Human anti-smallpox long-lived memory B cells are defined by dynamic interactions in the splenic niche and long-lasting germinal center imprinting

Summary

Memory B cells (MBCs) can persist for a lifetime, but the mechanisms that allow their long-term survival remain poorly understood. Here, we isolated and analyzed human splenic smallpox/vaccinia protein B5-specific MBCs in individuals who were vaccinated more than 40 years ago. Only a handful of clones persisted over such an extended period, and they displayed limited intra-clonal diversity with signs of extensive affinity-based selection. These long-lived MBCs appeared enriched in a CD21^hiCD20^hi IgG⁺ splenic B cell subset displaying a marginal-zone-like NOTCH/MYC-driven signature, but they did not harbor a unique longevity-associated transcriptional or metabolic profile. Finally, the telomeres of B5-specific, long-lived MBCs were longer than those in patient-paired naive B cells in all the samples analyzed. Overall, these results imply that separate mechanisms such as early telomere elongation, affinity selection during the contraction phase, and access to a specific niche contribute to ensuring the functional longevity of MBCs.

Graphical abstract.

Introduction

Immune memory is usually defined as the ability to respond faster and more potently to a re-exposure to the same or closely related pathogen, a key attribute of the adaptive T and B cell components of the immune system. The humoral arm of immune memory is composed of memory B cells (MBCs) and long-lived plasma cells (PCs), which emerge during T-dependent germinal center (GC) responses in which stringent antigen-driven selection takes place (Victora and Nussenzweig, 2012). In humans, MBCs are detected several decades after infection or vaccination (Amanna, 2007; Crotty et al., 2003; Yu et al., 2008) and can provide rapid protection even in the total absence of PCs and neutralizing antibodies (Bauer and Jilg, 2006).

Most studies on the fate of B cell memory have been performed using different fate-mapping mouse models. The current paradigm coming from these studies is that MBCs leave the GC at an earlier time point than PCs and display an overall lower affinity for the immunizing antigen (Viant et al., 2020, 2021; Weisel et al., 2016; Wong et al., 2020), this property allowing them to respond to variants of the initial pathogen. IgG+ B cells tend to dominate the GC reaction and display higher mutation loads (Pape et al., 2011; Sundling et al., 2021). However, both IgM+ and IgG+ MBCs have been observed to be stable for more than a year (Jones et al., 2015; Pape et al., 2018).

Studies tracking the generation of human MBCs following vaccination (Andrews et al., 2019; Goel et al., 2021; Kim et al., 2022; Matsuda et al., 2019; Wec et al., 2020) or natural infection (Davis et al., 2019; Gaebler et al., 2021; Sokal et al., 2021a, 2021b; Wang et al., 2021) have shown a sustained GC output of MBCs up to 6 months after challenge, these cells reaching maximal mutational load after that time point in most cases. Mid to high affinity antigen-specific MBCs could be tracked in the blood for over a year with evidence of a contraction phase before that time point in some (Andrews et al., 2019; Davis et al., 2019; Wec et al., 2020) but not all reports (Matsuda et al., 2019; Sokal et al., 2021b; Wang et al., 2021).

Recent deep single-cell analyses have further highlighted the heterogeneity of human MBCs, both between and inside major lymphoid organs, through markers like Tbet and CD11c for effector/memory B cells, FCRL5 in tonsils, and CD69, CD73, or CD45RB in the spleen and gut-associated lymphoid tissues (Glass et al., 2020; Johnson et al., 202; King et al., 2021; Sanz et al., 2019; Weisel et al., 2020, 2022; Zhao et al., 2018). Whether one of these subpopulations represents a unique long-lived MBC subset and whether a specific niche exists for these cells, alike long-lived PCs in the bone marrow (Nguyen et al., 2021), are still largely unknown.

Assessing the true longevity of human MBCs against common pathogens is, in most settings, hindered by the possibility for any given individual to re-encounter them during his lifetime. Smallpox, officially declared eradicated in 1980 (Fenner et al., 1988), represents one notable exception. Alongside smallpox eradication, anti-smallpox vaccination, based on the live vaccinia virus, was progressively discontinued during the 1970s. Any detectable anti-vaccinia memory B cell in an individual post 2010 would thus likely have been generated more than 40 years ago, without any re-stimulation with the immunizing antigen during that timeframe. Anti-vaccinia antibodies serum titers remain stable in human for more than 80 years after first vaccination (Amanna, 2007; Hammarlund et al., 2003; Taub et al., 2008), with several identified immunodominant epitopes. This includes the glycoprotein B5, one of the five proteins from the extracellular enveloped virion (EEV) form of vaccinia and known target of potent neutralizing antibodies (Aldaz-Carroll et al., 2005; Gilchuk et al., 2016; Lantto et al., 2011). Anti-vaccinia MBCs have similarly been followed longitudinally in the blood for up to 65 years post vaccination (Amanna, 2007; Crotty et al., 2003). Following a contraction phase taking place in the first couple of years after immunization, they reach a remarkably stable plateau that last for decades at around 0.1% of total circulating IgG+ MBCs. Functional analysis of such long-lived MBCs, however, is still lacking.

We reported previously that the spleen was enriched for these long-lived vaccinia-specific MBCs (Mamani-Matsuda et al., 2008). We describe here the isolation and functional study of human splenic anti-vaccinia MBCs, more than 40 years after vaccination. This enabled us to decipher the diversity and affinity of their repertoire as well as their transcriptional signature compared with other splenic MBC populations. Our results suggest an extended contraction phase, driven by affinity selection, and elongated telomeres acquired in GCs, together with signals provided by a specific splenic niche, as the molecular basis for the functional longevity of these cells.

Results

Long-lived splenic vaccinia-specific MBCs reside in the splenic CD73⁺CD21⁺CD27⁺ IgG⁺ B cell population

As a proxy to study long-lived MBCs, we elected to isolate vaccinia glycoprotein B5-specific cells from organ donors preenriched IgG⁺ splenic MBCs (Figures 1A and S1A). B5-stained cells represented between 0.003% and 0.07% of total IgG⁺ MBCs in the spleen of most organ donors born before 1970 and likely vaccinated during their childhood (Figure 1A). These frequencies are in line with the previously estimated pool of long-lived MBCs against all vaccinia antigens ranging between 0.02% and 0.3% of MBCs in the blood of most vaccinated donors and up to 0.6% in the spleen (Crotty et al., 2003; Mamani-Matsuda et al., 2008).

Figure 1. Long-lived splenic vaccinia-specific MBCs reside in the splenic CD73⁺CD21⁺CD27⁺ IgG⁺ B cell population.

(A) Experimental scheme for the isolation and retrospective validation of sorted of anti-vaccinia B5 MBCs from the spleen of organ donors.

(B) Corrected frequencies of B5-specific MBCs in live IgG⁺ CD27⁺ MBCs plotted against donor’s age. Donors born after 1977 (n = 8) and before 1970 (n = 56) are highlighted in red and white, respectively.

(C) Representative dot-plot of B5-staining (left) and corrected frequencies of B5-specific MBCs in enriched IgG⁺ or IgM⁺CD27⁺ live CD20⁺ splenic B cells (right, n = 5 donors).

(D) Representative overlaid IgG and CD27 staining (left) and frequencies of CD27⁺ in B5⁺IgG⁺ or total IgG⁺ live CD20⁺ splenic B cells (right, n = 18 donors).

(E) Representative overlaid CD11c and CD21 staining (left) and frequencies of CD21^lowCD11c⁺ in validated B5⁺IgG⁺ or total IgG⁺ live CD20⁺ splenic B cells (right, n = 10 donors).

(F) Representative CD73 staining (left) and graph frequencies of CD73⁺ in validated B5⁺IgG⁺ or total IgG⁺ live CD20⁺ splenic B cells (right, n = 6 donors).

(C-F) Wilcoxon matched-pairs signed rank test (**** p < 0.0001, ** p < 0.01, * p < 0.05).

See also Figure S1 and Table S1.

Starting from frozen vials of 30–60 × 10⁶ splenocytes, this usually allowed us to sort between 20 and 200 cells of interest, with only a handful of donors above 100 cells. These extremely low frequencies and numbers made tight gate positioning prior to cell-sorting impossible and any further analysis strongly susceptible to even minute contamination. We thus devised a single-cell culture strategy adapted from the Kitamura’s culture system (Nojima et al., 2011) to validate the specificity of sorted cells (Figures 1A and S1B). Throughout this study, this strategy was used to select cells for further VDJ analysis and correct gating strategies a posteriori for single-cell sequencing cell-sorting and FACS analysis of markers of interest using index-sorting information. Such system also allowed us to confirm the presence of true B5-specific B cells in the splenic IgG⁺ (Figure 1B; Table S1), but not IgM⁺CD27⁺ or IgA⁺ (Figures 1C and S1C), MBC compartment of close to 85% of donors born before 1970 whom we could test (56 out of 66). In addition, the estimated frequency of these cells appeared strikingly steady over three decades of life (mean = 0.012% ± 0.002 [SEM]), more than 40 years post initial vaccination. Conversely, no B5-specific IgG⁺ MBCs could be detected from the spleen of any of the 8 organ donors born after 1977 whom we could test, in line with previous ELISPOT or limited dilution assays on PBMCs and spleen of non-vaccinated individuals (Amanna and Slifka, 2006; Crotty et al., 2003; Mamani-Matsuda et al., 2008). These results confirm the exceptional longevity of such MBCs and the fact that cross-reactivation by related Orthopox-virus infections, although impossible to fully exclude (Gilchuk et al., 2016), is unlikely to explain such long-term maintenance in most donors.

Simple retrospective analysis of CD27 staining on all spleens for which we could reproducibly sort cells with over 80% purity showed a significant enrichment of B5-stained cells in the CD27⁺ subset of IgG⁺ MBCs (Figure 1D; p < 0.0001). Culture-validated B5-specific IgG⁺ MBCs were further found almost exclusively in the CD73⁺ and CD21⁺ subset of MBCs (Figures 1E and 1F), with little to no B5⁺ MBCs detected in the CD21^lowCD11c⁺ atypical/effector MBC compartment (Figure 1E). CD45RB and CD69 staining, on the other hand, showed strong donor-to-donor variations with no clear distinction between total and B5-specific MBCs inside a given donor (Figures S1D and S1E). This is in line with the overall variability in CD45RB/CD73 staining observed among spleens collected from over 40 patients aged 4–91 years old (Figures S1F–S1H), although one can note a steady increase in the CD45RB⁻CD73⁺ memory subset with age (Figure S1H), along with the expected decrease in naive B cells (Figure S1G).

Collectively, these results demonstrate the existence of a stable and long-lived population of anti-vaccinia MBCs in the CD73⁺CD21⁺CD27⁺ IgG⁺ MBC compartment of the spleen in vaccinated individuals.

Vaccinia-specific long-lived MBCs display evidence of extended clonal selection

To further characterize the repertoire of long-lived vaccinia-specific MBCs, we next performed V_H sequencing, clonal clustering, and functional analysis of MBC-derived monoclonal antibodies (mAbs) on validated single-cell culture wells using two complementary approaches. First, we single-cell sorted and cultured B5-specific MBCs in a slightly different culture system (see STAR Methods), which permits longer culture duration of memory-derived PCs and improved antibody yield in the supernatant, allowing for direct assessment of mAbs from culture supernatant (Sokal et al., 2021b; Watanabe et al., 2019). As a complementary approach, we sequenced light chains (see Table S2) and subsequently re-expressed the full IgG of the most representative member of each of the top clones from three donors.

B5-specific IgG⁺ MBCs did not display public use of given V_H among donors (Figure S2A) and consisted mostly of IgG1 (Figure S2B). However, in line with a T cell dependent and GC-derived MBC response, they displayed clear evidence of clonal expansion and somatic hypermutation (SHM) (Figures 2A and 2B). The repertoire of B5-specific MBCs notably showed extremely reduced diversity compared with the repertoire of MBCs a couple of months or years post vaccination against influenza (Andrews et al., 2019; Matsuda et al., 2019), yellow fever (Wec et al., 2020), or SARS-CoV-2 (Sokal et al., 2021b; Wang et al., 2021), with the top five clones accounting for 61%–84% of all sequences in all five donors analyzed (Figures 2A and S2C). The top clones for each donor thus represented each from 8 × 10⁴ to slightly over 5 × 10⁵ cells in a human spleen containing an estimated 70 × 10⁹ lymphocytes (Figure S2D) (Trepel, 1974).

Figure 2. Vaccinia-specific long-lived MBCs display evidence of extensive clonal selection.

(A) Clonal distribution forVμ sequencesfrom validated B5-specific IgG⁺ MBCs (n = 5 donors). Outer blue semicircular line indicates the proportion ofsequences belonging to the top five clones for each donor. The total number of sequences is indicated at the pie center. Colored dots highlight shared clones between donors.

(B) Number of mutations in the V_H segment of B5-specific IgG⁺ MBCs grouped by clone of size 5 or more for each donor. Mean ± SEM.

(C) Measured KDvalue(nM) foreachtested supernatants(white, n = 102) or re-expressed antibodies (red, n = 14)from B5-specific IgG⁺ MBCs. Mean KD (line) and total number of tested antibodies (above) per donor are shown.

(D) Representative lineage trees for B5-specific IgG⁺ MBC clones (dot size: number of identical V_H sequences; dot color: isotype). Sequence logo representing all CDR3 inside that clone, as well as V and J calls, is added on the bottom right. Affinities from tested supernatants are displayed on the right side of each tree in front of related sequence nodes.

(E) Representative amino acid V_H alignment for all non-identical sequences at the nucleotidic level from clone 11 (rank 1) of donor HD4.

See also Figure S2 and Table S2.

Diversity was further restricted at the intra-clonal level, as revealed by the narrow distribution of intra-clonal VH mutations (Figure 2B) and affinities (Figures 2C and 2D). Evolutionary tree of the major clones for each donor (Figures 2D and S2E) showed clear evidence of selective sweep and subclonal competition (Horns et al., 2019), and cells bearing the exact same V_H genes could reproducibly be sorted from multiple vials of the same donor in multiple cases. Most anti-B5 antibodies displayed medium to strong affinity against B5, with KD in the order of 10⁻⁸ to 10⁻¹⁰ M(Figure 2C), and all clones exhibited strong convergence at the level of affinity and CDR3 sequences (Figures 2D, 2E, and S2E; Table S2B).

Collectively, these results suggest extensive clonal selection and a role for BCR affinity in the initial selection and/or survival of long-lived MBCs.

Vaccinia-specific long-lived MBCs are enriched in a splenic CD21^hi IgG⁺ MBC subset

Selection of long-lived MBCs as a separate MBC lineage during the early stages of an immune response would most likely result in a unique transcriptional signature for this subset. To test this hypothesis, we next performed single-cell RNA sequencing (scRNA-seq) on single-cell-sorted B5-specific IgG⁺ MBCs, as well as paired total IgG⁺ MBCs and naive B cells from the spleen of six donors. The generated scRNA-seq dataset was further in-tegrated with a previously published dataset that included splenic MBCs from three adult organ donors and three immune thrombocytopenic purpura (ITP) adult patients as well as splenic naive, GC, and MBCs from four young patients with sickle cell disease (Crickx et al., 2021) (see Table S3A).

Unsupervised clustering analysis yielded 10 clusters (Figures 3A and 3B; see Tables S3B and S3C), which included five MBC clusters (clusters 0, 1, 3, 6, 7, and 8) and two clusters regrouping most sorted naive B cells (cluster 2) or most sorted GC B cells (cluster 5). Two additional clusters (clusters 4 and 9) displayed mostly markers linked to poor quality or apoptosis (see Table S3C) and were not further explored.

Figure 3. Vaccinia-specific long-lived MBCs are enriched in a splenic CD21^hi IgG⁺ MBC subset.

(A) UMAP and clustering of all (left) and individual sorted cell types (right) among the 8,326 scRNA-seq analyzed naive, GC, and total or vaccinia-specific memory B cells (n = 9 organ donors [healthy donors, HD]; n = 4 young donors [YD]; and n = 3 immune thrombocytopenia [ITP] patients).

(B) Relative cluster distribution for all sorted MBCs from indicated donor groups (top) and from donors for whom both total and vaccinia-specific MBCs were sorted (matched donors, n = 6 of 9 HD). Bar indicates mean with SEM.

(C) Expression (scaled normalized counts) of selected genes in cells from all clusters.

(D) Representative CD20 and CD21 staining in gated splenic naive and IgG⁺ MBCs, with frequencies and gating strategies for CD21^hiCD20^hi, CD21^intCD20^int, and CD21^loCD20^hi populations (left). Representative overlaid CD11c (right, up) and Tbet staining (right, bottom) in indicated splenic naive and IgG⁺ memory populations.

(E) Representative overlaid CD21 staining in gated circulating (PBMCs) or splenic naive, CD27⁺gD⁺ marginal zone (MZB) and IgG⁺ MBCs.

(F) Frequencies of naive, MZB, and IgG⁺ memory in B cells (left) and of indicated subpopulation in IgG⁺ MBCs (right) from total PBMCs or spleen of organ donors (n = 7).

(G) Frequencies of CD21^hiCD20^hi in splenic IgG⁺ MBCs plotted against donor’s age (n = 42 donors). Linear regression and 95% confidence intervals are shown.

(H) Representative overlaid CD21 and CD20 staining (left) and graph showing the frequencies of CD21^hiCD20^hi in validated B5⁺IgG⁺ or total IgG⁺ live CD20⁺ splenic B cells (right, n = 10 donors).

(I) Representative lineage trees for B5-specific IgG⁺ MBC clones (dot size: number of identical V_H sequence; dot color: index-sorting phenotype).

(B) Ordinary two-way ANOVA (top) and repeated measure two-way ANOVA (bottom) with all comparisons made compared with the “Mem HD” group.

(F) Multiple and (H) single Wilcoxon matched-pairs signed rank test with false discovery rate (FDR) correction for multiple comparisons) (**** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05).

See also Figure S3 and Table S3.

Closer analysis of the six MBC clusters showed two main resting MBC clusters (clusters 0 and 1), expressing TCF7 (Jenks et al., 2018) but separated by their mRNA expression levels of CR2 (CD21), MYC, AHNAK, or COCH (Figures 3C and S3A; see Table S3C). Another memory cluster (cluster 7) displayed a DN2/effector MBC signature (Jenks et al., 2018; Lau et al., 2017; Nellore et al., 2019), which included high expression of CD19 and MS4A1 (CD20), low expression of CD21 (CR2), and unique expression of TBX21 (Tbet), ITGAX (CD11c), and FCRL3/5 (Figure 3C; see Table S3C). The last three MBC clusters (cluster 3, 6, and 8) each displayed different flavors of activation or cytokine signaling signatures, including a TNF/NF-κB/immediate early genes signature in clusters 3 (mostly young donors’ MBCs) and 6 and an interferon/STAT1 signature in cluster 8 (Figure 3C; see Table S3C). The frequencies of MBCs falling into these three clusters, however, appeared highly variable among donors (see Table S3B).

Going against the hypothesis of a unique signature for long-lived MBCs, B5⁺ IgG⁺ MBCs did not segregate as a separate cluster in this analysis and could be found in all MBC clusters (Figure 3A). However, in line with their putative generation a few decades ago, most of these long-lived anti-Vaccinia MBCs were found in the two resting MBC clusters, most of them in cluster 0 with a small, albeit significant, enrichment in this cluster compared with total IgG⁺ MBCs (Figure 3B, p = 0.0007). Differential expression analysis between sorted B5⁺ IgG⁺ and total IgG⁺ MBCs did not highlight any gene signature independent of the overall cellular distribution inside identified clusters (data not shown), confirming the absence of a unique “longevity” signature in vaccinia-specific MBCs. Of note, fractions of long-lived anti-vaccinia MBCs did fall into each of the “activated” memory clusters. The frequencies of B5⁺ IgG⁺ MBCs in these clusters, however, mostly mirrored the ones of total MBCs from the same patients (Table S3B), in line with our previous results on CD69 staining on B5-specific MBCs (Figure S1D).

To validate such results in a broader cohort of donors, we next designed multi-parametric flow cytometry panels that would allow us to separate the main clusters found in our scRNA-seq analysis. Simply looking at CD21 (upregulated on cluster 0 MBCs) versus CD20 expression on the main subpopulations of splenic B cells revealed, in all donors, a substantial population of IgG⁺ MBCs displaying enhanced CD21 and CD20 expression (CD21^hiCD20^hi) compared with the bulk of naive B cells (Figure 3D), reaching fluorescence intensities like the one of IgD⁺CD27⁺ splenic marginal zone B cells (MZB) (Figure 3E). Unlike CD21^lowCD20^hi IgG⁺ MBCs, which also upregulate CD20, these cells did not express CD11c or Tbet (Figure 3D), appeared mostly absent from the blood (Figures 3E, 3F, and S3B) and accumulated in the spleen with age (Figures 3G and S3D). A similar CD21^hiCD20^hi population could also be detected in IgA and IgM-only memory subsets (Figure S3C), highlighting a putative common signature of splenic resident MBC populations. In contrast, CD21^lowCD20^hi DN2/effector MBCs (cluster 7) rarely exceeded a few percent of total splenic IgG⁺ MBCs and did not seem to increase with age (Figure S3D). No clear population of CD71⁺, CD25⁺, or CD95⁺ splenic MBCs could be detected in any of the donors but high frequencies of CD83⁺ IgG⁺ MBCs (marker of cluster 3), and matched B5⁺ IgG⁺ MBCs (HD72; Figures S3F and S3G), could be seen in a small number of donors and were often associated with high frequencies of GC B cells (in young adult donors) or PCs (in older donors) (Figure S3D). This, thus, likely reflects bystander signaling in the context of an ongoing immune response.

In line with our scRNA-seq data, in vitro validated B5⁺ IgG⁺ MBCs were mostly found in both CD21^intCD20^int and CD21^hiCD20^hi MBCs populations but were enriched in the later population (Figures 3H and S3E, p = 0.0371). If both populations of cells represented distinct MBC populations established early on during the GC response, one would expect divergence in the respective repertoire of CD21^hiCD20^hi and CD21^intCD20^int IgG⁺ B5⁺ cells after several decades. Re-interrogating our VDJ sequencing data (Figure 2) in the light of CD21 index-sorting information, however, revealed that for 26 of the 30 clones containing more than 5 members we could detect a mix of CD21^hiCD20^hi and CD21 ^intCD20^int cells (Table S2F), with overall similar frequencies of CD21^intCD20^int for all clones of a given donor (Figure S3H) and random distribution on the evolutionary tree of individual clones (Figures 3I and S3I).

Overall, these results argue in favor of a model in which cells can dynamically transit over time between the two major CD21^hi and CD21^int splenic MBC subsets. The enrichment of long-lived B5⁺ MBCs in the splenic memory CD21^hiCD20^hi compartment nonetheless suggests a survival benefit associ-ated with residing in particular splenic niches.

CD21^hi IgG⁺ splenic MBCs exhibit a NOTCH/MYC/ETV6-driven signature

To better understand the relationship between these two resting splenic MBC subsets, we next compared the transcriptional signature of cells in the corresponding scRNA-seq clusters (clusters 0 and 1) with the one of naive B cells (cluster 2), focusing on donors for which we had sorted both naive and MBCs to be able to properly account for potential donor-specific variations. Rather than separate transcriptional programs (Figures 4A, S4A, and S4B), both MBCs clusters appeared to share a common transcriptional signature that separated them from naive B cells with 133 commonly upregulated genes (gene module M2; Figure 4A) and over 400 commonly downregulated genes (gene module M7; Figure S4C). In addition to this core memory signature, cluster 0 (CD21^hi) MBCs further expressed a unique signature composed of 346 genes (gene modules M0 and M1; Figure 4A), with cluster 1 (CD21^int) MBCs mostly demonstrating intermediate expression of part of these genes as compared to naive B cells.

Figure 4. CD21^hi IgG⁺ splenic MBCs exhibit a NOTCH/MYC/ETV6-driven signature.

(A) scRNA-seq expression data for all donors with matched naive and memory populations and for all genes significantly upregulated in one of the two quiescent IgG⁺ memory clusters (clusters 0 and 1) in one of the three following comparisons: cluster 0 versus 1, cluster 0 versus 2, and cluster 1 versus 2. Row-scaled pseudobulk expression averaged by cluster and donor is displayed, and genes belonging to the HALLMARK_MYC_Targets_v1 (black lines) and RYAN_NOTCH_DIRECT genesets (red lines and gene labels; Ryan et al., 2017) are highlighted.

(B) Preranked geneset enrichment analysis (GSEA) results for all three comparisons performed in (A) and all genesets with a normalized enrichment score (NES) > 1.5 and FDR < 0.01 in one of the 3 comparisons.

(C) Representative enrichment plots for the comparison of cluster 0 versus cluster 1.

(D) Scaled expression (row-normalized area under the curve [AUC] score) of the topmost significantly upregulated SCENIC regulons in indicated clusters.

(E) Overlaid AUC score for MYC, ETV6, and HES1 regulons in cells from indicated clusters.

(F-J) Isolation of paired splenic naive, CD21^int, and CD21^hi IgG⁺ MBC samples from organ donors (n = 4) for bulk RNA-seq and ATAC-seq analysis. (F) Gating strategy, (G) PCA analysis of bulk RNA-seq expression data, (H) representative GSEA enrichment plot, (I) log2 of the fold change (Log2FC) in RNA expression for differentially expressed genes in the comparison of sorted CD21^hiCD20^hi and CD21 ^intCD20^int IgG⁺ MBCs, and (J) row-scaled RNA-seq expression data for all genes of the RYAN_NOTCH_DIRECT geneset detected as differentially expressed in one of the three comparisons: CD21^hi versus CD21^int, CD21^hi versus naives, and CD21^int versus naives. Log2FC in accessibility at the most differentially accessible NOTCH binding regions identified in our bulk ATAC-seq data for each individual gene are displayed on the right of the heatmap for all three comparisons.

See also Figures S4 and S5 and Table S3.

Gene set enrichment analysis (GSEA; Subramanian et al., 2005) confirmed the gradual acquisition of a distinct signature from cluster 2 naive B cells to clusters 1 and 0 MBCs, with no detectable enriched gene set unique to cluster 1 MBCs in the performed comparisons. Multiple metabolism-associated pathways, encompassing protein synthesis, glycolysis, oxidative phosphorylation, and upstream MYC activation, as well as a “NOTCH signaling in B cell lymphoma” signature (Ryan et al., 2017), appeared progressively upregulated from cluster 2 to clusters 1 and 0 (Figures 4B, 4C and S4D). The latest was associated with the upregulation of many known NOTCH2 target genes in B cells (Descatoire et al., 2014), including CR2 (CD21), HES1, HES4, DTX1, CD300A, or MYC (Figure 4A). Network inference analysis using the SCENIC pipeline (Aibar et al., 2017), equally highlighted MYC, ETV6, and HES1 as 3 of the top transcription factors (TFs) whose associated regulons appeared significantly over-expressed in cluster 0 MBCs, with low to intermediate expression in cluster 1 MBCs (Figures 4D and 4E).

To confirm these results in a context of increased sequencing depth and to validate our CD21/CD20-based gating strategy, we next performed bulk RNA and ATAC-seq on sorted CD21^hiCD20^hi and CD21^intCD20^int IgG⁺ MBCs and naive B cells (Figure 4F). Both bulk RNA-seq and ATAC-seq principal-component analysis (PCA) analysis demonstrated a clear separation along the first principal component between naive B cells and the two sorted subsets of MBCs, accounting for close to 50% of the overall variance (Figures 4G and S5A). Although the next principal components mostly highlighted donor-specific variations in each assay, sorted MBC subsets appeared nonetheless separated along the fifth (rPC5) and fourth principal component (aPC4) of each assay, with good correlation when looking at differentially expressed genes driving the separation at the epigenetic level (Figure S5B). Most importantly, GSEA analysis matched our scRNA-seq results, highlighting again NOTCH, MYC, and metabolic pathways-associated gene sets as progressively upregulated from naive to CD21^hi MBCs in the spleen (Figures 4H, S4E, and S4F), with known NOTCH, MYC, and ETV6 target genes mostly upregulated in CD21^hi IgG⁺ MBCs compared with CD21^int IgG⁺ MBCs (Figure 4I). The overall number of significantly differentially accessible regions, however, remained limited between both memory subsets (n = 41 with increased accessibility in CD21^hi, ATAC regions module M0; and n = 15 with decreased accessibility in CD21^hi, ATAC regions modules M4 and M5; Figure S5C). Among direct NOTCH target genes in B cells (Ryan et al., 2017), only HES4 did appear differentially accessible at the level of the chromatin (Figures 4J and S5D), and most other NOTCH or MYC target genes like CR2 (CD21) did not show such regulation (Figures 4J and S5E). This was not the case, however, when looking at both MBC subsets compared with naive B cells. Here, clear regulation at the epigenetic level appeared to explain differential expression of many NOTCH targets (Figure 4J), often in association with ETV6 or MYC, as evidenced for IL21R or TNFRSF13B (TACI) (Figures S5F and S5G).

Overall, these results suggest that temporal access to NOTCH ligands or specific cellular interactions rather than a unique epigenetic status regulates most of the transcriptional signature detected in CD21^hi IgG⁺ MBCs compared with CD21^int IgG⁺ MBCs. Different accessibility of NOTCH target genes between naive and MBCs suggests in contrast that they will respond differently to a similar NOTCH triggering.

CD21^hi IgG⁺ splenic MBCs share partial phenotype and homing potential with splenic marginal zone B cells

Among the signature highlighted by our combined single-cell and bulk RNA-seq transcriptomic analysis of splenic CD21^hi IgG⁺ MBCs figured numerous key B cell surface receptors associated with BCR signaling (both activator [CD19, MS4A1 (CD20), CR2 (CD21), CD27, or CD40] and inhibitor [CD22, CR1 (CD35), CD72, LAIR1, CEACAM1 (CD66a), CD84, LILR4A (CD85g), FCRL4, FCRL5, or CD300A]), survival (TNFRSF13B [TACI]), cell interaction, or migration (ICAM1 [CD54], ITGAL [CD11a]) (Figure 5A; Table S3). KLF2- and KLF2-induced genes (Hart et al., 2011; Lechner et al., 2021) were also downregulated in these CD21^hi IgG⁺ MBCs (Figure 5B), including many commonly expressed genes between cluster 1 IgG⁺ memory and naive B cells (Figure 4A; gene module 5: AHNAK, PELI1, RASGRP2, STK38, and EMP3). Many of these markers were previously described as upregulated on unswitched MZBs whose development also depends on NOTCH2 signaling (Descatoire et al., 2014) and KLF2 transcription factor downregulation (Hart et al., 2011).

Figure 5. CD21^hi IgG⁺ splenic MBCs share partial phenotype and homing potential with splenic marginal zone B cells.

(A) scRNA-seq expression (scaled normalized counts) of selected genes coding for integrins, chemokine receptors, and positive and negative regulators of BCR signaling in cluster 0 (CD21^hi), 1 (CD21^int), and 2 (naives).

(B) Enrichment plots for HART_KLF2^–/– FoB_DN geneset (Hart et al., 2011) in the comparison of cluster 0 versus cluster 1 (scRNA-seq) and sorted splenic CD21^hiCD20^hi versus CD21^intCD20^int IgG⁺ MBCs (bulk RNA-seq).

(C-E) Representative overlaid stainings (top) and median fluorescence intensities (medianFI) for indicated markers (bottom) in gated splenic naive, CD21^int, or CD21^hi IgG⁺ MBCs (gated as in Figure 3D), and IgD⁺CD27⁺ marginal zone B cells (MZB) (n = 7-10 HD donors). A non-parametric Friedman test with FDR correction for multiple comparisons (**** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05).

See also Figure S6.

Using multi-parametric FACS analysis, we could further validate most of the selected DEGs at the protein level. CD21^hi IgG⁺ MBCs displayed increased expression of integrins and adhesion molecules usually associated with retention in the marginal zone of the spleen (reviewed in Cyster and Allen, 2019), notably LFA-1 (CD11a/CD18) and ICAM1 (CD54), reaching expressions seen on MZB cells (Figure 5C). CXCR5 downregulation, also key for the relocation outside of the B cell follicle in the spleen (Cyster and Allen, 2019), appeared mostly common to all MBC populations (Figure 5C). CD21^hi IgG⁺ MBCs also displayed enhanced membrane expression of numerous BCR complex members listed above (Figure S6A), apart from CD85g, FCRL4,and FCRL5 whose expression could not be detected. Finally, NOTCH-regulated surface molecules, including NOTCH2 (Figure 5D), TACI (Figure 5E), and CD27 (Figure S6B) were also upregulated on CD21^hi compared with CD21^int IgG⁺ MBCs, as well as BAFFR and known MZB markers CD1c and TLR10. However, it should be noted that IgG⁺ MBCs still displayed key differences with MZB, notably reduced CD148 and enhanced CD80 and CD86 expression (Figure S6C) as well as markedly different integrin ß1/integrin ß7 expression profiles, with total IgG⁺ MBCs displaying a unique integrin ß1^hi/integrin ß7^low profile among splenic B cell subsets (Figure S6D). This likely reflects distinct differentiation pathways or tissue of origin and accessibility to specific niches in some organs like gut-associated lymphoid tissues (Gorfu et al., 2009).

Overall, these results validated the transcriptional signature identified for CD21^hi IgG⁺ MBCs and strongly suggested that CD21^hi IgG⁺ MBCs and MZB share common features related to activation status and relocation to the marginal zone outside the B cell follicle. Strong inter-donor variability could be seen for some of these markers, such as NOTCH2 (Figures 5D and S6E). This donor-specific variability, however, appeared correlated with overall CD21 expression on naive B cells (Figures S6F and S6G), suggesting inter-donor differences in the availability of NOTCH ligands in the spleen.

CD21^hi IgG⁺ splenic MBCs display enhanced metabolic activity

In addition to the NOTCH/splenic marginal zone signature, our transcriptomic data also pointed toward transient MYC activation and various increased metabolic pathways in IgG⁺ resting MBCs subsets, raising the question of a dedicated metabolic pathway associated with MBC longevity.

Metabolic profiling of IgG⁺ memory and naive B cells confirmed a moderate increase in basal glycolysis and oxidative phosphorylation (OXPHOS) (Figures 6A–6D), but without any clear increased dependency on OXPHOS in MBCs (Figures 6E, S7A, and S7B). IgG⁺ splenic MBCs also demonstrated a strong increase in glycolytic reserve and spare respiratory capacity compared with naive B cells (Figures 6A–6D), suggesting increased metabolic fitness. This was further associated with a significant increase in mitochondria amount and potential (Figure 6F), KREBS cycle metabolites (Figure 6G), and ex vivo glucose import (Figure 6H) in all resting IgG⁺ splenic MBCs, with an additional increase between CD21^hiCD20^hi and CD21^intCD20^int resting memory populations (Figures 6F and 6H). However, such overall increase in metabolic activity remained far from what can be reached upon in vivo activation as best reflected by glucose import (Figure 6I) and expression of metabolic enzyme (Figure S7C) in resting naive and MBC clusters compared with GC B cells in our young donor cohort.

Figure 6. CD21^hi IgG⁺ splenic MBCs display enhanced metabolic activity.

(A and B) (A) Representative Seahorse XF Gluco stress test’s extracellular acidification rate (ECAR) results and (B) calculated basal glycolysis, glycolytic activity, and glycolytic reserve for sorted splenic naive and IgG⁺ MBCs (n = 5 HD donors).

(C-E) (C) Representative Seahorse XF Mito stress test’s oxygen consumption rate (OCR) results; (D) calculated basal respiration, maximal respiration, and spare capacity; and (E) calculated OCR/ECAR ratio prior to oligomycin addition for sorted splenic naive and IgG⁺ MBCs (n = 5 donors). (A-E) Mean ± SEM.

(F) Representative overlaid histograms (top) and graph showing the fold change in TMRE and MitoID median fluorescence intensities (bottom) between splenic naive and CD21^hi or CD21^int IgG⁺ MBCs (gated as in Figure 3D, n = 8 HD donors).

(G) Row-scaled UHPLC-MS quantification of basal expression of KREBS cycle metabolites between splenic naive and total IgG⁺ MBCs (n = 3 HD donors).

(H) Representative overlaid histograms (left) and graph showing the fold change in 2-NBDG incorporation (right) between splenic naive and CD21^hi or CD21^int IgG⁺ MBCs (n = 8 HD donors).

(I) Representative overlaid histograms showing 2-NBDG incorporation in splenic naive, CD21^hi or CD21^int IgG⁺ memory, and GC B cells from young donors (n = 4 YD donors).

(J) Isotopomer distribution (from no labeled carbon [black, M⁺0] to six [yellow, M⁺6]) in all detected metabolites from splenic naive and total IgG⁺ MBCs at 0, 1, and 6 h post incubation with labeled ¹³C-glucose (n = 3 HD donors).

(K) Representative overlaid histograms (left) and graph showing the fold change in NileRed median fluorescence intensities (right) between splenic naive and CD21^hi or CD21^int IgG⁺ MBCs (n = 8 HD donors).

(L) Representative confocal image of NileRed stained CD21^hi IgG⁺ MBCs.

(M) Representative overlaid FSC and SSC fluorescence intensities in splenic naive and CD21^hi or CD21^int IgG⁺ MBCs.

(E) Wilcoxon matched-pairs signed rank test.

(B, D, F, H, and K) Repeated measure two-way ANOVA with FDR correction for multiple comparisons (****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05).

See also Figure S7.

Both Seahorse profiling (Figures S7D–S7G) and fluxomic experiments using labeled glucose and glutamine (Figures 6J and S7H) further demonstrated the limited use of metabolites downstream glycolysis and glutaminolysis in the KREBS cycle in both splenic naive and MBCs. We also could not detect any significant export of lactate in both human B cell subsets, as previously reported in murine B cells (Waters et al., 2018). Instead, progressive labeling of ATP and UTP in these experiments suggested a predominant use of glucose and glutamine as building blocks for cellular anabolism, notably ribose synthesis through the pentose phosphate pathways for glucose (Figure 6J) and pyrimidine biosynthesis for glutamine (Figure S7H).

The corollary of such results is a key dependency on fatty acid oxidation (FAO) for energy production in resting human B cell subsets. Supporting such hypothesis, etomoxir-mediated blockade of fatty acid import in the mitochondria did lead to a strong yet equally proportional reduction of OXPHOS in both naive and IgG⁺ MBCs (Figures S7I and S7J). Neutral lipids, stored in lipid droplets, were also detected in all subsets, with a gradual increase in both stored neutral lipids and overall cell size from naive to CD21^hi IgG⁺ MBCs (Figures 6K–6M). As etomoxir blockade of fatty acid oxidation led to a similar increase in the cytoplasmic content in neutral lipids in all subsets (Figure S7K), it, however, is likely that the size of the storage pool for such lipids is simply tailored to the ongoing metabolic needs of given B cell subpopulations.

Overall, our data suggest that the NOTCH/MYC signature of human splenic CD21^hi resting MBCs is mostly driving an overall moderately enhanced metabolism, but no qualitative change in metabolic pathways used. The absence of a clearly dedicated metabolic pathway for long-lived memory B cells was further confirmed by the absence of detectable differences in metabolic activity between B5⁺ IgG⁺ MBCs and total splenic IgG⁺ MBCs for all FACS-based assays that we could implement on such low cell numbers (Figures S7L and S7M).

IgG⁺ splenic MBCs acquire elongated telomeres that last for decades

An additional key requirement for long-term survival of any cell is to avoid entry into senescence. Telomeres, which protect chromosome extremities, are notably shortened upon each cellular division as well as under external stress (Rossiello et al., 2022), and this can lead to shortening of telomeres below a critical length, usually 4–5 kb. HSCs or tumor cells thus specifically reexpress enzymes participating to telomere elongation, in particular TERT, the catalytic subunit of the telomerase complex (Blackburn et al., 2015). None of these enzymes, apart from NHP2, appeared significantly over-expressed at the mRNA level in resting MBCs in our scRNA-seq dataset (Figure 7A). In contrast, strong expression of numerous members of the shelterin (dedicated to structure and protection of telomeres (de Lange, 2018)) and telomerase complex could be detected in actively dividing GC B cells, in line with the previously described telomerase activity in GC B cells (Hu et al., 1997; Martens et al., 2002; Norrback et al., 2001; Weng et al., 1997). TERT transcripts could not be detected in any of the assayed cell types in our scRNA-seq dataset.

Figure 7. IgG⁺ splenic MBCs exhibit elongated telomeres that last for decades.

(A) scRNA-seq expression (scaled normalized counts) of detected genes coding for members of the shelterin complex or linked to telomerase biogenesis/activity and MKI67 and signature score for G2M and S cell-cycle phases in cluster 0 (CD21^hi), 1 (CD21^int), 2 (naives), and 5 (GC).

(B) Representative TeloFish staining in a B5⁺ IgG⁺ splenic MBCs (top) and an internal control Muntjac cell (bottom).

(C) Median number of TeloFish spots detected per cell of each sorted cell type for each donor analyzed.

(Dand E) (D) Representative distribution of TeloFish spot intensities (ratio over average intensity of spots in Muntjac cells), and (E) variation in average intensity per spots as compared to naive B cells from the same donor in indicated subsets (n = 19 HD donors).

(F) Representative gel for telomere restriction fragment (TRF) analysis (left) and estimated telomere length (right) in indicated subsets (n = 4 HD donors).

(G) Average intensity per spots in indicated subsets compared with donors age (n = 19 HD donors). Linear regression and 95% confidence intervals are shown. (H-J) Representative TeloFish staining (H), representative distribution of TeloFish spots intensities (I), and variation in average intensity per spots as compared to naive B cells from the same donor in indicated subsets (J) (n = 6 HD donors).

(D and I) Kruskal-Wallis test and (C, E, F, and J) repeated measures one-way ANOVA with FDR correction for multiple comparisons (**** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05, n.s. non-significant).

To study telomere length in long-lived MBCs, in a setting that would allow studying only a few hundred cells and allow the resolution of B cell populations beyond the simple separation of CD27⁺ memory B cell and CD27⁻ naive B cells (Martens et al., 2002; Son et al., 2003; Weng et al., 1997), we turned to a FISH-based assay (Benyelles et al., 2019) (Figure 7B). Separating IgG⁺CD27⁺ MBCs from unswitched IgD⁺CD27⁺ marginal zone B cells revealed a striking difference between the two populations in the intensity, but not number, of imaged telomere spots (Figures 7C–7E). The significantly increased fluorescence intensity in IgG⁺ MBCs compared with naive B cells suggested elongated telomere length that was further confirmed using a classical telomere restriction fragment (TRF) assay (13.3 ± 0.4 kb versus 7.7 ± 0.6 kb; Figure 7F).

Analysis of nineteen donors aged 20 to 90 years old further showed a gradual decrease in telomere length with age in naive and IgG⁺ MBC populations (Figure 7G). B5⁺CD27⁺IgG⁺ MBCs displayed a slightly reduced telomere length compared with the bulk of CD27⁺IgG⁺ MBCs in each of the six donors that we could analyze (Figures 7H–7J), confirming their generation more than five decades ago. In all cases, however, B5⁺CD27⁺IgG⁺ MBCs maintained telomere length superior to the one of naive B cells from the same donor.

Collectively, these data suggest that telomeres are selectively elongated in GC-derived switched MBCs and that such long telomeres persist for decades, allowing long-term protection against senescence in these long-lived somatic cells.

Discussion

Immune memory in humans has been shown to extend well beyond decades, with some vaccination and natural infection settings known to offer life-long protection via serum antibodies or MBCs (Amanna, 2007; Bauer and Jilg, 2006). The ability to sort vaccinia B5-specific long-lived MBCs with high purity put us in a unique position to study MBCs generated decades ago and to address the question of whether these long-lived MBCs represent a separate subset of MBCs in the spleen with distinct survival mechanisms.

A first important result came from analyzing telomere length in splenic B cells. In all donors analyzed, and irrespective of age, IgG⁺ CD27⁺ MBCs had markedly longer telomeres than naive B cells, in contrast to what has been described in memory T cells (Fali et al., 2019; Martens et al., 2002; Rufer et al., 1999). Telomeres in anti-B5 B cells were slightly shorter than the ones of total IgG⁺ MBCs from the same donor; however, more than 40 years after being generated, these cells still displayed longer telomeres than their naive B cell counterparts. In the absence of telomerase activity, the length of telomeres in a somatic cell is mostly a marker of its proliferative history. It has been estimated using deuterium labeling that human CD8 memory T cells, generated by yellow fever vaccination, presented a doubling time of over 450 days (Akondy et al., 2017). We do not have these figures for human MBCs, but this number is similar to what has been described for MBCs in mouse models (Jones et al., 2015). The slow decline of telomere length with age in total IgG⁺ MBCs and the slightly shorter telomeres in presumably older anti-vaccinia MBCs are also in line with such a division rate. This slow erosion also favors a model in which early elongation at the naive to memory differentiation step, rather than periodical reactivation of telomerase throughout life, accounts for the telomere size seen in IgG⁺ switched MBCs. Strong telomerase activity in GC B cells has been reported, but results regarding whether this leads to elongated telomeres in generated MBCs were still controversial (Hu et al., 1997; Martens et al., 2002; Norrback et al., 2001; Son et al., 2003; Weng et al., 1997). Our data strongly suggest that the passage in GCs provides GC-derived MBCs with a reserve in telomere provision, in the 10–15 kb range. This reserve appears to last for decades and gives them, in theory, the capacity to divide extensively and differentiate into PCs if confronted to their cognate pathogen while having telomere length far above senescence-prone values. We did not detect any senescence signature in anti-B5 MBCs, and these cells readily proliferated and differentiated into PCs in our single-cell culture settings, as previously seen in limiting dilution or ELISPOT assays (Amanna and Slifka, 2006; Crotty et al., 2003; Mamani-Matsuda et al., 2008).

A second likely mark of GC imprinting appeared when analyzing the repertoire of anti-B5 MBCs. Only a handful of clones with limited intra-clonal diversity and a narrow range of intra-clonal affinities persisted over such an extended period, with mostly mid to high affinity antibodies (KD in the order of 10⁻⁸ to 10⁻¹⁰ M). We do not have a longitudinal follow-up on such a repertoire, but comparison with repertoires described in other vaccinal or natural infection settings up to a year post initial priming suggests that the affinity-driven selection process of the memory pool could extend well over the first year (Andrews et al., 2019; Sokal et al., 2021b; Wang et al., 2021; Wec et al., 2020). This result implies that the antigen be present and accessible during this period. Extended presentation of antigens by follicular dendritic cells (FDCs) has been hypothesized based on limited degradation of presented antigenic complexes at their cell surface (Heesters et al., 2013, 2014; Zinkernagel, 2018). Whether FDCs inside GCs represent the only location of antigenic depots in the context of vaccinia or other infection and vaccination settings, however, remains to be determined. It has also been reported in mice that many GC-derived MBCs present low-affinity to the original immunizing antigen (Viant et al., 2020). Our sorting strategy could have missed some of these low-affinity MBC clones, although human MBCs with affinities between 10⁻⁷ and 10⁻⁸ M have been sorted against proteins of approximatively the same size as B5, such as SARS-CoV-2 RBD, as a bait antigen (Sokal et al., 2021b).

A key question remains whether such selection of long-lived MBCs is also associated with an early imprinting of a unique transcriptional or metabolic profile. What seems to emerge from the analysis of MBCs generated decades ago is more along the way of a dynamic recirculation through a specific microenvironment providing the longevity signal. Like most switched MBCs, B5-specific long-lived MBCs were almost exclusively IgG⁺CD27⁺CD73⁺CD21⁺ and were mainly found in two resting splenic clusters expressing TCF7, a transcription marker associated with central memory (Jenks et al., 2018; Nellore et al., 2019), but not TBX21 (Tbet) (Johnson et al., 2020). These two subsets could be separated based on their intermediate or high expression of CR2 (CD21), with a constant circulation between them, as recently hypothesized by Kibler et al. based on bulk VDJ repertoire analysis and in vitro NOTCH triggering experiments (Kibler et al., 2021). This dynamic recirculation was best exemplified by the presence of B cells belonging to both clusters in most individual B5-specific clones, although B5-specific MBCs appeared overall enriched in the CD21^hi subset in all donors.

The transition to the CD21^hi subset appears associated with the gradual acquisition of a NOTCH/MYC signature akin to the one described for unswitched murine and human marginal zone B cells (Descatoire et al., 2014; Gaudette et al., 2021; Sintes et al., 2017), with nevertheless some sharp differences, the most notable being the absence of CD148 membrane tyrosine phos-phatase expression, the lack of a detectable mTOR signature, and a much higher expression of the CD80 and CD86 costimulatory molecules. In line with the growth-promoting role of the MYC transcription factor, CD21^hi IgG⁺ MBCs also displayed a moderate increase in cell size and basal metabolic activity. Such phenotypic, transcriptomic, and metabolic signatures for CD21^hi switched MBCs could apriori evokea “primed” state for this subset, as suggested in previous reports on similar CD21^hi B cell subsets (Ettinger et al., 2007; Gaudette et al., 2021; Kibler et al., 2021). However, several elements in the characterization of such CD21^hi splenic MBC population invite us to revisit this interpretation. First, the enrichment of anti-B5 MBCs in this subset suggests a key survival benefit associated with transiting through the CD21^hi cell state and its associated niche. This was also reflected at the level of TACI expression, a receptor for the BAFF cytokine, that has recently been linked to MBCs survival in mice (Müller-Winkler et al., 2021) and appeared to be under the direct control of NOTCH signaling in human splenic IgG⁺ MBCs. Second, parallel upregulation of both activating and inhibitory members of the BCR complex could be detected and no metabolic switch could be seen, unlike true human B cell activation (Waters et al., 2018). IgG⁺ splenic MBCs appear to mostly depend on fatty acid oxidation for their energy production, like their quiescent central memory CD8⁺ T cell counterpart (O’Sullivan et al., 2014), and additional glucose uptake seems mostly dedicated to fueling cellular anabolism. One could propose that these CD21^hi splenic MBCs are in a state of cellular idling (van Velthoven and Rando, 2019), balancing activation and restraint to efficiently maintain their responsiveness and fitness in the competitive microenvironment of the spleen while remaining quiescent.

As part of their distinct transcriptional signature, human CD21^hi IgG⁺ MBCs also displayed the integrin and chemokine profiles described for optimal retention of murine MZB in the splenic marginal zone (Lu and Cyster, 2002). Only indirect data exist regarding the precise positioning of IgG⁺ switched MBCs in the human spleen (Ettinger et al., 2007; Steiniger, 2015; Steiniger et al., 2005; Tangye et al., 1998), in part due to technical limitations in staining membrane IgG on histological sections. All reports, however, points toward a IgD⁻IgM⁻CD27⁺ splenic MBC population intermixed with unswitched IgD⁺CD27⁺ marginal zone B cells, CD3⁺ T cells, ILCs, CD11c⁺ DCs, neutrophils, and MAdCAM⁺ stromal cells in the perifollicular area of the human spleen (Ettinger et al., 2007; Magri et al., 2014; Puga et al., 2011; Steiniger et al., 2005, 2014; Tangye et al., 1998), which resembles the splenic marginal zone (MZ) described in mice and is often labeled as such (Steiniger, 2015). Our results suggest that this area of the human spleen, which provides notable access to ICAM-1/LFA-1-mediated interactions with stromal cells and BAFF and APRIL locally produced by neutrophils (Puga et al., 2011), could serve as a key niche for splenic switched MBCs survival in addition to contributing to their reactivation (Ettinger et al., 2007; Kibler et al., 2021).

In conclusion, we propose that the contraction of the MBC pool, probably lasting for a few years, seems to be based on their BCR affinity for the antigen. Their journey into GC imprints them with a long-lasting potential, as best exemplified by telomere elongation. This appears to be further complemented by regular transit into a specific splenic niche providing critical stimuli allowing them to stay functionally alert up to 60–70 years after their generation.

Limitations of the study

Major limitations of our work are inherent to the timing and technical challenges linked with the study of such long-lived MBCs. We have focused on an in-depth analysis of the response against the sole B5 antigen, and it cannot be excluded, at that stage, that different mechanisms might regulate the response against other epitopes. Additionally, studying MBCs generated more than 40 years ago has forced us to focus on donors aged 50 years or more. Many studies have already tracked the early steps of a primary MBC response in human. Our study suggests that extending such follow-up well over 1 year and to secondary lymphoid organs might be required to fully understand the mechanism at the basis of long-lived MBCs generation in humans and overall memory longevity in the context of emerging viral threats such as SARS-CoV-2 or monkeypox.

Star★Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
β7 (BV711)	BD Biosciences	Clone ID: FIB504; Cat# 744012; RRID: AB_2741930
BAFF-R (PE)	BioLegend	Clone ID: 11C1; Cat# 316905; RRID: AB_528982
CD1c (AF647)	BioLegend	Clone ID: L161; Cat# 331510; RRID: AB_1186032
CD3 (BUV395)	BD Biosciences	Clone ID: UCHT1; Cat# 563546; RRID: AB_2744387
CD3 (PerCP)	SONY	Clone ID: OKT; Cat# 2186690
CD3 (AF700)	BD Biosciences	Clone ID: UCHT1; Cat# 557943; RRID: AB_396952
CD11a (APC)	BioLegend	Clone ID: HI 111; Cat# 301212; RRID: AB_314150
CD11b(PE)	ImmunoTools	Clone ID: MEM-174; Cat# 21279114
CD11c (APC)	BD Biosciences	Clone ID: B-ly6; Cat# 560895; RRID: AB_2033936
CD11c (PE-Cy7)	BioLegend	Clone ID: Bu15; Cat# 337215; RRID: AB_2129791
CD11c BV605)	BD Biosciences	Clone ID: B-ly6; Cat# 563929; RRID: AB_2744276
CD14 (BUV395)	BD Biosciences	Clone ID: M φP9; Cat# 563561; RRID: AB_2744288
CD14 (PerCP)	BD Biosciences	Clone ID: M φP9; Cat# 345786; RRID: AB_2868812
CD14 (AF700)	BD Biosciences	Clone ID: M5E2; Cat# 557923; RRID: AB_396944
CD16 (BUV395)	BD Biosciences	Clone ID: 3G8; Cat# 563784; RRID: AB_2744293
CD16 (PerCP)	BioLegend	Clone ID: 3G8; Cat# 302030; RRID: AB_940380
CD16 (AF700)	SONY	Clone ID: 3G8; Cat# 2110130
CD18 (PE)	eBiosciences	Clone ID: CLB_LFA_1/1; Cat# 12-0187-7
CD19 (AF700)	BioLegend	Clone ID: HIB19; Cat# 302225; RRID: AB_493750
CD20 (APC-H7)	BD Biosciences	Clone ID: 2H7; Cat# 560853; RRID: AB_10561681
CD21 (BV711)	BD Biosciences	Clone ID: B-ly4; Cat# 563163; RRID: AB_2738040
CD21 (PE-Cy7)	BD Biosciences	Clone ID: B-ly4; Cat# 561374; RRID: AB_10681717
CD22 (PE)	BioLegend	Clone ID: HIB22; Cat# 302506; RRID: AB_2074593
CD24 (FITC)	BD Biosciences	Clone ID: ML5; Cat# 560992; RRID: AB_10562033
CD24 (PeCy7)	SONY	Clone ID: ML5; Cat# 2155600
CD25 (BUV737)	BD Biosciences	Clone ID: 2A3; Cat# 612806; RRID: AB_2870132
CD27 (BV421)	BD Biosciences	Clone ID: M-T271; Cat# 562513; RRID: AB_11153497
CD27 (APC)	BioLegend	Clone ID: M-T271; Cat# 356410; RRID: AB_2561957
CD27 (PeCy7)	SONY	Clone ID: M-T271; Cat# 2382060
CD29 (PE)	BioLegend	Clone ID: TS2/16; Cat# 303004; RRID: AB_314320
CD35 (PE)	BioLegend	Clone ID: E11; Cat# 333405; RRID: AB_2085021
CD38 (PerCP-Cy5.5)	BD Biosciences	Clone ID: HIT2; Cat# 551400; RRID: AB_394184
CD38 (BV421)	BioLegend	Clone ID: HIT2; Cat# 303525; RRID: AB_10900230
CD38 (APC)	BioLegend	Clone ID: HIT2; Cat# 303510; RRID: AB_314362
CD39 (PE-Cy7)	eBiosciences	Clone ID: eBioA1; Cat# 25-0399-41; RRID: AB_1582281
CD40 (PE)	IoTest	Clone ID: MAB89; Cat# PN IM1936U
CD45RB (AF647)	EXBIO Praha	Clone ID: MEM-55; Cat# A6-224-T100; RRID: AB_10734305
CD45RB (PE)	BioLegend	Clone ID: MEM-55; Cat# 310204, RRID: AB_314807
CD54 (FITC)	BioLegend	Clone ID: MEM-111; Cat#313103
CD66a (PE)	R and D Systems	Clone ID: 283340; Cat# FAB2244P; RRID: AB_2077348
CD69 (PE)	BioLegend	Clone ID: FN50; Cat# 310905; RRID: AB_314840
CD72 (PE)	BioLegend	Clone ID: 3F3; Cat# 316207; RRID: AB_2819945
CD73 (PE)	BioLegend	Clone ID: AD2; Cat# 344003; RRID: AB_1877224
CD73 (BV605)	BioLegend	Clone ID: AD2; Cat# 344023; RRID: AB_2650973
CD73 (BUV737)	BD Biosciences	Clone ID: AD2; Cat# 612813; RRID: AB_2870138
CD80 (PE)	BD Biosciences	Clone ID: L307.4; Cat# 557227; RRID: AB_396606
CD83 (PE)	BioLegend	Clone ID: HB15e; Cat# 305308; RRID: AB_314516
CD84 (PE)	BioLegend	Clone ID: CD84.1.21; Cat# 326007; RRID: AB_2074766
CD86 (PE)	BD Biosciences	Clone ID: 2331; Cat# 555658; RRID: AB_396013
CD148 (PE)	BioLegend	Clone ID: A3; Cat# 328708; RRID: AB_2174826
CD300a (AF647)	BD Biosciences	Clone ID: MEM-260; Cat# 566342; RRID: AB_2739699
CD305 (PE)	BD Biosciences	Clone ID: DX26; Cat# 550811; RRID: AB_393896
CXCR5 (AF488)	BD Biosciences	Clone ID: RF8B2; Cat# 558112; RRID: AB_397034
FVS575V	BD Biosciences	Cat# 565694; RRID: AB_2869702
IgA (FITC)	Dako	Polyclonal; Cat# F0057
IgA (PE)	Miltenyi Biotec	Clone ID: IS11-8E10; Cat# 130-113-476; RRID: AB_2733861
IgD (PE-CF594)	BD Biosciences	Clone ID: IA6-2; Cat# 562540; RRID: AB_11153129
IgD (FITC)	Thermo Fisher Scientific	Polyclonal; Cat# H15501; RRID: AB_2536563
IgG (BV510)	BD Biosciences	Clone ID: G18-145; Cat# 563247; RRID: AB_2738093
IgM (BV605)	BioLegend	Clone ID: RMM-1; Cat# 406523; RRID: AB_2563358
Notch2 (PE)	BioLegend	Clone ID: MHN2-25; Cat# 348303; RRID: AB_2153620
TACI (PE)	BioLegend	Clone ID: 1A1; Cat# 311906; RRID: AB_2203401
TLR10 (PE)	IMGENEX	Clone ID: 158C1114; Cat# IMG-386D; RRID: AB_1152239
Biological samples
Cryopreserved splenocytes and PBMCs from HD and YD donors.	This publication/Institut Necker Enfants Malades	N/A
Chemicals, peptides, and recombinant proteins
Vaccinia Virus (WR) B5R Protein with N-terminal Histidine Tag, Recombinant from baculovirus	BEI Resources Repository, NIAID, NIH	Cat# NR-456
Recombinant human IL-2	PeproTech	Cat#200-02
Recombinant human IL-4	PeproTech	Cat#200-04
Recombinant human IL-21	PeproTech	Cat#210-21
Recombinant human BAFF	PeproTech	Cat#310-13
Deposited data
Raw and analyzed scRNA-seq dataset	This paper	ArrayExpress: E-MTAB-11626
Raw and analyzed scRNA-seq dataset	Crickx et al. (2021)	ArrayExpress: E-MTAB-9955
Raw and analyzed RNA-seq dataset	This paper	ArrayExpress: E-MTAB-11622
Raw and analyzed ATAC-seq dataset	This paper	ArrayExpress: E-MTAB-11625
Single cell culture VDJ sequences	This paper	DDBJ/EMBL/GenBank: KFRA00000000-KFRE00000000 (BioProject: PRJNA821157)
Experimental models: Cell lines
MS40lo cell line	G. Kelsoe’s lab (Duke University)	Luo et al. (2009)
B21 cell line	This publication	N/A
Muntjac cell line	P. Revy’s lab (Institut Imagine)	Benyelles et al. (2019)
Software and algorithms
Flowjo v10.7.1	FlowJo, LLC	https://www.flowjo.com; RRID: SCR_008520
GraphPad Prism v9	GraphPad	https://www.graphpad.com; RRID: SCR_002798
Codon Code Aligner v9	Codon Code Corporation	https://www.codoncode.com/
R v4.0.2	R Foundation	https://www.r-project.org; RRID: SCR_001905
RStudio v1.3.1056	RStudio	https://rstudio.com; RRID: SCR_000432
IgBLASTn v1.16.0	NCBI	https://www.ncbi.nlm.nih.gov/igblast/; RRID: SCR_002873
HT Data analysis software 12.2.2.26	Sartorius/ForteBio	https://www.sartorius.com; RRID: SCR_003935
Icy (v2.3.0.0)	Icy	http://icy.bioimageanalysis.org/; RRID: SCR_010587
Fiji (ImageJ2 v2.3.0/1.53f)	Fiji	https://fiji.sc/; RRID: SCR_002285
Adobe Illustrator (CS6)	Adobe	https://www.adobe.com/products/illustrator.html; RRID: SCR_010279
Biorender	Biorender	https://biorender.com

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Pascal Chappert (pascal.chappert@inserm.fr).

Materials availability

No unique materials were generated for this study.

Experimental Model and Subject Details

Study participants

All healthy donors (HD) spleen and PBMCs samples used in this study were from organ donors who died from stroke or head trauma and were obtained from the Agence de la Biomຝecine (PFS07-009). None of them had lymphoma or autoimmune disease and donors with positive HIV, HCV or HBV serology or known immune pathology were excluded. While we didn’t have access to vaccination records from these donors, we postulated that donors born before 1970 would likely have received at least one shot of theanti-small-pox vaccine (vaccinia) as part of the existing vaccination requirements at the time. Conversely, donors born after 1980, and the official eradication of Smallpox, are unlikely to have been vaccinated and thus should not harbor smallpox/vaccinia-specific MBCs. All young donors (YD) spleen included in some of the analyses were from patients with sickle cell disease who were splenectomized between the age of 4 and 5 years. Information on the age, sex, and other clinical features of the cohort are provided in Table S1. Throughout the manuscript, all patients are referred to using the nomenclature presented in Table S1. No sample size calculation, randomization, or blinding were performed for this study.

This study was conducted in compliance with the Declaration of Helsinki principles and was approved by the Agence de la Bio-médecine and the Institutional Review Boards Comite de Protection des Personnes (CPP) Ile-de-France II, in the context of French laws regulating organ donation. Parents from young children with sickle cell disease provided written informed consent before the collection of splenic samples.

Method Details

Spleen sample processing

Immediately after splenectomy, splenic tissues were maintained at 4 °C until being crushed using a gentleMacs dissociator (Miltenyi). After filtration and 1:2 dilution by RPMI-1640 (Invitrogen), mononuclear cells were isolated by ficoll density-gradient (Ficoll Paque Plus, GE Healthcare) centrifugation at 2000 rpm for 20 min with the brakes off and then stored in liquid nitrogen. PBMCs were isolated from venous blood samples via standard density gradient centrifugation and then stored in liquid nitrogen.

B5 labeling with Alexa488 and Alexa647

Recombinant vaccinia virus (WR) B5R Protein (BEI, NR-546) (Aldaz-Carroll et al., 2005) was labeled with Alexa Fluor (AF) 488 (Life technologies, A30006) or AF647 (Life technologies, A30009), according to the manufacturer’s recommendations, with an additional spin filter purification step at the end.

Flow cytometry and cell sorting

Cells were slowly thawed in RPMI-1640 (Gibco)-2% FBS (Gibco) at 37 °C and washed twice prior to counting. For simple phenotyping, 2 x 10⁶ cells per condition were stained with the fluorochrome-conjugated antibody mix of interest for 40 min at 4 °C (1:100 for BV711 anti-η7, PE anti-BAFF-R, FITC anti-BCMA, AF647 anti-CD1c, AF700 anti-CD3, PerCP anti-CD3, BUV395 anti-CD3, PE anti-CD6, APC anti-CD11a, PE anti-CD11b, PeCy7 anti-CD11c, BV605 anti-CD11c, AF700 anti-CD14, PerCP anti-CD14, BUV395 anti-CD14, AF700 anti-CD16, PerCP anti-CD16, BUV395 anti-CD16, PE anti-CD18, AF700 anti-CD19, BV711 anti-CD21, PeCy7 anti-CD21, FITC anti-CD24, PeCy7 anti-CD24, BUV737 anti-CD25, PE anti-CD29, PE anti-CD35, PerCPCy5.5 anti-CD38, BV421 anti-CD38, PeCy7 anti-CD39, PE anti-CD40, PE anti-CD45RB, AF647 anti-CD45RB, APC anti-CD49d, FITC anti-CD54, PE anti-CD69, PeCy7 anti-CD71, BV786 anti-CD71, PE anti-CD72, BUV737 anti-CD73, BV605 anti-CD73, PE anti-CD80, PE anti-CD83, PE anti-CD84, AF647 anti-CD85g, PE anti-CD119, PE anti-CD148, PE anti-CD200, APC anti-CD300a, FITC anti-CXCR5, PE anti-FCRL5, FITC anti-IgD, BV605 anti-IgM, PE anti-Notch2, APC anti-OX40L, PE anti-TACI and PE anti-TLR10, 1:50 for APCH7 anti-CD20, BV421 anti-CD27, PeCy7 anti-CD27, APC anti-CD27, PE anti-IgA, PeCF594 anti-IgD and BV510 anti-IgG, 1:25 for PE anti-CD22, PE anti-CD66a, PE anti-CD86, PE anti-CD305, APC anti-FCRL4 and FITC anti-IgA and 1:10 for APC anti-CD11c and APC anti-CD38). Cells were then washed in RPMI-1640 and fixed in 2% PFA.

Samples were acquired using a LSR Fortessa SORP (BD Biosciences). Data were analyzed with FlowJo software. Detailed gating strategies for individual markers are depicted in Figure S1. To account for donor-specific variations in CD21 fluorescent intensities, CD21^hiCD20^hi and CD21^intCD20^int gates were always set-up so that the CD21^hiCD20^hi gate would encompass 5% of the naive B cell population in that sample (Figure 3D). For CD18, CD11b, NOTCH2, CD84, TLR10, TACI and BAFFR (Figures 5C–5E) and for CD22, CD35, CD40, CD72, CD305, CD66a and CD6 (Figures S6A-S6C), corrected median fluorescence intensities are reported, subtracting the intensity values obtained from corresponding isotype controls.

For cell sorting of major B cell population, cells were stained using the same protocol, without the final fixation step, and then sorted on a BD FACS cell sorter Aria III (BD Biosciences, 4-way purity mode). Throughout this study naive B cells were sorted as live CD3⁻CD14⁻CD16⁻CD20⁺CD27⁻IgD⁺CD24^midCD38^mid, marginal zone B cells (MZB) as live CD3⁻CD14⁻CD16⁻CD20⁺CD27⁺IgD⁺ and IgG⁺ switched MBCs (IgG⁺ memory) as live CD3⁻CD14⁻CD16⁻CD20⁺CD27⁺gG⁺, further subdivided as CD21^hiCD20^hi and CD21^intCD20^int (upper and lower 30% of CD21⁺ IgG⁺ MBCs respectively; Figure 4F) for RNA-seq and ATAC-seq analysis.

For B5⁺ MBCs sorting, cells were first stained for 40 min at 4 °C at 10⁸ cells/mL with PerCP-Cy5.5 anti-CD3/CD14/CD16 (1:100 each) and BV510 anti-IgG (1:50). IgG⁺CD3⁻CD14⁻CD16⁻ cells were first enriched on a BD FACSAria III cells sorter (Yield mode). Enriched IgG⁺ cells were further labeled with APC-H7 anti-CD20 (1:50), BV421 anti-CD27 (1:100), AF647 B5 (3 μg), AF488 B5 (5 μg) and 7-AAD (1:100). Single MBCs (B5⁻IgG⁺CD27⁺ or B5⁺IgG⁺CD27⁺) were then sorted directly on appropriate feeder cells at 37 °C on a BD FACSAria III (0-16-0 single cell purity mode). When comparing B5⁺ IgG⁺ with other B cell populations, IgG⁻ cells were enriched in parallel on the cell sorter and stained for IgD, CD24, CD27, CD38 and CD20 as described above. Naive, MZB and IgG⁺ MBCs were then sorted from their respective IgG⁺ or IgG⁻ fractions. To validate phenotype of interest on B5⁺ switched MBCs, additional markers were added to our staining panel on enriched IgG⁺ cells: BV711 anti-CD21 (1:100), PE anti-CD83/CD69 or CD45RB (1:100 each), BV605 anti-CD73 (1:100). Analysis was refined a posteriori using index sorting for each single cell validated in subsequent culture and ELISA. To assess the frequencies of B5⁺ MBCs in the IgM⁺CD27⁺ or IgA⁺ populations, the protocol was adapted to add BV421 anti-CD27, PE anti-IgA and BV605 anti-IgM antibodies (1:100 each) in the first staining mix and start with a parallel enrichment of the main isotype-switched or unswitched MBC populations (IgG⁺CD3⁻CD14⁻CD16⁻, IgA⁺CD3⁻CD14⁻CD16⁻ and IgM⁺CD27⁺CD3⁻CD14⁻CD16⁻ respectively).

For metabolic staining, 2 x 10⁶ cells per condition were either directly stained with the fluorochrome-conjugated antibody mix of interest for 40 min at 4 °C or, when also analysing B5⁺IgG⁺ MBCs, IgG⁺CD3⁻CD14⁻CD16⁻ cells were pre-enriched on a BD FACSAria III (Yield mode). For 2-NBDG incorporation, stained cells were first washed twice with glucose-free RPMI-1640. Cells were then incubated 30 min at 37 °C in glucose free RPMI with NBDG (2-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose) 30μM (Invitrogen, Cat# N13195). For mitochondria labeling, stained cells were incubated 30 min at 37 °C in RPMI 2% FCS with TMRE (tetramethylrhodamine, ethyl ester) at 25 nM (Thermo Fischer, Cat#T669) and Mito ID Red (1/10 000, Enzo, Cat# ENZ-51007-0100). Nile Red was directly added at 1 μg/ml (ThermoFisher, N1142) in the antibody mix. For etomoxir blockade experiments (Figure S6K), etomoxir (100μM) or control DMSO were added for indicated time prior to Nile Red staining.

Single-cell culture systems

hCD40L-expressing MS40L^lo feeder cells expressing CD40L were obtained from G. Kelsoe (Luo et al., 2009). hCD40L⁺hBAFF⁺hIL-21⁺ B21 feeder cells were specifically generated for this study from mouse fibroblast hCD40L-expressing L cells (Kneitz et al., 1999) (see below).

Single IgG⁺ or B5⁺IgG⁺ switched MBCs were single cell sorted in 96-well plates seeded 24h earlier with either MS40^lo feeders and appropriate cytokines (MS40^lo cultures) or mitomycin-treated B21 feeders (B21 cultures). B21 cultures were left at 37 °C for 14-21 days at which point supernatants were tested for IgG and B5 specificities and cell pellets were harvested in 50 μL of RLT buffer (Qiagen) supplemented with 1% 2-mercaptoethanol and subsequently stored at -80 °C until RNA extraction and VDJ sequencing. MS40^lo cultures were performed as previously described (Crickx et al., 2021). Briefly, cells were co-cultured at 37 °C with 5% CO₂ during 21 or 25 days in RPMI-1640 (Invitrogen) supplemented with 10% HyClone FBS (Thermo Scientific), 55 μM 2-mercaptoethanol, 10 mM HEPES, 1 mM sodium pyruvate, 100 units/mL penicillin, 100 μg/mL streptomycin, and MEM non-essential amino acids (all Invitrogen), with the addition of recombinant human BAFF (10 ng/ml), IL2 (50 ng/ml), IL4 (10 ng/ml), and IL21 (10 ng/ml; all Peprotech). Part of the supernatant was carefully removed at days 4, 8, 12, 15, 18, 21 and 25 and the same amount of fresh medium with cytokines was added to the cultures. After 25-29 days of single cell culture, supernatants were harvested and stored at -20 °C and cell pellets were harvested in 50 μL of RLT buffer (Qiagen) supplemented with 1% 2-mercaptoethanol and subsequently stored at -80 °C until RNA extraction and VDJ sequencing.

At the end of each parallel total IgG⁺ and B5-specific MBCs single-cell culture, IgG and B5 ELISA were used to distinguish non-B/poor quality B and non-specific B cell contaminants respectively (Figure S1B). Corrected frequencies of B5⁺ IgG⁺ were calculated as follows: %B5⁺_corrected = %B5⁺_Observed* (B5⁺_activation/IgG⁺_activation)* B5⁺_purity, where B5⁺_activation and IgG⁺_activation correspond to the frequencies of IgG⁺ wells at the end of the culture in sorted B5⁺IgG⁺ and control total IgG⁺ MBCs plates respectively and B5⁺_purity correspond to the frequency of B5⁺ wells among IgG⁺ wells from sorted B5⁺gG⁺ MBCs.

Generation of retroviral expression vectors for human BAFF (hBAFF) and human IL-21 (hIL-21)

5 μg of total RNA extracted from human splenic mononuclear cells were reverse transcribed with random hexamers at 42 °C, using the AffinityScript Multiple Temperature cDNA Synthesis Kit (Agilent) according to manufacturer’s instructions. hBAFF and hIL-21 ORFs were amplified by PCR on this cDNA with the following primers: hBAFF-For-BamHI AAGGATCCGCCGCCGATATGGA TGACTCCACAGA and hBAFF-Rev GGTCACAGCAGTTTCAATGC; hIL21-For-XhoI AACTCGAGCCGCCACCATGAGATCCAGT CCTGGCAAC and IL21-Rev-BglII AAAGATCTCAGGAATCTTCACTTCCGT. PCR were performed with Phusion DNA polymerase (New England Biolabs) under the following conditions: 30 s at 98 °C; 35 cycles 10 s at 98 °C, 30 s at 60 °C (hBAFF) or 62 °C (hIL-21), 1 min at 72 °C; 4 min at 72 °C. hBAFF and hIL-21 amplicons were cloned into TOPO-TA (Life Technologies) intermediate vector and subcloned into BglII/XhoI restriction sites of pMIG (Addgene # 9044, a gift from William Hahn) and XhoI/BglII sites of pRIPChp-aHA (Albagli-Curiel et al., 2007), respectively. Sequences were checked using Sanger sequencing.

Generation of feeder cells expressing hCD40L, hBAFF and hIL-21 (B21 feeders)

Nonreplicative ecotropic γ-retroviral particles were produced by transient lipofection (XtremeGene 9 DNA transfection reagent; Roche) of 293T cells grown in DMEM/F-12 supplemented with 10% FCS (Hyclone) with pMIG-hBAFF or pRIPChp-aHA-hIL-21, pLTR-env, and CMV-Gag/Pol vectors (ratio 2:1:1). At 48 hr after transfection, viral supernatants were filtered through 0.45 μm Minisart High Flow filters (Sartorius) and concentrated on centrifugal filter units (100K Amicon Ultra 15; EMD Millipore). hCD40L-expressing L cells were transiently transfected with pTarget-mCAT-1 expression vector (a gift from Yoshinao Kubo, University of Nagasaki, Japan) with an Amaxa Nucleofector 2b device (Lonza, solution R, program U-030), to allow for expression of SLC7A1 ecotropic receptor, and stably transduced 48 hr later with pMIG-hBAFF (MSCV-hBAFF-ires-EGFP) viral particles in the presence of 8 μg/mL polybrene (H9268; Sigma-Aldrich). hCD40L⁺hBAFF⁺ (GFP⁺) double-positive cells were FACS-sorted after staining with PE-coupled anti-human CD154 (Sony, 2154030; diluted 1:200), and long-term expression of membrane-bound hBAFF was checked by staining with an APC-coupled anti-hCD257 (Biolegend, 366507, diluted 1:100). hCD40L⁺hBAFF⁺ cells were subsequently transduced with pRIPChp-aHA-hIL-21 (pRIP-hIL-21-ires-mCherry) retroviral particles, using the same protocol. GFP⁺mCherry⁺cells were FACS-sorted to isolate hCD40L⁺hBAFF⁺hIL-21⁺ triple positive L cells hereafter named B21. hBAFF and hIL-21 secretion was checked by ELISA.

ELISA

Total IgG and B5-specific IgG from culture supernatants were measured using homemade ELISA. ELISA 96 well plates (Thermo Fisher) were coated with either goat anti-human Ig (10 μg/ml, Invitrogen, H17000) or recombinant B5 protein (2.5 μg/ml, BEI NR-546) in sodium carbonate during 1 hr at 37 °C. After plate blocking with 5% milk in PBS for 1 hr at RT, cell culture supernatants were added for 1 hr at RT. Plates were washed 5 times in PBS and were then incubated for 1 hr at RT with HRP-goat anti-human IgG (1 μg/ml, Immunotech) in 1x PBS buffer with 0.1 % BSAand 0.005% Tween 20 surfactant (Bio Rad). Plates were washed 5 times in PBS and incubated with KPL SureBlue™ TMB Microwell Peroxidase substrate (Eurobio, Cat# 5120-0077) for 5-15 minutes at room temperature. The reaction was stopped by adding 1.8 N H₂SO₄, and optical density was measured at a wavelength of 450 and 620 nm.

IgH sequencing of B5 specific clones

At the end of each single B cell culture, cell pellets were resuspended in 50μl RLT Buffer (Qiagen) and stored at -80 °C. Clones whose culture had proven successful (anti-IgG and B5 ELISA OD > 0.09 for B21 cultures or IgG concentration > 1 μg/mL for MS40^lo cultures) were selected and extracted using the RNeasy Micro kit (Qiagen, B21 cultures) or the NucleoSpin96 RNA extraction kit (Macherey-Nagel, MS40^lo cultures) according to the manufacturer’s instructions. A reverse transcription step was then performed using the SuperScript IV enzyme (ThermoFisher) in a 14 μl final volume (42 °C 10 min, 25 °C 10 min, 50 °C 60 min, 94 °C 5 min) with 4 μl of RNA and random hexamers (Thermofisher scientific). A PCR was further performed based on the protocol established by Tiller et al. (2008). Briefly, 3.5 μl of cDNA was used as template and amplified in a total volume of 40 μl with a mix of forward L-VH PCR1 primers (Table S2E) and reverse Cγ primer and using the HotStar® Taq DNA polymerase (Qiagen) and 40 (MS40^lo culture) or 50 cycles (B21 cultures) of PCR (94 °C 30 s, 58 °C 30 s, 72 °C 60 s). For MS40^lo cultures, PCR products were directly sequenced with the reverse primer CHG-D1 and read on ABI PRISM 3130XL genetic analyzer (Applied Biosystems). For B21 cultures, an additional 50 cycles of PCR (94 °C 30 s, 58 °C30 s, 72 °C 60 s) with a mix of forward L-VH PCR2 primers (Table S2E) was performed prior to sequencing. Sequences qualities were verified using CodonCode Aligner software (CodonCode Corporation).

Computational analyses of VDJ sequences

Processed FASTA sequences from cultured single-cell V_H sequencing were annotated using Igblast v1.16.0 against the human IMGT reference database. Clonal cluster assignment (DefineClones.py) and germline reconstruction (CreateGermlines.py) was performed using the Immcantation/Change-O toolkit (Gupta et al., 2015) on all heavy chain V sequences. Sequences that had the same V-gene, same J-gene, including ambiguous assignments, and same CDR3 length with maximal length-normalized nucleotide hamming distance of 0.15 were considered as belonging to the same clonal group. Mutation frequencies in V genes were then calculated using the calcObservedMutations() function from the Immcantation/SHazaM v1.1.0 R package. VH repartitions were calculated using the countGenes() function from the Immcantation/alakazam v1.1.0.999 R package. Phylogenetic trees were generated using the Immcantation/IgPhyML toolkit via the dowser R package (Immcantation/suite devel docker image (Hoehn et al., 2020) and further visualized using the AncesTree GUI (Foglierini et al., 2020). Prior to mutational tree reconstruction, the first 18bp of each V_H sequence was reverted to the reconstructed germline to avoid PCR2 primers-induced bias between B21 and MS40^lo culture. CDR3 sequence logos were generated using the ggseqlogo v0.1 R package. Further clonal analyses on all productively rearranged sequences were implemented in R. Graphics were obtained using R packages ggplot2 v3.3.5 and pheatmap v1.0.12.

Recombinant antibody production and purification

IgH and IgL chains from clones of interest were cloned as previously described (Tiller et al., 2008). Briefly, HEK293T cells were transfected with plasmid DNA encoding paired IgH and IgL chains using Jet Prime kit (Polyplus). Recombinant antibodies were purified from supernatants using protein-G beads using manufacturer’s instructions (GE Healthcare).

Affinity measurement using biolayer interferometry (Octet)

Biolayer interferometry assays were performed using the Octet Red96 (Sartorius). Anti-Human Fc Capture (AHC) biosensors (Sartorius Cat# 18-5060) were immersed in supernatants from single-cell MBC cultures or re-expressed monoclonal antibody at 25 °C for 500 seconds. Biosensors were equilibrated for 180 s in 1x PBS buffer with 0.1% BSA and 0.005% Tween 20 surfactant (PBS-BT) prior to measurement. Association was performed for 600s in PBS-BT with B5 protein at 100nM followed by dissociation for 900s in PBS-BT for each cycle. Biosensor regeneration was performed by alternating 30s cycles of regeneration buffer (glycine HCl, 10 mM, pH 2.0) and 30 s of PBS-BT for 3 cycles. Traces were reference sensor (unloaded sensor) and reference well (additional association cycle in the absence of B5) subtracted and curve fitting was performed using a local 1:1 binding model in the HT Data analysis software 12.2.2.26 (ForteBio).

Single-cell RNA-seq

The scRNA-seq dataset from 3 ITP patients (ITP1 to ITP3), 4 YD (YD1 to YD4) and 3 HD donors (HD1 to HD3) has been previously published (ArrayExpress database: E-MTAB-9955) (Crickx et al., 2021). A second dataset containing sorted naive, IgG⁺ and B5⁺IgG⁺ memory cell had been generated in parallel from 6 HD donors (HD4 to HD9) using the same protocol. Briefly, single-cell mRNA sequencing was performed according to an adapted version of the SORT-seq protocol (Muraro et al., 2016) with primers described in van den Brink et al. (2017). Cells were single-cell FACS-sorted (AriaIII) into 384-well plates containing 384 primers and Mineral oil (Sigma). After sorting, plates were snap-frozen on dry ice and stored at -80 °C. Each plate was designed to contain cells from each donor group. Of note, sorted naive/IgG⁺ and IgG⁺B5⁺ B cells included in the dataset described in this study were prepared using a fast thawing protocol and a subsequent apoptotic cell removal step using the Dead Cell Removal kit (Miltenyi). Sorted cells included in the previously published dataset (Crickx et al., 2021) were prepared using a slow thawing protocol at 37 °C (adapted from 10X genomics CG00039 demonstrated protocol; https://support.10xgenomics.com/single-cell-vdj/sample-prep/doc/demonstrated-protocol-fresh-frozen-human-peripheral-blood-mononuclear-cells-for-single-cell-rna-sequencing), without any additional step.

Further processing of cells, cDNA libraries generation, sequencing and reads alignment was performed at Single Cell Discoveries. For amplification, cells were heat-lysed at 65 °C followed by cDNA synthesis using the CEL-Seq2 protocol (Hashimshony et al., 2016) and robotic liquid handling platforms. After second strand cDNA synthesis, the barcoded material was pooled into libraries of 384 cells and amplified using in vitro transcription (IVT). Following amplification, the rest of the CEL-seq2 protocol was followed for preparation of the amplified cDNA library, using TruSeq small RNA primers (Illumina). The DNA library was paired-end sequenced on an Illumina Nextseq™ 500, high output, with a 2x75 bp Illumina kit. Read 1 was used for identification of the Illumina library barcode, 8 bp cell barcode and 6 bp UMI and read 2 was used to map to the reference transcriptome of Hg38 with STAR (Dobin et al., 2013). Data were demultiplexed and transcript counts were adjusted to the expected number of molecules based on counts, 4096 possible UMI’s and poissonian counting statistics as described in Grün et al. (2014). Mapping and generation of count tables was automated using the MapAndGo/starmap script (https://github.com/anna-alemany/transcriptomics/tree/master/mapandgo).

A total of 3867 MBCs, 1194 vaccinia-specific MBCs, 1812 naives B cells and 192 GC B cells were sequenced in both datasets (Table S3A). Initial quality control and cell filtering was performed using the R Scater package (McCarthy et al., 2017). We removed cells where more than 50% of transcripts mapped to ERCC spike-ins, cells with less than 2000 UMIs or 500 genes detected and kept only genes that were detected in at least 2% of any given sorted cell population. Index-sorting information and parallel in vitro validation using single culture were additionally used to optimized gating strategies and remove potential sort contaminants from further analysis in the context of IgG⁺ B5⁺ MBCs sorts. Overall, 8326 cells and 12466 genes were left for further analysis (Table S3A). Normalization, integration and clustering were then performed through the R Seurat v3 (Stuart et al., 2019) pipeline. Transcript counts were first normalized using the scTransform algorithm (Hafemeister and Satija, 2019), additionally correcting for potential bias related to the detected percentage of mitochondrial genes and selecting for the top 3000 variable features for downstream visualization and clustering analysis. After principal component analysis, sort-specific batch effects were removed using the Harmony algorithm (Korsunsky et al., 2019). The first 30 corrected principal component analysis dimensions were then used to construct a knn graph (k = 20 neighbors) and perform graph-based clustering (Louvain) with a resolution parameter of 0.6 as well as compute the UMAP coordinates for each cell. Final clustering steps as presented in the manuscript were focused solely on sorted MBCs. Dotplots displaying scRNA-seq gene expression (scaled normalized counts) were generated using the Seurat Dotplot() function. Size of dots represents the percentage of cells in the cluster in which transcripts for that gene are detected. Dot color represents the average expression level (scaled normalized counts) of that gene in the population. Heatmap displaying row-scaled pseudobulk expression averaged by cluster and donor were done combining the Seurat AverageExpression() function and the ComplexHeatmap package (Gu et al., 2016).

Differential expression analyses between clusters of interest were then performed using the general linear model (GLM) likelihood ratio test implemented in the edgeR package (Chen et al., 2016) using raw transcript counts of all genes detected in at least 10% of the cells from one of the compared groups and applying a final threshold of 0.25 for log-fold-change and 0.05 for adjusted P-value. Regulon identification and module activity scoring in cells was done using the pySCENIC workflow (Van de Sande et al., 2020) using version 9 of cisTarget motif collection (mc9nr). SCENIC regulon identification pipeline was run independently 20 times and regulons were then curated to remove regulons not detected, or gene not associated to a given regulon, in at least 80% of run and regulons with less than 10 remaining high confidence genes.

Preranked GSEA analyses were performed using the GSEA v4.1.0 App (https://www.gsea-msigdb.org/) using HALLMARK, PID and WP pathways downloaded from the Molecular Signatures Database (MSigDB) of the Broad institute in November of 2021. For each comparison, all genes detected in at least 10% of the cells from one of the groups were ranked based on edgeR log2FC and used as input. Summarized results from GSEA analysis are represented in Figures 4 and S4 as dot plots using the ggplot2 package with dot size representing the Normalized Enrichment Score (NES) and dot color representing the sign of the log fold change * log₁₀(False Discovery Rate value (FDR)). Only genesets with a NES > 1.5 and FDR < 0.01 in one of the 3 comparisons are shown. Top differentially expressed gene upregulated in cluster 0 versus cluster 1 in the leading-edge subset of each geneset are further highlighted on the right side of each plot.

All downstream analyses were performed with R v.3.5.2 on a macOS 10.13.6 (64 bit) system, using the following R packages and libraries: scater v.1.10.1, scTransform v. 0.2.1, harmony v0.99.9, Seurat v.3.1.2, reticulate v1.13, umap v0.2.4.1, edgeR v.3.32.1, ggplot2 v.3.2.1, ComplexHeatmap v2.6.2, SCENIC v 1.1.3 and AUCell v1.12.0.

RNA and ATAC-sequencing

CD21^hi (upper 30%) or CD21^int (lower 30%) IgG⁺ switched memory and naive B cells were sorted from 4 independent donors (Figure 4F). Assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATAC-seq) was performed on 50000 cells of each sorted population, according to published methods (Buenrostro et al., 2013). RNA was extracted from the remaining cells using an RNeasy Plus Microkit (Qiagen) and sent for library preparation and sequencing by Integragen (Evry, France). Libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina protocol according to supplier recommendations and sequenced (2 x 100 bp, paired-ends; Illumina NovaSeq™6000) at IntegraGen (Evry, France).

RNA-seq reads were aligned against the hg38 human genome using STAR (Dobin et al., 2013), keeping only reads with a single alignment. Raw counts for each sample were obtained using featureCounts.py from the SubRead package (Liao et al., 2013). Differential expression of filtered genes (HTSfilter, 11104 genes) was analyzed using DESeq2 v1.28.1 (Love et al., 2014) in R (version 4.0.0). Differentially expressed genes were selected with an adjusted value of P < 0.05 and a fold-change > 2 after Wald’s test and Benjamini-Hochberg correction. Principal component analysis (PCA) was performed on DESeq2-normalized read counts for the fold-change-filtered genes. Preranked GSEA analyses were performed as described above using all genes ranked based on DESeq2 residuals as input. Only genesets with a NES > 1.5 and FDR < 0.05 in one of the comparisons are reported in Figures 5 and S4.

ATAC-seq reads were aligned against the hg38 human genome, using Bowtie1 (Langmead et al., 2009). Open regions were determined with Model-Based Analysis of ChIP-Seq2 (MACS2) (Feng et al., 2012) and filtered with a value of q < 0.001. Read counts were determined by using featureCounts.py from the SubRead package (Liao et al., 2013), and differentially opened (DO) regions were identified with DESeq2 (Love et al., 2014) with a fold-change in accessibility > 2 and an adjusted value of P < 0.05 after Wald’s test and Benjamini- Hochberg correction. Generated libraries from one of the donors did not reach our quality control cut-offs and all ATAC-seq data from that donor were excluded from further analysis.

Graphical representations, data integration, and statistical analyses were generated with R.

Seahorse metabolic experiments

Human naive and IgG⁺ switched MBCs from 5 HD donors were sorted, washed with the appropriate buffer according to the experiment and seeded at a density of 200,000 to 300,000 cells per well in a XFe96 cell culture microplate. Cells were balanced for 1 h in relevant media prior to measurements. Each test was conducted according to the manufacturer’s recommendation and ECAR and OCR measured using aXF96 extracellular flux analyzer (Seahorse Bioscience, Agilent Technologies). Briefly, for mitochondrial respiration, unbuffered XF assay media (Agilent Technologies) supplemented with 2 mM glutamine, 10 mM glucose and 1 mM sodium pyruvate was used. After measurement of basal OCR, Oligomycin (ATP synthase inhibitor, 1 μM), FCCP (uncoupling agent measuring the maximal respiration capacity; 1 μM) and Antimycin A (ETC inhibitor; 1 μM) were sequentially injected. For glycolytic function, unbuffered XF assay media supplemented with 2 mM glutamine was used. After measurement of basal ECAR, glucose (10 mM), oligomycin (1 μM), and 2-deoxyglucose (2-DG, glycolytic inhibitor; 500 mM) were sequentially injected. For glutamine oxidation, unbuffered XF assay media supplemented with 10 mM glucose and 1 mM sodium pyruvate was used. After measurement of basal OCR, glutamine (2 mM) and BPTES (3 μM) were sequentially injected. For Glycolytic rate assay profile: Seahorse XF Base Medium without phenol red (Agilent technologies) with 2 mM glutamine, 10 mM glucose, 1 mM sodium pyruvate and 5.0 mM HEPES. After measurement of basal ECAR and OCR, rotenone (1 μM) and 2-DG (500 mM) were sequentially injected, and proton efflux rate was assessed.

For fatty acid oxidation (FAO) tests, cells were plated in FAO assay media (110 mM NaCl, 4.7 mM KCl, 1.25 mM CaCl2, 2 mM MgSO4, 1.2 mM NaH2PO4, 2.5 mM Glucose, 0.5 mM Carnitine, and 5 mM HEPES) supplemented either with BSA (100 μM) or BSA ⁺ etomoxir (200 μM). After measurement of basal OCR, oligomycin (1 μM), FCCP (1 μM), antimycin A (1 μM) were sequentially injected.

Isotopic profiling by LC-HRMS

Human naive and IgG⁺ switched memory from 3 HD donors were sorted, washed with glucose-free or glutamine-free RPMI-1640 and seeded at a density of 1,000,000 cells/mL in FACS tubes in glucose-free RPMI-1640 supplemented with 11mM ¹³C₆-D-Glucose (Sigma, Cat# 389374) and 2% FCS or glutamine-free RPMI-1640 supplemented with 2 mM ¹³C₅-L-Glutamine (Sigma, Cat# 605166) and 2% FCS respectively. 1 mL (1,000,000 cells) of each cell suspension was harvested at 0 hr, 1 hr and 6 hrs and cells were quenched by rapid filtration on a 0.2 μM polyamide membrane filter (Sartolon, Cat# 25007-47-N). The filter was snap frozen in liquid nitrogen at –196 °C and stored at -80 °C until extraction.

Subsequent extraction and analysis of metabolites was performed on the MetaToul metabolic platform (INSA Toulouse, FRANCE). Extraction was done with a solvent mixture of ACN/Methanol/H₂O (2:2:1) at –20 °C. Samples were then evaporated and resuspended in 120 μL of ultrapure H₂O before analysis. The analyses of metabolites were carried out on liquid anion exchange chromatography DionexTM ICS-5000⁺ Reagent-FreeTM HPICTM (Thermo Fisher ScientificTM, Sunnyvale, CA, USA) system coupled to QExactiveTM Plus high-resolution mass spectrometer (Thermo Fisher ScientificTM, Waltham, MA).

Liquid anion exchange chromatography was performed with the Thermo Scientific Dionex ICS-5000⁺ Reagent-Free HPIC system (Thermo Fisher Scientific) equipped with an eluent generator system (ICS-5000⁺EG, Dionex) for automatic base generation (KOH). Analytes were separated within 50 min, using a linear KOH gradient elution applied to an IonPac AS11- HC column (250 x 2 mm, Dionex) equipped with an AG11-HC guard column (50 x 2 mm, Dionex) at a flow rate of 0.38 ml/min. The gradient was as follows: equilibration with 7 mM KOH during 1.0 min; then KOH ramp from 7 to 15 mM, 1–9.5 min; constant concentration 10.5 min; ramp to 45 mM in 10 min; ramp to 70 mM in 3 min; ramp to 100 mM in 0.1 min; constant concentration 8.9 min; drop to 7 mM in 0.5 min; and equilibration at 7 mM KOH for 7.5 min. The column and autosampler temperatures were thermostated at 25 °C and 4 °C, respectively. The injected sample volume was 15 μl. Measures were performed in triplicates from separate cultures.

MS analyses were performed in negative FTMS mode at a resolution of 70 000 (at 400 m/z) in full-scan mode, with the following source parameters: the capillary temperature was 325 °C, the source heater temperature, 350 °C, the sheath gas flow rate, 50 a.u. (arbitrary unit), the auxiliary gas flow rate, 5 a.u., the S-Lens RF level, 50 %, and the source voltage, 2.75 kV. Metabolites were determined by extracting the exact mass with a tolerance of 5 ppm.

13C-profiler-emzed software was used for data processing. Isocor 2.0 was used for correction of natural abundance of occurring isotopes taking into account correction of the isotopic tracer for a 99% theoretical 13C purity in substrates.

TeloFISH

25000 naive, IgG⁺ switched memory or marginal zone B cells were directly sorted in individual wells of a Poly-L-Lysine coated 8 wells HTC slide (Thermo Scientific, Cat# 30-8H-RED-CE24) in an initial 1x PBS drop of 5 μL. 25000 Muntjac fibroblasts (Benyelles et al., 2019), used as reference, were further sorted on to each well and cells were left to sediment for 30 to 45 min at RT before fixation in Electron Microscopy grade 4% formaldehyde solution (Electron Microscopy Sciences, Cat#15714). For donor from whom B5⁺ IgG⁺ cells were also analyzed (n = 6), all B cell population were sorted after an initial enrichment step on a BD FACSAria III cell sorter (IgG⁺ versus IgG⁻CD3⁻CD14⁻CD16⁻, yield mode) and B5⁺IgG⁺ cells were sorted in a 1x PBS drop of 10μL alongside 200 Muntjac cells and left to sediment 5 to 10 min prior to fixation. A control well with 200 sorted B5⁻IgG⁺ cells was always performed. Cells were then washed three times in 1x PBS and the slides were dehydrated and air-dried. A hybridization mixture containing 70% formamide, 20 mM Tris, 100 nM TelC-Cy3 PNA probe (Panagene, Cat#F1002), and 1% blocking reagent (Roche, Cat# 11096176001) was added. DNA was denatured by heat at 82 °C for 5min. After hybridization at RT for 1.5 hrs, slides were washed successively with 70% formamide (2x15 min) and with 0.05 M Tris, 0.15 M NaCl, and 0.05% Tween-20 (3x5 min). Slides were then dehydrated in ethanol, air-dried, and mounted using DAPI-containing VECTASHIELD Antifade Mounting Medium (Eurobio, Cat# H-1200).

Imaging was performed on a ZEISS Apotome 2 microscope using a 63x objective. All image analyses were performed in Icy (icy.bioimageanalysis.org) using Spot detector algorithm (Olivo-Marin, 2002). Measured intensities are expressed as a ratio over the mean intensity detected for telomeres in spike-in Muntjac cells in the same image (at least 5 per image).

Telomeric Restriction Fragment

Naive, IgG⁺ switched memory and marginal zone B cells were sorted and stored as dry pellets at -80 °C. DNA was extracted using the AllPrep RNA/DNA Micro Kit (Qiagen, Cat# 80284). DNA(1 μg) was digested with HinfI and RsaI enzymes, resolved by a 0.7% agarose gel, and transferred to a nylon membrane. Hybridization was performed at 42 °C using EasyHyb solution (Roche) with γ-32P-labeled (TTAGGG)4 probe. After washes, membranes were exposed over a PhosphorImager (AGFA). PhosphorImager exposures of telomere-probed Southern blots were analyzed with the ImageJ program. The digitalized signal data were then transferred to Microsoft Excel and served as the basis for calculating mean TRF length using the formula L = Σ(ODi)/Σ(ODi/Li), where OD is the intensity of the hybridization signal and L is the length, in Kb, at the gel point i.

Illustration tool

All figures were assembled using Adobe Illustrator (CS6). The graphical abstract was created with Biorender.com.

Quantification and Statistical Analysis

Wilcoxon matched-pairs signed rank tests, Kruskal-Wallis tests, non-parametric Friedman tests, repeated measure one-way or twoway ANOVA tests were used to compare continuous variables as appropriate (indicated in Figures). Benjamini, Krieger and Yekutieli FDR correction was used for all multiple comparisons. A P-value ≤ 0.05 was considered statistically significant. In all figure panels, only significant P-values are reported, except for Figure 7C. Statistical analyses were all performed using GraphPad Prism 9.0 (La Jolla, CA, USA).

In brief.

Immune memory in humans has been shown to extend well beyond decades. Chappert et al. provide an extensive functional characterization of human splenic smallpox/vaccinia protein B5-specific MBCs, generated more than four decades ago, to decipher the distinct selection and survival mechanisms associated with MBCs longevity.

Highlight.

• Splenic vaccinia B5-specific MBCs can be isolated decades after last vaccination

• Long-lived B5-specific MBCs display signs of extensive affinity-based selection

• Long-lived B5-specific MBCs are enriched in a splenic CD21^hiCD20^hi IgG⁺ B cell subset

• Long-lived B5-specific MBCs harbor long-lasting elongated telomeres

Data and code availability

• All scRNA-seq, RNA-seq and ATAC-seq data have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) and are available as of the date of publication. Accession numbers are listed in the key resources table. Single cell culture VDJ sequencing data reported in Figures 2 and S2 are included in Table S2 and have been deposited as Targeted Locus Study projects at DDBJ/EMBL/GenBank available as of the date of publication. Accession numbers are listed in the key resources table. The version described in this paper is the first version, KFRA01000000-KFRE01000000.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.