Germinal center block exacerbates extrafollicular responses and 1 accelerates autoimmune disease progression in a murine lupus 2 model 3 4

Systemic lupus erythematosus and numerous other autoimmune diseases are characterized by affinity-matured, class-switched autoantibodies to nuclear antigens. Such antibodies are generally thought to arise in germinal centers (GCs). Several strategies to block GC formation and progression are currently being explored clinically. However, recent studies have suggested a key role for extrafollicular responses in driving the early events in autoimmune development. To investigate the relative contribution of these two pathways in autoimmune disease development, we leveraged a lupus murine model, where we could genetically block the GC pathway. We find that a B cell intrinsic block in GC formation accelerates extrafollicular responses and exacerbates autoimmune progression. The manifestations included higher levels of circulating, class-switched autoantibodies, as well as antibody- and complement-deposition in the kidney glomeruli. GC B cell cultures in vitro showed that loss of the GC transcription factor Bcl-6 prevents cellular expansion and accelerates plasma cell differentiation. This suggests that the in vivo phenotype was a direct consequence of rewiring of B cell intrinsic transcriptional programming. In a competitive scenario in vivo, in autoreactive mixed bone marrow chimeras, B cells harboring the genetic GC block contributed disproportionately highly to the plasma cell output. Taken together, this emphasizes the extrafollicular pathway as a key contributor to autoimmune pathogenesis and suggests that strategies aimed at blocking GCs should simultaneously target this pathway to avoid rerouting the pathogenic response. Highlights - Genetic GC block exacerbates autoimmune progression in a lupus model - An intrinsic GC block drives B cell differentiation into terminally differentiated plasma cells in vitro - B cells harboring a GC block competitively contribute to the plasma cell compartment in an autoreactive setting in vivo - Lupus mice with a GC block display immune complex deposition in kidney glomeruli that is indistinguishable from their wild-type counterparts

generally thought to arise in germinal centers (GCs). Several strategies to block GC formation 23 and progression are currently being explored clinically. However, recent studies have 24 suggested a key role for extrafollicular responses in driving the early events in autoimmune 25 development. To investigate the relative contribution of these two pathways in autoimmune 26 disease development, we leveraged a lupus murine model, where we could genetically block 27 the GC pathway. We find that a B cell intrinsic block in GC formation accelerates extrafollicular 28 responses and exacerbates autoimmune progression. The manifestations included higher 29 levels of circulating, class-switched autoantibodies, as well as antibody-and complement-30 deposition in the kidney glomeruli. GC B cell cultures in vitro showed that loss of the GC 31 transcription factor Bcl-6 prevents cellular expansion and accelerates plasma cell 32 differentiation. This suggests that the in vivo phenotype was a direct consequence of rewiring 33 of B cell intrinsic transcriptional programming. In a competitive scenario in vivo, in 34 autoreactive mixed bone marrow chimeras, B cells harboring the genetic GC block 35 contributed disproportionately highly to the plasma cell output. Taken together, this 36 emphasizes the extrafollicular pathway as a key contributor to autoimmune pathogenesis and 37 suggests that strategies aimed at blocking GCs should simultaneously target this pathway to 38 avoid rerouting the pathogenic response. 39 40 41 Highlights: 42 43 -Genetic GC block exacerbates autoimmune progression in a lupus model 44 -An intrinsic GC block drives B cell differentiation into terminally differentiated plasma cells 45 in vitro 46 -B cells harboring a GC block competitively contribute to the plasma cell compartment in an 47 4/31 inadvertent broadening of the response. An additional layer of control appears to be exerted 98 by a specialized subset of T regulatory cells, termed T follicular regulatory (TFR) cells (Fahlquist  99 Hagert & Degn, 2020). Nonetheless, these mechanisms appear to fail in autoimmune disease, 100 which frequently display rampant GC activity (Domeier et al., 2017;Luzina et al., 2001). 101 102 Hence, we hypothesized that the GC pathway is critical to the autoimmune process, and that 103 a genetic block of the GC pathway in vivo would prevent autoimmune development. To our 104 surprise, a global block in the GC pathway in vivo did not ameliorate autoimmune disease, but 105 rather exacerbated it. In an in vitro GC B cell culture system, GC blocked B cells expanded to 106 a lesser extent, but were found to more rapidly develop into PBs and PCs. In a competitive 107 scenario in vivo, GC blocked B cells competed efficiently with their wild-type counterparts, 108 disproportionate to their inability to participate in GCs. Our determination of the relative 109 contributions of the extrafollicular and GC pathways to autoimmune progression highlight a 110 critical role of extrafollicular responses in driving autoimmune development. 111

113
Mice 114 The Bcl-6 flx/flx strain (Hollister et al., 2013) and congenic B6.CD45.1 (B6.SJL-Ptprc a Pepc b /BoyJ) 115 were purchased from Jackson Laboratories (stock no. 023737 and 002014, respectively). 116 Aicda-Cre transgenic mice (Kwon et al., 2008) were kindly provided by Meinrad Busslinger, 117 The Research Institute of Molecular Pathology (IMP), Vienna. Aicda-Cre and Bcl-6 flx/flx strains 118 were intercrossed to generate Aicda-Cre+ and Aicda-Cre-Bcl-6 flx/flx littermates. 564Igi mice 119 (Berland et al., 2006) (B6.Cg-Igh tm1(Igh564)Tik Igk tm1(Igk564)Tik /J) were kindly made available by 120 Theresa Imanishi-Kari, Tufts University, and provided by Michael C. Carroll, Boston Children's 121 Hospital Mice were treated topically on the right ear 3 times per week for 4 weeks with 1 mg R848/mL 133 acetone using a cotton applicator or did not receive any treatment (Untreated/Unt Kit (Molecular Probes, S10453) according to manufactures instructions. In brief, antibody 205 7/31 (either "14D12" rat IgG2a to mouse MBL-C (Hycult Biotech), or "RTK2758" rat IgG2a isotype 206 control (Abcam)) was concentrated in antibody preparation buffer to a concentration of 2 207 mg/mL or above. Next, carbohydrates on the antibody were modified by the incubation with 208 β-galactosidase for 4 h at 37°C. Azide modification was achieved through incubation with 209 uridine diphosphate glucose-GalT enzyme overnight at 30°C. Antibody with modified 210 carbohydrates was purified and concentrated through a series of centrifugation steps using a 211 molecular-weight cutoff membrane concentrator, and the buffer was simultaneously 212 changed to 20 mM Tris, pH 7.0. Finally transformed and re-tested for normality. The following data sets were log-normally 358 distributed: Fig 1B, 1D, 1E, 1N (all 3 data sets), 1P (all 3 data sets), 3F, 5F, 5G, 5H. All other 359 datasets were normally distributed, except for the data in Figure 1L and 5F, which were 360 neither normally, nor log-normally distributed. However, a non-parametric t-test for the 361 isolated data of between each group in panel 1L and 5F showed similar results (for Fig 1L:  pathways in production of autoantibodies and cell frequencies in the absence of secondary 387 effects caused by organ failure, we decided to treat mice 3 times weekly for only 4 weeks (Fig.  388  1A). 389 390 After 4 weeks of treatment, the Bcl-6 flx/flx (Cre-) controls and the Aicda-Cre Bcl-6 flx/flx (Cre+) 391 mice showed similar, significant increases in spleen weight upon R848 treatment (Fig. 1B). 392 Anti-dsDNA autoantibodies of the IgG2c subtype measured in serum were dramatically 393 elevated upon R848 treatment compared with untreated animals. Surprisingly, there was a 394 trend towards higher levels in treated Cre+ mice compared with treated Cre-mice (Fig. 1C).

395
A similar trend towards an increase was seen in total IgG2c levels (Fig. 1D). No statistically 396 significant differences in total IgG1 and total IgG3 levels were seen upon treatment (Fig. 1E  397 and F). 398 399 To validate the effect of R848 treatment and the integrity of the GC block in Cre+ mice, we 400 carried out immunofluorescence staining of spleens to identify GC formation ( Fig. 1G-J). Using 401 the proliferation marker Ki67 and the naïve B cell marker IgD, it was evident that larger and 402 more frequent GCs were observed upon R848 treatment in Cre-animals (Fig. 1G, H and K).

403
Quantification revealed that this difference was statistically significant (Fig. 1L). Cre+ animals 404 did not display any baseline or R848-induced GCs ( Fig. 1I and J). However, in Cre+ treated 405 mice we did observe many proliferating cells at the T-B border and in the red pulp, likely 406 abortive primary foci and extrafollicular foci, respectively. 407 408 Flow cytometric analyses of inguinal LNs (IngLNs), mesenteric LNs (MesLN), and the spleen 409 were carried out ( Fig. 1M-Q, Fig S1). In treated animals, we saw slight increases in monocyte 410 and neutrophil frequencies in some of the tissues (Fig. 1M and N), and a slight increase in B 411 cell frequencies in the skin-draining IngLNs (Fig. 1O), which might be caused by the direct 412 stimulatory effect from the R848 treatment of the ear skin. We observed robust GC B cell 413 frequencies in Cre-R848 treated mice, compared to untreated littermates (Fig. 1P). No GC B 414 cells were found in Cre+ animals, further validating the fidelity of the GC block (Fig. 1P). 415 Surprisingly, despite this, we found a significant increase in PB and PC frequencies upon 416 12/31 treatment in both groups, and the level was significantly higher in the spleens of mice 417 harboring a GC block compared to Cre-R848-treated littermate controls (Fig. 1Q). This 418 observation corresponded well with the increase in plasma IgG2c autoantibody levels as well 419 as total IgG2c levels (Fig 1C and D). Taken together, this surprisingly indicated an exacerbated 420 autoimmune phenotype in GC block mice compared to GC sufficient mice upon R848 421 treatment. 422 14/31

439
To further understand the local effects of R848 treatment, we performed 440 immunofluorescence microscopy analyses of draining auricular lymph nodes (AurLNs) from 441 treated mice and untreated controls. We observed gross enlargement of the lymph nodes of 442 treated animals, with a robust induction of GCs in Cre-R848-treated mice (Fig. S2A). In 443 comparison, the Cre+ R848-treated mice had many proliferating cells outside the follicles (Fig.  444  S2A). These dividing cells in the AurLNs overlapped to some extent with the PC marker CD138, 445 pointing towards dividing extrafollicular PCs (Fig. S2B). To further understand the importance of the increased anti-dsDNA IgG2c autoantibodies in 454 Cre+ R848-treated mice, in the face of a complete absence of GCs, we implemented this 455 nanoparticle tracking approach ( Fig. 2A-C). We analyzed superoligomeric complexes in the 456 band from 90-130 nm ( Fig. 2D and E). In agreement with spMBL as a lupus marker, treated 457 mice tended towards higher levels, as compared to untreated mice, but interestingly, we also 458 observed a global trend towards higher levels in Cre+ compared to Cre-animals (Fig. 2F). 459 Although they did not reach statistical significance, these observations were well in line with 460 the previously noted increases in anti-dsDNA and total IgG2c antibodies upon R848 461 treatment, and in Cre+ compared to Cre-animals (Fig. 1C). In Cre-treated mice, GC B cell frequencies were positively correlated with the concentration 473 of spMBL particles in serum (Fig. 2G). Total IgG2c levels were similar between treated Cre+ 474 and Cre-mice (Fig. 2H) with a significant positive correlation between anti-dsDNA IgG2c and 475 the level of spMBL particles (Fig. 2I). A significant positive correlation was also observed 476 between the frequency of splenic GC B cells and the ratio between the spMBL and total IgG2c 477 levels in serum (Fig. 2J). Conversely, a significant inverse correlation was found between the 478 frequency of splenic GC B cells and the ratio between the spMBL and anti-dsDNA IgG2c 479 antibody concentrations measured in serum (Fig. 2K). Taken together, these findings now 480 took our original observations from the autoreactive B cell receptor knock-in model (564Igi) 481 into the epicutaneous R848 model on wild-type (Cre-) background. Importantly, in Cre+ mice, 482 there was an uncoupling of the concentration of spMBL particles and GC B cell levels, 483 indicating that the GC block failed to curb the production of spMBL complexes.

499 500
Enhanced immune complex deposition in kidneys of R848-treated mice 501 To understand the pathological importance of the GC block in R848-treated mice and the 502 elevated serum IgG2c, serum anti-dsDNA IgG2c and spMBL levels, we investigated whether 503 there were any pathological changes associated with lupus nephritis in the kidneys. First, we 504 carried out Periodic acid-Schiff (PAS) staining, which displayed no obvious kidney injury upon 505 R848 treatment (Fig. 3A). Apart from the presence of mesangial and capillary immune 506 deposits, histopathological changes associated with lupus nephritis may include increased 507 matrix or mesangial cellularity, endocapillary proliferation, thickening of capillary walls, 508 glomerular tuft necrosis, extracapillary proliferation (crescents), karyorrhexis, hyaline 509 thrombi (micronodular intracapillary aggregation of immune complexes), and glomerular 510 sclerosis (segmental or global), as well as, rarely, pathognomonic hematoxylin bodies 511 (Gasparotto et al., 2020;Weening et al., 2004). However, no significant histopathologic 512 findings were identified by light microscopy in any of the mice included in any of the groups. 513 In line with the PAS stainings, we did not observe any differences in glomerular nephrin levels 514 among the groups, suggesting normal glomerular podocytes (Fig. 3B). As we could not identify 515 any gross pathological changes nor changes in nephrin levels in the kidneys upon treatment, 516 this verified our short-term treatment strategy in terms of the goal to investigate early 517 immune-driven events in the absence of any secondary pathology. 518 519 To evaluate if immune complex deposition occurred in the kidney glomeruli, and if there were 520 any differences between GC-sufficient and deficient groups, we performed 521 immunofluorescence staining of kidney sections targeting total Ig, C3, and IgG2c (Fig 3B, 3C, 522 3E-G). The total Ig levels in glomeruli were clearly increased upon R848-treatment. 523 Interestingly, we found a trend towards an increase in the total Ig levels of Cre+ mice 524 compared with littermate R848-treated controls ( Fig. 3B and E). We also found that there was 525 a significant increase of antibodies of the pathogenic subtype IgG2c upon R848 treatment 526 ( Fig. 3C and G), and a trend towards an increase in C3 deposition upon R848-treatment ( Fig.  527 3C and F). However, differences between Cre+ and Cre-R848-treated groups were seen 528 neither in C3 nor IgG2c levels ( Fig. 3F and G). 529 530 Taken together, R848-treated mice displayed immune complex deposition in glomeruli, based 531 on an increased level of C3, total Ig and IgG2c (Fig. 3E-G). We corroborated these findings by 532 peroxidase-stainings, as a corollary to the immunofluorescence microscopy, and this 533 confirmed the glomerular changes in total Ig, C3 and IgG2c upon R848-treatment (Fig S3).    (Fig. 4A). Naïve B cells purified from 554 Cre-and Cre+ Bcl-6 flx/flx mice by negative magnetic-activated cell sorting were seeded onto 555 fibroblast feeder cells expressing CD40L, IL-21 and BAFF, and stimulated with IL-4 (Fig. 4A). 556 557 The combination of CD40L, BAFF, IL-21, and IL-4 stimulation has previously been shown to 558 induce a robust expansion of B cells with a GC-like phenotype, followed by differentiation compared to those derived from Cre+ mice ( Fig. 4B and E). This was mirrored by a similar 563 relative increase in PBs (B220+, CD138+) in Cre-cultures ( Fig. 4B and D), but a relative 564 decrease in PCs (B220neg, CD138+) ( Fig. 4B and C). Of interest, the total number of cells in 565 the live gate for Cre-cultures was approximately 4 times higher than that of Cre+ cultures 566 (Cre-: 140,000 vs. Cre+: 33,000, Fig. S4C). To understand this difference in cell numbers, we 567 investigated whether the Cre+, and hence Bcl-6 deficient, B cells had an increased propensity 568 to undergo apoptosis, because Bcl-6 has previously been reported to suppress P53 and inhibit 569 apoptosis in GC B cells (Phan & Dalla-Favera, 2004). Somewhat surprisingly, we found that 570 upon IL-4 stimulation, there was no difference in the frequency of dead cells (Fig. 4F and G), 571 a slight and significant drop in apoptotic cell frequency ( Fig. 4F and H), and a corresponding 572 increased relative frequency of live cells in Cre+ cultures compared to Cre-cultures on day 6 573 ( Fig. 4F and I). However, at day 10 there were no significant differences in live, apoptotic, nor 574 necrotic cell frequencies between Cre-and Cre+ cultures (Fig. 4J-M). Thus, apoptosis could 575 not account for the dramatic difference in resulting cell numbers between Cre+ and Cre- (Fig.  576  S4C). Taken together, this suggested that the higher overall cell numbers in Cre-cultures was 577 not simply a reflection of increased apoptosis among Bcl-6 deficient cells in Cre+ cultures, but 578 rather represented an improved intrinsic proliferative potential of the Bcl-6 sufficient cells. 579 580 In summary, our iGB experiments revealed a vigorous expansion of B cells and PCs in Cre-581 cultures, but less pronounced PC differentiation, whereas Cre+ cultures conversely displayed 582 a lesser degree of proliferation but more pronounced PC differentiation. This suggested that 583 B cells with an intrinsic GC block may differentiate quicker to PCs, and thereby lose their 584 proliferative capacity, a notion that is in line with the established function of Bcl-6 in 585 repressing upregulation of Blimp-1 (Vasanwala et al., 2002). 586   Our observation that Bcl-6 deficient B cells rapidly lost their replicative potential in vitro (Fig.  606  4) was somewhat at odds with our in vivo observations from the R848 model, which displayed 607 a global increase in PCs and autoantibodies (Fig. 1). This is because even if Bcl-6 deficient cells 608 more readily became PCs, their poorer capacity to expand compared to Bcl-6 sufficient cells 609 would be predicted to limit PC output. However There were no differences in the basic parameters when comparing Cre-control BM chimera 636 mice with Cre+ BM chimera mice, in which approximately 50% of the B cells harbored a GC 637 block (Fig. 5B-H). That is, aside from a very small but significant relative increase of CD8 T cells 638 in IngLN of Cre+ chimeras, we saw no statistically significant differences in anti-dsDNA IgG2c 639 (Fig. 5B), total anti-dsDNA Ig (Fig. 5C), B cell frequencies (Fig. 5D), CD4 and CD8 T cell 640 frequencies ( Fig. 5E and F, respectively), overall GC B cell frequencies (Fig. 5G) and PB/PC 641 frequencies (Fig. 5H) between the groups. This confirmed that the two groups of chimeras 642 were comparable and had robust GCB and PC compartments. In the total B cell compartment, 643 CD45.2 positive cells were present at levels comparable to that of CD45.1 cells in both Cre-644 (Fig. 5I) and Cre+ (Fig. 5J) chimeras. However, within the GC B cell gate, CD45.2 cells were 645 robustly represented in Cre-chimeras, but virtually absent in Cre+ chimeras ( Fig. 5I and J). 646 When quantifying this effect across chimeras and expressing as the ratio of CD45.2 of GCB 647 relative to CD45.2 of total B cells, it was clear that Bcl-6 deficient cells, as expected from their 648 genetic deficiency, were incapable of contributing to the GC compartment (Fig. 5K). However, 649 22/31 when similarly comparing PB/PC ratio over B cells, the cells harboring a GC block remained 650 able to contribute to the final PB/PC pool, albeit underrepresented relative to the competitor 651 cells (Fig. 5L).  GCs are believed to be the nexus of autoreactive responses in a range of autoimmune 669 diseases. Due to their role in potent antibody responses, memory generation, and long-lived 670 PC formation, there has been extensive interest in targeting GCs in autoimmune disease. The 671 strategy has proven useful in autoimmune models, but due to off-target effects, did not 672 initially progress through clinical trials ( we found blocking GCs did not lessen autoreactive manifestations, but in some cases 683 worsened these. Upon autoimmune induction, we observed a trend towards an increase in 684 anti-dsDNA antibodies of the IgG2c isotype (Fig. 1C) and total IgG2c antibody ( Fig. 1D) in GC 685 block mice, compared to WT. These changes were also mirrored by a significantly higher 686 frequency of PB/PCs in the spleens of GC block mice, despite a total absence of GCs. In 687 agreement with this, we observed robust deposition of immune complexes in the kidney 688 glomeruli of GC block mice, at least on par with that of GC sufficient controls (Fig. 3). This 689 indicated that the extrafollicular pathway could compensate, and in some cases even 690 augment, the autoreactive response. 691 692 To understand the B cell intrinsic effect of a GC block, we leveraged an induced GC B cell significantly increased propensity to undergo apoptosis ( Fig. 4H and L), rather, they much 697 more readily underwent terminal differentiation to PCs, and had a dramatically reduced 698 capacity to expand compared to their wild type counterparts (Fig. 4). However, this agreed 699 well with the established cross-regulation between Bcl-6 and Blimp-1, the master regulator 700 of the PC fate, also known as Prdm1 (Vasanwala et al., 2002). Although the increased 701 propensity for terminal PC differentiation was, in principle, well in line with our in vivo 702 observations, the lack of proliferative capacity was at the same time at odds with the dramatic 703 PC output in the mice harboring a GC block in B cells. This suggested that the PC differentiation 704 process in mice displaying a global GC block in B cells might be dysregulated, potentially as a 705 consequence of absence of GC-derived antibody feedback, as previously suggested for GC B 706 cells (Zhang et al., 2013). To address this possibility, we asked whether B cells with a GC block 707 would be precluded from contributing to the PC pool in a GC sufficient environment. Our 708 findings demonstrated that this was not the case, although the relative contribution of GC 709 block B cells to the PC pool was smaller than that of GC sufficient B cells (Fig. 5). However, 710 given their inability to expand in GCs, the magnitude of the contribution of GC block B cells to 711 the PC compartment in direct competition with GC-sufficient B cells was remarkable. In the 712 infectious setting, an early wave of extrafollicular PCs is crucial for the initial antibody 713 response. However, most PCs produced by the extrafollicular response undergo apoptosis 714 26/31 within a matter of days, and the global response becomes dominated by GC-derived 715 responses. In the chronic autoreactive setting, however, the continuous fueling of the 716 autoimmune process may continually renew this population. 717 718 We may speculate that the somewhat lower contribution of the extrafollicular PC 719 compartment in the mixed chimera model compared to that of the R848 model could be a 720 consequence of the markedly different time scales of the two experiments: the mixed 721 chimeras were analyzed 13 weeks post reconstitution, whereas the R848 mice were analyzed 722 4.5 weeks after commencement of treatment. This could be important, because at this point 723 it remains unclear whether the GC responses observed in our models contribute a 724 qualitatively different response to the autoimmune progression, e.g., through production of 725 memory B cells and long-lived PCs that may perpetuate and dominate the chronic response 726 over longer periods of time. By extension, the GC pathway may differentially allow epitope 727 spreading and inclusion of alternative antigens over time, as seen in human SLE patients 728 (Arbuckle et al., 2003). At least, it seems plausible that the longer the autoimmune process 729 has persisted, the more the long-lived GC responses and their derived memory output come 730 to dominate the process. However, conversely, the short-lived extrafollicular responses may 731 govern the early stages of the response and, as previously suggested, the very early break-of-732 tolerance driven by nucleic acid-containing antigens (Soni et al., 2020;Sweet et al., 2011). 733 734 Interestingly, due to its more potent nature, The GC reaction is also believed to be subject to 735 a much higher level of control, through a continued requirement for linked recognition in 736 successive rounds of diversity generation. Furthermore, a specialized subset of Tregs, TFRs, 737 exert a dominant negative level of control on the GC reaction (Fahlquist Hagert & Degn, 2020). 738 Hence it may be that the extrafollicular pathway in essence represents an evolutionary 739 'backdoor to autoimmunity', unguarded due to its relative insignificance in terms of high-740 quality, affinity-matured, and memory-inducing antibody responses. In this context, it is 741 fortunate that current CD40L targeting strategies block both the GC and extrafollicular 742 response. However, we suggest that future efforts should be aimed at further elucidating the 743 relative contributions of the extrafollicular and GC pathways. We may speculate that specific 744 targeting of the extrafollicular pathway would be a superior strategy, as it would 745 preferentially block the low quality and poorly controlled responses driving autoimmune 746 progression, while leaving intact the more stringently controlled and high-quality responses 747 that provide protection against infectious agents. Unfortunately, there is much more limited 748 knowledge regarding the biology of the extrafollicular responses, and no transgenic or 749 pharmacologic strategy allowing specific blockade of this pathway exists, making it difficult to 750 evaluate in animal models. 751 752 In summary, our findings here demonstrate that a complete or partial block of the GC 753 pathway is insufficient to curb autoreactive PC differentiation and might in some instances in 754 fact exacerbate the autoimmune progression. The GC commitment is controlled by the 755 expression level of the master transcriptional repressor, Bcl-6 (Robinson et al., 2020), which 756 regulates the GC fates across GC B cells, TFH cells and TFR cells. Interestingly, in the context of 757 the COVID-19 pandemic, it has been observed that Bcl-6 + GC B cells and Bcl-6 + TFH cells are 758 markedly diminished in SARS-CoV-2 infection (Kaneko et al., 2020). It has also been found that 759 critically ill SARS-CoV-2 patients display hallmarks of extrafollicular B cell activation and 760 shared B cell repertoire features previously described in autoimmune settings (Knight et al., 761