Temporal dynamics and genomic programming of plasma cell fates

Affinity-matured plasma cells (PCs) of varying lifespans are generated through a germinal center (GC) response. The developmental dynamics and genomic programs of antigen-specific PC precursors remain to be elucidated. Using a model antigen, we demonstrate biphasic generation of PC precursors, with those generating long-lived bone marrow PCs preferentially produced in the late phase of GC response. Clonal tracing using scRNA-seq+BCR-seq in spleen and bone marrow compartments, coupled with adoptive transfer experiments, reveal a novel PC transition state that gives rise to functionally competent PC precursors. The latter undergo clonal expansion, dependent on inducible expression of TIGIT. We propose a model for the proliferation and programming of precursors of long-lived PCs, based on extended antigen encounters followed by reduced antigen availability.


Introduction
Plasma cells (PCs) are considered to represent a terminally differentiated state generated by the encounter of B cells with antigens in the context of pathogens or vaccines. PCs constitutively secrete antibodies which can serve as a source of protective antibody responses [1][2][3][4][5] . Following antigen exposure, in the context of a T cell-dependent response, antigen-stimulated B cells interact with T follicular helper (Tfh) cells in secondary lymphoid organs (SLOs), e.g., spleen or lymph nodes, undergo clonal expansion, somatic hypermutation and affinity maturation in germinal centers (GCs), generating both memory B cells and PC precursors, the latter migrate through the bloodstream and home to the bone marrow (BM), where they undergo further maturation to terminally differentiate into PCs 6,7 .
The nature of signaling pathways and transcriptional programs that result in the generation of PC precursors in SLOs that are functionally competent to migrate through the bloodstream to the bone marrow still need to be thoroughly understood.
Antiviral antibody responses can be remarkably stable in humans, lasting decades in the case of varicella-zoster and measles viruses, but are less durable for influenza viruses 8 .
The durability of antibody responses to viral infections and vaccines reflects the longevity of PCs within the bone marrow (BM) 9,10 . The cellular and molecular mechanisms underlying the generation of short-lived versus long-lived PCs (SLPCs, LLPCs) are of heightened interest given the recent COVID-19 pandemic 11 . The longevity of PCs in the bone marrow could be dictated by the transcriptional programming of PC precursors emanating from the GC and/or by niches in the bone marrow that the PCs reside within.
The temporal dynamics of PC precursor generation have been inferred based on the emergence of antigen-specific PCs in the spleen as well as in the bone marrow, in the context of NP-specific B cell responses in murine models. A key study tracked the responses of NPspecific B cells in the spleen and bone marrow between 7 to 28 days post-immunization (d.p.i.) 12 . Splenic NP-specific IgG1 + antibody-secreting cells (ASCs) peaked at 14 d.p.i. and then declined, thereby primarily reflecting the generation of extrafollicular plasmablasts. In contrast, ASCs in the bone marrow manifested a nearly 5-fold increase between 14 and 28 d.p.i. 13 . These results suggested that bone marrow PC (BMPC) precursors are maximally generated in a temporally delayed manner during an ongoing GC response. The temporal dynamics of PC precursor generation have also been analyzed using an alternate NP-specific model. In this model, NP-reactive B cells isolated from B1-8 mice were transferred into AM14 transgenic Vk8R mice, prior to immunization with NP-CGG 12 . Two waves of ASC generation were noted in the spleen, with peaks at 11 and 38 (d.p.i.). Notably, the latter peak coincided with the maximal emergence of ASCs in the bone marrow that included LLPCs 12 . However, the nature of the PC precursors implicated by these studies and the mechanisms underlying their generation and/or expansion at later phases of the GC response remain to be delineated.
During PC differentiation, B cells undergo extensive genomic re-programming, which results in the repression of a large set of B cell genes and the activation of PC-specific as well hematopoietic progenitor and T cell genes 14,15 . This process is regulated by various Furthermore, stimulation of GC B cells with CD40L and IL-21 results in an increased frequency of BCL6 lo IRF4 hi cells 25 . The signaling interplay between CD40, IL-21R, and antigen recognition via the BCR, and how such stimuli dictate the generation of PC precursors as well as the fates of their PC progeny, remain to be explored. Thus, deeper analyses of the signaling and genomic regulatory mechanisms that underlie the generation of PC precursors generated during GC-dependent B cell responses will not only advance fundamental understanding of these distinctive cells but also enable their precise therapeutic manipulation so as to enhance vaccination strategies 26,27 and target pathogenic counterparts 28-30 .
We have previously utilized scRNA-seq in conjunction with BCR-seq to track NPspecific B cell responses in GCs and the dynamic genomic states associated with class switch recombination, somatic hypermutation and affinity maturation 21,31 . We now extend this approach to the emergence of PC precursors from the GC and tracking of B cell clones in the splenic and bone marrow compartments. We demonstrate a biphasic generation of BMPC precursors, with those generating long-lived PCs preferentially produced in the late phase of GC response. Clonal tracing in spleen and bone marrow compartments, coupled with adoptive transfer experiments, reveal a novel PC transition state that gives rise to transcriptionally distinct PC precursors. The proliferation of PC precursors is dependent on the inducible expression of TIGIT. We propose a signaling model for programming of longlived PC precursors based on extended antigen encounters followed by reduced antigen availability.

Temporal dynamics of PC precursors generated during a GC response
To analyze antigen-specific precursors of PCs that are generated during a germinal center response in the spleen and give rise to plasma cells in the bone marrow, we designed the adoptive transfer model system schematized in Fig. 1a. Mice were immunized with NP-KLH, LPS and alum. CD138 + cells isolated from splenocytes (Extended Data Fig. 1a) (Fig. 1a). Analysis of NP-specific antibody titers by ELISAs, and PCs by ELISPOTs, enabled dynamic monitoring of PC precursor activity as well as the durability 6 of the PCs generated from them. By initiating our analysis of PC precursors at 21 d.p.i., a week after the peak of the GC response 33 , we minimized the transfer of extrafollicular plasmablasts into the recipient animals. We used CD138 + splenocytes for adoptive transfer as they contain both B220 + and B220 -cells and therefore represent a continuum of PC differentiation states, including the PC precursors. Notably, the CD138 + cells were transferred into naïve animals, thereby selecting for GC-derived PC precursors that are competent to migrate to the bone marrow and give rise to PCs in the absence of antigen.
Analysis of the antibody titers in the recipient animals 21 days post-transfer (d.p.t.) suggested that the peak of PC precursor activity occurred at 35 d.p.i., much later than the peak of GC response (Fig. 1b) 1b,c). This suggested that PC precursors generated earlier in the GC response give rise to short-lived PCs, whereas those generated later give rise to longer-lived PCs. To directly test for the generation of PCs in the bone marrow and their longevity, we performed ELISPOT analyses in the recipient µMT mice at varying d.p.t.. As predicted by the antibody titers, the largest number of NP-specific ASCs were detected using PC precursors isolated at 35 d.p.i. (Fig. 1e). Importantly, in keeping with our expectation from the durability of antibody responses, these ASCs represented long-lived BMPCs as they were detectable at 120 d.p.t. (Fig. 1e). In contrast, such long-lived BMPCs were not detectable in adoptive transfers using PC precursors isolated at 21 d.p.i. We reasoned that splenic precursors generated at 21 d.p.i. may preferentially give rise to short-lived BMPCs. To test this possibility, we analyzed µMT recipients adoptively transferred with CD138 + cells from 21 d.p.i. at 7-60 d.p.t. for the durability of NP-specific PCs in the bone marrow. Notably, PCs were detectable as early as 7 d.p.t. but declined by 14 d.p.t. and were undetectable at 60 d.p.t. (Fig. 1f). These results demonstrate that SLPC precursors are generated earlier during a GC response, whereas their LLPC precursors are generated later. They suggest a temporal shift in the developmental programming of PC precursors, emanating from the GC, which results in BMPCs with longer durability at later times in the response. scRNA-seq reveals novel PC progenitor state and divergent PC clusters 7 To genomically delineate intermediates in the specification of BMPC precursors, we initially performed scRNA-seq and BCR-seq on splenic CD138 + splenocytes isolated at 35 d.p.i. in the GC response. This timepoint was selected as it represented maximal BMPC precursor activity based on the adoptive transfer experiments (Fig. 1). Computational analysis of the scRNAseq dataset using the ICGS2 pipeline 34 revealed cell clusters that could be annotated based on the expression of biologically informative marker genes (Fig. 2a  and Xbp1, were resolved into three clusters that were demarcated by the differential expression of Tigit, Slpi, and Lag3, respectively ( Fig. 2b and Extended Data Fig. 2b,c). ICGS2based clustering revealed a transitional PC cell state, hereafter termed PC progenitor, as these cells retained robust expression of B-lineage specific genes but appeared to be inducing expression of Irf4, Prdm1 and a large set of PC genes (Fig. 2b, Extended Data Fig.   2b). To examine developmental relationships between the PC progenitors and the cells contained in the GC as well as PC clusters, we performed single-cell trajectory analysis using Monocle2 35 . The various single-cell clusters could be organized into a continuous, well populated path with a branch at one end (Fig. 2c). GC B cells were clustered at the nonbranching end of the trajectory whereas the PC clusters were positioned at the other end, primarily before and after the branchpoint. Notably, PC progenitors were dispersed along the trajectory between the GC B cells and PCs. Pseudotime analysis of the Monocle2 generated trajectory, with the non-branching end serving as the origin, aligned with the inference that PC progenitors arise from GC B cells and generate distinct types of plasma cells (Extended Data Fig. 2d). Intriguingly, the Tigit and Slpi PCs appeared to represent one terminus and the Lag3 PCs another terminus in the trajectory (Fig. 2c). To further substantiate the developmental trajectory we performed analyses of the scRNAseq dataset using scVelo 36 . This analysis focused on the dynamics of spliced and unspliced transcripts. It reinforced the inference that PC progenitors arise from GC B cells and give rise to the various PC states (Fig. 2d). Accordingly, the PC progenitors manifested a lower PC gene signature score compared with their PC counterparts (Fig. 2e). Intriguingly, the PC compartment also displayed a higher mitotic gene expression score than the PC progenitors, suggesting that differentiated cells within the compartment undergo cell division after cell fate specification (Fig. 2f).

Tracing antigen-specific PC states emanating from the germinal center
Given the dynamic expression of the mitotic gene module in the PC progenitors and their differentiating progeny (Fig. 2f), we sought to track PC clones using BCR-seq. To enable the analysis of antigen-specific cells responding to NP-KLH, we first tabulated IGHV sequences that dominate the NP-specific GC B cell response at 14 d.p.i. 37 (Extended Data Fig. 3a Table 2). Using the combined IGHV sequences, we annotated antigen-specific cells in the 35 d.p.i. dataset. Importantly, the selected cells exhibited the spectrum of genomic states observed with the total CD138 + splenocytes (Fig. 3a) and were organized continuously along the Monocle2 developmental trajectory derived using the total CD138 + splenocytes (Fig. 3b). To analyze and trace NP-and presumptively KLH-specific clones across genomic states, we identified cells that contained identical V(D)J and VJ gene rearrangements in their heavy and light chain immunoglobulin loci, respectively (Supplementary Table 3 Table 3). The latter conclusion was reinforced by the analysis of somatic hypermutations in cells harboring the dominant NP-specific IGHV1-72*01 (Extended Data   Fig. 3e). This analysis revealed that PC progenitors and their differentiating counterparts accumulated somatic hypermutations, which increased in frequency from 21 d.p.i. to 35 d.p.i. and also evidenced affinity maturation, consistent with their GC origin (Extended Data Fig.   3f). We note that CellHarmony 38 was used to align the single cell genomic states of the CD138 + splenocytes at 21 d.p.i. with their counterparts at 35 d.p.i., using the latter as the reference set (Extended Data Fig. 3g). This enabled us to reveal differentially expressed

Functional and genomic analyses of GC-derived PC precursors
The coupled scRNA-seq and BCR-seq analyses of antigen-specific PC cell states and their trajectories, during a GC response, revealed a novel progenitor state as well as significant clonal expansion of differentiated PC progeny. However, the nature of the BMPC precursors and their genomic state(s) within the developmental trajectory, remained to be delineated.
ELISPOT analysis of these three subsets along with control cells suggested that the ability to secrete antibodies was acquired during the PC progenitor state, consistent with the upregulation of the PC gene module (Fig. 4c). To functionally assess which of those subsets contained BMPC precursors, we adoptively transferred the FACS-purified cells into µMT mice. NP-specific IgG1 titers, measured on 21 d.p.t., were substantially higher (>10-fold) in µMT recipients that were reconstituted with the B220 int CD138 + CD44 + CD11a + and B220 -CD138 + CD44 + CD11a + subsets compared to the B220 + CD138 + CD44 + CD11a + counterparts ( Fig. 4d). ELISPOT analysis of the bone marrow (60 d.p.t.) demonstrated that NP-specific IgG1 + ASCs were detected in comparable numbers between the B220 int CD138 + CD44 + CD11a + and B220 -CD138 + CD44 + CD11a + subsets. However, no such BMPCs were observed with the B220 + CD138 + CD44 + CD11a + cells (Fig. 4e). These results demonstrated that PC precursors, that are functionally competent to migrate and generate LLPCs in the bone marrow, are contained within three genomically distinct splenic PC clusters. Furthermore, the results distinguished PC progenitors from BMPC precursors as the former appeared to have acquired ASC function but not bone marrow homing/survival capabilities.
To complement the adoptive transfer system and also further resolve genomic states of splenic PC precursors that give rise to BMPCs, we performed clonal and genomic analyses  Table 5). Collectively, these results suggest that antigen-specific, Tigitexpressing splenic PCs, generated in the context of a GC-dependent immune response represent the dominant source of BMPC precursors.

TIGIT promotes PC precursor expansion and generation of PCs
The above results implied a cell-intrinsic function of TIGIT in the generation and/or proliferation of PC precursors. In line with the expression dynamics of the Tigit gene in the scRNA-seq datasets, TIGIT protein was highly expressed in splenic B220 int CD138 + cells, with reduced levels in B220 -CD138 + cells (Extended Data Fig. 6a). Of note, naïve and GC B cells did not express TIGIT. To analyze the cell-intrinsic functions of TIGIT, we generated mixed bone-marrow chimeras by transferring cells from CD45.1 Tigit +/+ (WT) and CD45.2 Tigit -/mice at a 50:50 ratio into irradiated WT recipients (Extended data Fig. 6b). CD45.2 Tigit +/+ bone marrow cells were also co-transferred with CD45.1 Tigit +/+ cells to generate control chimeras. Successful chimeras were confirmed by a 1:1 ratio of CD45.1 WT and CD45.2 Tigit -/-naive B cell populations in the spleen at 8 weeks after transplantation (Extended Data Fig.  6c), indicating that the deficiency of TIGIT did not impact the overall development of B cells.
Notably, the proportions of Tigit -/-cells were significantly decreased in the B220 int and B220 -PC compartments in the spleen and bone marrow (Extended Data Fig. 6d,e), demonstrating that TIGIT was required in a cell-intrinsic manner to maintain the steadystate frequencies of plasma cells and their precursors. To analyze the function of TIGIT in the generation of antigen-specific plasma cells, we immunized chimeric mice with NP-KLH and analyzed the wild-type and Tigit -/-splenic PC compartments at the indicated time points (Fig.   6a). Consistent with the observations in chimeric mice at steady state, the proportions of Tigit -/-cells in the splenic plasma cell precursor compartment were significantly decreased after immunization as were their differentiated PC progeny ( Fig. 6b-d). Importantly, the numbers of NP-specific ASCs generated from the Tigit -/-B cells were also significantly decreased in the spleen and bone marrow of the chimeras (Fig. 6e,f). Thus TIGIT functions in a cell-intrinsic manner to promote the generation and/or proliferation of PC precursors.
Given that TIGIT-expressing PC precursors appear to undergo greater clonal expansion ( Fig. 3f), we analyzed their proliferation index based on KI-67 staining (Extended Data Fig.   6f,g). Notably, a larger fraction of TIGIT + cells within the B220 int CD138 + CD44 + CD11a + PC precursor compartment were KI-67 + suggesting that TIGIT could function in controlling the proliferation of PC precursors. We tested this possibility by performing EdU incorporation assays in vivo with the immunized chimeric mice. The deficiency of TIGIT significantly impaired the proliferation of plasma cell precursors (Fig. 6g,h). Thus, TIGIT-mediated signaling in GC-derived PC precursors controls their expansion and the generation of BMPCs.

Discussion
We have attempted to address several fundamental questions concerning the nature of PC precursors that emanate from a secondary lymphoid organ, via a GC B cell response, and migrate to the bone marrow to give rise to PCs of varying longevity. Using the adoptive transfer of CD138 + splenocytes, isolated at varying times during an NP-KLH immunization response, into naïve B cell-deficient mice, we demonstrate a quantitative as well as qualitative change in PC precursor activity. Strikingly, the peak of PC precursor activity is observed at 35 d.p.i., 3 weeks after the peak of the GC response (14 d.p.i.). Furthermore, such precursors preferentially give rise to longer-lived PCs in the bone marrow. Consistent with our findings, temporal analysis of NP-specific ASCs in spleen and BM, during a NP-CGG response, suggested a late (29-47 d.p.i.) emergence of long-lived PCs that coincided with their maximum output in the splenic compartment 12,13 . It should be noted that the two experimental systems generate concordant results despite the substantial differences in approaches. The former involves time-resolved and quantitative analysis of newly generated NP-specific ASCs in spleen and bone marrow compartments of immunized animals, but did not involve the characterization and functional testing of PC precursors. Our approach, involving the adoptive transfer of PC precursors isolated from immunized animals into naïve We note that in the time-stamp model, LLPCs were tracked up to 70 days. We demonstrate using our adoptive transfer system that PC precursors generated at 21 d.p.i. can give rise to PCs that are detectable between 50-80 total days, after immunization. Notably, by analyzing PC precursors that arise much later in the GC response at 35 d.p.i. we show that they give rise to longer lived PCs in the bone marrow, up to a total of 155 days after immunization. The collective evidence is consistent with both quantitative and qualitative changes in the PC precursor compartment, as the germinal center response wanes due to diminished antigen availability.
Though the adoptive transfer experiments of CD138 + splenocytes enabled temporal and quantitative monitoring of PC precursor activity, they did not reveal the identity of the cells nor their genomic states. To do so unequivocally, we performed scRNA-seq and BCRseq of CD138 + splenocytes and subsets so as to link their genomic states with developmental capacities. The conclusions of these experiments were then validated in immunized but otherwise unperturbed mice, by clonal tracing of antigen-specific PC precursors that were generated in the spleen and gave rise to PCs in the bone marrow. Genomic analysis of the The foregoing analysis led us to consider the possibility that the PC progenitors, may represent the sought after PC precursors, whose activity is temporally and quantitatively monitored in the adoptive transfer experiments. However two complementary sets of experimental approaches ruled out this possibility and instead demonstrated that the dominant set of PC precursors are differentiated progeny of the PC progenitors. The majority of the PC precursors reside within a splenic PC cluster that express, in an inducible manner, the inhibitory T cell immunoreceptor TIGIT. This conclusion is based on two key findings: (i) that the B220 + subset of CD138 + CD44 + CD11a + splenocytes within which PC progenitors reside have undetectable PC precursor activity in the adoptive transfer experiments; in contrast, the B220 int subset of CD138 + CD44 + CD11a + splenocytes, containing the majority of PC-Tigit* cells, have appreciable PC precursor activity; and (ii) clonal tracing demonstrates that splenic PC-Tigit* cells are the predominant source of PC lineages in the bone marrow.
We note that in the adoptive transfer experiments, we were not able to quantitatively distinguish between the PC precursor activity of the three PC clusters revealed by the scRNA-       Fig. 2. ScRNA-seq analysis of PC genomic states and trajectories a, Heatmap generated using cluster-specific marker genes delineated by the MarkerFinder algorithm in AltAnalyze (Methods) for CD138 + splenocytes isolated at 35 d.p.i. (Fig. 1a) and profiled using 5'-end droplet based scRNA-seq. Columns in heatmap represent cells (n=8,813); rows represent markerfinder genes (n=413). Cell clusters were generated using ICGS2 in AltAnalyze             Fig. 2. scRNA-seq analysis of PC genomic states and trajectories a, Comb plots displaying the incidence and amplitude of indicated GC B cell, DZ GC and cell cycle genes in designated cluster as in Fig. 2a. b, Plots displaying the incidence and amplitude of indicated PC genes in each cluster as above. c, Plots displaying the incidence and amplitude of indicated marker genes distinguishing the three PC clusters as above. d, Pseudotime analysis of cells in the Monocle 2 trajectory (Fig. 2e)   C57BL/6J WT B220 -CD138 + CD44 + CD11a + B220 int CD138 + CD44 + CD11a + B220 + CD138 + CD44 + CD11a + BCR and scRNA-seq Extended Data Fig. 5. Clonal tracking of antigen-specific PC precursors that migrate from spleen to bone marrow a, Experimental design enabling clonal tracking of PC precursors that migrate from spleen to bone marrow of NP-KLH immunized mice (35 d.p.i.). Coupled scRNA-seq and BCR-seq was performed on indicated cells (including PC progenitors), within each compartment, isolated by flow cytometry. b, Circos plot displaying clones and their genomic states in spleen and bone marrow. Colored bars denote distinctive ICGS2 delineated genomic states in spleen and bone marrow. Colored lines represent clones that contain cells with identical V(D)J rearrangements that span two or more genomic states. c, Heatmap displaying the frequencies of clones spanning indicated genomic states in the spleen and the bone marrow.
Extended Data Figure 6 CD45.  . g, Quantification of the frequency and mean fluorescence intensity of KI-67 + cells in B220 int subset as in (f). Each symbol represents an individual mouse (d,e,g). Statistical significance was tested by two-tailed t-test (d,e,g). **p<0.01; and ****p<0.0001.
Adoptive transfer of splenic CD138 + cells and subsets. Carrier cells for the adoptive transfer were prepared by making a single cell suspension of spleen collected from μMT mice. For adoptive transfer of total CD138 + splenocytes, 3 x 10 5 enriched CD138 + cells and 1 x 10 6 μMT splenocytes were injected retro-orbitally into μMT recipients. For splenic CD138 + subsets transfer, 1 x 10 5 flow sorted CD138 + subsets and 1 x 10 6 μMT splenocytes were injected retro-orbitally into μMT recipients. Calculation of cell cluster frequencies in CD138 + splenic subsets. CD138 + splenic subsets were flow sorted based on the indicated cell surface markers and processed for scRNA-seq. Cell clusters within the CD138 + splenic subsets were identified using the outputs from cellHarmony. The frequency of cell clusters within the mixed group of purified B220 + CD138 + CD44 + CD11a + and B220 int CD138 + CD44 + CD11a + splenocytes were calculated by subtracting the proportion of cell clusters contained within the unmixed B220 -CD138 + CD44 + CD11a + cells processed using scRNA-seq.
Differential gene expression analyses. Differential gene expression analysis was performed using cellHarmony between the reference (CD138 + splenocytes -35 d.p.i.) and the query cluster. Genes with an absolute fold > 1.2 and empirical Bayes t-test p-value < 0.05 (Benjamini Hochberg corrected) were considered differentially expressed. A chi-square test was performed to assess differential cell type frequency within.
Pseudotime trajectory and RNA velocity analyses. Cell-by-gene counts and cell type labels file were filtered for cells from LZ and DZ GC B cell, PC progenitor and distinct PC clusters, and were used as input for Monocle2 pseudotime analysis using default options. LZ GC B cells were indicated as the root of the analysis. The counts data was modeled with negative binomial distribution using Monocle 2 ('expressionFamily=negbinomial.size').
Monocle 2 was allowed to select its own genes for pseudotime estimation based on differential gene analysis across the filtered B-lineage cells ('fullModelFormulaStr = ~Groups'). The reverse graph embedding (RGE) ('method' in reduceDimension) method was set to "DDRTree" as recommended 5 . The Monocle2 cell state that contained the maximum number of LZ B cells was set as the "root" state to calculate the pseudotime of all cells.
For RNA velocity analysis, the cell per gene matrix generated from the cell ranger, filtered for LZ and DZ GC B cells, PC progenitors and distinct PC clusters, was provided as an input for Velocyto tool to generate a loom file that contained the spliced and unspliced counts for each gene in each cell 6 . The loom file was then preprocessed using the default parameters described in scVelo package (Version 0.2.5) 7 and Scanpy (Version 1.9) packages 8 .
Gene signature score. To calculate signature scores of G2/M genes, an algorithm from our previous study 2 was adopted to control the variation of quality and complexity of scRNA-seq data of individual cells. To calculate the gene signature score of PCs, this algorithm was modified based on the top 50 upregulated genes analyzed in BMPCs by bulk RNA-seq along with four important PC genes 9 .
BCR analysis of 5'-end scRNA-seq. Single-cell VDJ sequencing data were processed using Cellranger VDJ (v.3.1.0 -5 Prime V1; v.6.1.2 -5 Prime V1.1) reference for mouse from 10x Genomics. Cell barcode associations from CellRanger VDJ and linked gene expression were used for coupling the BCR and transcriptomes, respectively. Cells that did not contain both the BCR and gene expression features were excluded for all downstream clonal analyses. The somatic hypermutation rate and affinity maturation were determined for these cells as KLH-specific IGHV genes (Supplementary table 2). The R code defining the clones is provided at (https://github.com/kairaveet/bcr-clones).

Statistical analysis.
For the biological datasets, we used Prism version 9 (GraphPad) for differential testing. Pairwise statistical tests were performed using a two-tailed Student's ttest, whereas multicondition comparisons were performed using a one-way ANOVA with Tukey's multiple comparison test or Kruskal Wallis with Dunn's multiple comparison test were used based on the distribution, as indicated. For comparison of multiple measures of same samples for more than two or more timepoints, repeated measures two-way ANOVA with Tukey's multiple comparison test was used.