting inference regarding anatomical origin, tissue residency, or local differentiation. Single-cell profiling depth is also limited relative to the diversity of antigen-specific B cell states, and rare or transient populations may therefore be under-sampled. In addition, our dataset is transcriptome- and BCR-based and lacks matched protein-level measurements, limiting direct linkage between transcriptional state and surface phenotype. Finally, although clonal relationships and transcriptional programs support distinct differentiation states, direct experimental validation will be required to define their functional contributions to antibody production and protection. Future studies should define the upstream signals that drive TBX21-associated effector-like B cell differentiation following mucosal vaccination, including whether local innate sensing contributes to this program. Paired mucosal and systemic sampling will be essential to determine where these states originate and how they traffic between compartments.

Resource availability

Lead contact

Requests for further information and resources and reagents should be directed to and will be fulfilled by lead contact, Anoma Nellore (anoma.nellore@nyulangone.org).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Gene Expression Omnibus (GEO) Sequencing and BCR repertoire data have been deposited at and are publicly available as of the date of publication. RNA-seq data have been deposited at GSE314199 and are publicly available as of the date of publication. Single cell gene expression and BCR data have been deposited at GSE319537.
• •Custom scripts used for bulk BCR repertoire analyses are available from the lead contact upon request. This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper are available from the lead contact upon request

Acknowledgments

We acknowledge Sergei Koralov PhD and Benjamin TenOever PhD for their thoughtful comments, edits and suggestions. We additionally acknowledge the members of the UAB Single Cell Core (Shanrun Liu) and Flow Cytometry Facility (Hanumanthu Sagar) for their technical assistance. This work was funded by an R21 to A.N., a CCTS grant no. UM1TR004771 to A.N., and a U19AI109962 grant to T.R.

Author contributions

A.N., F.E.L., and T.D.R. conceived the idea for the project and secured the initial funding. A.N. designed the experiments that were performed by A.N., E.Z., C.D.S., J.T.G., J.L., B.M., and F.Z. B cell tetramers were developed and produced by J.A. Human samples used in this study were obtained via the Alabama Vaccine Research Clinic, directed by P.A.G. Bioinformatic analyses were performed by A.D.R., C.D.S., P.M., and C.F. All other data were analyzed by A.N. A.N. wrote the manuscript and prepared final figures. Critical feedback on the project and manuscript was provided by R.S.H., R.D., V.E.N., A.D.R., T.D.R., F.E.L., P.A.G., and C.D.S.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-IgD	BD	Cat# 555778; RRID:AB_396113
Anti-IgD	BD	Cat# 561490; RRID:AB_10679356
Anti-IgD	BioLegend	Cat# 348226; RRID:AB_2561619
Anti-CD19	BioLegend	Cat# 302234; RRID:AB_2561668
Anti-CD19	BD	Cat# 560727; RRID:AB_2562599
Anti-CD27	BioLegend	Cat# 356418; RRID:AB_2562599
Anti-CD3	BioLegend	Cat# 344808; RRID:AB_10640736
Anti-CD3	BD	Cat# 560158; RRID:AB_1645478
Anti-CD4	BioLegend	Cat# 344607; RRID:AB_1953236
Anti-CD14	BD	Cat# 561384; RRID:AB_10611720
Anti-FcRL5	BioLegend	Cat# 340304; RRID:AB_2104588
Anti-CD27	BioLegend	Cat# 340304; RRID:AB_2104588
Anti-CXCR5	BioLegend	Cat# 356904; RRID:AB_2561813
Anti-T-bet	BioLegend	Cat# 644814; RRID:AB_10901173
Anti-T-bet	BioLegend	Cat# 644812; RRID:AB_2200540
Anti-CD21	BioLegend	Cat# 354908; RRID:AB_2561544
Anti-CD27	BioLegend	Cat# 356418; RRID:AB_2562599
Anti-Rabbit IgG (H+L)	ThermoFisher	Cat# A-21245; RRID:AB_2535813
AffiniPure Goat Anti-Human IgG (H+L)	Jackson ImmunoResearch	Cat# 109-005-003; RRID:AB_2337532
Peroxidase AffiniPure Goat Anti-Human IgG (H+L)	Jackson ImmunoResearch	Cat# 109-035-088; RRID:AB_2337584
Alkaline Phosphatase AffiniPure Goat	Jackson ImmunoResearch	Cat# 109-055-003; RRID:AB_2337599
Biological Samples
PBMCs from influenza-vaccinated subjects	Alabama Vaccine Research Clinic (AVRC), UAB	N/A
Chemicals, Peptides, and Recombinant Proteins
Igepal CA-360	Sigma-Aldrich	Cat# I8896
Neuraminidase (C. perfringens)	Sigma-Aldrich	Cat# N5631
7-AAD (7-aminoactinomycin-D)	ThermoFisher	Cat# A1310
LIVE/DEAD Fixable Near-IR Dead Cell Kit	ThermoFisher	Cat# L10119
LIVE/DEAD Fixable Red Dead Cell Kit	ThermoFisher	Cat# L23102
Formalin solution, 10% neutral buffered	Sigma-Aldrich	Cat# HT501128
2-Mercaptoethanol	ThermoFisher	Cat# BP176-100
TotalSeq-C hashtag C0251	Biolegend	Cat# 394661
TotalSeq-C hashtag C0252	Biolegend	Cat# 394663
TotalSeq-C hashtag C0253	Biolegend	Cat# 394665
TotalSeq-C hashtag C0254	Biolegend	Cat# 394665
Critical Commercial Assays
AviTag Biotin-Protein Ligase Kit	Avidity	Cat# BirA-500
Chemiluminescent Substrate for Alkaline Phosphatase	Moss Inc.	Cat# ChemiAP-RK100
EasySep Human B Cell Negative Selection Kit	StemCell Technologies	Cat# 17954
EasySep Human Pan-B Cell Negative Selection Kit	StemCell Technologies	Cat# 19554
Chromium Next GEM Chip G Single Cell Kit, 8 rxns	10x Genomics	Cat# 1000120
Chromium Next GEM Single Cell 3′ GEM, Library & Gel Bead Kit v3.1	10x Genomics	Cat# 1000268
Library Construction Kit	10x Genomics	Cat# 1000190
Chromium 3′/5′ Feature Barcode Kit	10x Genomics	Cat# 1000256
RNeasy Mini Kit	Qiagen	Cat# 74104
SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing	Takara	Cat# 634888
NexteraXT DNA Library Preparation Kit	Illumina	Cat# FC-131-1024
HiFi Polymerase Mastermix	KAPA Biosystems	Cat# KK2601
Deposited Data
RNA-seq	This paper	GEO: GSE314199
Single-cell and BCR sequencing (GEX, VDJ)	This paper	GEO: GSE319537
Recombinant DNA
Ca09 Influenza HA (H1)	GeneArt	NC_026433.1
SW13 Influenza HA (H3)	GeneArt	EPI543739
Michigan 15 Influenza HA (H1)	GeneArt	MK622940.1
Singapore Influenza HA (H3)	GeneArt	Not specified
Brisbane Influenza HA (H1)	GeneArt	Not specified
Kansas Influenza HA (H3)	GeneArt	Not specified
Software and Algorithms
FlowJo v9.9.3 / v10.2	FlowJo LLC	N/A
GraphPad Prism v7.0a / v8.0	GraphPad Software	N/A
STAR aligner	Dobin et al.41	https://github.com/alexdobin/STAR
GenomicRanges (R/Bioconductor)	Lawrence et al.53	https://bioconductor.org/packages/GenomicRanges
edgeR v3.24.3	Robinson et al.54	https://bioconductor.org/packages/release/bioc/html/edgeR.html
CellRanger v7.1 (multi-pipeline)	10x Genomics	https://www.10xgenomics.com/support/software/cell-ranger
Seurat v5.1.0 (R v4.4.1)	Satija Lab	https://satijalab.org/seurat
SoupX	Young & Behjati55	https://github.com/constantAmateur/SoupX
DoubletFinder	McGinnis et al.56	https://github.com/chris-mcginnis-ucsf/DoubletFinder
IMGT High-V-QUEST v3.6.3 (reference release 202030-2)	IMGT57	https://www.imgt.org/HighV-QUEST/
MATLAB R2024b	The MathWorks Inc.	RRID:SCR_001622
R v4.41	R Core Team	N/A

Experimental model and study participant details

Human subjects

The UAB Institutional Review Board approved all human study protocols. Subjects self-identified as healthy and reported having missed the prior season’s influenza vaccine. Subjects were recruited and consented through the Alabama Vaccine Research Clinic (AVRC).

Subjects were enrolled to receive their first seasonal influenza vaccine as one of the following:

• •2015-2016 Fluzone (Sanofi-Pasteur, inactivated influenza vaccine, IIV)
• •2015-2016 FluMist (AstraZeneca, live attenuated influenza vaccine, LAIV)
• •2018-2019 FluMist (AstraZeneca)
• •2018-2019 Fluzone (Sanofi-Pasteur)
• •2019-2020 FluMist (AstraZeneca)
• •2019-2020 Fluzone (Sanofi-Pasteur)

Blood was drawn on days 0, 7, 14, 21, and 120 (± 1 week) from subjects enrolled in the 2015-2016 season, and on days 7, 14, and 21 from subjects who received all other vaccine formulations.

A subset of subjects underwent a prime-boost regimen under UAB IRB-approved clinical trial protocol NCT04080245:

• •Subjects who received the 2018-2019 FluMist (N = 3) were boosted with the 2018-2019 Fluzone 56 days later, with blood drawn 63 days after boost.
• •Subjects who received the 2019-2020 FluMist (N = 2) were boosted with an additional dose of 2019–2020 FluMist 56 days later, with blood drawn 63 days after boost.

For the single-cell RNA-seq cohort, six LAIV-inoculated and four IIV-inoculated subjects were enrolled, with blood drawn at days 7, 14, and 21 post-immunization.

Method details

Lymphocyte and plasma isolation

Peripheral blood was drawn into K2-EDTA tubes (BD Bioscience). Peripheral blood mononuclear cells (PBMCs) and plasma were isolated by density gradient centrifugation over Lymphocyte Separation Medium (CellGro). Red blood cells were lysed with ammonium chloride solution (StemCell). Plasma and PBMCs were either used immediately for single-cell sequencing or aliquoted and stored at −80°C for downstream assays.

Recombinant influenza HA protein production and B Cell tetramer generation

Coding sequences (amino acids 18–524) of influenza HA ectodomains were synthesized (GeneArt, Regensburg, Germany) from vaccine antigen–matched strains. HA ectodomains were cloned into the pCXpoly(+) vector modified with a 5′ human CD5 signal sequence and a 3′ GCN4 isoleucine zipper trimerization domain (GeneArt), followed by either a 6XHIS tag (HA-6XHIS construct) or an AviTag (HA-AviTag construct).¹¹

The HA-6XHIS and HA-AviTag constructs for each HA were co-transfected at a 2:1 ratio using 293Fectin Transfection Reagent into FreeStyle 293-F Cells (ThermoFisher Scientific). Transfected cells were cultured in FreeStyle 293 Expression Medium for 3 days, and supernatant was recovered by centrifugation.

Recombinant HA molecules were purified by FPLC using a HisTrap HP Column (GE Healthcare) and eluted with 250 mM imidazole. Purified HA was biotinylated using biotin-protein ligase (Avidity), and biotinylated proteins were tetramerized with fluorochrome-labeled streptavidin (Proenzyme). Labeled tetramers were purified by size exclusion chromatography on a HiPrep 16/16 Sephacryl S-300 column (GE Healthcare).

Flow cytometry

Single-cell suspensions were blocked with 2% human serum and stained with antibody panels described in Table S2A. 7AAD or LIVE/DEAD Fixable Dead Cell Stain Kits (Molecular Probes/ThermoFisher) were used to discriminate live cells.

To eliminate non-specific HA tetramer binding, cells were treated at 37°C with 0.5 U/mL neuraminidase (C. perfringens, Sigma) to remove sialic acid, washed, and then stained with HA tetramers. Intracellular proteins were detected by staining cell-surface markers, fixing cells in 10% neutral buffered formalin solution (Sigma), and staining permeabilized cells (0.1% IGEPAL, Sigma) with intracellular antibodies.

Stained cells were acquired using either a FACSCanto II (BD Bioscience) or a BD Symphony flow cytometer (Invitrogen, ThermoFisher). Analysis was performed using FlowJo v9.9.3 or FlowJo v10.2.

Cell sorting

B cell subsets were sort-purified for bulk RNA-seq or single-cell VDJ and GEX analyses using a FACSAria (BD Biosciences) in the UAB Comprehensive Flow Cytometry Core. Table S2B lists the flow panels used for subset sort-purification.

RNA-seq library preparation and data analysis

B cell subsets were sorted directly into RLT buffer (Qiagen) with 1% mercaptoethanol and snap-frozen in liquid nitrogen. RNA was extracted using the QuickRNA Micro Prep Kit (Zymo), and cDNA was prepared using the SMART-Seq v4 cDNA Synthesis Kit (Takara). Sequencing libraries were constructed using 200 pg cDNA as input for the NexteraXT kit with NexteraXT indexing primers (Illumina).

Libraries were quality assessed, pooled, and sequenced using 75 bp paired-end chemistry on a NextSeq500 at the UAB Heflin Genomics Core. Sequencing reads were mapped to the hg38 human genome using STAR⁴¹ with default settings, using the UCSC KnownGene table as the reference transcriptome. Reads overlapping exons were tabulated using the GenomicRanges⁵³ package in R/Bioconductor.

Genes expressed at ≥3 reads per million in all samples from one group were considered detected and used as input for edgeR v3.24.3⁵⁴ to identify differentially expressed genes (DEGs). P-values were FDR-corrected using the Benjamini-Hochberg method,⁵⁸ with genes of FDR < 0.05 and absolute log2 fold change > 1 considered significant. Expression data were normalized to reads per kilobase per million mapped reads (RPKM). Samples included either 3 LAIV samples (day 14 timepoint) or 6 IIV samples (day 14 timepoint) as biological replicates per B cell subset.

Single-cell RNA library preparation and sequencing

Six LAIV-inoculated and four IIV-inoculated subjects were enrolled, with blood drawn at days 7, 14, and 21 post-immunization. B cells were negatively selected using column purification (EasySep Pan-B Cell Purification Kit, StemCell Technologies), treated with neuraminidase, stained with HA tetramers, and labeled with antibodies conjugated to TotalSeq-C oligomer hashtags (Table S4, tab 4).

Single-cell suspensions were processed on the 10x Genomics workflow for cell capture and scRNA gene expression (GEX), BCR, and Feature Barcoding library preparation, using the Chromium Single Cell 5′ Library and Gel Bead Kit (Next GEM 5′ Kit v2) and the Single Cell 5′ Feature Barcode Library Kit (10x Genomics), following manufacturer’s instructions.

Single-cell transcriptome and repertoire profiling

Raw sequence reads were processed using CellRanger (version 7.1) with the multi-pipeline and default parameters for demultiplexing, gene expression quantification, and BCR sequence assembly. Reference datasets used for GEX and VDJ annotation were refdata-gex-GRCh38-2024-A and refdata-cellranger-vdj-GRCh38-alts-ensembl-7.1.0, respectively (10x Genomics).

CellRanger output was analyzed in R (version 4.4.1) using the Seurat package (version 5.1.0). Count data were log-normalized and transformed, and variable genes were detected; uniform manifold approximation and projection (UMAP) was performed with default parameters using Seurat with 15 principal components. Ambient RNA levels were detected and corrected using SoupX.⁵⁶ Doublets were estimated and removed using DoubletFinder.⁵⁷ Transcriptionally similar clusters were identified using shared nearest neighbor (SNN) modularity optimization at a resolution of 0.8.

High-confidence VH consensus sequences were analyzed using IMGT High-V-QUEST (version 3.6.3; reference release 202030-2) to retrieve VDJ annotation and nucleotide CDR3 sequences.⁵⁸ Corresponding VDJ sequence annotation data were mapped back to known GEX barcode data for further analysis in Seurat.

Gene set enrichment analysis (GSEA) on single-cell data (clusters 0, 1, 2, 3, 4, 8 vs. FcRL5+ over FcRL5-negative differentially expressed genes) was performed using parameters of 0.1 minimum percent expression and adjusted p < 0.05.

HA ELISAs

Plasma from vaccinated subjects was serially diluted on EIA/RIA ELISA plates (Costar) previously coated with recombinant HA. HA-specific IgG antibodies were detected using HRP-conjugated anti-human IgG secondary antibodies (Jackson ImmunoResearch) and developed using ABTS with acid stop. Absorbance was measured at 415 nm using a SpectraMax M2 (Molecular Devices). Arbitrary units (AU) were calculated by normalizing OD values to a reference control serum included on each plate.

Quantification and statistical analysis

Comparisons between two matched groups (e.g., plasmablasts at day 0 vs. day 7 within the same subject) were performed using the Wilcoxon matched-pairs signed rank test for non-normally distributed variables. Comparisons between unmatched groups were performed using the Mann-Whitney test.

Strength and direction of association between two variables were assessed using the D'Agostino-Pearson normality test, followed by Spearman correlation testing with one-tailed t-testing.

Data were considered statistically significant at p ≤ 0.05. Statistical analyses were performed using GraphPad Prism version 7.0a or 8.0 (GraphPad Software). Scatterplots and other visualizations were generated using MATLAB (R2024b, The MathWorks Inc., Natick, MA) or R v4.41.