key: cord-0788728-9ei6ubx6
authors: Lederer, Katlyn; Castaño, Diana; Atria, Daniela Gómez; Oguin, Thomas H.; Wang, Sidney; Manzoni, Tomaz B.; Muramatsu, Hiromi; Hogan, Michael J.; Amanat, Fatima; Cherubin, Patrick; Lundgreen, Kendall A.; Tam, Ying K.; Fan, Steven H.Y.; Eisenlohr, Laurence C.; Maillard, Ivan; Weissman, Drew; Bates, Paul; Krammer, Florian; Sempowski, Gregory D.; Pardi, Norbert; Locci, Michela
title: SARS-CoV-2 mRNA vaccines foster potent antigen-specific germinal center responses associated with neutralizing antibody generation
date: 2020-11-21
journal: Immunity
DOI: 10.1016/j.immuni.2020.11.009
sha: b446a8baf5498f5fb8dced2118201de912cd5f8b
doc_id: 788728
cord_uid: 9ei6ubx6

The deployment of effective vaccines against SARS-CoV-2 is critical to eradicate the COVID-19 pandemic. Many licensed vaccines confer protection by inducing long-lived plasma cells (LLPC) and memory B cells (MBC), cell types canonically generated during germinal center (GC) reactions. Here, we directly compared two vaccine platforms −mRNA vaccines and a recombinant protein formulated with an MF59-like adjuvant− for their ability to quantitatively and qualitatively shape SARS-CoV-2-specific primary GC responses over time. We demonstrated that a single immunization with SARS-CoV-2 mRNA, but not with the recombinant protein vaccine, elicited potent SARS-CoV-2-specific GC B and T follicular helper (Tfh) cell responses as well as LLPC and MCB. Importantly, GC responses strongly correlated with neutralizing antibody production. mRNA vaccines more efficiently induced key regulators of the Tfh cell program and influenced the functional properties of Tfh cells. Overall, this study identifies SARS-CoV-2 mRNA vaccines as strong candidates for promoting robust GC-derived immune responses.

The coronavirus disease 2019 (COVID-19) outbreak was declared a pandemic in March 56 have been published thus far reporting a limited characterization of the immune responses 94 induced by the SARS-CoV-2 vaccine candidates (Corbett et al., 2020b; Gao et al., 2020; 95 Jackson et al., 2020; Laczkó et al., 2020; Mulligan et al., 2020b; Sahin, 2020; Smith et al., 2020; 96 Yu et al., 2020; Zhang et al., 2020) . However, the lack of a direct comparison and universal 97 metrics for the assessment of the immune responses elicited by the different vaccines make it 98

challenging to conclude what vaccines can induce superior immune responses. Additionally, no 99 study has deeply investigated the capacity of the SARS-CoV-2 vaccines to elicit GC responses 100 and the ensuing quality of Ab responses. This is a fundamental gap in our knowledge given the 101 crucial role of GCs in regulating B cell responses and high-quality Ab generation. 102

Herein, we systematically compared SARS-CoV-2 mRNA vaccines encoding the 103 receptor biding domain (RBD) and full length spike protein of SARS-CoV-2 with recombinant 104 SARS-CoV-2 RBD protein (rRBD) formulated with AddaVax, an MF59-like adjuvant. Our in-105 depth analysis was aimed at shedding light on the ability of the two different vaccine platforms 106 to mold GC responses and the functional properties of Tfh cells. Overall, our study revealed a 107 superior ability of the SARS-CoV-2 mRNA vaccines to elicit SARS-CoV-2-specific GC B cell 108 responses, which were associated with an efficient production of nAbs after a single or a booster 109 immunization. Moreover, it proved that the mRNA vaccines could efficiently promote the Hundreds of millions of people will have to be vaccinated to combat the COVID-19 pandemic, 121 and vaccination strategies deploying one or few immunizations are an attractive option if capable 122 of eliciting efficient immune responses. mRNA vaccines have been previously shown to promote 123 potent or protective Ab responses upon a single immunization (Laczkó et al., 2020; Pardi et al., 124 J o u r n a l P r e -p r o o f 2018b). Conversely, purified/recombinant proteins formulated with adjuvants often require 125 multiple immunizations to achieve sufficient titers of protective Abs. In light of the importance 126 of GCs for the generation of high-quality persistent Ab responses, we hypothesized that a single 127 immunization with SARS-CoV-2 mRNA vaccines (Laczkó et al., 2020) can result in superior 128 GC formation in comparison to a SARS-CoV-2 recombinant protein formulated with the MF59-129 like adjuvant AddaVax. To test our hypothesis, BALB/c mice were immunized once 130 intramuscularly (i.m.) with 30 µg of an mRNA-LNP vaccine encoding a SARS-CoV-2 stabilized 131 full length spike glycoprotein with deleted furin cleavage site (full S ∆ furin mRNA), or with a 132 second mRNA-LNP encoding the SARS-CoV-2 receptor binding domain (RBD mRNA), a 133 prominent target for neutralizing Abs on the full S protein (Laczkó et al., 2020) . To confirm the superior ability of SARS-CoV-2 mRNA vaccines to foster the formation of GCs 153 following a single immunization, we performed microscopy studies on inguinal LN from mice 154 immunized 7 days earlier with RBD mRNA or rRBD-AddaVax ( Figure 1D -E). GCs are defined 155 J o u r n a l P r e -p r o o f by microscopy as clusters of GL7 + cells forming around a network of follicular dendritic cells 156 (FDC, CD21/35 hi cells) (Heesters et al., 2014) , and surrounded by a mantle of IgD + naïve B cells 157 (De Silva and Klein, 2015) . GCs are anatomically segregated from T cell areas (CD3 + ). A global 158 view of LN sections revealed a robust induction of GCs by the RBD mRNA vaccine as opposed 159 to minimal GC formation in rRBD-AddaVax immunized mice that resembled the one observed 160 in naïve mice ( Figure 1D ). A previously published study also described small GC formation by 161 microscopy following one immunization with a protein antigen in MF59 (Liang et al., 2017) . 162

The high level of colocalization of GL7 + cells and CD21/35 hi FDC, together with the presence of 163 infiltrating CD3 + cells, suggested that the structures observed in the LN of RBD mRNA 164 immunized mice were bona fide GCs ( Figure 1E ). Therefore, SARS-CoV-2 mRNA vaccines, but 165 not rRBD-AddaVax, elicited potent GC formation upon a single immunization. 166

and contract by day 28. 169 We sought to determine if GC B cells induced by the SARS-CoV-2 mRNA vaccines were 170 antigen-specific. To this aim, we generated two fluorescently labelled rRBD probes to track 171 RBD-specific GC B cells induced by the different vaccines via flow cytometry. A large fraction 172 of the GC B cell response was RBD-specific in RBD mRNA immunized mice 7 days post 173 immunization ( Figure 2A-B) . RBD-specific GC B cells were also induced by full S ∆ furin 174 mRNA, although at significantly lower levels than in RBD mRNA immunized mice ( Figure 2A -175 B) . Since the full S protein includes additional epitopes besides the ones contained in RBD, we 176 measured full S-specific GC B cells by taking advantage of two fluorescently labelled full S 177 probes ( Figure S2A CoV-2 mRNA vaccines elicited elevated titers of SARS-CoV-2-specific IgG as early as day 14 248 post immunization and the titers remained high through at least day 60 following vaccination. 249

Despite of an almost complete lack of GC induction, mice receiving the rRBD-AddaVax 250 generated rRBD-specific Ab responses, although SARS-CoV-2-specific IgG titers rose more 251 slowly than following mRNA vaccination and remained lower in rRBD-AddaVax at day 60 post 252 immunization ( Figures 4A and S4A ). This observation confirmed that rRBD-AddaVax 253 preparation was not inactive and induced a delayed antigen-specific immune response. SARS-254

CoV-2-specific IgG1 accounted for the majority of the antigen-specific IgG responses in the 255 mice immunized with rRBD-AddaVax, most of which had negligible IgG2a and IgG2b 256 responses . In contrast, SARS-CoV-2 mRNA immunized mice 257 displayed high levels of SARS-CoV-2-specific IgG2a and IgG2b titers, together with IgG1 at all 258 time points . 259 LLPC persistently secrete Abs without the need for antigen re-exposure and are responsible for 260 maintaining protective levels of circulating Abs that can prevent or hinder a re-infection by 261 pathogens (Sallusto et al., 2010) . We compared bone marrow SARS-CoV-2 specific Ab 262 secreting cells (ASC) elicited by the different vaccines 60 days post immunization. Consistent 263 with the serum IgG data above, mice immunized with rRBD-AddaVax generated RBD-specific 264 IgG + LLPC, despite minimal GC responses. However, mice immunized with mRNA vaccines 265 had substantially higher RBD-specific IgG + LLPC numbers (Figures 4E and S4E) suggesting 266 more efficient generation of multiple facets of long-term humoral immunity by these vaccines. 267

To investigate the apparent discrepancy between the inability to form GCs and the presence of 268 detectable Ab and LLPC responses of rRBD-AddaVax immunized mice, we performed a 269 qualitative analysis of neutralizing Ab (nAb) responses. We tested nAbs in sera from animals 270 immunized 14 or 60 days earlier with the various SARS-CoV-2 vaccines by performing in vitro 271 microneutralization assays with authentic SARS-CoV-2. Remarkably, in contrast to mice 272 immunized with SARS-CoV-2 mRNA vaccines where nAbs were present in the sera at both time 273 points tested, no detectable nAbs were induced by rRBD-AddaVax after a single immunization 274 ( Figure 4F ). Comparable data were obtained with an in vitro pseudoneutralization assay ( Figure  275 4G). The level of nAbs detected by the microneutralization assay were associated to SARS-CoV-276 2-specific IgG responses, but showed a weaker correlation with RBD-specific LLPC , suggesting that some of the ASC detected at this time point could reflect lower 278 J o u r n a l P r e -p r o o f quality or non-neutralizing responses. Since SARS-CoV-2 mRNA immunized mice generated 279 both elevated nAbs and GC B cells, we hypothesized the existence of a strong association 280 between these two parameters. As anticipated, total and SARS-CoV-2 -specific GC B cells 281 strongly correlated with the levels of nAbs at day 14 post immunization . 283

Altogether, these data strongly indicate that efficient GC responses are fundamental to obtain 284 high quality SARS-CoV-2 neutralizing Abs, and that a potent GC B cell induction in mice 285 immunized once with SARS-CoV-2 mRNA vaccine, but not with rRBD-Addavax, is connected 286 to efficient SARS-CoV-2 nAb production. 287 288 SARS-CoV-2 mRNA vaccines are superior in comparison to rRBD-AddaVax vaccine at 289 inducing SARS-CoV-2-specific Tfh cells. 290

Tfh cells are key regulators of GC responses (Crotty, 2019; Vinuesa et al., 2016) . Knowing that 291

Tfh cell responses normally peak between day 7-9 in mice (Baumjohann et al., 2011; Botta et al., 292 2017), we determined the frequencies and absolute numbers of Tfh cells induced by the mRNA 293 vaccines and rRBD-AddaVax in draining LN 7 days post immunization. Tfh cells were measured 294 by flow cytometry as CD4 + CD44 hi CD62Lcells expressing the Tfh cell signature markers 295 CXCR5 and Bcl-6 ( Figure 5A ) or CXCR5 and PD-1 ( Figure S5A ). In line with the potent GC B 296 cell induction driven by the SARS-CoV-2 mRNA vaccines ( Figure 1A S5C) (closer to the GC peak for this vaccine group). This was an anticipated outcome based on 302 the poor GC B cell induction measured in response to rRBD-AddaVax ( Figure 1A-B) . 303

Additionally, we found a robust correlation between the numbers of Tfh cells, defined either as 304 CXCR5 + Bcl-6 + or CXCR5 + PD-1 + cells, and total or RBD-specific GC B cells . Of note, Tfh cell absolute numbers were significantly correlated with nAbs after a single-306 dose immunization ( Figures 5E and S5E) . 307

To determine whether the kinetic of Tfh cell induction mirrored GC B cell responses, we 308 evaluated Tfh cell frequencies and absolute numbers over time following immunization with 309 J o u r n a l P r e -p r o o f SARS-CoV-2 mRNA vaccines or Luc mRNA vaccine. We found that Tfh responses driven by 310 SARS-CoV-2 mRNA vaccines peaked at day 7 post immunization and then waned over time 311 ( Figure 5F ). At day 28 post immunization, elevated Tfh cell responses were no longer detectable. 312

The correlation between Tfh cells and RBD + GC B cells suggested that the Tfh cells induced by 313 the SARS-CoV-2 mRNA vaccines were SARS-CoV-2-specific. To directly measure antigen-314 specific polyclonal Tfh cell populations we successfully established a robust in vitro assay for 315 the detection of SARS-CoV-2-specific Tfh cells. In this assay, the stimulation with a SARS-316

CoV-2 peptide pool, aided by CD28 costimulation, was able to induce a detectable production of Collectively, our study directly shows that, even when analyzed in an antigen-specific fashion, 325 SARS-CoV-2 mRNA immunized mice presented a more robust generation of Tfh cells compared 326 to rRBD-AddaVax. 327 Tfh cells are a functionally heterogeneous population (Morita et al., 2011; Locci et al., 2013; 331 Weinstein et al., 2016) . To shed light on the functional characteristics of the Tfh cells induced by 332 the vaccines under investigation, we evaluated the capacity of the mRNA vaccines and rRBD-333

AddaVax to shape the Tfh cell potential to produce IFN-γ and IL-4 upon in vitro stimulation. To 334 this aim, mice were immunized with rRBD-AddaVax or with SARS-CoV-2 mRNA vaccines. 335 Seven days following immunization, lymphocytes from draining LN were stimulated in vitro 336 with PMA and ionomycin, and IFN-γ and IL-4 production by Tfh cells was measured by ICS 337 To better understand the biological consequences of the observed Tfh functional polarization, we 351 calculated the ratio of IgG1 to IgG2a and IgG1 to IgG2b SARS-CoV-2 specific Abs in the sera 352 of mice immunized 60 days earlier with rRBD-AddaVax or the SARS-CoV-2 mRNA vaccines 353 ( Figure 6F ). In mice, Th2 biased responses are associated with IgG1 production. Conversely, 354

Th1 polarized responses are linked to IgG2 Ab production (Reinhardt et al., 2009; Snapper and 355 Paul, 1987; Stevens et al., 1988) . In agreement with the functional polarization of Tfh cells, we 356 found higher IgG1/IgG2 Ab ratios in mice immunized with rRBD-AddaVax in comparison to the 357 mRNA vaccines ( Figure 6F ). 358

To gain insights on the mechanisms by which mRNA vaccines foster Tfh cell responses, we 359 evaluated the expression of distinctive transcription factors and key molecules associated with 360 the follicular program and function of Tfh cells. We evaluated IL-21 + Tfh cells by ICS following 361 PMA/ionomycin in vitro stimulation. IL-21 is a cytokine produced by Tfh cells that regulates 362 proliferation and differentiation of GC B cells into PC (Crotty, 2019; Vinuesa et al., 2016) . Our 363 analysis revealed a higher frequency of IL-21 + Tfh cells driven by SARS-CoV-2 mRNA vaccine 364 in comparison to rRBD-AddaVax ( Figure 6G -H). Ascl-2 is a transcription factor (TF) regulating 365 the initiation of Tfh cell development (Liu et al., 2014) and has been shown to promote the 366 expression of CXCR5, a chemokine receptor involved in Tfh homing to B cell follicles (Crotty, 367 2019; Vinuesa et al., 2016) . qPCR analysis of Tfh cells isolated from mice immunized 7 days 368 earlier with SARS-CoV-2 mRNA vaccines revealed a higher expression of Ascl-2 in comparison 369 to Tfh cells from rRBD-AddaVax immunized mice ( Figure 6I ). Consistent with the heightened 370 expression of Ascl-2, CXCR5 was expressed at higher levels in Tfh cells from SARS-CoV-2 371 mRNA immunized mice when compared to Tfh cells induced by rRBD-AddaVax ( Figure 6J ). 372

Differently, the Tfh signature TF Bcl-6 (Crotty, 2019; Vinuesa et al., 2016) , which is not 373 influenced by Ascl-2 (Liu et al., 2014) , was expressed at comparable levels on the Tfh cells 374 driven by the two different vaccine platforms ( Figure S6L ). The co-stimulatory molecule ICOS 375 plays a central role in establishing the Tfh program and serves as migration receptor to promote 376 the follicular homing of Tfh cells (Crotty, 2019; Xu et al., 2013) . At the protein level, ICOS 377 expression was enhanced in Tfh cells of mRNA vaccinated mice ( Figure 6K We next questioned whether the difference in the magnitude of secondary RBD-specific GC 399 responses between groups was paralleled by quantitative and qualitative differences in Ab 400 responses post boost. While both rRBD-AddaVax and RBD mRNA groups had detectable levels 401 of RBD-specific IgG titers following the second immunization, the RBD mRNA group had a 402 superior induction of RBD-specific IgG compared to the rRBD-AddaVax group ( Figure 7C ). 403

Next, we measured nAbs by SARS-CoV-2 microneutralization assays before and after the 404 booster immunization. In line with the data shown in Figure 4 , only RBD mRNA induced 405 substantial nAb responses 4 weeks after a single immunization ( Figure 7D ). Differently, 406 following a booster immunization both RBD mRNA and rRBD-AddaVax immunized animals 407 were able to produce detectable nAb levels. There was, however, a marked difference in the 408 magnitude of nAb responses that mirrored RBD-specific GC B cell responses, with the group of 409 animals immunized with RBD mRNA presenting nAb responses two-logs higher than the rRBD-410

AddaVax group ( Figure Altogether, these data indicate that rRBD-AddaVax induces negligible GC formation and nAb 415 production after a single-dose immunization, along with larger but mostly non-RBD-specific GC 416 B cell responses that are associated with low-levels nAb production after a booster 417 immunization. Conversely, SARS-CoV-2 mRNA vaccine elicits powerful SARS-CoV-2-specific 418 GC responses as well as a robust nAb production that can be further enhanced by a booster 419 immunization. 420 421 422

Recent studies, have revealed that multiple SARS-CoV-2 vaccine approaches are 424 capable, to varying extent, of mediating the production of nAbs that can neutralize SARS-CoV2 425 in vivo and/or in vitro (Anderson et al., 2020; Corbett et al., 2020b; Gao et al., 2020; 426 Jackson et al., 2020; Laczkó et al., 2020; Mulligan et al., 2020a; Sahin, 2020; Smith et al., 2020; 427 Yu et al., 2020; Zhang et al., 2020) . Outstanding questions were left unanswered thus far in the 428 SARS-CoV-2 vaccine field: what is a good metric to predict nAb formation? Are GCs important 429 for nAb formation? And how do different vaccine platforms compare to each other? Our study, 430 built on a systematic comparison between two vaccine platforms, nucleoside-modified mRNA-431 LNP and recombinant protein formulated with the MF59-like adjuvant AddaVax (rRBD-432 AddaVax), evaluated quantitatively and qualitatively the GC responses to the novel virus SARS-433

CoV-2 upon immunization. We found that SARS-CoV-2 mRNA vaccines had a superior 434 capacity, in comparison to rRBD-AddaVax, to elicit potent SARS-CoV-2 specific GC B cells 435 responses after the administration of a single vaccine dose. Importantly, we demonstrate here 436 that GC B cells and Tfh cells strongly correlated with the production of nAbs. In fact, mice 437 immunized with SARS-CoV-2 mRNA vaccines had robust GC responses coupled with in vitro 438 neutralization of SARS-CoV2 virus. Conversely, a single immunization with rRBD-AddaVax 439 resulted in minimal GC responses and lack of nAb production over time. The notion that 440 effective nAb formation deeply relies on GC reactions has been widely studied in the HIV field 441 (Havenar-Daughton et al., 2017) . It is well established that the generation of broadly nAbs 442 (bnAbs) capable of neutralizing most HIV strains depend on a hard-core GC-dependent affinity 443 maturation process lasting several years. The generation of HIV bnAbs is an extreme example 444 for many reasons, including the elevated rate of mutation of HIV that forces bnAbs to acquire an 445 unusually high degree of SHM and to target less immunodominant conserved epitopes (Havenar-446 Daughton et al., 2017) . Many viruses do not require such drastic GC responses to confer 447 protection. Nevertheless, studies on other viruses such as influenza virus, revealed that most 448 human Abs to influenza are heavily mutated, and these mutations are likely critical to mediate 449 broad protection against the virus (Victora and Wilson, 2015) . Hence, it is not unexpected to 450 observe here a strong connection between GC responses, which harbor SHM and the affinity 451 maturation process, and the capacity of the SARS-CoV-2 mRNA vaccines to foster a high level 452 of nAbs after a single immunization. Future studies will be required to establish the degree of 453 SHM and affinity maturation resulting from immunization with SARS-CoV-2 mRNA vaccines, 454 and how SHM translate to in vivo protection. 455

The correlation between nAbs and Tfh cells following a single immunization also 456 hints at the importance of T cell help for the generation of protective SARS-CoV-2 specific Ab 457 responses. This finding is in line with recently reported correlations of antigen-specific CD4 T 458 cells (Grifoni et al., 2020) and circulating Tfh cells (Mathew et al., 2020) with SARS-CoV-2-459 specific Abs in COVID-19 donors. Of note, increased frequencies of Th1 and Th2 biased 460 circulating Tfh cells were found in humans to be associated with the highest plasma neutralizing 461 activity in COVID-19 patients (Juno et al., 2020) . This is an important finding strengthening the 462 idea that the functional profile of Tfh cells might be relevant in shaping the quality of 463 neutralizing response to SARS-CoV-2. Our study demonstrated that the bona fide Tfh cells 464 induced by SARS-CoV-2 mRNA vaccines had a mixed Th1-Th2 functional profile (especially 465 clear in RBD mRNA immunized BALB/c mice) that was characterized by the production of 466 IFN-γ and IL-4. This is opposed to the Th2 skewed Tfh cell profile triggered by rRBD-AddaVax. 467

In agreement with these data, we found that the rRBD-AddaVax vaccine favors Th2-type Ab 468 responses (high RBD-IgG1 titers, low RBD-IgG2 titers), while mRNA vaccines are superior at 469 molding mixed Th1-Th2 type Ab responses (high RBD-IgG1 and RBD-IgG2 titers). Multiple 470 studies have demonstrated that mRNA vaccines for SARS-CoV-2 trigger CD4 T cells responses 471 that are dominated by Th1 cytokine production (Corbett et al., 2020b; Laczkó et al., 472 2020; Zhang et al., 2020) . In light of IL-4 importance in maintaining the survival of GC B cells 473 and regulating GC reactions (Crotty, 2019; Vinuesa et al., 2016) , it is expected that Tfh cells, 474 differently from conventional CD4 T cells, might exhibit some degree of IL-4 production even in 475 a Th1-inducing milieu. This differential behavior of conventional CD4 T cells and Tfh cells can 476 be explained by alternative transcriptional requirement for the regulation of IL-4 production in 477 these different cell types. Indeed, Gata3 regulates IL-4 production in conventional CD4 T cell, 478

while Tfh cell biology does not depend on Gata3 (Nurieva et al., 2008) and the SLAM-Sap 479 pathway is instead responsible for the modulation IL-4 production by Tfh cells (Yusuf et al., 480 2010) . Overall, the fact that mRNA vaccines generate Tfh cells with a mixed Th1/Th2 profile 481 along with Th1 biased conventional CD4 T cells might be a desirable feature to potentially avoid 482 vaccine-associated enhanced respiratory disease (VAERD) (Graham, 2020) . In fact, VAERD has 483 been connected to immune responses skewed toward a Th2 profile in other preclinical and 484 clinical studies on different respiratory viruses. 485

At least two reasons could explain the superior ability of SARS-CoV-2 mRNA 486 vaccines. First, mRNA encoding SARS-CoV2 antigens are translated into proteins directly in the 487 host cells (Pardi et al., 2018b) and it was shown for the Luc mRNA control vaccine that 488

Luciferase expression is detectable for at least 10 days following injection (Pardi et al., 2018a; 489 2015) . This could lead to a more prolonged antigen availability along with continuous 490 presentation of antigens via MHC class II molecules in comparison to the administration of a 491 recombinant protein, which in turn could foster greater GC responses. In keeping with this 492 hypothesis, prolonged antigen administration via osmotic pumps results in improved GC B cell 493 and Tfh generation in non-human primate and murine HIV vaccine models (Cirelli et al., 2019; 494 Tam et al., 2016) . It is worth highlighting that a possible extended protein production driven by 495 mRNA vaccines did not cause the formation of overt overextended GC reactions. Our analysis of 496 GC responses over time demonstrated a deep quantitative reduction of Tfh and GC B cells after 4 497 weeks post immunization. However, it is interesting to notice that the frequency of SARS-CoV-498 2-specific GC B cells was still elevated at this time point, possibly suggesting that low-level GC 499 activity might still be occurring and contributing to the elevated SARS-CoV-2 specific IgG titers 500 and nAbs found 60 days post immunization. In various experimental models, sensitive 501 techniques have indeed shown that antigen-specific GCs might be present for a prolonged time 502

and that late GC responses could contribute to the clonal diversity of B cell responses (Allen et 503 al., 2007) . A second less explored alternative/complementary explanation for the superior GC 504 activity of mRNA vaccines could be that a component of this vaccine platform is endowed with 505 an intrinsic Tfh cell adjuvanticity and/or is more efficient at priming naïve CD4 T cells. 506

Overall, our study provides a detailed quantitative and qualitative overview of the 507 GC responses induced by two types of SARS-CoV-2 vaccines and uncovers a strong connection 508 between SARS-CoV-2 nAb generation and SARS-CoV-2-specific GC reactions. With multiple 509 vaccines currently being considered as candidates to fight the COVID-19 pandemic, this study 510 sets a benchmark that can improve the evaluation of the immune responses induced by SARS-511

CoV-2 vaccine candidates in future pre-clinical and clinical studies. 512 513

While the points discussed above could explain the different performance of the two vaccine 515 platforms in regulating GC primary responses upon a single immunization, we do not rule out the 516 possibility that sequential doses of booster immunizations with rRBD-AddaVax might give rise 517 to improved GC responses and nAb production, as indicated by our analysis of secondary GC B 518 cell and Tfh cell responses at 10 days post boost. A late nAb response to rRBD-AddaVax 519 following booster immunizations, which might be suggested by the observed delayed kinetics of 520 GC B and Tfh cells after a single immunization, could also be plausible and cannot be excluded 521 by our investigation. Additionally, as suggested by a recently reported Phase 1 clinical trial 522 (Keech et al., 2020) , the usage of SARS-CoV-2 full S as immunogen, which is a much longer 523 protein containing a higher number of epitopes, combined with stronger adjuvants such as 524

Matrix-M1, might likely lead to superior GC responses and nAb formation in comparison to 525 rRBD-AddaVax, even during primary responses. Hence, it is possible that repeated 526 immunizations with different recombinant SARS-CoV-2 protein/adjuvant combinations might 527 represent a successful approach to elicit GC responses and nAbs in humans. Future studies will 528 be needed to address all the points discussed above. In (A-C) n = 9 mice per group were analyzed. Data are combined from three independent 586 experiments. Mean ± SEM is shown, and each data point represents an individual mouse. One-587 way-ANOVA with Bonferroni correction or unpaired two-tailed Mann-Whitney U tests was 588 conducted according to the distribution of the data. In (D-E) n = 4 mice per group from two 589 independent experiments were analyzed and representative samples were displayed. * p ≤ 0.05, 590 ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. See also Figure S1 , Table S2 and Table S5 . One-way-ANOVA with Bonferroni correction or unpaired two-tailed Mann-Whitney U tests was 606 conducted according to the distribution of the data. In (C-D), statistics were calculated versus 607 Luc mRNA group. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. See also Figure S2  608 and Table S2 . In (A-B) n = 8 mice for RBD-mRNA immunization, and n = 9 mice per group for other 623 immunization conditions. In (C-F) n = 9 mice per group were analyzed. In (A-F) data was 624 combined from three independent experiments. Mean ± SEM is shown, and each data point 625 represents an individual mouse. One-way-ANOVA with Bonferroni correction or unpaired two-626 tailed Mann-Whitney U tests was conducted according to the distribution of the data. * p ≤ 0.05, 627 ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. See also Figure S3 , Table S2 and Table S3 . Mean ± SEM is shown, and each data point represents an individual mouse. In (F-G) geometric 646 mean ± geometric SD is shown, and each data point represents an individual mouse. LOD: Limit 647 of detection (dotted line). One-way-ANOVA with Bonferroni correction or unpaired two-tailed 648

Mann-Whitney U tests was conducted according to the distribution of the data. * p ≤ 0.05, ** p ≤ 649 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. See also Figure S4 . In (A-E) n = 9 mice per group were analyzed. Data are combined from three independent 668 experiments. In (F) n = 9 mice per group were analyzed at day 7 and day 14. Data are combined 669 from three independent experiments. n=10 mice per group were analyzed at day 28, and data are 670 combined from two independent experiments. In (G-H) n = 10 mice per group were analyzed. 671

Data are combined from four independent experiments. In (B) and (F-H) , data was graphed as 672

Mean ± SEM. One-way-ANOVA with Bonferroni correction or unpaired two-tailed Mann-673 Whitney U tests was conducted according to the distribution of the data. For kinetics in (F), day 674 0 represents the average of 18 naïve animals, and statistics were calculated versus Luc mRNA 675 group. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001. See also Figure S5 and Table S1 . Ab titers are shown as geometric mean ± geometric SD. Each data point represents an individual 697 mouse. One-way-ANOVA with Bonferroni correction or unpaired two-tailed Mann-Whitney U 698 tests was conducted according to the distribution of the data. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 699 0.001, **** p ≤ 0.0001 See also Figure S6 and Table S1 . 

Immunogen preparation and immunization 745 SARS-CoV-2 RBD protein (rRBD) was produced in 293F cells, as described previously 746 (Amanat et al., 2020; Stadlbauer et al., 2020) . Briefly, 600 million cells were transfected with 747 200 µg of purified DNA encoding codon-optimized RBD of SARS-CoV-2 using ExpiFectamine 748 293 transfection kit (Gibco, #A14525). The manufacturer's protocol was followed and cells were 749 harvest on day 3. Cells were spun at 4000g for 10 minutes and sterile-filtered with a 0.22 µm 750 filter. Supernatant was incubated with Ni-NTA resin (Qiagen cat#30230) for 2 hours. After 2 751 hours, this mixture was loaded onto columns and the protein was eluted using elution buffer with 752 high amounts of imidazole. Protein was concentrated using 10 kDa Amicon centrifugal units 753 (Millipore Sigma cat#UFC901024) and re-constituted in PBS. Concentration was measured 754 using Bradford reagent (Bio-Rad cat#5000201) and a reducing sodium dodecyl sulphate-755 polyacrylamide gel electrophoresis (SDS-PAGE) was run to check the integrity of the protein. 756 SARS-CoV-2 full length S (full S) protein was prepared as described above for SARS-CoV-2 757 rRBD. 758

Full-length spike ∆furin (full S ∆furin, RRAR furin cleavage site abolished) and RBD mRNA 759 vaccines were designed based on the full spike (S) protein sequence of SARS-CoV-2 (Wuhan-760

Hu-1, GenBank: MN908947.3) (Laczkó et al., 2020) . mRNA coding sequences of RBD, full S 761 ∆furin and firefly luciferase (Luc) were codon-optimized, synthesized, produced and 762 encapsulated in lipid nanoparticles (LNP) as previously reported (Freyn et al., 2020) . 763 rRBD was diluted in phosphate buffered saline (PBS, Corning, 21-031-CV) and combined in a 764 1:1 ratio with AddaVax (InVivoGen, vac-adx-10). mRNA-LNP were diluted in PBS prior to 765 inoculation. Vaccines were injected into the gastrocnemius muscle using a 0.5 mL 28 G x 1/2" 766 insulin syringe (BD Biosciences, 329461). For booster immunizations, the same dose of the 767 respective vaccine was injected into the same site as the primary immunization. All staining steps were carried out at 4°C in FACS buffer (PBS with 2% heat inactivated FBS). 798

Single cell suspensions were Fc blocked with anti-CD16/CD32 monoclonal antibody (mAb) 799 prior to staining. 800

Tfh cell: Cells were incubated with biotinylated CXCR5-biotin for 1 hour, washed and then 801 incubated with a cocktail of fluorescently labeled anti-mouse mAbs, streptavidin, and Fixable 802 Viability dye eFluor780 for 30 minutes. Cells were washed with FACS buffer, then fixed and 803 permeabilized in FoxP3/Transcription Factor Staining Buffer Set (eBioScience, 00-5523-00) 804 according to manufacturer's instructions before intranuclear staining with mAbs (Table S1) Then, they were stained with a cocktail of fluorescently labeled anti-mouse mAbs containing 808 streptavidin, Fixable Viability dye eFluor780, RBD-or full S-PE, and RBD-or full S-Alexa 809

Fluor 647 (Table S2) (Table S3) Table S5 . After washing in PBS, slides were mounted using ProLong TM Diamond 856 antifade (Invitrogen), and imaged using a Zeiss LSM710 confocal microscope. Tile scan images 857 were taken using a 10X magnification lens with 10 % overlap. Enlarged images were taken using 858 a 20X magnification lens. Images were processed and analyzed using Zeiss Blue 3.1 (Zeiss) and 859 containing Earl's salts and L-glutamine supplemented with 100 U/mL Penicillin, 100 µg/mL 894 Streptomycin, 2% FBS, 1 mM sodium pyruvate, and 1X MEM nonessential amino acids). Assay 895 controls were treated and diluted in the same manner as experimental samples. 100 TCID 50 of 896 authentic SARS-CoV-2 (isolate USA-WA1/2020, BEI Resources NR-52281) was added to each 897 sample and incubated for 1 hour. Cells were washed 1 time in PBS. Sample/virus (100µL) was 898 added to cells and incubated for 4 days. Cells were fixed in 10% Neutral-buffered formalin 899 (VWR 16004-128) and stained with 0.1% Crystal Violet (Sigma C38886-500g). Neutralization is 900 demonstrated by absence of cytopathic effect in Vero E6 monolayers. Data is reported as the 901 geometric mean of the reciprocal dilution factor of two replicates and defined as Minimal 902 effective concentration (MEC). An improved fluorescent assay was performed to generate the 903 neutralization data in Figure 7 and S7 with the following modifications to the original assay. 904

Vero E6 cells (4x10 4 ) were used to coat the plates. Each serum sample, prepared as described 905 above, was incubated with 100 TCID 50 fluorescent SARS-CoV-2 (Xie et al., 2020) Antibody neutralization assay using VSV∆G-RFP SARS-CoV-2: all mouse sera were heat-927 inactivated for 30 minutes at 55 ⁰C prior to use in neutralization assay. Vero E6 cells stably 928 expressing TMPRSS2 were seeded in 100 µL at 2 x 10 4 cells/well in a 96-well collagen coated 929 plate. The next day, 2-fold serially diluted serum samples were mixed with VSV∆G-RFP SARS-930

CoV-2 pseudotype virus (100-400 focus forming units/well) and incubated for 1 hour at 37 ⁰C. 931

The mouse anti-VSV Indiana G 1E9F9 was also included in this mixture to neutralize any 932 potential VSV-G carryover virus at a concentration of 600 ng/mL (Absolute Antibody, 933

Ab01402-2.0). Cells were infected with the respective serum-virus mixture for 23-24 hours. 934

Then, cells were washed and fixed with 4% paraformaldehyde before visualization on an S6 935 FluoroSpot Analyzer (CTL). Individual infected foci were enumerated, and the values compared 936 to control wells without serum. The focus reduction neutralization titer 50% (FRNT50) was 937 measured as the greatest serum dilution at which focus count was reduced by at least 50% 938 relative to control cells that were infected with pseudotype virus in the absence of mouse serum. 939

FRNT50 titers for each sample were measured in at least two technical replicates performed on 940 separate days. 941 942

Both inguinal and popliteal LN were harvested, pooled per mice, and processed as described 944

above. Cells were incubated with anti-mouse CD16/CD32 and CXCR5-biotin antibodies as 945 outlined in the Tfh staining. The following cocktail of anti-mouse mAbs was used to stain the 946 cells and identify Tfh and naïve CD4 T cells: B220 Alexa Fluor 700, CD4 PerCP-Cy5.5, CD44 947 FITC, and PD-1 PE-Cy7; together with Streptavidin BV421 and Fixable Viability dye 948 eFluor780. Cells were washed and resuspended in FACS buffer and naïve CD4 T (live B220 -949 CD4 + CD44 -) and Tfh (live B220 -CD4 + CD44 hi CXCR5 + PD-1 hi ) cell populations were sorted on 950 FACSAria Fusion (BD Biosciences). The purity of the sorted cells was immediately checked 951 using the same instrument and was 97% or higher. Total RNA was extracted with RNeasy plus 952 micro kit (Qiagen, 74034) following manufacturer's instructions. 953

Total RNA was retrotranscribed into cDNA with SuperScript II reverse transcriptase (Invitrogen, 954 18064014). Quantitative real-time PCRs (qPCR) for Ascl2 and Actb were performed using the 955 following primer pairs: Actb: 

Germinal-center organization and cellular 977 dynamics

SARS-CoV-2 Vaccines: Status Report

A 984 serological assay to detect SARS-CoV-2 seroconversion in humans

Safety and Immunogenicity of SARS-CoV-2 mRNA-1273 Vaccine in Older 993 Adults

Cutting Edge: Distinct waves of BCL6 995 expression during T follicular helper cell development

Dynamic regulation of 998 T follicular regulatory cell responses by interleukin 2 during influenza infection

Slow Delivery Immunization Enhances HIV Neutralizing Antibody 1006 and Germinal Center Responses via Modulation of Immunodominance

SARS-CoV-2 mRNA Vaccine Development Enabled by Prototype Pathogen

Evaluation of the mRNA-1273 Vaccine against SARS-CoV-2 in 1031 Nonhuman Primates

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

WHO Declares COVID-19 a Pandemic

Dynamics of B cells in germinal centres

A Multi-Targeting, Nucleoside-Modified mRNA 1041 Influenza Virus Vaccine Provides Broad Protection in Mice

Development of an inactivated vaccine candidate for SARS-CoV-1046 2

Identification of a T follicular helper 1050 cell subset that drives anaphylactic IgE

Rapid COVID-19 vaccine development

Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with COVID-19 1056 Disease and Unexposed Individuals

Tfh cells and HIV bnAbs, an 1058 immunodominance model of the HIV neutralizing antibody generation problem

Follicular dendritic cells: dynamic antigen 1061 libraries

An mRNA Vaccine against SARS-1068

Humoral and circulating follicular helper T cell responses in recovered patients with 1073 COVID-19

Phase 1-2 Trial 1078 of a SARS-CoV-2 Recombinant Spike Protein Nanoparticle Vaccine

A single immunization with nucleoside-modified 1087 mRNA vaccines elicits strong cellular and humoral immune responses against SARS-CoV-2 1088 in mice

Vaccine priming is restricted to draining lymph nodes and controlled by adjuvant-1092 mediated antigen uptake

1095 Induction of Robust B Cell Responses after Influenza mRNA Vaccination Is Accompanied 1096 by Circulating Hemagglutinin-Specific ICOS+ PD-1+ CXCR3+ T Follicular Helper Cells. 1097 Front Immunol 8

Transcription factor achaete-scute homologue 2 initiates follicular T-1101 helper-cell development

International AIDS Vaccine Initiative 1104

Human circulating PD-1105 1+CXCR3-CXCR5+ memory Tfh cells are highly functional and correlate with broadly 1106 neutralizing HIV antibody responses

Oil-in-Water Emulsion MF59 Increases 1110

Cell Differentiation and Persistence in Response to Vaccination

The UPenn COVID Processing Unit †, Betts, M

Deep immune profiling of COVID-19 patients reveals distinct 1119 immunotypes with therapeutic implications

Human blood CXCR5(+)CD4(+) T cells are counterparts of 1125 T follicular cells and contain specific subsets that differentially support antibody secretion

Phase 1/2 study of COVID-1131 19 RNA vaccine BNT162b1 in adults

Phase 1/2 Study to Describe 1136 the Safety and Immunogenicity of a COVID-19 RNA Vaccine Candidate

Adults 18 to 55 Years of Age

Generation of T follicular helper cells is 1140 mediated by interleukin-21 but independent of T helper 1, 2, or 17 cell lineages

MF59 is a safe and potent vaccine adjuvant that enhances protection 1143 against influenza virus infection

Simply put: Vaccination saves lives

Nucleoside-modified mRNA vaccines induce potent T follicular helper and germinal center 1153 B cell responses

Zika virus protection by a 1160 single low-dose nucleoside-modified mRNA vaccination

mRNA vaccines -a new era in 1162 vaccinology

Expression kinetics of nucleoside-modified mRNA 1165 delivered in lipid nanoparticles to mice by various routes

Cytokine-secreting follicular T cells shape 1167 the antibody repertoire

COVID-19 vaccine 1170 BNT162b1 elicits human antibody and TH 1 T-cell responses

From vaccines to memory and back

Phenotype and functions of memory Tfh cells in 1174 human blood

Pandemic 1176 Preparedness: Developing Vaccines and Therapeutic Antibodies For COVID-19

Immunogenicity of a DNA vaccine candidate for COVID-19

Interferon-gamma and B cell stimulatory factor-1 reciprocally 1189 regulate Ig isotype production

SARS-CoV-2 Seroconversion in Humans: A Detailed 1193 Protocol for a Serological Assay, Antigen Production, and Test Setup

Regulation of antibody isotype secretion by subsets of antigen-specific 1197 helper T cells

CCR6 1200 Defines Memory B Cell Precursors in Mouse and Human Germinal Centers, Revealing 1201 Light-Zone Location and Predominant Low Antigen Affinity

Sustained antigen availability during germinal 1205 center initiation enhances antibody responses to vaccination

Germinal center selection and the antibody response to 1208 influenza

Vaccine BNT162b2 Selected for a Pivotal Efficacy Study

TFH cells progressively differentiate to regulate the germinal center response

WHO, World Health Organization. 2020. DRAFT landscape of COVID-19 candidate vaccines -1221

An Infectious cDNA Clone of SARS-CoV-2

Follicular T-helper cell recruitment governed by bystander B cells and ICOS-1229 driven motility

DNA vaccine 1238 protection against SARS-CoV-2 in rhesus macaques

Germinal center T follicular helper cell IL-4 production is dependent on 1241 signaling lymphocytic activation molecule receptor (CD150)

MEK inhibitors activate Wnt signalling 1245 and induce stem cell plasticity in colorectal cancer

A thermostable 1250 mRNA vaccine against COVID-19