mRNA-based influenza vaccine expands the B cell response breadth in humans Authors: Hanover C Matz, Tae-Geun Yu, Kritika Dixit, Caroline Kikawa, Julian Q Zhou, Garazi Pena Alzua, Lowrey Peyton, Anders Madsen, Fangjie Han, Ariana Ghez Farrell, Robert Hoelzl, Aaron J Schmitz, Stephen C Horvath, Hunter K Keplinger, Benjamin S Strnad, Mark J Hoegger, William D Middleton, Michael K Klebert, Nina H Lin, Raffael Nachbagauer, Florian Krammer, Robert Paris, Jesse D Bloom, Jackson S Turner, Rachel M Presti, Jiwon Lee, Ali H Ellebedy Published: 2026 ABSTRACT AbstractConventional influenza virus vaccines induce antibody responses of limited breadth. Whether mRNA-based influenza virus vaccines can induce a superior germinal center (GC) response in humans remains unclear. Here we assessed B cell responses in an observational study of cohorts of healthy young adults receiving a licensed, split-virion or investigative mRNA-based quadrivalent seasonal influenza virus vaccine over two consecutive seasons. mRNA-based vaccines consistently elicited higher antibody titers and frequencies of memory B cells. In the draining lymph nodes, mRNA vaccination stimulated sustained GC reactions that persisted for at least 26 weeks after vaccination in 5 of 13 participants across the two seasons. Proteomic analysis of serum IgG repertoire showed that mRNA vaccination increased the number of vaccine-elicited serum IgG clonotypes and promoted intraclonal expansion within pre-existing clonotypes. B cell lineage analyses further indicated that expanded serum clonotypes map to GC B cell-associated sub-branches, consistent with ongoing GC-driven evolution underlying intraclonal expansion. This repertoire remodeling was accompanied by increased binding breadth against antigenically divergent influenza viruses. These findings reveal a key role for persistent GC responses in broadening the repertoire of vaccine-induced antibodies.Subject terms: RNA vaccines, Germinal centres Main Vaccines have reduced or eradicated the burden of many infectious diseases1. However, for rapidly evolving pathogens such as the seasonal influenza virus, designing fully efficacious vaccines that induce broadly neutralizing, durable antibody responses remains challenging2. One component of the adaptive immune response to target for improved vaccine efficacy is the germinal center (GC) reaction. GCs are microanatomical structures that form in secondary lymphoid organs after engagement of antigen and cognate B cells and T cells3, facilitate the Darwinian selection of high-affinity antigen-specific B cell clones and ultimately enhance antibody responses4,5. Additionally, GCs contribute to long-term protection by producing bone marrow plasma cells, which continuously secrete antibodies6,7, and memory B cells (MBCs), which rapidly differentiate into antibody-secreting plasmablasts (PBs) after antigen re-exposure8. Understanding how to design vaccines that effectively engage the GC reaction to promote both broadly neutralizing antibodies and prolonged immune memory would greatly advance our ability to combat antigenically variable pathogens. The influenza virus is a significant public health problem, causing 290,000–650,000 annual global deaths9. Seasonal influenza virus vaccines remain the most effective method for reducing disease burden, with vaccines typically targeting the glycoprotein hemagglutinin (HA) used for viral entry into host cells10. However, due to antigenic drift11,12, vaccines must be reformulated yearly13. Additionally, antigenic imprinting contributes to the generation of recall antibody responses that can be less effective against circulating viral strains14,15. These drawbacks underscore the need to determine how different vaccine platforms influence the GC response to influenza virus vaccination. The COVID-19 pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), demonstrated that lipid nanoparticle-encapsidated mRNA-based vaccination is an effective alternative to conventional protein-based and inactivated virus-based vaccines in a primary immune response16,17. However, influenza virus vaccination induces secondary recall responses, and it is not known whether mRNA-based vaccination in humans against influenza virus produces a superior GC response compared to conventional vaccines. In mice, potent GCs are induced by mRNA vaccines to influenza virus18,19. In the context of SARS-CoV-2, mRNA vaccines have been found to induce strong GC responses in mice and humans16,20. Memory responses play a critical role in SARS-CoV-2 booster immunizations21,22, resulting in GCs that can broaden the cross-reactivity of antigen-specific clones23. Furthermore, previous research has shown that both mRNA-based vaccines (to SARS-CoV-2) and inactivated virus-based vaccines (to influenza virus) can produce persistent GCs in humans, in some cases up to 6 months after vaccination16,17,24, but how these persistent GCs reshape the antigen-specific secreted IgG repertoire remains to be fully determined. We sought to characterize human B cell responses to an investigational mRNA-based quadrivalent seasonal influenza virus vaccine (mRNA-1010)25. Ultrasound-guided fine needle aspirations (FNAs) were used to directly sample the GC compartment in draining axillary lymph nodes. Our results demonstrated that mRNA-based vaccination induces robust recall MBC responses and sustained GCs compared to conventional split-virion vaccination. Proteomic analysis of serum IgG clonotypes (Ig-seq) revealed increased diversity of the serological IgG repertoires following mRNA-1010 vaccination, driven by both induction of new clonotypes and intraclonal expansion within pre-existing clonotypes. This was accompanied by increased binding breadth of circulating antibodies against influenza virus variants. Results mRNA-based seasonal influenza virus vaccination induces a robust humoral response We conducted an observational study (WU397) of 75 healthy adults (ages 20–50, median age 30) over the 2022–2024 Northern Hemisphere influenza virus seasons, with 38 participants receiving mRNA-1010 encoding the HA glycoproteins and 37 participants receiving split-virion quadrivalent influenza virus vaccine (Fluarix) corresponding to the recommended virus strains (Supplementary Table 1). Note that these vaccines include slightly different HAs, as the World Health Organization recommends different strains for egg-produced vaccines like Fluarix than vaccines that do not require viral production in eggs like mRNA-1010 (refs. 26,27). Vaccine formulations for the 2023–2024 Northern Hemisphere season were the same as for the 2022–2023 season, except for an update in the A(H1N1)pdm09 virus component (A/Wisconsin/67/2022 for cell culture- and recombinant-based vaccines and A/Victoria/4897/2022 for egg-based vaccines)28. Blood samples were collected at baseline and at 1, 2, 4, 8, 17 and 26 weeks after vaccination. A subset of these participants (mRNA-1010, n = 13; Fluarix, n = 15) enrolled for FNA specimens, collected at baseline and at 2, 8, 17 and 26 weeks after vaccination (Fig. 1a). HA-specific PB responses in the blood were measured by enzyme-linked immunosorbent spot (ELISpot) assay against all four vaccine HA proteins for participants in the 2022–2023 cohort. Frequencies of mean HA-specific IgG+ and IgA+ PBs per million peripheral blood mononuclear cells (PBMCs) peaked in both vaccine cohorts at 1 week after vaccination (Fig. 1b). At peak, participants receiving mRNA-1010 displayed significantly higher frequencies of IgG+ and IgA+ A/H1-specific and A/H3-specific PBs than participants who received Fluarix (Fig. 1c). IgM+ PB responses were greater in participants receiving Fluarix (Extended Data Fig. 1a,b). No significant difference was observed in frequencies of influenza B IgG+ and IgA+ HA-specific PBs between the cohorts (Fig. 1c). Correspondingly, we observed significantly higher fold changes in IgG plasma antibody binding titers to A/H1 and A/H3 as measured by enzyme-linked immunosorbent assay (ELISA) at 4 weeks after vaccination in participants receiving mRNA-1010 than in participants receiving Fluarix in the 2022–2024 influenza season (Fig. 1d). These significantly higher titers persisted up to the final time point collected (week 17 or 26). Significantly higher fold changes in IgG binding titers were observed to the influenza B virus HAs in the mRNA-1010 cohort at 4 weeks after vaccination, but not the final time point (Fig. 1d). In the 2022–2023 cohort, hemagglutination inhibition (HAI) titers against A/H1N1 and A/H3N2 increased relative to baseline at 4 weeks after vaccination, but no significant difference was observed in fold change of HAI titers at 4 weeks or 17/26 weeks after vaccination over baseline between mRNA-1010 and Fluarix recipients (Extended Data Fig. 1c). We detected significantly higher fold changes in circulating HA-specific MBCs by flow cytometric analysis (the gating strategy is provided in Extended Data Fig. 1d) in the 2022–2024 mRNA-1010 cohort at 4 and 17/26 weeks after vaccination over baseline (Fig. 1e). Overall, mRNA-1010 induced more robust peak antibody responses than split-virion vaccination. Humoral response to mRNA-based seasonal influenza virus vaccination. a, WU397 2022–2024 study design; LN, lymph node. b, ELISpot quantification of mean HA-binding IgG- and IgA-secreting PBs in blood in 2022–2023 participants receiving mRNA-1010 (red, n = 14) or Fluarix (black, n = 15). Numbers of HA-binding PBs were quantified against the four recombinant-based vaccine HAs and averaged; LoD, limit of detection. c, ELISpot quantification of HA-binding IgG- and IgA-secreting PBs at 1 week after vaccination in 2022–2023 participants receiving mRNA-1010 (red, n = 14) or Fluarix (black, n = 15). Horizontal bars represent geometric means. d, Plasma IgG titers at baseline, 4 and 17/26 weeks after vaccination (left) and fold change in plasma IgG titers at 4 and 17/26 weeks over baseline (right) against the four recombinant-based vaccine HAs in participants receiving mRNA-1010 or Fluarix in the 2022–2024 influenza season. Numbers on the left panels represent geometric mean titers for each time point, and numbers on the right panels represent mean fold change. e, Frequencies (top) and fold change over baseline (bottom) in HA-specific MBCs as a percentage of CD19+ cells in blood by flow cytometry for participants receiving mRNA-1010 or Fluarix in the 2022–2024 season. Numbers represent geometric mean values (top) and mean values (bottom). In c–e, P values were determined by two-tailed Mann–Whitney U-test. In d and e, for participants that did not complete a blood collection at week 26, samples from 17 weeks after vaccination were used for analysis (mRNA-1010, n = 4; Fluarix, n = 3); A/H1, A/H1N1; A/H3, A/H3N2; B/Yam, B/Yamagata/16/88-like lineage; B/Vic, B/Victoria/2/87-like lineage. Source data Humoral response to mRNA-1010. a, ELISpot quantification of mean HA-binding IgM-secreting PBs in blood in 2022-2023 mRNA-1010 (red, n = 14) and Fluarix (black, n = 15) participants. Numbers of HA-specific PBs were quantified against the four vaccine HAs and averaged. b, ELISpot quantification of HA-binding IgM-secreting PBs at 1 week post vaccination in 2022-2023 participants. Horizontal bars represent geometric means. P values determined by two-tailed Mann-Whitney U test. c, HAI titers for the four recombinant-based vaccine virus strains in 2022-2023 mRNA-1010 (red, n = 14) and Fluarix (black, n = 15) participants at baseline, 4, and 17/26 weeks. Numbers on panels represent geometric mean HAI titers for each time point. d, Flow cytometry gating strategy for HA-specific MBCs from PBMCs. Source data Sustained GC response to mRNA vaccination We next characterized the dynamics of GC responses in participants receiving mRNA-1010 or Fluarix. Both types of vaccine are injected into the deltoid muscle, which drains primarily to the lateral axillary lymph nodes. We used ultrasonography to identify and guide FNAs of accessible axillary lymph nodes on the side of immunization. The same lymph node was sampled in each participant at each time point. Flow cytometric analysis detected high proportions of HA-specific GC B cells in participants with active GC responses after vaccination with mRNA-1010 (Fig. 2a and Extended Data Fig. 2a). Across the 2022–2023 and 2023–2024 vaccination seasons, only in participants who received mRNA-1010 (5/13) did we detect HA-specific GC responses at 26 weeks (approximately 6 months) after vaccination (Fig. 2b). In neither season were sustained GC responses observed in participants who received Fluarix. This was remarkably different from previous studies24,29 in which participants had not received a seasonal influenza virus vaccine within the last 3 years, yet influenza virus-specific GCs were detected after administration of Flucelvax up to 13 weeks after vaccination (2/7, ~28% of participants). Participants in the current study were not excluded based on prior seasonal influenza virus vaccination history, suggesting that the high frequency of persistent GCs detected in participants receiving mRNA-1010 (~38%) was not impeded by prior responses, in contrast to the lack of persistent GCs in participants receiving Fluarix. GC response to mRNA-1010. a, Representative flow cytometry plots of BCL6 and HA-probe staining on CD3−CD19+CD20+IgDloBCL6+CD38int live singlet lymphocytes in FNA samples at baseline, 2, 8, 17 and 26 weeks after vaccination. Top, representative participant who received mRNA-1010 (red, 06); bottom, representative participant who received Fluarix (black, 05). b, Frequencies of HA-specific GC B cells determined by flow cytometry in FNA samples from participants who received mRNA-1010 (red, left) or Fluarix (black, right) in the 2022–2024 season. c, Representative uniform manifold approximation and projection (UMAP) plots of transcriptional clusters of B cells from baseline (sorted MBCs, FNA), 1 (sorted PBs), 2 (FNA), 8 (FNA), 17 (FNA) and 26 (sorted MBCs, FNA) weeks after vaccination (mRNA-1010 participant 29). Each dot represents a cell, colored by phenotype as defined by transcriptomic profile. Total numbers of cells are shown at the bottom right; ABC, activated B cell; GCB, GC B cell. d, UMAP plots of transcriptional clusters of B cells for all samples from participants receiving mRNA-1010 (28, 29) or Fluarix (05, 09) as in c, with HA-specific clones as determined by mAb ELISA mapped onto transcriptional clusters. Numbers of HA-specific cells (red) are shown at the bottom right, whereas numbers of HA-specific clones for all participants are shown in Supplementary Table 8; ND, no data. Source data Flow cytometry gating strategies for analysis of B cell responses in vaccinees. a, Flow cytometry gating strategy for HA-specific GC B cells from FNAs. b, Gating strategy for sorting PBs and MBCs from peripheral blood. PBs were sorted as CD3−CD19+IgDloCD20loCD38+ live singlet lymphocytes; MBCs were sorted as CD3−CD19+IgDlo live singlet lymphocytes. To further investigate the B cell response, single-cell RNA-sequencing (scRNA-seq) analysis was performed on week 0 and week 17 or 26 (depending on sample availability) sorted total MBCs from PBMCs, week 1 sorted total PBs from PBMCs and whole FNA samples from all sampled time points for four participants receiving mRNA-1010 (06, 17, 28 and 29) and three participants receiving Fluarix (05, 09 and 22) with detectable GC responses in the 2022–2023 cohort (gating for MBC and PB sorting is shown in Extended Data Fig. 2b). This allowed for identification of B cell subtypes (PB, GC, MBC and so on) based on transcriptomic profiles (Fig. 2c, Extended Data Fig. 3a–d and Supplementary Tables 2–7) and inference of B cell clonal relationships based on paired heavy and light chain B cell receptors (BCRs). BCRs from the same clone shared heavy and light chain V and J genes, complementarity-determining region 3 (CDR3) lengths and 85% similarity among the nucleotide sequences of their heavy chain CDR3. Identification of HA-specific B cell clones in lymph nodes and blood. a and c, UMAPs showing scRNA-seq transcriptional clusters of total cells (a) and of B cells (c) from PBs and MBCs sorted from blood and FNA of draining axillary lymph nodes combined. b, and d, Dot plots for the marker genes used for identifying annotated clusters in a and c, respectively. e, UMAP plots of transcriptional clusters of B cells for all samples from mRNA-1010 (06, 17) and Fluarix (22) participant with HA-specific clones as determined by mAb ELISA mapped onto transcriptional clusters as in Fig. 2c. Numbers of HA-specific cells (red) are shown at the bottom right. Numbers of HA-specific clones for all participants are shown in Supplementary Table 8. Source data Clonally distinct BCRs from the PB, GC B cell and lymph node plasma cell (LNPC) compartments were expressed as monoclonal antibodies (mAbs; n = 2,989) and tested for HA specificity by ELISA (Supplementary Table 8). HA-specific B cell clones were visualized by overlaying on the B cell subtypes, with most HA-specific BCRs derived from week 1 PBs (Fig. 2d and Extended Data Fig. 3e). Assignment of HA specificity to the scRNA-seq BCR data allowed us to analyze the clonal overlap between the GC and PB compartments in participants receiving mRNA-1010 and Fluarix. As the response to influenza virus vaccination is primarily a recall of HA-specific MBCs29,30, the week 1 PB compartment predominantly represents clones with prior antigen experience. B cell clones detected in both the GC (all time points) and week 1 PB compartment are thus most likely derived from MBCs. Most HA-specific GC B cell clones in all of the participants receiving mRNA-1010 (median = 87%) and one participant receiving Fluarix (median = 20%) overlapped with the week 1 PB compartment (Fig. 3a and Supplementary Table 9). Although most (>50%) GC clones in all of the participants who received mRNA-1010 overlapped with the week 1 PB compartment, some GC clones were not identified as part of the recall response. Although it is possible that recalled clones were simply not detected at our sequencing depth, it is also possible these clones represent naive primary responses being recruited into GCs and providing further antibody repertoire diversification. GCs induced by mRNA-1010 recruit low-frequency MBCs and increase SHM. a, Clonal overlap of sequences between PBs sorted from PBMCs 1 week after vaccination and GC B cells from all FNA time points among HA-specific clones for participants who received mRNA-1010 (06, 17, 28 and 29, top) and Fluarix (05, 09 and 22, bottom). Chord width corresponds to clonal population size; numbers of HA-specific clones are in Supplementary Table 9. Percentages are of GC B cell clones overlapping with PBs and PB clones overlapping with GC B cells; W1, week 1. b, Nucleotide mutation frequency in the immunoglobulin heavy chain variable gene (IGHV) and light chain variable gene (IGLV) regions for clonally related week 1 PBs and week 26 GC B cells (n = 42) from participants receiving mRNA-1010. P values were determined by two-tailed Wilcoxon matched pairs signed rank test; W26, week 26. c, Representative flow cytometry plots of IgG and IgA staining on CD3−CD19+CD20+IgDloBCL6+CD38intHA+ live singlet lymphocytes in FNA samples at 8 weeks after vaccination. Left, representative mRNA-1010 participant (red, 06). Right, representative Fluarix participant (black, 05). d, Frequencies of IgG+ (top) and IgA+ (bottom) HA-specific GC B cells determined by flow cytometry in FNA samples from participants receiving mRNA-1010 (red, n = 6) or Fluarix (black, n = 11) in the 2022–2023 influenza season. e, Proportions of isotypes of HA-specific cells from scRNA-seq data; BCR specificity was determined by mAb ELISA, and cells were identified in transcriptional scRNA-seq clusters from participants who received mRNA-1010 (06, 17, 28 and 29) or Fluarix (05, 09 and 22). Left, all B cells in the scRNA-seq data. Right, only GC B cells in the scRNA-seq data. Source data To determine if the persistent GCs identified in participants receiving mRNA-1010 altered somatic hypermutation (SHM) and affinity maturation in these HA-specific overlapping clones, we identified 21 pairs of clonally related BCRs present as week 1 PBs and week 26 GC B cells across three of the four participants receiving mRNA-1010 in the scRNA-seq data (no such paired clones were detected in participants who received Fluarix due to the absence of detectable HA-specific week 26 GC B cells in this cohort). In these 21 paired clones, we observed significantly increased immunoglobulin heavy chain variable gene (IGHV) and light chain variable gene (IGLV) nucleotide mutation frequencies in the week 26 GC B cell compartment compared to the week 1 PB compartment, indicating that these clones had undergone further SHM in GCs (Fig. 3b). However, we did not detect significant increases in their binding affinities by biolayer interferometry (BLI) or ELISA (Extended Data Fig. 4a,b). Analysis of germinal center antibody clones from mRNA-1010 participants. a, Equilibrium dissociation constant (KD) of Fabs (n = 42) derived from paired week 1 PB and week 26 GC B cell clones of mRNA-1010 participants interacting with immobilized HA protein measured by BLI. b, Area under the curve (AUC) values for mAbs (n = 42) derived from paired week 1 PB and week 26 GC B cell clones of mRNA-1010 participants as measured by binding ELISA against HA protein. In a and b, clones were tested against HA strains based on predetermined binding specificity by mAb ELISA; clones with multiple binding specificities were tested individually against each HA strain. c, Phylogenetic trees of representative HA-specific IgA GC B cell clones from scRNA-seq data of mRNA-1010 participants. Left (397-29) represents a clone likely derived from an IgA+ MBC; right (397-06) represents a clone likely derived from an IgG+ MBC that class switched to IgA. IGHA, immunoglobulin heavy chain alpha constant region; IGHG, immunoglobulin heavy chain gamma constant region. PB, plasmablast; MBC, memory B cell; GC, germinal center B cell; LNPC, lymph node plasma cell. Source data We further examined the flow cytometry data for evidence that low-frequency MBCs may be engaged in mRNA-1010 vaccine-induced GC responses. Among HA-specific GC B cells in FNA samples, we observed detectable frequencies of both IgG+ and IgA+ cells in participants who received mRNA-1010, whereas the corresponding populations in participants who received Fluarix were predominantly IgG+ (Fig. 3c). Analysis of the kinetics of GC responses demonstrated that IgA+ HA-specific GC B cells were detectable in participants who received mRNA-1010 in the 2022–2023 cohort up to 26 weeks after vaccination (Fig. 3d). Consistent with this observation, IgA+ GC B cells were detected by scRNA-seq in participants receiving mRNA-1010 but not in participants receiving Fluarix (Fig. 3e). Phylogenetic analysis of the IgA clones suggested they potentially arose from both IgA+ MBCs and IgG+ MBCs that had class switched to IgA (Extended Data Fig. 4c). As the ELISpot PB data demonstrated that IgA+ HA-specific PBs displayed, on average, a fivefold lower frequency than IgG+ HA-specific PBs (Fig. 1c), and other studies suggest that IgA+ influenza virus-specific MBCs are lower frequency than IgG+ MBCs29,31, these results collectively suggest that mRNA-1010 engages low-frequency MBCs in GC responses. Higher number of vaccine-elicited clonotypes in serum following mRNA vaccination Given the sustained GC responses and engagement of low-frequency MBCs elicited by mRNA-1010, we next examined how these cellular dynamics reshape the secreted IgG repertoire at the clonotypic level using Ig-seq integrated with high-throughput sequencing of BCRs (BCR-seq32; Fig. 4a). For in-depth analysis, we focused on the serological IgG repertoire specific to the A/H3 component of the vaccine (A/Darwin/9/2021) as A/H3N2 is typically the most antigenically variable vaccine component year to year, and we observed a pronounced difference in A/H3-specific plasma binding titers between the two vaccine cohorts (Fig. 1d). We analyzed 16 participants by Ig-seq, including 8 who received mRNA-1010 (06, 14, 17, 21, 23, 28, 29 and 31) and 8 who received Fluarix (05, 09, 19, 20, 22, 24, 27 and 32) from the 2022–2023 cohort. A/H3-specific serum IgG was affinity purified from sera collected at weeks 0, 4 and 17/26 and analyzed by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Spectra were assigned by matching CDRH3 peptides to participant-specific BCR databases generated by BCR-seq. mRNA-1010 diversifies the serological repertoire and promotes intraclonal expansion. a, Experimental design and analytical workflow integrating proteomic analysis of serum immunoglobulins (Ig-seq) with high-throughput sequencing of B cell transcripts (BCR-seq) to delineate the A/H3-specific serological IgG repertoire; NGS, next-generation sequencing. b, Representative heat maps showing the relative serum abundance of IgG clonotypes comprising the A/H3-specific serological repertoire across time points (397-31, mRNA-1010; 397-27, Fluarix). Each column represents an individual IgG clonotype, with color indicating the relative amount in serum (arbitrary units (AU)), calculated by scaling the anti-A/H3 serum IgG binding titer by the relative abundance of each clonotype determined by proteomic analysis. Clonotypes are classified as pre-existing (detected at week 0) or vaccine elicited (not detected at week 0). c, Number of serum IgG clonotypes detected at each time point for individual participants (mRNA-1010, n = 8; Fluarix, n = 8). Lines connect longitudinal samples within participants. d, Number of vaccine-elicited A/H3-specific IgG clonotypes detected in week 4 serum. Each data point represents an individual participant (mRNA-1010, n = 8; Fluarix, n = 8). e, BASELINe-estimated selection pressure (Σ) for vaccine-elicited clonotypes in CDRs (top) and framework regions (FWRs; bottom). Positive Σ values indicate antigen-driven positive selection. f, Representative examples of intraclonal CDRH3 expansion within pre-existing clonotypes (397-06, mRNA-1010; 397-05, Fluarix). Newly detected postvaccination CDRH3 peptide variants are shown, with amino acid replacements highlighted in red. Asterisks (*) indicate sequences identified in the week 1 PB compartment by scRNA-seq; Pre-vax, before vaccination; Post-vax, after vaccination. g, Intraclonal CDRH3 diversity fold change for pre-existing clonotypes, calculated as the ratio of unique CDRH3 peptides detected after vaccination (post-vax) to those detected before vaccination (pre-vax). Each data point represents an individual clonotype (mRNA-1010, n = 495; Fluarix, n = 450) pooled across participants within each group. h, Fraction of pre-existing clonotypes showing intraclonal expansion after vaccination (mRNA-1010, n = 8; Fluarix, n = 8). i, Relative abundance of expanded clonotypes among the pre-existing clonotypes across time points for individual participants. Lines connect longitudinal samples within participants (mRNA-1010, n = 8; Fluarix, n = 8). j, Lineage trees of GC-associated pre-existing IgG clonotypes. Tips are colored by serum IgG detection status (detected before vaccination, after vaccination or not detected in serum). GC B cell sequences identified by scRNA-seq are marked with red circles. Branch lengths correspond to SHM per site; UCA, unmutated common ancestor. For c, d, h and i, each data point represents an individual participant, and horizontal bars indicate means (c, d, h and i) or geometric mean (g) with the values shown. P values were determined by a two-sided Mann–Whitney U-test (c, d, h and i), by two-sided comparison of BASELINe posterior probability density functions using the testBaseline function34 (e) or by a two-sided linear mixed-effects model with vaccination type as a fixed effect and donor as a random effect (g), with degrees of freedom estimated using Satterthwaite’s approximation67. Panel a created in BioRender; Lee, J. https://biorender.com/s0otztv (2026). Source data We subsequently categorized individual IgG clonotypes as ‘pre-existing’ or ‘vaccine-elicited’ based on their presence or absence in prevaccination serum (week 0), respectively (Fig. 4b). We note that ‘vaccine-elicited’ does not imply arising from naive responses but rather denotes clonotypes that became newly detectable in serum after vaccination, likely reflecting activation of previously established MBCs differentiating into antibody-secreting cells in the context of recall responses8. Pre-existing clonotypes comprised most of the week 4 A/H3-specific serum IgG repertoire in both cohorts (83.9% in mRNA-1010 and 86.2% in Fluarix; Extended Data Fig. 5a), consistent with recall-dominated seasonal influenza vaccination responses8,29,33. Delineation of A/H3−specific serum IgG clonotypes and CDRH3 diversification following vaccination. a, Relative abundance of pre-existing serum IgG clonotypes at week 4 post vaccination. b, Number of pre-existing A/H3-specific IgG clonotypes detected at week 4 post vaccination. c, Number of vaccine-elicited A/H3-specific IgG clonotypes detected at week 17/26 post vaccination. d and e, Number of unique intraclonal CDRH3 peptide variants detected after vaccination for vaccine-elicited (d) or pre-existing (e) serum IgG clonotypes. f, CDRH3 diversity index of the A/H3-specific serological repertoire (effective number of species; exp[Shannon entropy]) across time points (weeks 0, 4, and 17/26). Lines connect longitudinal samples within participants with values above indicating cohort means at each time point. In a-c and f, each data point represents an individual participant (mRNA-1010, n = 8; Fluarix, n = 8), with horizontal bars indicating means (values shown above). In d and e, each data point represents an individual clonotype pooled across participants within each cohort (mRNA-1010, n = 475; Fluarix, n = 258 for d; mRNA-1010, n = 495; Fluarix, n = 450 for e), with horizontal bars indicating geometric means (values shown above). P values were determined by two-sided Mann-Whitney U test (c and f), or by a two-sided linear mixed-effects model with vaccination type as a fixed effect and donor as a random effect (e), with degrees of freedom estimated using Satterthwaite’s approximation67. Source data Across participants, we identified between 67 and 186 total A/H3-specific serum IgG clonotypes per individual, with the mRNA-1010 cohort exhibiting greater clonotypic diversity at weeks 4 and 17/26 than the Fluarix cohort (Fig. 4c). Despite similar contributions of pre-existing clonotypes in terms of abundance, mRNA-1010 induced a significantly greater number of vaccine-elicited clonotypes at week 4 (Fig. 4d), whereas the number of pre-existing clonotypes was comparable between cohorts (Extended Data Fig. 5b). Although this increase diminished over time, it remained modestly higher at week 17/26 (Extended Data Fig. 5c). Notably, vaccine-elicited clonotypes showed stronger evidence of antigen-driven selection, with increased positive selection in CDRs and no corresponding increase in framework regions, as assessed by BASELINe34 (Fig. 4e). Together, these data indicate that mRNA-1010 increases the clonotypic diversity of the secreted IgG repertoire, consistent with enhanced recruitment and selection of diverse MBC-derived lineages in the context of sustained GC responses. Intraclonal expansion of pre-existing serum IgG clonotypes after mRNA vaccination Beyond changes in the number of clonotypes, we next asked whether vaccination altered diversification within clonotypes. To quantify this diversification, we enumerated, for each pre-existing serum IgG clonotype, the number of unique CDRH3 peptide variants identified by Ig-seq (intraclonal CDRH3 diversity). Notably, among highly abundant individual pre-existing clonotypes identified in serum after vaccination, many clonotypes from participants who received mRNA-1010 were drastically expanded with increased diversity in the CDRH3 peptides within single IgG clonotypes (Fig. 4f,g). By contrast, pre-existing clonotypes identified in participants who received Fluarix largely remained unchanged in intraclonal CDRH3 diversity at week 4. Several of these newly detected CDRH3 peptide variants could also be identified in the week 1 PB compartment of our scRNA-seq data. Given that week 0 CDRH3 peptides of pre-existing clonotypes represent constitutively secreted IgG from bone marrow plasma cells elicited by prior exposure, this intraclonal CDRH3 expansion could arise from reactivation of clonally related MBCs and, potentially, continued diversification through GC re-entry and selection24,29,35. When examining CDRH3 composition within each serum IgG clonotype, vaccine-elicited clonotypes exhibited comparable intraclonal CDRH3 diversity between cohorts (Extended Data Fig. 5d), whereas pre-existing IgG clonotypes from participants who received mRNA-1010 displayed higher intraclonal CDRH3 diversity after vaccination (Extended Data Fig. 5e). The frequency of pre-existing clonotypes exhibiting intraclonal expansion (defined by a higher number of unique CDRH3 peptides at week 4 than at week 0) was substantially higher in recipients who received mRNA-1010, with an approximately threefold higher frequency, on average, than Fluarix (Fig. 4h). Moreover, pre-existing clonotypes with expanded CDRH3 peptides contributed more prominently to the postvaccination serum repertoire in participants who received mRNA-1010 than participants who received Fluarix, and this pattern was maintained through the week 17/26 time point (Fig. 4i). The serum IgG repertoire from participants who received mRNA-1010 showed a marked increase in the CDRH3 diversity index, measured as the effective number of species (exponential of Shannon entropy), incorporating both the number and abundance of unique CDRH3 peptides at each time point36–38 (Extended Data Fig. 5f). Together, these data indicate that mRNA-1010 not only increases the number of serum clonotypes but also updates pre-existing IgG clonotypes by eliciting a broader spectrum of intraclonal variants that constitute a major fraction of the postvaccination serum IgG response. GC-linked lineage structure underlies intraclonal expansion To further investigate whether intraclonal expansion detected in serum reflected GC-associated evolution, we performed B cell clonal lineage analysis on A/H3-specific serum IgG clonotypes exhibiting expanding CDRH3 diversity and clonally related to GC B cells identified in the scRNA-seq data. Using these criteria, we identified three IgG clonotypes from participants who received mRNA-1010 and none from participants who received Fluarix, likely reflecting the rarity of such clonotypes39,40 together with the low frequency of IgG+ HA-specific GC B cells in Fluarix (Fig. 3d). For these clonotypes, we inferred clonal lineage trees by integrating BCR-seq data with scRNA-seq data and mapped CDRH3 peptides detected in serum by Ig-seq onto lineage members (Fig. 4j). These pre-existing clonotype lineages contained discrete branches corresponding to CDRH3 peptide detected before vaccination (week 0) and expanded CDRH3 peptide variants detected after vaccination. Importantly, GC B cell sequences identified by scRNA-seq (red circles), spanning weeks 8 to 26 after vaccination, localized to sub-branches corresponding to postvaccination CDRH3 peptide variants, providing a direct connection between sustained GC responses after mRNA-1010 vaccination and the emergence of diversified, clonally related IgG variants in the circulating serological repertoire. Increased antibody binding and neutralization breadth as a result of mRNA vaccination To test whether diversification of the serological repertoire was linked to the magnitude of the humoral response, we first compared repertoire remodeling measured by Ig-seq with bulk anti-A/H3 serology. The fraction of newly detected CDRH3 peptide variants after vaccination, reflecting the combined contribution of vaccine-elicited clonotypes and intraclonal diversification, positively correlated with the change in anti-A/H3 serum titer at week 4 relative to week 0 (Spearman’s ρ = 0.694; Fig. 5a). Antibody binding and neutralization breadth after mRNA vaccination. a, Correlation between the change in serum anti-A/H3 titer at 4 weeks after vaccination relative to week 0 and the fraction of newly detected CDRH3 peptide variants after vaccination, quantified by Ig-seq. Each data point represents an individual participant. ρ indicates Spearman’s correlation coefficient. The trend line represents an ordinary least squares linear regression fit, and the shaded area indicates the 95% confidence interval for the regression line. The P value was determined by a two-sided Spearman rank correlation test. b,c, Median log2 (fold change) in MFI of binding of plasma samples at 4 weeks after vaccination over baseline for participants in the 2022–2023 cohort who received mRNA-1010 (red, n = 14) or Fluarix (black, n = 15) to beads coated with a panel of A/H1 (b) and A/H3 (c) glycoproteins. d,e, Median log2 (fold change) in neutralization titer of plasma samples at 4 weeks after vaccination over baseline for participants in the 2022–2023 cohort treated with mRNA-1010 or Fluarix. For b–e, homologous vaccine virus strains are indicated in bold. Shaded areas represent interquartile ranges. P values were determined by Mann–Whitney U-test. To control for multiple hypothesis testing, a two-stage step-up false discovery rate correction was applied to the P values; *q < 0.05. Source data To determine if the increased diversification of the serological repertoire induced an increase in the cross-reactivity of the plasma antibody response in the 2022–2023 season, we next measured the binding of plasma samples against antigenically diverse A/H1N1 and A/H3N2 influenza virus HA proteins coated on multiplex fluorescent beads and covering more than 50 years of viral evolution41–44. Comparing log2 (fold change) in median fluorescence intensity (MFI) over baseline (week 0), we observed significantly higher fold changes in plasma antibody binding at 4 weeks after vaccination in the mRNA-1010 cohort against 9 A/H1 and 12 A/H3 strains tested (Fig. 5b,c). At 26 weeks after vaccination, significantly higher antibody binding was not observed for the A/H1 strains but persisted in the mRNA-1010 cohort against the antigenically variable A/H3 strains (Extended Data Fig. 6a,b). Additionally, we tested the cross-reactivity of the individual mAbs derived from the four participants treated with mRNA-1010 (06, 17, 28 and 29) and three participants treated with Fluarix (05, 09 and 22) previously determined to be A/H1 and/or A/H3 specific by ELISA against panels of seven group 1 (A/H1 and A/H5) and seven group 2 (A/H3 and A/H7) influenza virus HA proteins representing multiple decades of viral evolution41–46. Measuring the fold change in MFI relative to a negative-control antigen, SARS-CoV-2 receptor binding domain (RBD), mAbs from both mRNA-1010 and Fluarix recipients exhibited cross-reactivity across group 1 and group 2 HAs, including A/H5N1 and A/H7N9 avian influenza virus HAs (Extended Data Fig. 6c,d). Additionally, mAbs derived from the mRNA-1010 cohort could be binned for positive HA binding based on an MFI of greater than three times the binding MFI of the respective mAb to SARS-CoV-2 RBD. Neither A/H1- nor A/H3-specific mAbs that bound from one to six variant HA strains exhibited statistically significant differences in IGHV nucleotide mutation frequency (Extended Data Fig. 6e,f). However, mAbs specific for A/H3 that bound all six A/H3 proteins and A/H7 exhibited significantly higher levels of SHM than those mAbs lacking cross-reactivity across the full group 2 HA panel (Extended Data Fig. 6f). Those mAbs specific to A/H1 that recognized A/H5 trended toward higher SHM but lacked statistical significance, likely due to a more limited sample size (Extended Data Fig. 6e). These data suggest that rates of SHM may play a role in expanding antibody binding breadth. Antibody binding breadth induced by mRNA vaccination. a and b, Median log2 fold change in MFI of binding of plasma samples at 17/26 weeks post vaccination over baseline for 2022-2023 mRNA-1010 (red, n = 14) and Fluarix (black, n = 15) participants to beads coated with a panel of A/H1 (a) and A/H3 (b) glycoproteins. Samples collected at 17 weeks post vaccination were used for patients that did not complete a blood draw at 26 weeks (mRNA-1010, n = 2, Fluarix, n = 2). Shaded areas represent interquartile ranges. P values were determined by two-tailed Mann-Whitney U test. To control for multiple hypothesis testing, a two-stage step-up false discovery rate correction was applied to the P values (*q < 0.05). c and d, Fold change in binding MFI of individual mAbs tested against Group I (c) and Group II (d) HAs relative to binding MFI to SARS-CoV-2 RBD. A/H1-specific mAbs tested, mRNA-1010 (n = 264) and Fluarix (n = 74); A/H3-specific mAbs tested, mRNA-1010 (n = 349) and Fluarix (n = 34). Bars represent median values. For a-d, homologous vaccine virus strains are indicated in bold. e and f, IGHV nucleotide mutation frequency of individual mRNA-1010 mAbs binned by positive MFI binding ( > 3 times respective MFI binding to SARS-CoV-2 RBD) of Group I (n = 263) (e) and Group II (n = 344) (f) HAs. P values determined by two-tailed Mann Whitney U test. Source data Although the results of the bead assay indicated an increase in antibody binding breadth to A/H1N1 and A/H3N2 at the peak of the response, we wanted to determine if there was a concomitant increase in neutralization breadth. Testing plasma antibody neutralization at the peak of the response and the final time point collected against homologous (A/Darwin/9/2021) and heterologous (A/Thailand/8/2022) A/H3N2 viruses in a conventional microneutralization assay, we observed an increase in neutralization titers to both viruses in the 2022–2023 mRNA-1010 and Fluarix cohorts (Extended Data Fig. 7a,b). However, there was no significant difference in the fold change over baseline in mRNA-1010 neutralization titers compared to Fluarix for either the homologous or heterologous A/H3N2 viruses (Extended Data Fig. 7c,d). As using this technique was logistically difficult to compare many virus strains, we turned to a high-throughput sequencing-based assay to measure neutralization breadth against multiple A/H1N1 and A/H3N2 viruses47–49, including a number of viruses that circulated in late 2023 during the timeframe the vaccine was designed to cover. We observed a significantly higher increase in plasma neutralization titers in the mRNA-1010 cohort at 4 weeks after vaccination against 11 of 13 A/H1N1 viruses tested (Fig. 5d). We also tested plasma neutralization against 83 different A/H3N2 viruses. The mRNA-1010 cohort exhibited higher increases in plasma neutralization titers against most strains at 4 weeks after vaccination, although these titer increases did not achieve statistical significance (Fig. 5e). Similar results were observed in plasma neutralization titers against A/H1N1 and A/H3N2 viruses for the final time point collected (Extended Data Fig. 7e,f). Collectively, our data demonstrate that mRNA-1010 vaccination induces greater diversification of the serological repertoire, which translates to higher total serum binding titers and increased binding breadth against diverse influenza virus strains. Neutralization breadth of plasma antibody titers after mRNA vaccination. a and b, Microneutralization titers of 2022-2023 mRNA-1010 (n = 14) and Fluarix (n = 15) participants against A/Darwin/9/2021 (homologous, a) and A/Thailand/8/2022 (heterologous, b) A/H3N2 viruses. Numbers represent geometric mean titers. c and d, Fold change in microneutralization titers at 4 weeks and 17/26 weeks post vaccination. Numbers represent average fold change. e and f, Median log2 fold change in neutralization titer of plasma samples at 17/26 weeks post vaccination. For e and f, homologous vaccine virus strains are indicated in bold. Shaded area represents interquartile ranges. P values were determined by two-tailed Mann-Whitney U test. To control for multiple hypothesis testing, a two-stage step-up false discovery rate correction was applied to the P values. For a-f, samples collected at 17 weeks post vaccination were used for patients that did not complete a blood draw at 26 weeks (mRNA-1010, n = 2, Fluarix, n = 2). Source data Discussion We observed that mRNA-based seasonal influenza vaccination is an effective alternative to a licensed split-virion vaccine. In participants receiving mRNA-1010, vaccination resulted in higher peak and final time point antibody titers to A/H1N1 and A/H3N2 seasonal influenza viruses over two vaccination seasons. We observed significantly higher peak antibody titers for influenza A and B virus HAs across the 2022–2024 vaccination seasons in participants that received mRNA-1010 than in participants that received a currently licensed standard-dose vaccine, Fluarix. These results agree with the findings of phase III randomized clinical trials for mRNA-1010. Initial trials demonstrated noninferiority for influenza A virus strains but not influenza B virus strains50, but optimizing the mRNA-1010 formulation improved immune responses to influenza B virus strains51. Subsequently, in a phase III trial, mRNA-1010 met prespecified noninferiority success criteria for immunogenicity endpoints versus licensed standard-dose and high-dose vaccines51. Additionally, we observed that mRNA-1010 induced persistent GCs up to 26 weeks after vaccination, whereas no persistent GCs were detected in individuals who received Fluarix. Previous studies from our group demonstrated sustained GCs 13 weeks after vaccination with Flucelvax, but these participants had not received an influenza virus vaccine within 3 years (ref. 24). The higher frequency of persistent GCs in participants who received mRNA-1010 than in participants who received Fluarix suggests that the response generated by mRNA-1010 is not impeded by pre-existing immunity. This is a potential benefit of mRNA-based seasonal influenza virus vaccination, as pre-existing antibody titers have long been suggested to limit the effectiveness of annual vaccination52. Although the antibody titers observed in this study did not result in significant differences in HAI between 2022–2023 cohorts, our data suggest that mRNA-1010 better engages a diverse pool of low-frequency MBCs, leading to remodeling of the secreted, HA-specific serum IgG repertoire. We observed increased clonotypic representation after mRNA-1010 vaccination, including a higher number of vaccine-elicited serum IgG clonotypes, together with pronounced intraclonal CDRH3 expansion within pre-existing clonotypes. Lineage analyses further linked expanded serum members to GC B cell-associated sub-branches, supporting the notion that sustained GC activity contributes to the emergence of further diversified IgG variants in circulation. Collectively, these processes result in greater breadth of the influenza virus HA-specific serum IgG response. We observed greater binding breadth of mRNA-1010-induced antibodies at both a polyclonal plasma level and across individual mAbs, with evidence of increased SHM contributing to diversification within the repertoire. Increased binding breadth has been observed for individual antibodies specific to SARS-CoV-2 in the context of mRNA vaccine booster immunizations23 and the plasma antibodies of individuals who received mRNA vaccines compared to SARS-CoV-2 infection53. Combined with the results of our study, these data broadly point to the value of mRNA-based vaccines in stimulating broad binding immunity to pathogen variants. Furthermore, additional studies using a multivalent mRNA vaccine against 20 known influenza virus subtypes in animal models54 and sequential immunization of mRNA lipid nanoparticle vaccines in mice55 demonstrate the capacity of mRNA vaccines for generating optimal cross-protection. In our study, boosting of antibody titers to future strains of influenza virus A/H1N1 and A/H3N2 by mRNA-1010 was observed, suggesting that mRNA-based vaccines have the potential to overcome the constant problem of influenza virus antigenic drift10. Although increased neutralization breadth was not statistically significant for antigenically variant A/H3N2 strains, likely due to sample size limitations, other effector functions beyond neutralization are possible for circulating antibodies, including opsonization and phagocytosis, complement activation and antibody-dependent cellular cytotoxicity56. Significant improvements in binding breadth could expand these protective functions across variant influenza virus strains, providing additional mechanisms to combat antigenic drift. The mechanisms underlying the robust humoral response and persistent GCs induced by mRNA-1010 are still not fully elucidated. It is possible that mRNA vaccines deliver greater amounts of antigen to the draining lymph nodes, resulting in both the stimulation of low-frequency MBCs and prolonged antigen duration to drive GC persistence. It is also possible that the self-adjuvanting properties of lipid nanoparticle mRNA vaccines57,58 robustly activate either professional antigen-presenting cells and/or B cells for a more potent immune response. The form of antigen presented after mRNA-1010 vaccination (that is, membrane-bound) may also contribute to increased antigen valency and thus stimulation of low-frequency and/or low-affinity MBCs. As an mRNA-based vaccine produces HA antigens directly in human cells after administration, it is also possible that the antigens more accurately match the glycosylation patterns of HA antigens produced by influenza virus infection. This would be an advantage over chicken egg-based vaccines such as Fluarix and other cell-based vaccines, which rely on non-human cells for antigen production and may contribute to the improved binding breadth of the antibodies induced by mRNA-1010. Differences in N-linked glycosylation can affect the immune response to the influenza virus59, and different vaccine formulations can alter the structure of HA antigens60. Broadly, our results demonstrate that although there is a biological ceiling to antibody affinity with repeat antigen exposure, MBCs may re-engage in GCs for further SHM to diversify the antibody repertoire. Sustained GCs stimulated by mRNA-1010 vaccination potentially enhance this expansion of the MBC repertoire by increasing the opportunity for clones to undergo repeated rounds of mutation. This diversification may provide anticipatory mutations to combat rapidly evolving pathogens such as influenza virus61,62. Increased mutation frequencies in antibodies specific to SARS-CoV-2 have been observed to improve resistance to antigenic mutation in the SARS-CoV-2 RBD63 as well as increasing neutralization potency64. Increased mutation can also improve the neutralization breadth of antibodies specific to human immunodeficiency virus65. The functional benefit of MBC lineages diversified by SHM is apparent based on our Ig-seq and serological binding breadth analyses. The presence of pre-existing clonotypes with expanded CDRH3 diversity and abundant newly detectable secreted vaccine-elicited clonotypes in the serological repertoire of participants who received mRNA-1010 but not in those that received Fluarix demonstrates that the diversification of the available pool of secreted antibodies likely contributes to the improved binding breadth observed in the responses of the mRNA-1010 cohort. Thus, it is possible that repeat doses over multiple influenza seasons of a vaccine such as mRNA-1010 may broaden the influenza virus-specific MBC repertoire via SHM in persistent GCs, in contrast to conventional inactivated vaccines, which, lacking broad MBC stimulation or sustained GCs, may narrow the antigenic landscape of recall responses. Although much of our data support the role of broader recall of MBCs after mRNA-1010 vaccination, our GC sequencing and Ig-seq data suggest that there is the potential for recruitment of primary responses to additionally diversify the antibody repertoire. Clones not identified in the recall response but present in the GC or as vaccine-elicited clones in serum can potentially contribute to further increases in antibody binding breadth. As a vaccine that elicits both strong protection against antigenically drifting seasonal influenza virus strains and broad binding against divergent strains is recommended for developing a universal influenza virus vaccine66, the results suggest that mRNA-based vaccines would greatly contribute to advancing this goal. Methods Experimental model and study participant details Study WU397 was approved by the Institutional Review Board of Washington University in St. Louis (2208058), and written consent was obtained from all participants. Seventy-five participants were enrolled. Participants were aged 20–50 years old. Forty-one participants were women, and 34 participants were men. No statistical methods were used to predetermine sample size. Investigators were not blinded to experiments and outcome assessments. Prior vaccination against the influenza virus in previous seasons was not included as an exclusion criterion. Participants had not received a seasonal influenza virus vaccine within the last 6 months and had not received any vaccines within the last 28 days. Blood samples were obtained by standard phlebotomy. PBMCs were isolated using Vacutainer CPT tubes (BD); the remaining red blood cells were lysed with ammonium chloride lysis buffer (Lonza). Cells were immediately used or cryopreserved in 10% DMSO in fetal bovine serum (FBS). Three attending radiologists with expertise in ultrasound performed the ultrasound-guided FNA of axillary lymph nodes. Ultrasound of the axilla was performed with a high-frequency (11- to 14-MHz) linear transducer to identify the most accessible lateral axillary node closest to the central axillary vessels. Characteristic features and dimensions of the selected baseline node were documented in annotated images and cine clips, including size, vascularity, distance to central axillary vessels and distance to skin, to facilitate identification and FNA of the same lymph node at follow-up. Six passes, each consisting of approximately 60 needle throws, were made using 25-gauge needles, which were flushed with 3 ml of RPMI 1640 supplemented with 10% FBS and 100 U ml−1 penicillin/streptomycin, followed by three 1-ml rinses. Red blood cells were lysed with ammonium chloride lysis buffer (Lonza). Cells were washed twice with PBS supplemented with 2% FBS and 2 mM EDTA and immediately used or cryopreserved in 10% DMSO in FBS. Vaccines and antigens Fluarix quadrivalent influenza virus vaccine (Northern Hemisphere 2022–2023 and 2023–2024 seasons) was purchased from GlaxoSmithKline Biologicals. Seasonal quadrivalent mRNA-1010 vaccine (Northern Hemisphere 2022–2023 and 2023–2024 seasons) was provided by Moderna. For ELISpot assays, 293F-expressed recombinant HA proteins derived from A/H1N1 (A/Wisconsin/588/2019), A/H3N2 (A/Darwin/6/2021), B/Yamagata/16/88-like lineage (B/Phuket/3073/2013) and B/Victoria/2/87-like lineage (B/Austria/1359417/2021) were provided by Moderna. For ELISAs, 293F-expressed recombinant HA proteins were provided by Moderna (A/Wisconsin/588/2019, B/Phuket/3073/2013) or were purchased from SinoBiological (A/Wisconsin/67/2022, 40940-V08H2; A/Darwin/6/2021, 40868-V08H; B/Austria/1359417/2021, 40862-V08H). For flow cytometry staining, 293F-expressed recombinant HA proteins (A/Wisconsin/67/2022, 40940-V08H2; A/Wisconsin/588/2019, 40787-V08H1; A/Darwin/6/2021, 40868-V08H; B/Austria/1359417/2021, 40862-V08H; B/Phuket/3073/2013, 40498-V08H1) were purchased from SinoBiological. Recombinant HA was biotinylated using an EZ-Link Micro NHS-PEG4-Biotinylation kit (Thermo Scientific); excess biotin was removed using 7-kDa Zeba desalting columns (Pierce). For BLI, recombinant 6×His-tagged HA proteins from influenza virus strains (A/Wisconsin/588/2019, IT-003-00117ΔTMp; A/Darwin/6/2021, IT-003-00443ΔTMp; B/Phuket/3073/2013, IT-003-B11p; B/Austria/1359417/2021, IT-003-B23ΔTMp) expressed in 293F cells were purchased from Immune Technology. For Ig-seq, recombinant 6×His-tagged HA protein from H3N2 (A/Darwin/9/2021, 40859-V08H1) expressed in 293F cells was purchased from SinoBiological. For multiplex fluorescence bead assays, 293F-expressed recombinant HA proteins from the following influenza virus strains were purchased from Immune Technology: A/H3N2 strains: A/Thailand/8/2022 (IT-003-00445ΔTMp), A/Darwin/6/2021 (IT-003-00443ΔTMp), A/Cambodia/e0826360/2020 (IT-003-00441ΔTMp), A/South Australia/34/2019 (IT-003-00438ΔTMp), A/Hong Kong/2671/2019 (IT-003-00440ΔTMp), A/Singapore/INFIMH-16-0019/2016 (IT-003-00434ΔTMp), A/Switzerland/8060/2017 (IT-003-00435ΔTMp), A/Victoria/361/2011 (IT-003-00423∆TMp), A/Brisbane/10/2007 (IT-003-0042ΔTMp) and A/Fujian/411/2002 (IT-003-0046∆TMp); A/H1N1 strains: A/Victoria/4897/2022 (IT-003-00120ΔTMp), A/Wisconsin/67/2022 (IT-003-00121ΔTMp), A/Sydney/5/2021 (IT-003-00119ΔTMp), A/Wisconsin/588/2019 (IT-003-00117ΔTMp), A/Guangdong-Maonan/SWL1536/2019 (IT-003-00115ΔTMp), A/Michigan/45/2015 (IT-003-00105∆TMp), A/California/04/2009 (IT-003-00106∆TMp), A/New_Caledonia/20/1999 (IT-003-001ΔTMp), A/Puerto_Rico/8/1934 (IT-003-0010∆TMp); A/H7N9 strain: A/Shanghai/1/2013 (IT-003-0073ΔTMp) and A/H5N1 strain: A/Hubei/1/2010 (IT-003-00519∆TMp). Additionally, 293F-expressed recombinant HA proteins from the following influenza virus strains were purchased from eEnzyme: A/H3N2 strains: A/Texas/50/2012 (IA-H3-050WP), A/Perth/16/2009 (IA-H3-0045t) and A/Hong Kong/1/1968 (IA-H3-00418t); A/H1N1 strains: A/Hawaii/70/2019 (IA-H1-H19WP) and A/Brisbane/2/2018 (IA-H1-B18WP). A/Croatia/10136RV/2023 (41058-V08H) HA was purchased from SinoBiological. A/Netherlands/01317/2025 HA and SARS-CoV-2 RBD were produced in-house by recombinant expression in 293F cells. Viruses Influenza virus strains (A/Wisconsin/588/2019, A/Darwin/6/2021, B/Phuket/3073/2013 and B/Austria/1359417/2021) were provided by R. Webby (St. Jude Children’s Research Hospital). A/Thailand/8/2022 and A/Darwin/9/2021 were used for in vitro microneutralization, and these reassortant vaccine virus strains were sourced from the National Institute for Biological Standards and Control. ELISpot Direct ex vivo ELISpot assays were performed to determine total and vaccine-binding IgG-, IgA- and IgM-secreting cells in PBMCs. ELISpot plates were coated overnight at 4 °C with 3 µg ml−1 recombinant HA protein and 10 µg ml−1 anti-human immunoglobulin-κ and immunoglobulin-λ light chain (Cellular Technology). Secreting cells were detected using three-color FluoroSpot IgA/IgG/IgM ELISpot kits (Cellular Technology) according to the manufacturer’s instructions. ELISpot plates were analyzed using an ELISpot counter (Cellular Technology). ELISAs ELISAs were performed in MaxiSorp 96-well plates (Thermo Fisher Scientific). Wells were coated with 1 µg ml−1 recombinant HA antigens or bovine serum albumin (BSA) in PBS (100 µl) and incubated at 4 °C overnight. Plates were blocked with 0.05% Tween-20 and 10% FBS in PBS (blocking buffer). Plasma samples were tested at a 1:30 starting dilution in blocking buffer, followed by seven additional threefold serial dilutions. Recombinant mAbs were diluted to 10 or 2 μg ml−1 in blocking buffer and added to the plates. Plates were incubated for 90 min at room temperature, followed by three washes with 0.05% Tween-20 in PBS. Goat anti-human IgG horseradish peroxidase (Jackson ImmunoResearch) was diluted 1:2,500 in blocking buffer and added to the plates. Plates were incubated for 90 min at room temperature, followed by three washes with 0.05% Tween-20 in PBS and three washes in PBS. Peroxidase substrate (SigmaFAST o-phenylenediamine dihydrochloride, Sigma-Aldrich) was used to develop the plates. Reactions were stopped by the addition of 1 M HCl. Optical density measurements were taken at 490 nm. The half-maximal binding dilution for plasma was calculated using nonlinear regression (GraphPad Prism v10). The threshold of positivity for recombinant mAbs was set as three times the optical density of background binding to BSA. Recombinant mAbs that demonstrated low cross-reactivity at 10 μg ml−1 were further tested at 2 μg ml−1. For testing the avidity of mAbs, ELISAs were performed as described earlier, but wells were coated with 0.1 µg ml−1 recombinant HA antigens (100 µl). Recombinant mAbs were diluted to 30 µg ml−1 in blocking buffer, followed by seven additional threefold serial dilutions. Area under the curve was calculated using GraphPad Prism v10. Flow cytometry and cell sorting For MBC analysis from WU397 participants, cryopreserved PBMCs were thawed and incubated for 30 min on ice with biotinylated recombinant HA proteins, purified CD16 (3G8, BioLegend, 1:100), CD32 (FUN-2, BioLegend, 1:100) and CD64 (10.1, BioLegend, 1:100) in 2% FBS and 2 mM EDTA in PBS (P2). Cells were washed twice and stained for 30 min on ice with CD38–BB700 (HIT2, BD Horizon, 1:500), CD20–Pacific Blue (2H7, 1:400), CD19–BV750 (HIB19, 1:100), IgD–PE (IA6-2, 1:200), streptavidin–BV650 (1:400), CD3–FITC (HIT3a, 1:200) and Zombie NIR (all BioLegend) diluted in Brilliant Staining buffer (BD Horizon). Cells were washed twice with P2. Samples were resuspended in P2 and acquired on an Aurora using SpectroFlo v3.3 (Cytek). Flow cytometry data were analyzed using FlowJo v10.1 (Treestar). For analysis, FNA single-cell suspensions were incubated for 30 min on ice with biotinylated recombinant HA proteins, purified CD16 (3G8, BioLegend, 1:100), CD32 (FUN-2, BioLegend, 1:100), CD64 (10.1, BioLegend, 1:100) and PD-1–BB515 (EH12.1, BD Horizon, 1:100) in P2. Cells were washed twice and stained for 30 min on ice with IgG–BV480 (goat polyclonal, Jackson ImmunoResearch, 1:100), IgA–FITC (M24A, Millipore, 1:500), CD8a–A532 (RPA-T8, Thermo, 1:100), CD38–BB700 (HIT2, BD Horizon, 1:500), CD71–PE-Cy7 (CY1G4, 1:400), CD20–Pacific Blue (2H7, 1:400), CD4–Spark Violet 538 (SK3, 1:400), CD19–BV750 (HIB19, 1:100), IgD–BV785 (IA6-2, 1:200), CXCR5–PE-Dazzle 594 (J252D4, 1:50), CD14–Spark UV 387 (S18004B, 1:100), CD27–PE-Fire810 (O323, 1:200), IgM–BV605 (MHM-88, 1:100), CD3–APC-Fire810 (SK7, 1:50) and Zombie NIR (all BioLegend) diluted in Brilliant Staining buffer (BD Horizon). Cells were washed twice with P2, fixed for 1 h at 25 °C using a True Nuclear fixation kit (BioLegend), washed twice with True Nuclear Permeabilization/Wash buffer, stained with Ki-67–BV711 (Ki-67, BioLegend, 1:200), Blimp1–PE (646702, R&D, 1:100), FOXP3–Spark 685 (206D, BioLegend, 1:200) and BCL6–R718 (K112-91, BD Horizon, 1:200) for 1 h at 25 °C and washed twice with True Nuclear Permeabilization/Wash buffer. Samples were resuspended in P2 and acquired on an Aurora using SpectroFlo v3.3 (Cytek). Flow cytometry data were analyzed using FlowJo v10.1 (Treestar). For sorting PBs and MBCs from WU397 participants, cryopreserved PBMCs collected at baseline (MBCs), 1 week (PBs) and 26 weeks (MBCs) after vaccination were stained for 30 min on ice with CD20–Pacific Blue (2H7, 1:400), IgD–PerCP-Cy5.5 (IA6-2, 1:200), CD19–PE (HIB19, 1:200), CD38–BV605 (HIT2, 1:100), CD3–FITC (HIT3a, 1:200) and Zombie NIR (all BioLegend) diluted in P2. Cells were washed twice, and PBs (live singlet CD3−CD19+IgDloCD20loCD38+) or MBCs (live singlet CD3−CD19+IgDlo) were sorted using a Bigfoot (Invitrogen) into PBS supplemented with 0.05% BSA and immediately processed for scRNA-seq. scRNA-seq samples Sorted PBs, sorted MBCs and lymph node FNA samples were processed using the following 10x Genomics kits: Chromium Next GEM Single Cell 5′ kit v2 (PN-1000263), Chromium Next GEM Chip K Single Cell kit (PN-1000286), BCR Amplification kit (PN-1000253) and Dual Index kit TT Set A (PN-1000215). Chromium Single Cell 5′ Gene Expression Dual Index libraries and Chromium Single Cell V(D)J Dual Index libraries were prepared according to manufacturer’s instructions. Both gene expression and V(D)J libraries were sequenced on a NovaSeq 6000 (Illumina), targeting a median sequencing depth of 50,000 and 5,000 read pairs per cell, respectively. Processing of 10x Genomics single-cell BCR reads Demultiplexed pair-end FASTQ reads were preprocessed using Cell Ranger v.6.0.1 as previously described17 (Supplementary Tables 2–5). Initial germline V(D)J gene annotation was performed on the preprocessed BCRs using IgBLAST v.1.18.0 (ref. 68) with the deduplicated version of IMGT/V-QUEST reference directory release 202150-3 (ref. 69). Isotype annotation was pulled from the ‘c_call’ column in the ‘filtered_contig_annotations.csv’ files outputted by Cell Ranger. Further sequence-level and cell-level quality controls were performed as previously described17. Checks against potential cross-sample contamination were performed by examining the presence of any pairwise overlap between samples in terms of BCRs with both identical unique molecular identifiers (UMIs) and identical V(D)J nucleotide sequences. For each group of cells that originated from different samples but whose BCRs had identical UMI and V(D)J sequences, there tended to be only one cell that also had corresponding transcriptomic data. As such, for each group, when there was exactly one cell with transcriptomics-based annotation, only that cell was kept; when there was more than one cell, none of the cells was kept. Together, 112 cells were removed from the BCR data. Individualized genotypes were inferred based on sequences that passed all quality controls using TIgGER v.1.0.0 (ref. 70) and used to finalize V(D)J annotations. Sequences annotated as non-productively rearranged by IgBLAST were removed from further analysis. Clonal lineage inference for single-cell BCR data B cell clonal lineages were inferred on a by-individual basis based on productively rearranged sequences as previously described17. Briefly, paired heavy and light chains were first partitioned based on common V and J gene annotations and CDR3 lengths. Within each partition, pairs whose heavy chain CDR3 nucleotide sequences were within 0.15 normalized Hamming distance from each other were clustered as clones. Following clonal inference, full-length clonal consensus germline sequences were reconstructed using Change-O v.1.2.0 (ref. 71). Single-cell BCR analysis A B cell clone was considered HA specific if it contained any sequence corresponding to a recombinant mAb that was synthesized based on the single-cell BCRs and that tested positive for binding via ELISA. Clonal overlap between B cell compartments was visualized using circlize v.0.4.13 (ref. 72). SHM frequency was calculated for each heavy chain sequence using SHazaM v.1.1.0 (ref. 71) by counting the number of nucleotide mismatches from the germline sequence in the variable segment leading up to CDR3. Phylogenetic trees for HA-specific B cell clones containing IgA GC B cells were constructed with heavy chains on a by-participant basis using IgPhyML v1.1.3 (ref. 73) and the HLP19 model74. Trees were visualized using ggtree v3.10.1 (ref. 75). Processing of 10x Genomics single-cell 5′ gene expression data Demultiplexed pair-end FASTQ reads were first preprocessed on a by-sample basis, and samples involved in a given analysis were subsequently subsampled to the same effective sequencing length and aggregated using Cell Ranger v.6.0.1 as previously described17. Quality control was performed on the aggregate gene expression matrix consisting of 681,188 cells and 36,601 features using SCANPY v.1.8.2 (ref. 76). Briefly, to remove presumably lysed cells, cells with mitochondrial content greater than 30% of all transcripts were removed. To remove likely doublets, cells with more than 8,000 features or 80,000 total UMIs were removed. To remove cells with no detectable expression of common endogenous genes, cells with no transcript for any of a list of 34 housekeeping genes17 were removed. The feature matrix was subset, based on their biotypes, to protein-coding, immunoglobulin and T cell antigen receptor genes that were expressed in at least 0.05% of the cells in any sample. The resultant feature matrix contained 16,539 genes. Finally, cells with detectable expression of fewer than 200 genes were removed. The same quality control criteria were applied when samples were analyzed on a by-participant basis for selection of B cells for expression as mAbs. All expressed cells were included in the analysis combining all participants, regardless of their quality control metrics in the combined iteration. After quality control, there were a total of 621,877 cells from 109 single-cell samples (Supplementary Tables 6 and 7). Single-cell gene expression analysis Transcriptomic data were analyzed using SCANPY v.1.8.2 (ref. 76) as previously described17 with minor adjustments suitable for the current datasets. Briefly, overall clusters were first identified using Leiden graph clustering with a resolution of 0.12 (Extended Data Fig. 3a). UMAPs were faceted by participant and inspected for convergence to assess whether there was a need for integration. Cluster identities were assigned by examining the expression of a set of marker genes29 for different cell types (Extended Data Fig. 3b). To remove potential contamination by platelets, 383 cells with a log-normalized expression value of >2.5 for PPBP were removed. Cells from the overall B cell cluster were further clustered to identify B cell subsets using a Leiden graph clustering resolution of 0.95 (Fig. 2c, Extended Data Fig. 3c and Supplementary Tables 6 and 7). Cluster identities were assigned by examining the expression of a set of marker genes29 for different B cell subsets (Extended Data Fig. 3d) along with the availability of BCRs. Clusters 6 and 22 were further clustered at a resolution of 0.10 to differentiate naive cells, MBCs and PBs/LNPCs. Cells found in the PB/LNPC clusters that came from blood samples were labeled PBs, whereas those that came from FNA samples were labeled LNPCs. Cells found in the GC B cell clusters but that came from blood samples and that had a PB-like expression profile were labeled PBs. Although clusters 10 and 17 clustered with B cells during overall clustering, they were labeled ‘B and T’ as these cells tended to have both BCRs and relatively high expression levels of CD2 and CD3E. Cluster 23 showed no marked expression level of any marker gene and was labeled ‘unassigned’. The ‘B and T’ and ‘unassigned’ clusters were subsequently excluded from the final B cell clustering. Heavy chain SHM frequency and isotype usage of the B cell subsets were inspected for consistency with expected values to further confirm their assigned identities. Selection of single-cell BCRs for expression Single-cell gene expression analysis was first performed on a by-participant basis. The number of B cell clones to be expressed was determined by balancing cost and maximizing coverage. For clones found in the week 1 PB compartment but not in the GC B cell or LNPC compartment at any time point, one week 1 PB per clone was selected from every such clone with a clone size of at least four cells. For clones found in the GC B cell or LNPC compartment at any time point but not in the week 1 PB compartment, one GC B cell or LNPC per clone was selected from every such clone with a clone size of at least three cells from all participants except participant 397-06 and from every such clone from participant 397-06. Additionally, for every such clone from participants 397-17 and 397-29 that persisted through week 26 in the GC B cell or LNPC compartment to be expressed, one week 26 GC B cell or LNPC per clone was selected from every such clone from participants 397-17 and 397-29 that had a clone size of one or two cells and that persisted through week 26 in the GC B cell or LNPC compartment. For clones found in both the week 1 PB compartment and the GC B cell or LNPC compartment at any time point, one week 1 PB or one GC B cell or LNPC per clone was selected from every such clone from all participants except participant 397-05 and from every such clone with a clone size of at least three cells from participant 397-05. Last, for every HA-specific clone from participants 397-06, 397-17 and 397-29 that was found in both the week 1 PB and the week 26 GC B cell compartments, if not already expressed, one week 1 PB and/or week 26 GC B cell was selected so that every such clone would have one week 1 PB and one week 26 GC B cell expressed. For selection, where there were multiple choices in terms of compartments and/or time points, a compartment and/or time point was first randomly selected. Among cells with the matching compartment and/or time point, the cell with the highest heavy chain UMI count was then selected, breaking ties based on IGHV SHM frequency. In all selected cells, native pairing was preserved. The selected BCRs were curated as previously described17 before synthesis. Recombinant monoclonal antibodies and fragment antigen binding production Selected pairs of heavy and light chain sequences were synthesized by GenScript and sequentially cloned into IgG1, immunoglobulin κ-chain/immunoglobulin λ-chain and fragment antigen binding (Fab) expression vectors. Heavy and light chain plasmids were co-transfected into Expi293F cells (Thermo Fisher Scientific) for recombinant mAb production, followed by purification with protein A agarose resin (GoldBio). Expi293F cells were cultured in Expi293 Expression Medium (Gibco) according to the manufacturer’s protocol. HAI assay HAI assays were performed using the 2022–2023 cell culture-based vaccine virus strains (A/Wisconsin/588/2019, A/Darwin/6/2021, B/Phuket/3073/2013 and B/Austria/1359417/2021). One volume of serum was treated with four volumes of receptor-destroying enzyme (Seiken) at 37 °C overnight before inactivation at 56 °C for 1 h. In a 96-well U-bottom plate (Greiner Bio-One), serial twofold dilutions of 25 µl of receptor-destroying enzyme-treated sera from a 1:2 dilution in PBS (resulting in a 1:10 dilution of sera) were incubated with 25 µl (4 hemagglutinating units) of influenza virus in duplicate wells for 1 h. Subsequently, 50 µl of 1.0% (vol/vol) turkey erythrocytes (Lampire Biological Laboratories) was added to each well. After a 30-min incubation, individual HAI titers were read as the reciprocal of the highest dilution at which 100% hemagglutination was inhibited. The geometric mean HAI titer was calculated for each participant, and titers of <10 were assigned a value of 5 for calculation purposes. All serum samples were tested for nonspecific binding to turkey erythrocytes before performing the assay. In vitro microneutralization assay To test neutralization of A/Thailand/8/2022 and A/Darwin/9/2021 (A/H3N2 reassortant vaccine strains), mAbs were threefold serially diluted at a starting concentration of 10 µg ml−1 in infection medium (1× MEM with tosyl phenylalanyl chloromethyl ketone-treated trypsin). They were then mixed with 100 times 1.64 × 105 and 3.5 × 105, 50% tissue culture infectious dose of the respective viruses and incubated for 1 h at room temperature in a total volume of 100 µl, testing each antibody in duplicate. Virus-only controls were prepared by mixing viruses with infection medium. Virus–antibody mixes, virus-only mixes and medium only were then added to MDCK cells seeded at 25,000 cells per well in 96-well plates (Corning) the day prior. Following a 1-h incubation, cells were washed with PBS, and 100 µl of infection medium with serially diluted antibodies was added. After 48 h of incubation at 37 °C, an HA assay with guinea pig blood was performed to detect the presence of the virus. Sequencing-based neutralization assays Sequencing-based neutralization assays were performed as described in prior work48,49 with a new combined subtype library containing viruses expressing HAs from both (pdm09)A/H1N1 and A/H3N2 subtypes. Briefly, barcoded influenza viruses were generated using reverse genetics methods and plasmids encoding non-HA segments (PB1, PB2, PA, NA, M, NP and NS) from A/WSN/1933 and a barcoded HA from a recently circulating influenza virus. These strains were then equally balanced as a virus library pool and used to assay human sera neutralization titers. To prevent potentially confounding interactions with sialic acids present in human sera, serum samples were treated with a 1:4 dilution of receptor-destroying enzyme and heat inactivation. For the neutralization assay, sera were used at an initial 1:40 dilution and serially diluted 2.7-fold across seven rows of a 96-well plate. The final row of wells was used as the no-serum controls. The virus library was then diluted, added to each well and incubated at 37 °C with 5% CO2 for 1 h. After incubation, 1.5 × 105 MDCK cells overexpressing 2,6-sialyltransferase were added to each well and incubated under the same conditions for ~16 h before the cells were lysed, and barcodes were deep sequenced. Further details on the sequencing-based neutralization assay protocol are available at 10.17504/protocols.io.ewov1962klr2/v1. To process the sequencing data to neutralization titers, we used the modular analysis pipeline seqneut-pipeline and custom code available at https://github.com/jbloomlab/flu_seqneut_with_Ellebedy_lab. BLI Kinetic binding studies were performed on an Octet-R8 (Sartorius) instrument. His tags were removed from Fabs using the thrombin cleavage site preceding the tag. Fabs were treated with biotin-tagged thrombin protease (Sigma-Aldrich) for 2 h at room temperature, followed by removal of remaining His-tagged Fabs with HisPur Ni-NTA resin (Thermo Scientific). The thrombin protease was removed via Streptavidin Sepharose High Performance affinity resin (Cytiva). Anti-Penta-His (HIS1K) sensor tips (Sartorius) were pre-equilibrated in HEPES-buffered saline (0.15 M sodium chloride, 10 mM HEPES and 3 mM EDTA, pH 7.6) with 0.05% Tween-20 and 1% BSA (kinetic buffer), followed by loading of HA proteins (10 µg ml−1) to 0.5 nm. Thrombin-cleaved Fabs diluted in kinetic buffer (1,000 nM, six twofold serial dilutions) were loaded for 120 s and dissociated for 500 s in kinetic buffer. A HIS1K sensor dipped in kinetic buffer was used as a reference sensor. Kinetic parameters of reference-subtracted kinetic traces were calculated with Octet BLI analysis software v12.1 using a global fit 1:1 binding model. A/H3 binding titer for Ig-seq analysis Half-maximal effective concentration values from the ELISAs were used to determine the A/H3 A/Darwin/9/2021-specific serum binding titers. First, Costar 96-well ELISA plates (Corning, 07-200-721) were coated overnight at 4 °C with 4 µg ml−1 recombinant H3 (SinoBiological, 40859-V08H1) and blocked with a blocking solution containing 1% BSA and 0.05% Tween-20 in PBS (pH 7.4). After blocking, serially diluted serum samples were bound to the plates for 2 h at room temperature, followed by incubation with 1:5,000-diluted anti-human Fab secondary antibody–horseradish peroxidase (Sigma, A0293) for 1 h. The plate was rinsed with a washing solution containing 0.1% Tween-20 in PBS three times between every step. For detection, 50 μl of TMB substrate (Thermo Scientific) was added before quenching with 50 μl of 1 M H2SO4. Absorbance was measured at 450 nm using a microplate reader. Half-maximal effective concentration values were derived from a curve fitting function in GraphPad Prism. High-throughput sequencing for Ig-seq analysis VH amplicon was prepared as previously described32,77. Briefly, total RNA from bulk PBMCs (1 × 107 cells) was prepared from week 1 postvaccination samples and reverse transcribed according to the manufacturer’s instructions using SuperScript IV enzyme (Invitrogen) and oligo(dT) primer (Invitrogen). VH transcripts were subsequently amplified using the FastStart High Fidelity PCR System (Roche) with gene-specific primers78. Paired VH:VL sequencing of B cells from each participant was conducted as previously described79,80. PBMCs were isolated as single cells and emulsified with poly(dT)-conjugated magnetic beads to capture heavy and light chain mRNA. These magnetic beads underwent emulsion overlap extension PCR with reverse transcription. All sequences obtained from heavy chain amplicons and paired VH:VL amplicons were merged and annotated using MiXCR 2.1.5 (ref. 81) and used to construct participant-specific search databases as previously established80,82. High-throughput sequences of VH were included for all donors analyzed by Ig-seq. For donors with FNA-derived data (05, 06, 09, 17, 22, 28 and 29), BCR sequence databases were further supplemented with paired VH:VL sequences obtained from both BCR-seq and scRNA-seq. For one additional participant treated with Fluarix (20), paired VH:VL sequences were supplemented from BCR-seq. Purification of total IgG from serum and subsequent digestion into F(ab′)2 Each serum sample was first passed through a 3-ml Protein G agarose (Thermo Fisher, 20397) affinity column in gravity mode. Flow-through was collected and passed through the column three times. The column was washed with 20 ml of PBS before elution with 5 ml of 100 mM glycine-HCl (pH 2.7). The eluted solution, containing total IgG, was immediately neutralized with 0.75 ml of 1 M Tris-HCl (pH 8.0). Purified IgG was digested into F(ab′)2 with 25 μg of IdeS per 1 mg of IgG for 5 h on a rotator at 37 °C and incubated with Strep-Tactin agarose (IBA-Lifesciences, 2-1206-025) for 1 h to remove IdeS. Antigen enrichment of F(ab′)2 and MS sample preparation Recombinant A/H3 (SinoBiological, 40859-V08H1) was immobilized on NHS-activated agarose resin (Thermo Fisher, 26197) by overnight rotation at 4 °C. The coupled agarose resins were washed with PBS, and unreacted NHS groups were blocked with 1 M Tris-HCl (pH 7.5) for 30 min at room temperature. The resins were further washed with PBS and packed into a 0.8-ml centrifuge column (Thermo Fisher, 89868). For each sample, F(ab′)2 was incubated with the individual antigen affinity columns for 2 h on a rotator at room temperature. Flow-through was collected, and the column was washed with 5 ml of PBS. H3-specific F(ab′)2 was eluted with 1% (vol/vol) formic acid in 0.5-ml fractions. Elution fractions were pooled and concentrated under vacuum to a volume of ~10 μl and neutralized using 2 M NaOH. The neutralized elution samples and flow-through samples were denatured with 50 μl of 2,2,2-trifluoroethanol and 5 μl of 100 mM dithiothreitol at 55 °C for 1 h and alkylated by incubation with 3 μl of 550 mM iodoacetamide for 30 min at room temperature in the dark. Alkylation was quenched with 892 μl of 40 mM Tris-HCl, and protein was digested with trypsin (1:30 (wt/wt) trypsin/protein) for 16 h at 37 °C. Formic acid was added to 1% (vol/vol) to quench the digestion, and the sample volume was concentrated to 150 μl under vacuum. Peptides were then purified using C18 spin columns (Thermo Scientific, 89870), washed three times with 0.1% formic acid and eluted with a 60% acetonitrile and 0.1% formic acid solution. C18 eluate was concentrated under vacuum centrifugation and resuspended in 50 μl in 5% acetonitrile and 0.1% formic acid. LC–MS/MS analysis Samples were analyzed by LC–MS/MS on an Easy-nLC 1200 (Thermo Fisher Scientific) coupled to an Orbitrap Fusion Tribrid (Thermo Scientific). Peptides were first loaded onto an Acclaim PepMap RSLC NanoTrap column (Dionex, Thermo Scientific) before separation on a 75 μm × 15 cm Acclaim PepMap RSLC C18 column (Dionex, Thermo Scientific) using a 1.6–76% (vol/vol) acetonitrile gradient over 90 min at 300 nl min−1. Eluting peptides were injected directly into the mass spectrometer using an EASY-Spray source (Thermo Scientific). The instrument was operated in data-dependent mode with parent ion scans (MS1) collected at 120,000 resolution. Monoisotopic precursor selection and charge state screening were enabled. Ions with a charge of ≥+2 were selected for collision-induced dissociation fragmentation spectrum acquisition (MS2) in the ion trap, with a maximum of 20 MS2 scans per MS1. Dynamic exclusion was active with a 15-s exclusion time for ions selected more than twice in a 30-s window. Each sample was run three times to generate technical replicate datasets. MS/MS data analysis Participant-specific search databases for MS data acquisition were generated using all VH sequences obtained through BCR-seq and scRNA-seq from each participant involved in this study. VH sequences were grouped into clonotypes on the basis of single-linkage hierarchical clustering, with cluster membership requiring ≥90% identity across the CDRH3 amino acid sequence as measured by edit distance as previously described83. These clustered VH sequences were then combined with a database of background proteins, which included a consensus human protein database (Ensembl 73, longest sequence/gene), decoy VL sequences for individual participants and a list of common protein contaminants (MaxQuant) to construct the peptide search database. Peptide spectra were searched against the database using SEQUEST (Proteome Discoverer 2.4; Thermo Scientific). Searches considered fully tryptic peptides only, allowing up to two missed cleavages. A precursor mass tolerance of 5 ppm and fragment mass tolerance of 0.5 Da were used. Modifications of carbamidomethyl cysteine (static), oxidized methionine and formylated lysine, serine or threonine (dynamic) were selected. High-confidence peptide-spectrum matches (PSMs) were filtered at a false discovery rate of <1% as calculated by Percolator (q < 0.01, Proteome Discoverer 2.4; Thermo Scientific). Isoleucine/leucine sequence variants were collapsed into single peptide groups. For each scan, PSMs were ranked first by posterior error probability, then q value and finally XCorr. Only unambiguous top-ranked PSMs were kept; scans with multiple top-ranked PSMs (equivalent posterior error probability, q value and XCorr) were designated ambiguous identifications and removed. The average mass deviation for each peptide was calculated as previously described84. Peptides with an average mass deviation of >1.7 ppm were removed. Peptide abundance was calculated from the extracted-ion chromatogram (XIC) peak area, as previously described83. For each peptide, a total XIC area was calculated as the sum of all unique peptide XIC areas of associated precursor ions. The average XIC area across replicate injections was calculated for each sample. For each dataset, the eluate and flow-through abundances were compared, and peptides with greater than or equal to fivefold higher signal in the elution sample were considered to be antigen specific. Clonotype indexing and peptide-to-clonotype mapping High-confidence peptides identified through MS/MS analysis were mapped to clonotype clusters. Peptides that uniquely mapped to a single clonotype were considered ‘informative’, and clonotypes that detected two or more PSMs were kept as high-confidence identifications. The abundance of each antibody clonotype was calculated by summing the XIC areas of the informative peptides mapping to four or more amino acids of the CDRH3 region. The relative amount in serum of each clonotype was calculated by multiplying its relative abundance by the serum titer for that sample. Estimation of selection pressure Selection pressure was estimated with BASELINe as previously described34,85. For each IgG clonotype, a representative matched B cell clone was selected from clones corresponding to detected serum CDRH3 peptides. To ensure high-confidence analysis, the clone with the greatest number of available clonal sequences was chosen, prioritizing clones with scRNA-seq data when present. To maintain comparability, analyses were restricted to participants with both bulk BCR-seq and FNA-derived scRNA-seq data. Heavy chain sequences from the selected clone were collapsed into a representative sequence86. Observed and expected mutations were computed using the HH_S5F targeting model, and posterior probability density functions were computed and visualized with calcBaseline and groupBaseline functions with the focused test used as the selection score87. Statistical significance was assessed with the testBaseline function. Diversity index Representative CDRH3s matched to peptide sequences identified by MS/MS analysis were used to measure the diversity of CDRH3 within the repertoire. To compare the diversity quantitatively, we calculated the effective number of species (exponential of Shannon entropy, 1D) as previously suggested37,38: 1D=exp(−∑i=1Spiln(pi)), where pi is the frequency of the ith clone, and S is total number of clones in the repertoire. Phylogenetic analysis of serum IgG clonotypes High-throughput BCR sequences obtained for Ig-seq were processed with the Immcantation toolkit suite in R v4.4.0. Paired-end single-chain reads were first merged with PEAR v0.9.6 (ref. 88). Low-quality reads were removed, primer regions were masked, and duplicate reads were collapsed using scripts adapted from pRESTO v0.7.2 (ref. 89). Only sequences with at least two reads for VH amplicons were kept and merged with paired VH:VL reads and scRNA-seq reads. Non-productive sequences were filtered using the Change-O package v1.3.0 and were annotated using IgBLAST68. The Immcantation suite pipeline was used to reconstruct the lineage trees71. Novel V genes were detected with TIgGER70. Using SHazaM v1.3.0 (ref. 71), clonal threshold was determined using length-normalized Hamming distance. Clonal assignment was performed using single-linkage hierarchical clustering with SCOPer v1.3.0 (ref. 90). The lineage trees were inferred using IgPhyML v2.0.0 with Dowser v2.3.1 (ref. 91), implementing partitioned maximum likelihood for mixed datasets containing single-chain and paired-chain BCR sequences92. For representative lineages in Fig. 4j, up to 50 unique sequences from non-GC B cells were randomly downsampled to improve visual clarity. Multiplex fluorescence bead assay Recombinant HA proteins, SARS-CoV-2 RBD and BSA were incubated overnight at 4 °C with beads of different fluorescence intensity peaks of the 7K and 8K Blue Particle Array Kit (Spherotech) at 40 µg ml−1, with the exception of recombinant HA A/Perth/16/2009, which was used at 10 µg ml−1. Beads were washed twice with 0.05% Tween-20 in PBS, resuspended in plasma samples diluted 1:200 or mAbs diluted to 10 µg ml−1 in 0.05% Tween-20 and 10% FBS in PBS and incubated for 30 min. Beads were washed twice with 0.05% Tween-20 in PBS, stained with IgG–KIRAVIA Blue 520 (M1310G05, BioLegend, 1:100), incubated for 30 min, washed twice with 0.05% Tween-20 in PBS, resuspended in 2% FBS and 2 mM EDTA in PBS and acquired on an Aurora using SpectroFlo v3.3 (Cytek). Data were analyzed using FlowJo v10.1 (Treestar). Fold change in MFI was calculated for each week 4 or week 17/26 sample by dividing its MFI by the MFI of the corresponding baseline sample. Fold change in mAb MFI binding to specific HAs was calculated relative to the corresponding MFI of that mAb for SARS-CoV-2 RBD. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41590-026-02569-5. Supplementary information Supplementary Tables 1–9. Reporting Summary Ig-seq supplementary MS data. Source data Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Acknowledgements We thank all study participants for providing samples, members of the Washington University School of Medicine Infectious Disease Clinical Research Unit for WU397 study coordination (study coordinators D. Carani, A. Haile, J. Hajare, R. Thompson and J. Wing; pharmacist M. Royal), the staff of the Center for Clinical Research Imaging at Washington University School of Medicine, P. Rudick and M. Reiss for assistance with sample collection and the Genome Technology Access Center in the Department of Genetics at Washington University School of Medicine. The WU397 study was reviewed and approved by the Washington University Institutional Review Board (approval 2208058). Extended data Author contributions A.H.E. conceived and designed the study. M.K.K. and R.M.P. wrote and maintained the Institutional Review Board protocol, recruited participants and coordinated sample collection. H.C.M., F.H., K.D. and H.K.K. processed specimens. B.S.S., M.J.H. and W.D.M. supervised lymph node evaluation before FNA and performed FNA. H.C.M. and F.H. performed ELISpot assays. H.C.M. and K.D. performed ELISAs. A.M. performed HAI assays. H.C.M. and K.D. performed flow cytometry, and J.S.T. performed cell sorting. J.Q.Z. analyzed scRNA-seq and BCR repertoire data. A.J.S., S.C.H., F.H. and H.C.M. generated and characterized monoclonal antibodies. H.C.M. performed BLI assays. H.C.M. performed multiplex bead assays. G.P.A. and R.H. performed neutralization in cell culture. A.G.F. and C.K. designed and performed the sequencing-based neutralization assays, and C.K. and J.D.B. analyzed the data. T.-G.Y., L.P. and J.L. designed and performed BCR-seq and Ig-seq experiments and analyzed the data. H.C.M., A.H.E., N.H.L., R.N. and R.P. analyzed data. A.H.E. and F.K. supervised experiments and obtained funding. H.C.M., A.H.E., T.-G.Y. and J.L. composed the manuscript. All authors reviewed and edited the manuscript. Peer review Peer review information Nature Immunology thanks Kim Good-Jacobson, Sophie Valkenburg and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: P. Jauregui, in collaboration with the rest of the Nature Immunology team. Funding This work was supported in part by funding from the National Institutes of Health (NIH) National Institute of Allergy and Infectious Diseases (NIAID) and Moderna. The Genome Technology Access Center in the Department of Genetics at Washington University School of Medicine is partially supported by National Cancer Institute (NCI) Cancer Center Support Grant P30 CA91842 to the Siteman Cancer Center from the US NIH. A.H.E. discloses support for the research of this work from the NIAID (grants U01AI141990 and U01AI144616; contracts 75N93021C00014 and 75N93019C00051). H.C.M. discloses support from the NIAID (T32AI007172) and NCI (T32CA009547). J.L. discloses support for the research of this work from the NIH (grants P20GM113132 and P01AI089618). T.-G.Y. discloses support from the National Research Foundation of Korea funded by the Ministry of Education (RS-2024-00411420). J.D.B. discloses support for the research of this work from the NIAID (grant R01AI165821; contract 75N93021C00015). C.K. discloses support from the NIAID (T32AI083203). J.D.B. is an investigator of the Howard Hughes Medical Institute. F.K. discloses support for the research of this work from the NIAID (contract 75N93019C00051). The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official view of the NIH or NIAID. Data availability 10x Genomics raw sequencing data and the transcriptomics count matrix have been deposited at Sequence Read Archive and Gene Expression Omnibus under BioProject PRJNA1049392. Processed transcriptomics and BCR data have been deposited at Zenodo at 10.5281/zenodo.13929847 (ref. 93). The proteomics data reported in this paper are archived through MassIVE (https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp) under accession code MSV000100975. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD077942. Additional data supporting this manuscript have been deposited at Washington University Medicine’s Digital Commons Data@Becker Repository to facilitate public access. A Data Transfer and Use Agreement may be required to access individual-level data. The URL to the dataset is 10.17632/srys9y54kx. Source data are provided with this paper. Competing interests The laboratory of A.H.E. and the Infectious Disease Clinical Research Unit received funding under sponsored research agreements from Moderna related to the data presented in the current study. The laboratory of A.H.E. received funding from Emergent BioSolutions and AbbVie that is unrelated to the data presented in the current study. A.H.E. has received consulting and speaking fees from InBios International, Fimbrion Therapeutics, RGAX, Mubadala Investment Company, Moderna, Pfizer, GSK, Danaher, Third Rock Ventures, Goldman Sachs and Morgan Stanley and is the founder of ImmuneBio Consulting. A.J.S., J.S.T. and A.H.E. are recipients of a licensing agreement with Abbvie that is unrelated to the data presented in the current study. N.H.L., R.N. and R.P. are employees of and shareholders in Moderna. The Icahn School of Medicine at Mount Sinai has filed patent applications regarding influenza virus vaccines on which F.K. is listed as inventor and F.K. has received royalty payments from some of these patents. F.K. is co-founder and scientific advisory board member of Castlevax. F.K. has consulted for Merck, GSK, Sanofi, Curevac, Gritstone, Seqirus and Pfizer and is currently consulting for 3rd Rock Ventures and Avimex. The laboratory of F.K. is also collaborating with Dynavax on influenza vaccine development and with VIR on influenza virus therapeutics. J.D.B. is on the scientific advisory boards of Apriori Bio, Invivyd, Aerium Therapeutics and the Vaccine Company. J.D.B. consults for GlaxoSmithKline and Pfizer. J.D.B. receives royalty payments as inventors on Fred Hutch licensed patents related to incorporating barcodes into the influenza genome and viral deep mutational scanning. The other authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Hanover C. Matz, Tae-Geun Yu. These authors jointly supervised this work: Jiwon Lee, Ali H. Ellebedy. Contributor Information Jiwon Lee, Email: jiwonjlee@korea.ac.kr. Ali H. Ellebedy, Email: ellebedy@wustl.edu Extended data is available for this paper at 10.1038/s41590-026-02569-5. Supplementary information The online version contains supplementary material available at 10.1038/s41590-026-02569-5. References Associated Data Supplementary Materials Supplementary Tables 1–9. Reporting Summary Ig-seq supplementary MS data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Data Availability Statement 10x Genomics raw sequencing data and the transcriptomics count matrix have been deposited at Sequence Read Archive and Gene Expression Omnibus under BioProject PRJNA1049392. Processed transcriptomics and BCR data have been deposited at Zenodo at 10.5281/zenodo.13929847 (ref. 93). The proteomics data reported in this paper are archived through MassIVE (https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp) under accession code MSV000100975. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD077942. Additional data supporting this manuscript have been deposited at Washington University Medicine’s Digital Commons Data@Becker Repository to facilitate public access. A Data Transfer and Use Agreement may be required to access individual-level data. The URL to the dataset is 10.17632/srys9y54kx. Source data are provided with this paper.