key: cord-0996212-yumqn2d7
authors: Cavazzoni, Cecilia B.; Hanson, Benjamin L.; Podestà, Manuel A.; Bechu, Elsa D.; Clement, Rachel L.; Zhang, Hengcheng; Daccache, Joe; Reyes-Robles, Tamara; Hett, Erik C.; Vora, Kalpit A.; Fadeyi, Olugbeminiyi O.; Oslund, Rob C.; Hazuda, Daria J.; Sage, Peter T.
title: Follicular T Cells Optimize the Germinal Center Response to SARS-CoV-2 Protein Vaccination in Mice
date: 2022-02-01
journal: Cell Rep
DOI: 10.1016/j.celrep.2022.110399
sha: a136270a8209f69f65dfc2ec2ee9d9aa915d5fab
doc_id: 996212
cord_uid: yumqn2d7

Follicular helper T (Tfh) cells promote, whereas follicular regulatory T (Tfr) cells restrain, germinal center (GC) reactions. However, the precise roles of these cells in the complex GC reaction remain poorly understood. Here, we perturb Tfh or Tfr cells after SARS-CoV-2 spike protein vaccination in mice. We find Tfh cells promote the frequency and somatic hypermutation (SHM) of Spike-specific GC B cells and regulate clonal diversity. Tfr cells similarly control SHM and clonal diversity in the GC, but do so through limiting clonal competition. In addition, deletion of Tfh or Tfr cells during primary vaccination results in changes in SHM after vaccine boosting. Aged mice, which have altered Tfh and Tfr cells, have lower GC responses presenting a bimodal distribution of SHM. Together, these data demonstrate GC responses to SARS-CoV-2 spike protein vaccines require a fine balance of positive and negative follicular T cell help to optimize humoral immunity.

High affinity antibodies are thought to be a direct result of the germinal center (GC) reaction, which promotes somatic hypermutation (SHM) and affinity maturation (Victora and Nussenzweig, 2012) (McHeyzer-Williams et al., 2012) . Class switch recombination (CSR) has also been attributed to the GC reaction, although CSR may be induced before B cells enter GCs (Roco et al., 2019) . Follicular helper T (Tfh) cells are essential for GC responses by providing cytokine and costimulatory signals (Crotty, 2019) . Patients with mutations in Tfh effector molecules develop a primary immunodeficiency-like disease (Tangye et al., 2013 , Kotlarz et al., 2014 . Limited Tfh help has been thought to promote affinity maturation and clonal expansion through competition between B cells in the GC (Crotty, 2019) . In support of this, enhancing Tfh-B interactions through targeting antigen to B cells promotes dark zone cycling and clonal expansion (Victora et al., 2010 , Gitlin et al., 2015 . However, the effect of enhanced Tfh-B interactions may have positive or negative effects on selection of high affinity clones (Victora et al., 2010 , Gitlin et al., 2015 . Moreover, induction of metabolic flux in B cells results in diminished affinity maturation (Ersching et al., 2017) . However, untangling the roles of Tfh help from enhanced antigenic signals into B cells in these systems is difficult. Using an alternative strategy of reduced MHC expression on B cells, a recent study showed that the peptide:MHC concentration controls GC entry, but not affinity maturation (Yeh et al., 2018) . The precise roles of Tfh cells in optimizing GC responses remain poorly understood.

B cell responses can be fine-tuned by regulatory mechanisms in a process called humoral immunoregulation. Follicular regulatory T (Tfr) cells can gain access to the B cell follicle and restrain B cell responses (Sage and Sharpe, 2020 , Fonseca et al., 2019 , Wing et al., 2020 .

Alterations in Tfr cells have been associated with autoimmune disease and contribute to age-J o u r n a l P r e -p r o o f related defects in vaccination (Sage and Sharpe, 2020) . Using a Tfr-deleter strategy, we showed that deletion of Tfr cells early before GC formation led to enhanced GC expansion and augmentation of autoantibodies (Clement et al., 2019) . Other studies in which Bcl6 is deleted from Treg cells (including Tfr cells) have confirmed the role for Tfr cells in controlling autoantibodies (Gonzalez-Figueroa et al., 2021 , Wu et al., 2016 , Fu et al., 2018 . However, these studies have also suggested subtle (or positive) roles for Tfr cells in foreign antibody , Gonzalez-Figueroa et al., 2021 , Wu et al., 2016 , Fu et al., 2018 .

In settings of SARS-CoV-2 infection, the frequencies of some subsets of Tfh cells correlate with anti-SARS-CoV-2 antibody responses, and defects in Tfh cells may contribute to mortality (Juno et al., 2020 , Kaneko et al., 2020 . Tfr cells inversely correlate with SARS-CoV-2 antibody during infection (Gong et al., 2020) . However, the precise roles of Tfh and Tfr cells in controlling GC responses remain poorly understood. Understanding how Tfh and Tfr cells control complex GC dynamics is critical for the development of strategies to enhance vaccine efficacy to SARS-CoV-2 and other emerging pathogens. To address these questions, we utilized recently developed Tfh and Tfr deleter mice as well as models of aging to alter levels of follicular T cells during SARS-CoV-2 spike protein vaccination. We also incorporated single GC B cell culture assays along with BCR sequencing to determine the consequences of altering follicular T cells at the GC level. We found Tfh cells were required for optimal SARS-CoV-2 Spike specific B cell presence in GCs and for SHM, but controlled clonal diversity. We also found that Tfr cells contributed to SARS-CoV-2 spike specific GC frequencies, and promoted SHM and expansion of Spike-specific clones. Although Tfh and Tfr cells are classically thought as stimulatory and inhibitory mediators of vaccine responses, J o u r n a l P r e -p r o o f respectively, we posit that these cells have dynamic and complex roles in optimizing B cell responses.

Tfh cells have been implicated as being essential for GC responses, although the distinct roles of Tfh cells in controlling GC dynamics in a polyclonal environment are poorly understood.

Moreover, Tfh (and Tfr) cell frequencies have been linked to SARS-CoV-2 antibody responses (Kaneko et al., 2020 , Juno et al., 2020 . To understand follicular T cell and GC responses in more detail in the context of SARS-CoV-2 vaccination, we vaccinated wild type mice with a SARS-CoV-2 prefusion spike trimer using Addavax as an adjuvant and harvested draining lymph nodes (dLN). We found an increase in the total follicular T cell population whether we identified follicular T cells as CD4 + ICOS + CXCR5 + or CD4 + PD-1 hi CXCR5 + cells, the latter likely indicating a "GC"phenotype (Fig. 1A) . We subdivided total follicular T cells (CD4 + PD-1 hi CXCR5 + CD19 -) into FoxP3 -Tfh cells and FoxP3 + Tfr cells and found that Tfh cells expanded after vaccination, whereas Tfr cells did not change (Fig. 1B) . This made the Tfr percentage of all follicular T cells lower compared to control non-vaccinated mice, although this did not reach statistical significance. FAS + GL7 + CD19 + GC B cells and IgG1 + CD38class switched B cells were substantially increased in the dLN compared to unvaccinated mice ( Fig.   1C-D) . The SARS-CoV-2 vaccine also generated serological IgG responses to S1, RBD and S2 domains of SARS-CoV-2 Spike (Fig. 1E) . To understand the specificity of GC B cells in these settings, we used a single GC B cell culture assay to investigate the specificity, affinity and BCR J o u r n a l P r e -p r o o f sequence of GC B cells at the single cell (e.g. clonal) level (Kuraoka et al., 2016) . These assays are advantageous since specificity can be obtained through sensitive ELISA rather than antigen probes which may only detect high affinity clones. Using this assay, we cultured individual GC B cells from mice immunized with the SARS-CoV-2 vaccine 14 days prior. Day 14 corresponds to the peak of the GC response (Fig. S1) . Presence of IgG was screened by ELISA and IgG + wells were assessed SARS-CoV-2 spike reactivity. We found ~24% of IgG + GC B cells were specific for CoV-2 Spike (Fig. 1F-G) . These data demonstrate that a SARS-CoV-2 Spike protein vaccine can generate Tfh cells as well as robust Spike-specific GC responses.

The precise roles of Tfh cells in mediating clonal dynamics within GCs have remained poorly understood due to a lack of tools. We recently developed a "Tfh-DTR" mouse to delete Tfh cells with administration of diphtheria toxin (DT) (Mohammed et al., 2021) . The Tfh-DTR contains both Cd4 Cre and Cxcr5 LoxSTOPLoxDTR alleles which posits DTR on the surface of CD4 + T cells expressing CXCR5, including Tfh cells. Tfh-DTR mouse also deletes Tfr cells, but we refer to it as the Tfh-DTR for simplicity (Mohammed et al., 2021) . We vaccinated control (Cd4 Cre Cxcr5 wt ) or Tfh-DTR (Cd4 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the SARS-CoV-2 Spike protein vaccine, administered DT, and assessed responses on day 14. The dLN from control mice contained CD4 + PD-1 + CXCR5 + follicular T cells which were almost completely missing from Tfh-DTR mice ( Fig. 2A) . Deletion was specific for follicular T cells and was equally robust for Tfh and Tfr cells ( Fig. 2A-B) . The frequency of GC B cells was diminished by ~50% in Tfh-DTR mice indicating Tfh cells are essential for robust GC responses (Fig. 2C) . At this early time point we J o u r n a l P r e -p r o o f did not find alterations in the concentration of Spike-specific IgG in serum, although RBDspecific IgG was slightly attenuated ( Fig. 2D and Fig. S1A ).

To assess how Tfh cells control GC responses in more detail we vaccinated control or Tfh-DTR mice, deleted Tfh cells and, on day 14 performed single GC B cell culture assays. We screened cultures for presence of IgG and then Spike-reactivity (Fig. 2E) . We found that in control mice 42.85% of IgG + GC B cells were specific for SARS-CoV-2 Spike. In contrast, only 11.63% of IgG + GC B cells were specific for Spike in Tfh-deleted mice ( Fig. 2E-G) . These data suggest Tfh cells may be required for Spike-specific B cell entry into GCs, although altered survival and proliferation are also likely factors. Moreover, some GC B cell clones from Tfh-DTR mice showed evidence of lower affinity (i.e. higher KD values) (Fig. 2F) . To assess whether Tfh cells are required for SHM we performed similar experiments in which control or Tfh-DTR mice were vaccinated and single GC B cells were sorted at day 14 and immediately processed for BCR sequencing. When we assessed the total number of mutations in Igh segments we found no substantial differences between control or Tfh-DTR mice ( Fig. 2H and Fig. S1B ). However, when we subdivided clones (defined as same V-J, CDR-H3 length, and at least 80% identity of amino acid sequence) based on extent of expansion we found expanded clones in control mice had an increased mutations compared to singletons, which was not found in Tfh-DTR mice (Fig.   2I ). In particular, highly expanded clones (found 4 or more times) had lower SHM in Tfh-DTR compared to control mice. These data suggest Tfh cells are required for SHM during clonal expansion of B cells within GCs. We also assessed extent of clonal expansion. We found control mice had some evidence of clonal expansion, including an RBD-specific clone (VH2-9-1/JH4) found in a previous study (Alsoussi et al., 2020) (Fig. 2J) . Additional SARS-CoV-2 clones were annotated from our single GC B cell culture assays (Table S1) . GC B cells from the Tfh-DTR J o u r n a l P r e -p r o o f mouse had more clonal expansion, some of which were identical in VH/JH segment usage, CDR-H3 length and CDR-H3 amino acid sequence to Spike-specific clones. To assess clonal diversity, we calculated the N75 index using two different identity cutoffs for clonal assignment (Mesin et al., 2020) ). An 80% identity cutoff identifies and groups clonotypes that potentially diverged from a common ancestor, whereas a 100% identity cutoff strictly discriminates between clonotypes. We found Tfh-DTR mice had a lower N75 index at both 100 and 80% identity indicating less clonal diversity compared to control mice. Since we found lower frequencies of Spike-specific B cells (as determined by culture assays) but also less clonal diversity (as determined by ex vivo BCR sequencing) in GCs, these data suggest in the absence of Tfh cells there are fewer Spike-specific B cell clones in GCs, but these clones are able to expand to a greater extent. Therefore, Tfh cells promote the presence and SHM of SARS-CoV-2 Spike B cells in GCs but also promote clonal diversity.

The role of Tfr cells in controlling GC responses to foreign antigens has remained controversial, with some studies finding Tfr cells limit, whereas others promote, GC B cells (Sage and Sharpe, 2020) . To study the role of Tfr cells in regulating GC clonal dynamics, we used Tfr-DTR mice which allows deletion of Tfr cells with DT (Clement et al., 2019) (Mohammed et al., 2021) . We vaccinated control (Foxp3 Cre Cxcr5 wt ) or Tfr-DTR (Foxp3 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the SARS-CoV-2 Spike protein vaccine and administered DT to delete Tfr cells. DT administration resulted in a severely attenuated frequency of Tfr cells but normal frequencies of Tfh, CXCR5 -PD1 + FoxP3 -T conventional cells, CXCR5 -PD1 + FoxP3 + Treg cells, and B cells (Fig. 3A-B) . We J o u r n a l P r e -p r o o f found a ~1.5 fold increase in the frequency of GC B cells and a ~2-fold increase in IgG1 + CD38class switched B cells in Tfr-DTR, compared to control mice ( Fig. 3C-D) . We did not find substantial differences in anti-SARS-CoV-2 Spike IgG, and slightly elevated anti-SARS-CoV-2 RBD IgG in sera ( Fig. 3E and Fig. S2A ).

To study how Tfr regulation alters the clonal dynamics of SARS-CoV-2 Spike proteinspecific B cell responses in GCs, we performed single GC B cell culture assays. We found ~25% of GC B cell clones from control mice were SARS-CoV-2 Spike-specific, with almost half being specific for potentially neutralizing (S1 and RBD) epitopes (Fig. 3F) . In contrast, in Tfr-DTR mice only ~12% of GC B cells were specific for SARS-CoV-2 Spike, and of those cells, only a small fraction were specific for S1 and RBD epitopes (Fig. 3F) . Interestingly, roughly half of the SARS-CoV-2 Spike-specific GC B cells in Tfr-DTR mice did not have detectable reactivity to RBD, S1 or S2. It is possible that these cells are specific for conformational epitopes in the Spike trimer. We did not find any substantial reactivity of the clones to autoantigens (Fig. S2B) . When we assessed Spike-specific clones for affinity we found lower affinities in Tfr-DTR mice (Fig.   3G ). These data suggest that, despite Tfr cells restraining GC B cell differentiation, Tfr cells promote the relative abundance of vaccine-specific B cells in the GC reaction, most likely by limiting clonal competition. To determine how Tfr cells alter somatic hypermutation of GC B cells, we performed BCR sequencing on total GC B cells sorted ex vivo. We did not find significant differences in the total number mutations in GC B cells from control or Tfr-DTR mice (Fig. 3H) . However, when we separated the cells based on clonal expansion, we found expanded clones in Tfr-DTR mice had less mutations compared to control mice with many germline sequences ( Fig. 3H and Fig. S2C) . The most expanded clone in Tfr-DTR mice acquired few mutations and was found in two separate mice. Although clonality may skew mutational analyses J o u r n a l P r e -p r o o f we found multiple clones in the 4+ category in Tfr-DTR mice. When we assessed clonal diversity, we found lower diversity in some Tfr-DTR mice (using 100% identity cutoff) (Fig.   3I ). However, the difference in diversity was less substantial using an 80% cutoff (p=0.397 versus p=0.0735 at the 100% identity). Interestingly, we did not find any Spike-specific clones in Tfr-DTR mice based on sequence, but found some in control mice. Taken together, these data indicate that Tfr cells disproportionately restrain non-vaccine specific GC B cells compared to SARS-CoV-2 clones, and that by limiting the total number of clones within the GC, Tfr cells are able to promote SHM and affinity maturation of Spike-specific B cells.

Many classical vaccination strategies have diminished effectiveness in the elderly due to agerelated changes in the immune system. These changes include altered/defective Tfh cells and expansion of fully suppressive Tfr cells (Sage et al., 2015 , Webb et al., 2021 , Lefebvre et al., 2016 . To understand how aging alters SARS-CoV-2 Spike-specific GC responses, we vaccinated 8-week old "young" and 80-week old "aged" mice with our adjuvanted SARS-CoV-2 Spike protein vaccine. On day 14 after vaccination, we found that aged mice had substantial increases in frequencies of Tfh and Tfr cells ( Fig. 4A and Fig. S3A ). We found evidence that Tfh cells were responding to antigen in aged mice because they expressed the cell cycle marker Ki67+ (Fig. S3B) . We also found increased frequencies of CXCR5 -FoxP3 + Treg cells and total CD19 + B cells (Fig. 4B) . We found a 2-fold decrease in the frequency of CD19 + GL7 + FAS + GC B cells as well as a similar reduction in the frequency of IgG1 + CD38class switched B cells in aged mice ( Fig. 4C and Fig. S3B ). Moreover, we found profound reductions in SARS-CoV-2 J o u r n a l P r e -p r o o f Spike-and RBD-specific IgG in serum of aged mice compared to young mice (Fig. 4D) . These data demonstrate reduced humoral immunity and GC responses in aged mice.

To determine how aging alters GC clonal dynamics, we performed single GC B cell cultures. We found ~50% of IgG + GC B cells in young mice were specific for SARS-CoV-2 Spike ( Fig. 4E-F) . Surprisingly, we found slightly increased frequencies of SARS-CoV-2 Spikespecific GC B cells in aged mice. These data suggest that increases in Tfr cells (and altered Tfh cells) during aging may contribute to diminished GC B cell frequencies overall, but may allow vaccine-specific GC B cell responses by limiting inter-clonal competition. Next, we assessed BCR repertoire analysis on total sorted single GC B cells. We did not find any substantial changes in total SHM in aged mice, although we did find a much broader and bimodal distribution of mutations with more germ-line sequences but also more highly mutated clones ( Fig. 4G and Fig. S3C ). We also assessed the clonal diversity of GC B cells. Young mice had some clonal expansion, including clones specific for SARS-CoV-2 Spike protein (Fig. 4H) . We found that aged mice had similar clonal expansion and clonal diversity compared to young mice.

Together, these data indicate that changes in Tfr and Tfh cells during aging alter GC optimization resulting in a broader distribution of somatic hypermutation without subsequent changes in clonal diversity.

To determine if regulation of early GC responses by Tfr cells translated into alterations after secondary challenge we used a vaccine boosting strategy. We vaccinated control (Foxp3 Cre Cxcr5 wt ) or Tfr-DTR (Foxp3 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the SARS-CoV-2 Spike protein J o u r n a l P r e -p r o o f vaccine and administered DT only until day 11. We boosted mice at day 30 with the same adjuvanted SARS-CoV-2 vaccine and assessed mice 8 days later (day 38 since initial vaccination) (Fig. 5A) . This strategy allowed for absence of Tfr cells during priming but presence during boosting. We included adjuvant in the vaccine boost to generate robust secondary GCs and to maintain consistency in the vaccine formulation. The frequencies of Tfh and Tfr cells were similar between control and Tfr-DTR mice 8 days after vaccine boosting, suggesting Tfr cells have successfully repopulated (Fig. 5B) . The frequencies of CD19 + GL7 + FAS + GC B cells as well as CD19 + IgG1 + CD38class switched B cells were both slightly elevated in Tfr-DTR mice compared to control mice, although this did not reach statistical significance ( Fig. 5C-D) . Vaccine boosting resulted in substantial increases in SARS-CoV-2 Spike-and RBD-specific antibodies serum of control mice (Fig. 5E ). However, we did not find any consistent differences in the amount of SARS-CoV-2 IgG in serum between Tfr-DTR and control mice.

We performed single GC B cell culture assays with GC B cells from vaccine-boosted mice (as in Fig. 5C ). We found the frequency of SARS-CoV-2 Spike-specific GC B cells of all IgG + GC B cells was slightly elevated in Tfr-DTR mice, but this did not reach statistical significance ( Fig. 5F-G) . However, only ~1/4 of these Spike-specific GC B cells were directed towards S1/RBD epitopes, in contrast to control mice in which ~1/2 of Spike-specific GC B cells showed binding to S1/RBD epitopes. When we measured the affinities of Spike-specific clones we found that clones from Tfr-DTR mice had overall similar affinities to those from control mice ( Fig. 5H) . We assessed the clones for autoreactivity and found increased autoreactivity of clones in early Tfr-deleted mice, but this did not reach statistical significance (Fig. S5A) . We also performed ex vivo BCR sequencing to determine if somatic hypermutation and clonal expansion J o u r n a l P r e -p r o o f differed between control and Tfr-DTR mice. We did not find substantial changes in the average number of VH mutations in control versus Tfr-DTR mice when we assessed all GC B cells ( Fig.   5I and Fig. S4B ). However, when we separated cells based on extent of clonal expansion, we found lower mutations for both singletons and greatly expanded clones in Tfr-DTR compared to control mice. We also assessed clonal diversity but did not find substantial differences between control and Tfr-DTR mice (Fig. 5J ). Together these data indicate that early Tfr humoral immunoregulation during primary GCs can affect secondary GCs after vaccine boosting by optimizing somatic hypermutation.

We also assessed the contribution of Tfr cells at the time of vaccine boosting to GC B cell responses in secondary GCs. We vaccinated control (Foxp3 Cre Cxcr5 wt ) or Tfr-DTR (Foxp3 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the adjuvanted SARS-CoV-2 Spike protein vaccine and boosted on day 30. We then administered DT from day 30 until day 36 to delete Tfr cells only during vaccine boosting. We harvested draining lymph nodes and serum on day 38 (8 days after boost) (Fig. 5K) . We found minor increases in Spike-specific serum IgG in Tfr deleted mice, but this did not reach statistical significance (Fig. 5L) . We also assessed Spike-specific IgG+ GC B cells using our single GC B cell culture assays but found no substantial differences in the frequency of Spike-specific GC B cells (Fig. 5M) . These data indicate that the roles of Tfr cells in controlling clonal dynamics in primary GCs may not be the same in secondary GCs after vaccine boosting.

Vaccine Boosting J o u r n a l P r e -p r o o f Next, we determined if optimization of the GC reaction by Tfh cells early during primary GCs altered secondary GC responses after vaccine boosting. To do this, we vaccinated control (Cd4 Cre and Cxcr5 wt ) or Tfh-DTR (Cd4 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the adjuvanted SARS-CoV-2 recombinant Spike-protein vaccine and administered DT only until day 11, boosted mice with the same SARS-CoV-2 vaccine on day 30, and assessed draining lymph nodes on day 38 (Fig.   6A ). We found that Tfh cells had reconstituted by day 38 confirming the absence of Tfh cells only during primary responses (Fig. 6B) . However, Tfr cells were slightly attenuated at day 38.

This difference likely reflects biological changes due to loss of Tfh cells and not lack of the ability of Tfr cells to reconstitute because Tfr reconstitution in Tfr-DTR mice was complete with a similar deletion schedule. We found GC B cell frequencies were similar between control and Tfh-DTR mice suggesting no changes in GC B cell differentiation (Fig. 6C) . When we assessed SARS-CoV-2 Spike and RBD-specific IgG in serum, we found that boosting resulted in substantial increases in antibody responses that were comparable between control and Tfh-DTR mice (Fig. 6D) . These data demonstrate that early Tfh help in primary GCs does not affect GC B cell frequencies nor augmentation of serological antibody after secondary vaccination.

To understand how early Tfh help alters GC dynamics after secondary challenge we performed single GC B cell cultures. We found that ~23% of IgG + GC B cells were specific for SARS-CoV-2 Spike in control mice and ~17% in Tfh-DTR mice (Fig. 6E) . We found antibodies produced by Spike-specific single GC B cell clones in Tfh-DTR mice had lower affinities than antibodies from control mice, although only a few clones were able to be tested (Fig. 6F) . We also performed ex vivo single GC B cell Igh sequencing to assess somatic hypermutation and clonal expansion. We found no difference in the number of mutations for IgM + antibodies but did find a reduced number of mutations in IgG + GC B cells in Tfh-DTR compared to control J o u r n a l P r e -p r o o f mice ( Fig. 6G and Fig. S5A) . We further separated GC B cells based on expansion and found that non-expanded and greatly expanded (>4) clones had evidence of lower number of mutations in Tfh-DTR compared to control mice. We also assessed clonal expansion and clonal diversity and found a similar distribution of clones in control and Tfh-DTR mice (Fig. 6H) . Together these data indicate that early Tfh help is important for optimal somatic hypermutation and affinity maturation from secondary GCs after vaccine boosting.

We also performed experiments in which we deleted Tfh cells during vaccine boosting.

We vaccinated control (Cd4 Cre and Cxcr5 wt ) or Tfh-DTR (Cd4 Cre Cxcr5 LoxSTOPLoxDTR ) mice with the adjuvanted SARS-CoV-2 Spike protein vaccine which were boosted on day 30. After boosting, DT was administered to delete Tfh cells. We harvested draining lymph nodes and serum on day 38 (8 days after boost) (Fig. 6I) . We did not find any differences in Spike-specific nor RBD-specific serum antibody between control and Tfh-deleted mice (Fig. 6J) . To assess Spike-specific IgG GC B cells, we performed single GC B cell cultures. We found that secondary GCs (at day 38) had slight increases in Spike-specific B cells when Tfh cells were deleted at the time of boosting. Nonetheless, both control and deleter mice presented higher frequency of Spike-specific GC B cells than controls at day 8 (only given one dose of the vaccine at day 30) (Fig. 6K) . This is in contrast to primary GCs which showed decreases in Spike-specific GC B cells when Tfh cells were deleted. These studies suggest that Tfh cells may have different roles in optimizing Spike-specific GC B cells during primary and secondary GCs.

The role of Tfh cells in mediating the dynamics of the GC reaction have been studied previously by modulating antigen presentation of, or modulating metabolic flux in, GC B cells. Although J o u r n a l P r e -p r o o f these alterations may be consistent with Tfh help, Tfh cells may alter GC responses through a number of mechanisms. Moreover, the role of Tfh cells in mediating GC clonal dynamics have remained controversial. More recently, a role for Tfh cells in GC contraction by upregulation of Foxp3 (and possibly by differentiating into Tfr-like cells) has also been described (Jacobsen et al., 2021) . Therefore, we utilized a deletion strategy to largely eliminate Tfh cells at specific times. We found that Tfh cells promoted somatic hypermutation in response to SARS-CoV-2 Spike protein vaccination. This was more evident in expanded clones, which is consistent with these clones receiving more stimulatory signals. Interestingly, the decrease in somatic hypermutation when Tfh cells were deleted during primary vaccination persisted after vaccine boosting, even though Tfh cells have repopulated by this time. It is possible that the loss of Tfh cells during primary vaccination resulted in memory B cells with attenuated SHM that re-entered secondary GCs after vaccine boosting. This may partially explain the reduced SHM, since recent studies suggest that GC-experienced memory B cells account for only a small proportion of B cells in secondary GCs after vaccine boosting (Mesin et al., 2020) . An alternative possibility is that absence of Tfh cells during primary responses prevented memory Tfh cells from forming, and diminished SHM after boosting may be due to altered functionality between de novo formed and memory expanded Tfh cells. Moreover, the lack of Tfh-help may have favored extrafollicular responses. Surprisingly, we found that Tfh cells simultaneously control diversity and clonal expansion in response to a SARS-CoV-2 vaccine. We hypothesize that the profound reduction of Tfh cells in the Tfh-deleter during primary vaccination results in substantially increased clonal competition for help allowing only a few clones to dominate and expand, resulting in reduced diversity. Near-germline SARS-CoV-2 neutralizing antibodies have been found both in humans and mice. Some of these clones appear to be specific for SARS-CoV-2 J o u r n a l P r e -p r o o f Spike based on heavy chain sequence. Interestingly, although we found the lowest affinities for SARS-CoV-2 Spike in antibodies derived from GC B cells in Tfh deleted mice, we also found some high affinity clones with affinities comparable to the ones found in control mice.

The roles of Tfr cells in controlling GC clonal dynamics have been controversial (Sage and Sharpe, 2020) . We previously showed that deletion of Tfr cells before GC formation results in increased GC B cells leading to augmented antibody responses (including vaccine specific and autoantibodies) but lower vaccine specific antibody affinity (Clement et al., 2019) . Here we find that deletion of Tfr cells resulted in increased GC frequencies but diminished SARS-CoV-2

Spike-specific contribution in GCs, particularly for S1 and RBD specific clones. We hypothesize that Tfr cells set high activation thresholds on GC B cells and the reduction in SARS-CoV-2

Spike-specific B cells in the GC during Tfr deletion is due to increased clonal competition of non-vaccine specific clones. This contributes to the diminished frequency of somatic hypermutation, possibly by eliminating the ability of B cells to test newly mutated BCRs.

Interestingly, the decrease in somatic hypermutation persisted during secondary vaccine responses even though Tfr cells were able to reconstitute by the time of secondary vaccination. This may be due to less mutated B cells emerging as memory cells during primary GCs that may enter secondary GCs later in Tfr-DTR mice.

Immunological aging is a complex process that includes multiple changes in the immune system such as diminished naïve lymphocyte repertoires, thymic involution, defective cytokine responses and expansion of Treg cells (Frasca et al., 2020) . Age-related defects in humoral immunity include expansion of fully suppressive Tfr cells as well as Tfh cell dysfunction (Sage et al., 2015) . We found that during SARS-CoV-2 Spike protein vaccination, aged mice have substantial decreases in GC frequency and also substantially less RBD-specific antibody in J o u r n a l P r e -p r o o f serum. Surprisingly, despite the decrease in GC frequency, SARS-CoV-2 Spike specific B cells were relatively able participate in GCs. However, GC B cells in aging had alterations in SHM with a more diverse and bimodal distribution. This effect is likely due to the combined effects of Tfh and Tfr functions. For instance, the high activation threshold due to expanded Tfr cells and the lack of adequate Tfh help likely explain the abundance of germline sequences. However, it is likely that the low competition setting combined with high threshold for activation due to Tfr expansion may allow a very small number of clones to expand and mutate substantially.

Together this work demonstrates that GC clonal dynamics are highly complex and both positive and negative signals by Tfh and Tfr cells, respectively, contribute to optimal GC reactions that balance affinity maturation and clonal diversity. It may be possible to fine-tune this balance to enhance vaccine responses depending on whether highly mutated oligoclonal antibodies or less mutated highly diverse antibodies are beneficial.

One limitation may be that deletion of Tfh and Tfr cells may not be complete. In addition, our inability to integrate expansion kinetics with epitope specificity and affinity limit the ability to separate entry from proliferation of GC B cells. (F) Assessment of SARS-CoV-2 Spike specificity for IgG + GC B cells using single GC culture assays. Number in circles indicates total number of IgG + GC B cells analyzed. "ND"= not determined to be specific for S1, RBD or S2. "S1"= Specific for S1 but not RBD. "RBD"= specific for RBD. "S2" Specific for S2. KD indicates equilibrium dissociation constant for Spikespecific B cell clones measured by biolayer interferometry. (B) Frequencies of CD4 + PD-1 + CXCR5 + FoxP3 + Tfr, CD4 + PD-1 + CXCR5 + FoxP3 -Tfh, CD4 + PD-1 + CXCR5 -FoxP3 -T cells, CD4 + PD-1 + CXCR5 + FoxP3 + Treg cells, and CD19 + B cells. (E) Serological analysis of SARS-CoV-2 Spike or RBD-specific IgG.

(F) Percentage of SARS-CoV-2 Spike-specific cells as a percentage of IgG + GC B cells from single cell GC cultures, along with individual epitope specificity. Number in center of circle plot indicates total number of IgG + GC B cells analyzed. "ND"= not determined to be specific for S1, RBD or S2. "S1"= Specific for S1 but not RBD. "RBD"= specific for RBD. "S2" Specific for 

(A) Assessment of total CD4 + PD-1 + CXCR5 + follicular T (left), CD4 + PD-1 + CXCR5 + FoxP3 -Tfh (middle) and CD4 + PD-1 + CXCR5 + FoxP3 + Tfr (right) cells from draining lymph nodes of young (8 weeks old) or aged (80 weeks old) mice that were vaccinated with an adjuvanted SARS-CoV-2 Spike protein vaccine and harvested on day 14.

(B) Percentage of CD4 + FoxP3 + CXCR5 -Treg cells (left) and total CD19 + B lymphocytes (right).

CD19 + IgG1 + CD38class switched B cells as a percentage of all CD19 + B cells (right).

(D) Serological analysis of SARS-CoV-2 Spike or RBD-specific IgG.

IgG+GC B cells analyzed. "ND"= not determined to be specific for S1, RBD or S2. "S1"= Specific for S1 but not RBD. "RBD"= specific for RBD. "S2" Specific for S2. 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Peter Sage (psage@bwh.harvard.edu).

The materials are listed in the Key Resources Table, and are available from the Lead Contact.

All data reported in this paper will be shared by the lead contact upon reasonable request. 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.

C57bl/6J mice were from Jackson Laboratories. Tfh-DTR mice are defined as Cd4 Cre Cxcr5 LoxSTOPLoxDTR mice and have been published previously (Mohammed et al., 2021) . Cd4 Cre Cxcr5 wt littermates were used as a control for Tfh-DTR mice. Tfr-DTR mice are defined as Foxp3 IRES-CreYFP Cxcr5 IRES-LoxP-STOP-LoxP-DTR mice and have been published previously (Clement et al., 2019 , Mohammed et al., 2021 . Foxp3 IRES-CreYFP Cxcr5 wt littermates were used as a control for Tfr-DTR mice. Mouse progenies were routinely screened for leakiness of the Foxp3 IRES-CreYFP allele by flow cytometry as described (Clement et al., 2019) . Mice were males and females, 6-8 weeks old and aged mice were 80 weeks old. All mice were used according to Brigham and Women's Hospital Institutional Animal Care and Use Committee and National Institute of Health guidelines. 

96-well plates (Nunc MaxiSorp, Thermo) were coated with 100 µL of SARS-CoV-2 Spike prefusion stabilized trimer (ACROBiosystems, SPN-C52H9) or SARS-CoV-2 Spike protein subunits (S1, S2 and RBD) at 2µg/mL, overnight at 4°C. Then, after 1h of blocking (PBS-1%BSA), 100 µL of diluted serum samples or culture supernatants were incubated for 1h at room temperature. 96-well half area plates (Microlon, Greiner) were used for supernatants. All J o u r n a l P r e -p r o o f volumes were reduced by half in this case. Serum dilution was 1:500 or 1:2,000 (for day 10 or 14 after vaccination) and 1:10,000 (for day 8 after boosting vaccination). Supernatants were not diluted. Standard curves were obtained by serial dilution of an RBD-specific monoclonal antibody (Sino Biological, 40591-MM43). Secondary antibody anti-mouse IgG conjugated to AP (Southern Biotech) was diluted 1:1,000 in PBS-BSA 1% and incubated for 1h at room temperature. Reactions were developed with AP substrate and absorbance was measured at 405 nm. Anti-nuclear antigen/HEp-2 cell-lysate ELISAs were performed using the ANA-HEp2

Screen ELISA kit (Tecan, RE70151), as described above.

Single GC B cells were sorted, as described above, into 96-well round-bottom plates (Corning) containing 1x10 3 NB21.2D9 cells/well as previously described (Kuraoka, 2016) with minor modifications, such as not adding IL-4. Cells were cultured in OptiMEM (GIBCO) supplemented with 10% heat-inactivated fetal bovine serum (GIBCO, 16140071), 2 mM L-glutamine, 1 mM sodium pyruvate, 50 μM 2-mercaptoethanol, 100 U penicillin, and 100 μg/ml streptomycin. IgG secretion was assessed by ELISA after 6 days of culture and IgG+ supernatants were collected after 9 days of culture when cells were frozen in TCL lysis buffer supplemented with 1% 2mercaptoethanol for Igh sequencing. One 96-well plate from each mouse was sorted. The frequency of Spike specific cells was calculated from the total of IgG+ wells. For the Spike specific frequency on a per mouse basis, only mice with at least 8 IgG+ wells were included.

Germinal center B cells from draining lymph nodes were stained with anti-B220 (Biolegend, RA3-6B2), CD38 (Biolegend, Clone 90), T-and B-cell activation antigen (BD Biosciences, GL-J o u r n a l P r e -p r o o f 7) CD138 (Biolegend, 281-2) and CD4 (Biolegend, for 30 minutes at 4°C in PBS with 1% FBS and 1mM EDTA. GC B cell population was defined as B220 + CD38 lo/-GL-7 + CD138 -CD4 -. Single cells were sorted into 96-well PCR plates containing 5 µl/well of TCL lysis buffer (Qiagen) with 1%β-mercaptoethanol for Igh sequencing. RNA extraction was performed using SPRI beads (Tas et al., 2016) . RNA was reverse transcribed into cDNA using oligo (dT) primer.

Igh transcripts were amplified (Tiller et al., 2009 ) and PCR products were barcoded and sequenced utilizing MiSeq (Illumina) Nano kit v.2, as previously described (Mesin et al., 2020) .

Paired-end sequences were assembled with PandaSeq (Masella et al., 2012) and processed using FASTX toolkit. Igh sequences were submitted to IMGT HighV-QUEST for V(D)J rearrangement assignment (Brochet et al., 2008) . To identify VH mutations both IMGT (Lefranc et al., 2009 ) and Vbase2 (Retter et al., 2005) databases were used and, in case of discrepancy, IgBLAST (Ye et al., 2013) was used. VH mutation analyses were restricted to cells with productively rearranged Igh genes. Functional rearrangements were grouped into clones defined by same VH and JH segments, same CDR-H3 length and at least 80% similarity in CDR-H3 amino acid sequences.

Bio-layer interferometry (BLI) was performed on an Octet RED96 instrument (ForteBio) to determine binding affinities of IgG from single GC B cell culture supernatants. Antibodies were immobilized in mouse Fc capture biosensors (Anti-Mouse IgG Fc Capture (AMC)). Association was measured by immersing biosensors loaded with IgG from culture supernatants in wells containing SARS-CoV-2 Spike recombinant trimer (75nM, 150nM, 300nM or 1µM in kinetics buffer provided by the manufacturer) for 600s. Dissociation was monitored after transfer of the biosensors into distinct wells containing kinetics buffer for 600s. KD values are the ratio between J o u r n a l P r e -p r o o f the association (Ka) and dissociation (Kd) constants and were determined with the best concentration in which a local fit 1:1 binding indicated adequate goodness of fit (X 2 and R 2 ) both by the Octet data analysis software (ForteBio) and Prism 9 (GraphPad).

Statistical tests were performed using Prism 9 (GraphPad) utilizing Student's two-tailed unpaired t test for normalized data, or Mann Whitney test for non-normal data as indicated in figure legends. Frequency distributions of VH mutation numbers were plotted as curves utilizing smoothing splines with number of knots of 5. All measurements were taken from distinct samples, except for memory experiments in which the same mice were bled before and after boost. (A) Detection of ANA reactivity using a HEp-2 ELISA in IgG + single GC B cell culture supernatants. Supernatants screened for SARS-CoV2 Spike protein specificity were also tested for binding to autoantigens.

(B) VH mutation analysis for singletons and expanded (found at least twice) clones. Graphs show distribution of sequences by number of mutations (bin size=2). Data represents the raw data used to generate regressed data in Fig.   5I .

(C) Numbers of sequences and VH mutation analyses per mouse.

Data from 2 independent experiments (total of 105 single cell supernatants, as in Figure 5F ) (A), or from n=2 mice per group from one experiment (B-C).

J o u r n a l P r e -p r o o f (A) VH mutation analysis for singletons and expanded (found at least twice) clones. Graphs show distribution of sequences by number of mutations (bin size=2). Data represents the raw data used to generate regressed data in Fig.   6G .

(B) Numbers of sequences and VH mutation analyses per mouse.

(C) Frequency of Tfh cells in dLN on day 8 after SARS-CoV-2 vaccine boosting and DT treatment. Experiment was performed as in Figure 6I .

Data from n=2-3 mice per group from one representative experiment (A-B), or from n=2-5 mice per group from one representative experiment (C).

J o u r n a l P r e -p r o o f 

A Potently Neutralizing Antibody Protects Mice against SARS-CoV-2 Infection

IMGT/V-QUEST: the highly customized and integrated system for IG and TR standardized V-J and V-D-J sequence analysis

Follicular regulatory T cells control humoral and allergic immunity by restraining early B cell responses

T Follicular Helper Cell Biology: A Decade of Discovery and Diseases

Germinal Center Selection and Affinity Maturation Require Dynamic Regulation of mTORC1 Kinase

T follicular regulatory (Tfr) cells: Dissecting the complexity of Tfr-cell compartments

Age-related factors that affect B cell responses to vaccination in mice and humans

Deficiency in T follicular regulatory cells promotes autoimmunity

HUMORAL IMMUNITY. T cell help controls the speed of the cell cycle in germinal center B cells

Peripheral CD4+ T cell subsets and antibody response in COVID-19 convalescent individuals

Follicular regulatory T cells produce neuritin to regulate B cells

Expression of Foxp3 by T follicular helper cells in end-stage germinal centers

Humoral and circulating follicular helper T cell responses in recovered patients with COVID-19

Human IL-21 and IL-21R deficiencies: two novel entities of primary immunodeficiency

Complex Antigens Drive Permissive Clonal Selection in Germinal Centers

Age-related impairment of humoral response to influenza is associated with changes in antigen specific T follicular helper cell responses

IMGT, the international ImMunoGeneTics information system

CD4+ follicular regulatory T cells optimize the influenza virus-specific B cell response

PANDAseq: paired-end assembler for illumina sequences

Restricted Clonality and Limited Germinal Center Reentry Characterize Memory B Cell Reactivation by Boosting

Follicular T cells Mediate Donor Specific Antibody and Rejection after Solid Organ Transplantation

VBASE2, an integrative V gene database

The multifaceted functions of follicular regulatory T cells

Defective TFH Cell Function and Increased TFR Cells Contribute to Defective Antibody Production in

The good, the bad and the ugly -TFH cells in human health and disease

Visualizing antibody affinity maturation in germinal centers

Cloning and expression of murine Ig genes from single B cells

Germinal centers

Germinal center dynamics revealed by multiphoton microscopy with a photoactivatable fluorescent reporter

Ageing promotes early T follicular helper cell differentiation by modulating expression of RBPJ

Control of foreign Ag-specific Ab responses by Treg and Tfr

Follicular regulatory T cells repress cytokine production by follicular helper T cells and optimize IgG responses in mice

IgBLAST: an immunoglobulin variable domain sequence analysis tool

Germinal center entry not selection of B cells is controlled by peptide-MHCII complex density

Spike-specific circulating T follicular helper cell and cross-neutralizing antibody responses in COVID-19-convalescent individuals

We thank Dr. Dan Barouch and the other members of the Massachusetts Consortium for Pathogen Readiness (Mass CPR) for helpful discussions. We thank Dr. Garnett Kelsoe for supplying the NB21 cell line. This work was made possible through grant support by the