key: cord-0992741-sfxzngra
authors: Scheid, Johannes F.; Barnes, Christopher O.; Eraslan, Basak; Hudak, Andrew; Keeffe, Jennifer R.; Cosimi, Lisa A.; Brown, Eric M.; Muecksch, Frauke; Weisblum, Yiska; Zhang, Shuting; Delorey, Toni; Woolley, Ann E.; Ghantous, Fadi; Park, Sung-Moo; Phillips, Devan; Tusi, Betsabeh; Huey-Tubman, Kathryn E.; Cohen, Alexander A.; Gnanapragasam, Priyanthi N.P.; Rzasa, Kara; Hatziioanno, Theodora; Durney, Michael A.; Gu, Xiebin; Tada, Takuya; Landau, Nathaniel R.; West, Anthony P.; Rozenblatt-Rosen, Orit; Seaman, Michael S.; Baden, Lindsey R.; Graham, Daniel B.; Deguine, Jacques; Bieniasz, Paul D.; Regev, Aviv; Hung, Deborah; Bjorkman, Pamela J.; Xavier, Ramnik J.
title: B cell genomics behind cross-neutralization of SARS-CoV-2 variants and SARS-CoV
date: 2021-04-24
journal: Cell
DOI: 10.1016/j.cell.2021.04.032
sha: 9d42f246c777e2f512d94518f1c51d41f1fac39f
doc_id: 992741
cord_uid: sfxzngra

Monoclonal antibodies (mAbs) are a focus in vaccine and therapeutic design to counteract SARS-CoV-2 and its variants. Here, we combined B cell sorting with single-cell VDJ and RNA-seq and mAb structures to characterize B cell responses against SARS-CoV-2. We show that the SARS-CoV-2-specific B cell repertoire consists of transcriptionally distinct B cell populations with cells producing potently neutralizing antibodies (nAbs) localized in two clusters that resemble memory and activated B cells. Cryo-electron microscopy structures of selected nAbs from these two clusters complexed with SARS-CoV-2 spike trimers show recognition of various receptor-binding domain (RBD) epitopes. One of these mAbs, BG10-19, locks the spike trimer in a closed conformation to potently neutralize SARS-CoV-2, the recently arising mutants B.1.1.7 and B.1.351, and SARS-CoV and cross-reacts with heterologous RBDs. Together, our results characterize transcriptional differences among SARS-CoV-2-specific B cells and uncover cross-neutralizing Ab targets that will inform immunogen and therapeutic design against coronaviruses.

(mAbs) were isolated using phage display library techniques (Prabakaran et al., 2006; Sui et al., 2004) and Epstein-Barr virus transformed B cells (Corti et al., 2015; Traggiai et al., 2004) . Since then, high-throughput single-cell RNA-seq (scRNA-seq) of B cells has allowed simultaneous characterization of their clonal landscape and associated transcriptional profiles (Neu et al., 2019) . When combined with functional testing and structural characterization of selected mAbs, this integrated approach should allow us to learn more about transcriptional pathways involved in the generation of efficient antiviral antibody (Ab) responses and the roles of different B cell subpopulations (Horns et al., 2020; Mathew et al., 2020; Neu et al., 2019; Waickman et al., 2020) .

Recent efforts to develop therapeutic mAbs against SARS-CoV-2 were aided by structures that have revealed how the SARS-CoV-2 spike binds to its angiotensin-converting enzyme 2 (ACE2)

receptor , specificities of polyclonal Ab responses in COVID-19 convalescent individuals (Barnes et al., 2020b) , and commonalities among receptor-binding domain (RBD)binding mAbs (Barnes et al., 2020a; Tortorici, 2020; Yuan et al., 2020) . Collectively, these structures guide choices of mAb pairs for treatment cocktails, while informing structure-based engineering experiments to improve mAb potencies and/or resistance to viral mutations.

Furthermore, recent mapping of neutralizing SARS-CoV-2 mAbs that target conserved spike epitopes (Lv et al., 2020; Piccoli et al., 2020) has the potential to guide structure-based immunogen design to elicit cross-reactive mAbs against zoonotic coronaviruses with spillover potential.

J o u r n a l P r e -p r o o f 5 Here, we use scRNA-Seq to investigate SARS-CoV-2 spike-specific B cell responses in 14 subjects who had recovered from COVID-19. We matched the VDJ sequence and transcriptional profiles with functional studies from 92 mAbs and identified two transcriptional clusters (TCs) from which the majority of neutralizing Abs (nAbs) were isolated. We structurally characterized six of the most potently nAbs derived from B cells in these two TCs, including BG10-19 that reaches between adjacent RBDs on a single spike trimer, locking it in a conformation that cannot bind ACE2 in a manner distinct from previously-described mAbs (Barnes et al., 2020a; Tortorici, 2020) . BG10-19 potently neutralized SARS-CoV-2, the UK variant B.1.1.7 (Davies et al., 2021) , the South African variant B.1.351 (Tegally et al., 2020) as well as the heterologous SARS-CoV pseudotyped viruses. Furthermore, characterization of mAbs belonging to the VH3-53/VH3-66-encoded class (Barnes et al., 2020b; Wu et al., 2020a; Yuan et al., 2020) showed common binding modes for mAbs with short (<14 amino acids) and long (>15 amino acids) heavy chain complementarity-determining region 3 (CDRH3) loops, providing new insights into this recurring class of SARS-CoV-2 nAbs.

To understand the development of B cell responses after SARS-CoV-2 infection, we enrolled 14 subjects who had recently recovered from COVID-19. Subjects were diagnosed in March 2020, none required hospitalization, and the time between diagnosis and enrollment ranged between 31 and 61 days ( Table S1 ). 12 of 14 subjects were diagnosed with COVID-19 using PCR-based testing. The remaining two subjects were diagnosed based on serum reactivity to RBD in ELISA, clinical symptoms and history of recent exposure (Table S1 ).

J o u r n a l P r e -p r o o f 6 To evaluate serum neutralizing activity, we used a pseudotyped virus with SARS-CoV-2 spike (S) protein Schmidt et al., 2020 ) (STAR Methods). We detected serum neutralization in 11 of 14 (79%) subjects ( Figure S1A , Table S1 ). ID 50 titers ranged from 51 to 655, and no correlation was seen between serum neutralization and time since diagnosis, age or gender of the subjects (Figure S1B -D, Table S1 ).

To characterize the B cell response against the SARS-CoV-2 spike, we sorted a total of 6,113 B cells from 14 subjects that bound to the SARS-CoV-2 S or RBD using fluorescence-activated cell sorting (FACS) (Figures 1A, B) . The frequency of SARS-CoV-2 Sor RBD-binding B cells ranged from 0.05-0.3% of CD19 + B cells (Figures 1A, B) . We profiled sorted cells by 5' directed scRNA-Seq for both mRNA and paired VDJ profiling, recovering matched single-cell VDJ and transcriptome profiles in 129 -649 cells for SARS-CoV-2 S and 47 -487 cells for RBD per donor ( Figure 1C) . We identified clonally related B cells in each donor, by accounting for V and J gene similarities and sequence similarity between the CDRH3 amino acid sequences (Nouri and Kleinstein, 2018) . Expanded B cell clones with at least 2 clonally related cells, accounted for 7-45% of sorted B cells in the 14 subjects ( Figure 1D ).

We tested if our B cell repertoires include cells that underwent somatic hypermutation and class switch recombination, two steps in the generation of high affinity mAbs (Victora and Nussenzweig, 2012) . We found that the fractions of IgG + , IgM/IgD + and IgA + B cells varied substantially among subjects (9-48%, 48-87% and 3-12% respectively) ( Figure S1E ). When we investigated clonal expansion within each immunoglobulin (Ig) isotype, we found that in 8 of 14

subjects IgM+ clones were statistically significantly more expanded than either IgA+ clones or IgG+ clones or both ( Figure S1F ). Inferring levels of heavy chain somatic mutations (from RNA, J o u r n a l P r e -p r o o f 7 STAR Methods), the total number of mutations was lower in SARS-CoV-2 S-and RBD-binding B cells than in historic memory B cell (MBC) controls (Rubelt et al., 2012) (Figure S1G ) with IgM + and IgD + cells showing significantly lower levels compared to IgG + and IgA + B cells ( Figure   S1H ). Lower levels of somatic mutations were especially observed in IgM+ B cells with low clonal expansion as 9/14 subjects showed a statistically significant positive correlation between clone size and number of mutations in their IgM+ repertoire ( Figure S1I ).

We next tested if our sorted B cells were enriched for Ig genes VH3-53 and VH3-66 with short CDRH3 regions, features which are preferentially found among class 1 nAbs (Barnes et al., 2020b; Wu et al., 2020a; Yuan et al., 2020) . 13 of the 14 recovered donors showed a higher fraction of combined VH3-53/VH3-66 in B cell repertoires against SARS-CoV-2 S, RBD or both compared to historic MBC controls (Rubelt et al., 2012) . This difference was statistically significant in 10 of 14 (71%, FDR < 0.1, Two-proportions z-test) subjects ( Figure 1E ) and remained significant in 10 of 14 (71%, FDR < 0.1, Two-proportions z-test) subjects when considering only mAbs with CDRH3 regions shorter than 14 amino acids. Among other Ab features CDRH3 length and hydrophobicity were slightly increased and CDRH3 charge was decreased compared to historic controls, and these differences existed for both S-and RBDbinding B cells (Figures S1J-L) and across Ig isotypes (Figures S1M-O) . Thus, the B cell repertoire that binds to SARS-CoV-2 S and SARS-CoV-2 RBD in recovered individuals is enriched for heavy chain genes that are associated with class 1 nAbs.

To functionally evaluate selected mAbs, we chose 4 of the 14 subjects with the highest serum neutralization titers ( Figure S1A and Table S1 ) and produced a total of 92 mAbs from these J o u r n a l P r e -p r o o f 8 donors ( Table S2 ). The selection of mAbs included 72 representatives from most of the expanded B cell clones in these subjects and 20 singlets (Figures S2A-D) . All of the selected mAbs were expressed as IgG1s in order to allow for direct comparison of binding and neutralization and were initially evaluated for binding by ELISA. 56 of 92 mAbs (61%) showed strong or intermediate binding to either SARS-CoV-2 S, RBD, or both in ELISA (Figure 2A and Table S3 ). 42 of 56 strong or intermediate binders (75%) bound to both SARS-CoV-2 S and RBD, 9 of 56 (16%) only bound to SARS-CoV-2 S, and 5 of 56 (9%) only to RBD (Figure 2A and Table S3 ).

Polyreactivity, non-specific binding to unrelated antigens, is a feature of Abs that is selected against throughout B cell development (Wardemann et al., 2003) , but can be generated during affinity maturation (Tiller et al., 2007) . Ab responses to HIV-1 envelope trimer, for example, display high levels of polyreactivity, which was suggested to be a means to increase Ab avidity (Mouquet et al., 2010) and binding to divergent HIV-1 envelope strains (Prigent et al., 2018) . At the same time, polyreactivity negatively affects half-life and clinical utility of mAbs Shingai et al., 2014) .

To assess polyreactivity, we tested all 92 isolated mAbs for binding to ssDNA, dsDNA, insulin, bacterial lipopolysaccharide (LPS) and streptavidin-APC by ELISA (Wardemann et al., 2003) .

We included streptavidin-APC as an antigen to assess any potential off-target binding of B cells to the staining reaqent used in FACS (STAR Methods). 11 of 92 (12%) mAbs and 6 of 56 (11%) intermediate or strong binders to SARS-CoV-2 S or RBD showed reactivity against two or more of the four polyreactivity antigens in ELISA ( Figure S2E and Table S3 ), significantly less J o u r n a l P r e -p r o o f 9 than polyreactivity frequencies detected in healthy MBCs (22.7%, (Tiller et al., 2007) ) or HIVspecific B cells (75% (Mouquet et al., 2010) ). In addition, none of our mAbs were found to be reactive in a polyreactivity ELISA assay against baculoviorus lysate (Hotzel et al., 2012) (STAR Table S3 ). We conclude that most mAbs isolated from SARS-CoV-2 S-and RBDbinding B cells are not polyreactive when expressed as IgG1.

We next screened all 92 mAbs for neutralizing activity in a pseudotyped virus neutralization assay Schmidt et al., 2020) (above and STAR Methods). 27 of 92 mAbs (29%) showed neutralizing activity when tested up to a concentration of at least 25 µg/ml, and nAbs were identified from all 4 selected subjects ( Figure S2F and Tables S2 and S3) . IC 50 values ranged from 3 ng/ml to 25µg/ml with a median of 99 ng/ml ( Figure S2F and Table S2) .

Neutralizers showed at least intermediate binding in ELISA to both SARS-CoV-2 S trimer and RBD, with the exception of BG4-14, a weak neutralizer with binding by ELISA only to RBD (Tables S2 and S3) .

Neutralizing activity was detected for 21 of 42 mAbs (50%) selected from IgG + cells compared with 4 of 35 (11%) from IgM + /D + cells and 2 of 15 (13%) from IgA + cells (χ 2 = 15.9, P value 0.0004) (Tables S2 and S3 ). When we tested the 27 neutralizers in a neutralization assay against authentic SARS-CoV-2 we found the results between both SARS-CoV-2 neutralization assays to be comparable (correlation coefficient (r) for comparison between IC 50 values 0.701, P value < 0.0001) ( Figure 2B and (Davies et al., 2021; Tegally et al., 2020) show decreased sensitivity to some SARS-CoV-2 mAbs, polyclonal sera from recovered COVID-19 donors (Wibmer et al., 2021) and sera from SARS-CoV-2 mRNA vaccinees (Liu et al., 2021; Wang et al., 2021; Wu et al., 2021) . We produced pseudotyped SARS-CoV-2 viruses carrying the reported spike mutations in found these variants (STAR 

mAbs that target conserved epitopes among different betacoronaviruses are subject of intense investigation given their potential utility in future coronavirus outbreaks. A small set of mAbs that were isolated from SARS-CoV-infected individuals have been shown to cross-react with SARS-CoV-2 and vice versa .

To evaluate potential cross-reactivity of our mAbs to other coronaviruses, we tested all 92 mAbs for binding to SARS-CoV and MERS-CoV spike (S) protein and RBD in ELISA ( Figure 2D and Table S3 ). Two SARS-CoV-2 non-neutralizing mAbs showed strong cross binding to MERS J o u r n a l P r e -p r o o f (BG4-23 and BG1-13), one of which (BG4-23) also showed strong cross binding to SARS-CoV (Table S3) . Three SARS-CoV-2 neutralizers (BG1-28, showed strong cross-binding to SARS-CoV (Table S3) . Of these, only BG10-19 neutralized SARS-CoV potently in a pseudovirus neutralization assay with an IC 50 value of 3 ng/ml compared with 20 ng/ml for S309, a cross neutralizing mAb isolated from a SARS-CoV patient ( Figure 2E ) . BG10-19 also neutralized SARS-CoV-2 more potently than S309 with an IC 50 of 9 ng/ml compared to 79 ng/ml (Table S2 and ). BG1-28 and BG10-14 did not neutralize SARS-CoV and neither did a selection of SARS-CoV-2 neutralizers that showed intermediate cross binding to SARS-CoV in ELISA ( Figure 2E and Table S3 ).

Secreted IgA and IgM can, if expressed with a J chain, multimerize and therefore increase overall avidity depending on the density and accessibility of the antigen binding sites . Consistent with this, some IgA dimers of SARS-CoV-2 nAbs show increased neutralizing potency .

To test the effect of dimerization on neutralizing activity in our mAbs we expressed 13 of the15 mAbs which were derived from IgA+ B cells (Tables S2 and S3) (Figures S2J, K) . We conclude that the neutralizing activity of some SARS-CoV-2 mAbs can be increased through expression as monomeric or dimeric IgA.

To understand the mechanism of BG10-19-mediated neutralization of SARS-CoV-2 and SARS-CoV, we determined a 3.3Å single-particle cryo-electron microscopy (cryo-EM) structure of a complex between SARS-CoV-2 S trimer S3 and Table S5 ).

The BG10-19 -S structure revealed S trimers adopting a closed conformation bound to three BG10-19 Fabs (Figures 3A and S3A-D) . BG10-19 recognizes a quaternary epitope comprising interactions that bridge two neighboring RBDs and the N165 NTD -glycan on the adjacent NTD ( Figure 3B) . Specifically, BG10-19 uses five of its six complementarity-determining region (CDR) loops to interact with a proteoglycan epitope focused atop the RBD α1 (residues 338-347) and α2 (residues 364-374) helices, with additional contacts with residues 436-450 ( Figure   3C -E). The N343 RBD -glycan (modeled as a complex-type pentasaccharide) interfaces with both CDRH3 and CDRL2 loops, including contacts with the core fucose moiety in a manner similar to the cross-reactive mAb S309 ( Figure 3E ) . The CDRH2 and CDRH3 loops mediate the majority of RBD contacts (~760Å 2 of ~1090Å 2 total paratope buried surface area (BSA)), establishing extensive polar and van der Waals interactions with RBD residues ( Figure   3F ,G). Collectively, interactions mediated by CDRH1-3 and CDRL2 loops establish the primary epitope recognized by BG10-19, which does not overlap with the ACE2 receptor-binding motif J o u r n a l P r e -p r o o f 13 (RBM) ( Figure S3E ) and accounts for 87% of epitope BSA on the S trimer (~925Å 2 of peptide BSA and ~220Å 2 glycan BSA).

However, unlike the binding mode of S309, BG10-19 adopts a pose that positions the light chain CDRL1 loop and FWR3 atop a neighboring "down" RBD, which contributes an additional ~150Å 2 BSA to the total quaternary epitope ( Figure 3A-D) . This binding orientation is distinct from previously-described nAbs that utilize interactions of long CDRH3s to mediate bridging interactions to stabilize the S trimer in a closed state (Barnes et al., 2020a; Tortorici, 2020) .

Interestingly, BG10-19 interactions with the neighboring RBD also prevent sampling of the "up" RBD conformation ( Figure S3C ). Indeed, SARS-CoV-2 S trimer binding to immobilized sACE2-CH3 (Tada et al., 2020) was blocked in the presence of BG10-19 Fab, while a RBD -BG10-19

complex was capable of binding sACE2 in SPR experiments ( Figure 3H ). These data suggest that BG10-19 utilizes a neutralization mechanism that inhibits exposure of the ACE2 RBM, by locking the S trimer into a closed conformation.

Sequence conservation at the BG10-19 epitope explains the potent cross-neutralizing activity against SARS-CoV ( Figure 2E ), in that 23 of 29 residues are strictly or conservatively substituted between SARS-CoV-2 and SARS-CoV RBDs ( Figure S3F ). Given the conserved nature of the BG10-19 epitope, we characterized the potential of BG10-19 to cross-react with zoonotic sarbecoviruses. Bat coronavirus strains WIV1-CoV and SCH014-CoV are clade 1 sarbecoviruses and ACE2-tropic (Li et al., 2003) . When we assessed BG10-19 cross-reactivity to these viruses, we observed binding to WIV1-CoV RBD at the same levels as SARS-CoV-2 and SARS-CoV, but no neutralizing activity ( Figure S3G -J). The lack of neutralizing activity against WIV1-CoV was surprising, given its 95% amino acid sequence identity with SARS-CoV (Cohen et al., 2020; Pinto et al., 2020) . Overall, our findings suggest that the SARS-CoV-2 trimer-specific Ab response includes rare mAbs such as BG10-19 with highly potent crossneutralizing activity against SARS-CoV.

As mentioned above, our scRNA-seq approach pairs VDJ sequences and expression profiles of B cell populations sorted from 14 convalescent subjects, and can therefore provide insights into the expression states of SARS-CoV2 specific B cells. These B cells were selected based on expression of CD20, CD19 and SARS-CoV-2 S or RBD binding alone without selection for other surface markers (STAR Methods and above). We profiled 6,113 sorted B cells, revealing 6 distinct transcriptional cell clusters (TCs) (Figures 4A, B ) that had representation from all donors, though with enrichment of some donors in specific clusters ( Figure S4A and see below). These clusters were neither dependent on binding to SARS-CoV-2 S or RBD ( Figure   S4B ) nor on cell cycle phase ( Figure S4C ). CD44 and CXCR4 were expressed in cells from all TCs, consistent with their frequent expression among mature B cells ( Figure S4D ) (Kremmidiotis and Zola, 1995; Nagasawa et al., 1996; Nie et al., 2004) . CD38 expression levels were low throughout all clusters, in keeping with our selection of B cells expressing surface Ig binding to SARS-CoV-2 S or RBD (Ellebedy et al., 2016) ( Figure S4D ). The cell clusters were distinguished by increased expression of specific marker genes ( Figure 4A ), such as genes associated with different B cell populations, including naïve and MBCs, consistent with isotype, mutation status and clonal expansion features of these cells. Specifically, the TC1 marker genes FCER2 (CD23) and TCL1A are known to be highly expressed in mature naïve B cells ( Figure   4A , Table S4 ) (Horns et al., 2020) , consistent with an enrichment in TC1 of IgD + and IgM + B cells (36% and 63% of cells, respectively, Figure S4E ) with low levels of inferred somatic mutations ( Figure S4F ) and minimal clonal expansion ( Figure S4G, H) . In TC3 and TC4, CD27

and CD80, both expressed in MBCs (Moroney et al., 2020; Zuccarino-Catania et al., 2014) were highly expressed (Figure S5A) , and the majority of cells in these clusters were IgG1 + (53% and 73% of cells, respectively, Figure S4E ), somatically mutated ( Figure S4F ) and more clonally expanded than cells in TC1 ( Figure S4G, H) . Cells in TC3 and TC4 also highly expressed CXCR3, a chemokine receptor found on some class switched B cells that is believed to facilitate migration to sites of inflammation (Moroney et al., 2020; Muehlinghaus et al., 2005) and CD70, which is upregulated on stimulated B cells where its interaction with CD27 on effector T cells plays an important role in antiviral T cell responses (Izawa et al., 2017; van Gisbergen et al., 2011 ) ( Figure 4A and Figure S5A ).

In the weeks following vaccination or infection with influenza or infection with Ebola, antigen specific B cells can present as Ab secreting cells, MBCs or "activated B cells" (ABCs), which wane after several weeks and show relatively higher expression levels of CD52, TLR10, CD19

and CD20 (Ellebedy et al., 2016) . To test if either TC3 or TC4 include ABCs, we examined the distribution of expression levels in each cell cluster of CD52, TLR10, CD19 and CD20 and 10 other genes that are expressed at higher levels in ABCs compared to MBCs ( Figure S5A ) (Ellebedy et al., 2016) . All 14 genes were significantly higher expressed (one-tailed t-test, FDR < 0.01) in TC4 than in TC3 ( Figure S5A ). Thus, SARS-CoV-2 S-and RBD-binding B cell repertoires in recently recovered subjects include ABCs (TC4) and MBCs (TC3) that are both mostly IgG1 + , show expansion of B cell clones and somatic hypermutation. We also detected IgD + and IgM + mature naïve B cells (TC1) with low levels of somatic hypermutation, minimal clonal expansion and high expression of CD23 and TCLA1.

To relate mAbs with high levels of binding to or neutralization of SARS-CoV-2 to B cells in certain TCs, we identified the clusters from which the 92 mAbs we produced and tested were J o u r n a l P r e -p r o o f 16 derived. Because we selected mAbs for testing based on representation of expanded clones (72/92) or randomly selected singlets that were mostly IgG+ or IgA+ (20/92), all TCs are not equally represented among the 92 mAbs (Table S3) . Nevertheless, 20 of 33 (60%) neutralizers, 14 of 15 (94%) potent neutralizers (monomeric IC 50 ≤ 0.1 µg/ml) and 5 of 10 (50%) nonneutralizing high binders were derived from cells in TCs 3 and 4, whereas 36 of 49 (73%) of the non-neutralizing low binders belonged to TCs 0 and 2 ( Figures 4D and S2J , Tables S2 and S3). The majority of neutralizers from TCs 3 and 4 were derived from singlets or small clones and did not have clonal members in other TCs among the cells we sampled ( Figure S5B , Table   S3 and Data S1). We further tested our clusters for enrichment in class 1 nAbs (VH3-53/VH3-66 antibodies with short CDRH3 regions (Barnes et al., 2020a) and see above). VH3-53/VH3-66

Abs were overall significantly enriched in TC3 and TC4 compared to TCs 1, 2 and 5 ( Figures   4E and S5C , two-tailed t-test, P ≤ 0.0001), and 8/8 evaluated VH3-53/VH3-66 nAbs with short CDRH3 regions were derived from B cells that belonged to either TC3 or TC4 (Tables S2 and S3). We conclude that the cells from which we isolated strongly binding and neutralizing mAbs against SARS-CoV-2 frequently had transcriptional profiles consistent with MBCs and ABCs.

While MBCs and antibody secreting plasma cells are two separate B cell compartments (Leyendeckers et al., 1999) , they can originate from the same germinal center reaction, MBCs can become plasma cells upon stimulation (Victora and Nussenzweig, 2012) and antibodies against HIV-1 envelope derived from MBCs have been detected in serum from matched patients (Scheid et al., 2009 patients.

Given the enrichment of VH3-53/VH3-66-encoded Abs in TC3 and TC4, we selected two mAbs which were coded by cells in these clusters with distinct CDRH3 lengths for structural characterization. We solved a 3.0Å crystal structure of Fabs from BG4-25 (VH3-53-encoded with 12 aa CDRH3) and the SARS-CoV mAb CR3022 (Tian et al., 2020) in complex with SARS-CoV-2 RBD ( Figure S6A and Table S6 ). Consistent with the binding mode of class 1 nAbs (Barnes et al., 2020b; Yuan et al., 2020) , BG4-25 recognizes an RBD epitope that overlaps with >90% of residues in the ACE2 RBM ( Figure S6B ), which is only fully accessible with "up" RBD conformations. Two V-gene encoded regions in class 1 nAbs prominently contribute to epitope recognition; the 31 SNY 33 CDRH1 and 53 SGGS 56 CDRH2 sequence motifs, which take part in extensive hydrogen bond interactions at the RBD interface (Tan et al., 2021; Wu et al., 2020b; Yuan et al., 2020) . Analysis of inferred somatic mutations in 93 VH3-53/VH3-66 mAbs coded in our collection of SARS-CoV-2 binding B cells with CDRH3 less than 14 amino acids (IMGT definition; (Lefranc et al., 2015) ) revealed frequent inferred mutations in these motifs, including Although unliganded Fab structures often exhibit a disordered CDRH3 (e.g., 1RZI, 1RZF), it is unusual for an Ab bound to an antigen to exhibit a disordered CDRH3. From an examination of 731 Ab-antigen structures with resolutions of 3.5Å or better in the Structural Antibody Database (SABDab; (Dunbar et al., 2014) ), we found only 6 with missing residue numbers between heavy chain residues 95 to 107, implying a disordered CDRH3 (PDBs 3LH2, 4JDT, 7JWB, 5ANM, 4M8Q, 3LHP). None of these complexes involved conventional Ab-antigen pairs; instead, they were germline forms of Abs, the epitope was presented in a scaffold, or only the VH domain J o u r n a l P r e -p r o o f 19 was involved in binding. This suggests that the orientation adopted by BG1-22 is not one that promotes CDRH3-mediated interactions with the antigen, as is classically observed in Abantigen structures, but instead simply accommodates the longer CDRH3 length by displacing much of the loop to outside the Ab-antigen interface. Taken together, these results provide further insight into class 1 nAbs and suggests that longer CDRH3s, while infrequent, are not a restriction to V-gene mediated interactions at the RBD interface of this Ab class.

To further understand the specificity of RBD-targeting, we determined cryo-EM structures of Cryo-EM structures of BG7-20 -S and BG1-24 -S complexes revealed RBD-targeting similar to nAbs that belong to the class 2 binding mode (Barnes et al., 2020a) . This class of SARS-CoV-2 nAbs recognizes up and down RBD conformations, overlaps with the ACE2 RBM, has secondary interactions with neighboring "up" RBDs, and has the potential for intra-protomer J o u r n a l P r e -p r o o f 20 avidity effects. Consistent with class 2 nAbs, BG7-20 and BG1-24 show a similar epitope focused along the RBD ridge that overlaps with residues involved in ACE2 binding and includes contacts with E484, F486, and Q493. The binding pose of BG1-24 promotes stabilization of the N165 NTD glycan, adding to the observation that class 2 nAbs can involve interprotomer glycan contacts . Interestingly, the N-glycan interaction is mediated by a hydrophobic Met-Phe sequence at the tip of CDRH2, a common feature of VH1-69 Abs (Chen et al., 2019) .

This feature has been attributed to facilitating broad neutralization by Abs against influenza and Hepatitis C (Guthmiller et al., 2020) (Chen et al., 2019) and likely explains BG1-24's polyreactivity ( Figure S2E and Table S3 ).

With . However, after several passages, plaque purified viruses harboring the G339R and L441P mutations showed partial escape from BG10-19 ( Figure 7J ).

We note that neither of these two mutations have so far been sequenced and deposited on the GISAID database (Elbe and Buckland-Merrett, 2017) . Thus, we conclude that BG10-19's SARS-CoV-2/SARS-CoV cross neutralization and ability to tolerate single RBD mutations within its epitope makes it an attractive therapeutic candidate.

Abs play an indispensable role in antiviral responses both through their ability to neutralize (Corti and Lanzavecchia, 2013) and by engaging other components of the immune system J o u r n a l P r e -p r o o f 22 through interactions with their Fc regions (Bournazos et al., 2020) . Different viruses perturb Ab responses through characteristic mechanisms. HIV-1, for example, constrains efficient Ab responses through narrow structural pathways to broad neutralization (Scheid et al., 2011) and by causing B cell exhaustion (Moir et al., 2008) .

In this study, we show from a comprehensive in-depth analysis of SARS-CoV-2 binding B cells from convalescent individuals that the repertoire of SARS-CoV-2 binding B cells includes MBCs and ABCs, as well as mature naïve B cells. Interestingly, we found clonal cells to be present in TC0, which is enriched in IgM + B cells with inferred somatic mutations and low expression of CD27. Despite showing clonal expansion and somatic mutations, mAbs produced from this cluster were mostly low binding and non-neutralizing. We speculate that this cluster might contain cells from an early extrafollicular B cell response as also observed in influenza infection (Lam and Baumgarth, 2019) . On the other hand, high binding and potent neutralizing activity were mostly detected in mAbs isolated from ABCs and MBCs that shared transcriptional phenotypes across different individuals.

Immunologic correlates for protection from SARS-CoV-2 after vaccination or prior exposure are not yet defined, but studies of other respiratory viruses suggest that serum neutralization could play an important role in protective immunity against SARS-CoV-2 (Kulkarni et al., 2018; Verschoor et al., 2015) . Consistent with other SARS-CoV-2 studies, we did not detect any intradonor correlation between serum neutralization and the potency of mAbs ), but we found a strong correlation between serum neutralization and the relative size of Additionally, high-resolution structures of VH3-53/VH3-66-class mAbs provided further understanding of the rules that govern potent neutralization and viral escape from this recurring antibody class and showed that CDRH3 length may not be a limitation to VH gene segmentmediated interactions at the RBD interface. Collectively, these structures and insights into the cellular processes behind the induction of potent, cross-reactive nAbs will not only aid us in our battle to control the current COVID-19 pandemic through the use of safe and effective mAb treatments but will also provide additional criteria for the evaluation of humoral immune responses elicited from candidate vaccines against emerging zoonotic viruses with pandemic potential. 

A provisional patent for the novel SARS-CoV-2 mAbs described in this study has been filed. The values of 2 and above are considered strongly reactive and are indicated in red (Tiller et al., 2007) . Results for highly-, moderately-and non-polyreactive control mAbs ED38, JB40 and mGO53 are included as indicated (Wardemann et al., 2003) .

values from SARS-CoV-2 pseudovirus neutralization assay in µg/ml (y-axis) for all mAbs isolated from subjects BG1, BG4, BG7 and BG10 and produced as IgG1 ( (Table S2) Shannon entropy values calculated based on the general form of the diversity index proposed by (Hill, 1973) , improved by resampling strategies in (Chao et al., 2014; Chao et al., 2015) and

implemented in the Alakazam R package (Gupta et al., 2015) . 

Further information and requests may be directed to, and will be fulfilled by the lead author, Ramnik J. Xavier (xavier@molbio.mgh.harvard.edu)

All reagents generated in this study are available upon request from the Lead Contact.

Antibody 

All work with human samples was performed in accordance with approved Institutional Review Board protocols (IRB) which were reviewed by the IRB at Brigham and Women's Hospital, Boston. Subjects who had recovered from COVID-19 (Table S1) the mAbs were diluted to a concentration of 1µg/ml in dilution buffer and duplicate 12 two-fold serial dilution curves were generated. One known positive and two known negative samples were included on each plate as controls.

A standard curve based on absorbances from the mAb CR3022 (Tian et al., 2020 ) dilution series included with each plate was used to estimate Ab abundance in test samples and allow for comparison of results across batches. Estimated Ab abundance in test samples was compared to the background signal from a cohort of pre-pandemic serum samples that served as negative controls. Serum samples with Ab abundance greater than 3 standard deviations (SD) above the mean of the pre-pandemic serum samples were considered to be positive and samples with Ab abundance less than 3 SD above the mean of the pre-pandemic serum samples were considered negative. (Davies et al., 2021) . Pseudotyped particles were generated and neutralization assays were performed as previously described (Crawford et al., 2020a ) (Crawford et al., 2020b) . Briefly, the genes encoding the respective spike proteins were cotransfected with Env-deficient HIV backbone to create pseudotyped lentiviral particles. For neutralization assays, 4-or 5-fold serially diluted purified IgG was incubated with SARS-CoV pseudotyped virus for 1 hour at 37ºC. The virus/Ab mixture was added to 293TACE2 target cells and incubated for 48 hours at 37ºC, then cells were lysed and luciferase activity was measured using Britelite Plus (Perkin Elmer). Relative luminescence units (RLUs) were normalized to values derived from cells infected with pseudotyped virus in the absence of Ab. Data were fit to a 5-parameter nonlinear regression in AntibodyDatabase .

60-80 ml of blood from each donor were processed using Ficoll Paque Plus (GE Healthcare) in order to isolate peripheral blood mononuclear cells (PBMCs) according to the manufacturer's instructions. After PBMC isolation we immediately proceeded with isolation of CD20 + B cells with either 1 µg/ml biotinylated SARS-CoV-2 spike trimer or 1 µg/ml biotinylated SARS-CoV-2 RBD (See below for protein expression; biotinylation was performed using avitag technology (Avidity) following the manufacturer's instructions). Cells were simultaneously stained with FITC mouse anti-human CD19 Ab (BD, 340864) and incubated for 20 minutes at 4°C before they were washed and resuspended in PBS with 5% fetal bovine serum (FBS). Cells were then stained with streptavidin-coupled APC (Biolegend #405207) for 5 minutes at 4°C and washed and resuspended in PBS with 5% FBS. Antigen binding B cells were then sorted using a Sony MA900 cell sorter by gating on live cells in the forward scatter and side scatter (Figure 1A, B) and on CD19-FITC and SARS-CoV-2 S-APC or RBD-APC double-positive cells. After sorting, cells were washed and counted using a hemocytometer and microscopy, before resuspending up to 10,000 cells in a volume of 32µl for 5' single cell RNA-Seq (see below).

Cells were separated into droplet emulsions using the Chromium Next GEM Single-cell 5′

Solution ( ( Figure S2I ).

mAb concentrations were determined measuring absorbance at 280nm using Nanodrop 2000c (Thermo Scientific) or IgG specific ELISA as previously described (Tiller et al., 2008) . mAb (Tang et al., 2014) and S227.14 (Absolute Antibody #Ab00263-10.0) for SARS-CoV S and SARS-CoV RBD (Rockx et al., 2008) . Antigen specific ELISA results are expressed as AUC using Graphpad PRISM software. Polyreactivity ELISAs were performed as previously described (Tiller et al., 2008) with the following modification: In addition to the antigens ssDNA, dsDNA, LPS and insulin (Tiller et al., 2008) , streptavidin-coupled APC (Biolegend #405207) was used as a fifth antigen in order to assess potential off target binding against this reagent used J o u r n a l P r e -p r o o f 50 for cell sorting (see above). As described previously (Tiller et al., 2008) , polyreactivity is defined as reactivity to 2 or more antigens among the antigens single stranded DNA, double stranded DNA, lipopolysaccharide (LPS) or insulin. As previously described, mAbs ED38, JB40 and mGO53 (Wardemann et al., 2003) were used as strongly polyreactive, intermediately polyreactive and non-polyreactive control mAbs respectively. In ELISA assays including IgG, IgA monomers and IgA dimers, HRP-conjugated goat anti-human kappa and lambda chain Abs (Biorad #STAR127P and #STAR129P) were used as secondary Abs at a 1/5000 dilution.

Off-target mAb binding to baculovirus (BV) particles generated in Sf9 insect cells was tested as previously described (Hotzel et al., 2012) . A solution of 1% baculovirus in 100mM sodium bicarbonate buffer pH 9.6 was adsorbed to a 384-well ELISA plate (Nunc Maxisorp) using a Tecan Freedom Evo2 liquid handling robot and the plate was incubated overnight at 4ºC.

Following blocking with 0.5% BSA in PBS, 1µg/ml of IgG was added to the blocked assay.

Plates were incubated for 3 hours at room temperature. mAb binding was detected using an HRP-conjugated anti-Human IgG (H&L) secondary Ab (Genscript). ELISA was developed using SuperSignal ELISA Femto Maximum Sensitivity Substrate (Thermo Scientific). Anti-HIV mAbs NIH45-46 (Scheid et al., 2011) and 45-46m2 (Diskin et al., 2013) were used as positive controls and mAb 3BNC117 (Scheid et al., 2011) was used as negative control. Measurements were performed in quadruplicate and OD values within 1.5-fold the negative control were considered to be negative.

Vero E6-TMPRSS2 were seeded at 10,000 cells per well the day prior to infection in CellCarrier-384 ultra microplate (Perkin Elmer). mAb samples were tested in 4-fold 9-point dilution spots starting at a highest concentration of 100 µg/mL. Serial diluted mAbs were mixed separately J o u r n a l P r e -p r o o f 51 with diluted SARS-CoV-2 virus and incubated at 37°C with 5% CO 2 for 1 hour. mAb-virus complexes were added to the cells in triplicate. Plates were incubated at 37°C with 5% CO 2 for 48 hours. After that, plates were fixed and inactivated using 4% paraformaldehyde in PBS for 2 hours at room temperature. Plates were then washed and incubated with diluted anti-SARS-CoV/SARS-CoV-2 nucleoprotein mouse Ab (Sino) for 1.5 hours at room temperature. Plates were subsequently incubated with Alexa488-conjugated goat anti-mouse (JacksonImmuno) for 45 mins at room temperature, followed by nuclear staining with Hoechst 33342 (ThermoFisher).

The fluorescence images were recorded and analyzed using Opera Phenix™ High Content

Screening System. The half-maximal inhibitory concentrations (IC 50 ) were determined using four parameters logistic regression (GraphPad Prism 8.0).

Expression and purification of SARS-CoV-2 6P stabilized S trimers and

constructs encoding the sarbecovirus RBDs were conducted as previously described (Cohen et al., 2020) . Briefly, constructs were purified from supernatants of transiently transfected Expi293F cells (Gibco) by Ni 2+ -NTA affinity and size exclusion chromatography (SEC). Peak fractions were identified by SDS-PAGE, pooled, and stored at 4˚C. IgGs were expressed, purified, and stored as described (Barnes et al., 2020b) . Fabs were generated by papain digestion using crystallized papain (Sigma-Aldrich) in 50 mM sodium phosphate, 2 mM EDTA, 

Purified Fab and S 6P trimer were incubated at a 1.1:1 molar ratio per protomer on ice for 30 minutes prior to deposition on a freshly glow-discharged 300 mesh, 1.2/1.3 UltrAuFoil grid.

Immediately before 3 µl of complex was applied to the grid, fluorinated octyl-malotiside was added to the Fab-S complex to a final detergent concentration of 0.02% w/v, resulting in a final complex concentration of 3 mg/ml. Samples were vitrified in 100% liquid ethane using a Mark IV Vitrobot after blotting for 3 s with Whatman No. 1 filter paper at 22˚C and 100% humidity.

Data collection and processing followed a similar workflow to what has been previously described in detail (Barnes et al., 2020a) . Briefly, micrographs were collected on a Talos Arctica transmission electron microscope (Thermo Fisher) operating at 200 kV for all Fab-S complexes.

Data were collected using SerialEM automated data collection software (Mastronarde, 2005) and movies were recorded with K3 camera (Gatan). Data collections parameters are summarized in Table S5 . For all data sets, cryo-EM movies were patch motion corrected for beam-induced motion including dose-weighting within cryoSPARC v2.15 (Punjani et al., 2017) after binning super resolution movies. The non-dose-weighted images were used to estimate CTF parameters using cryoSPARC implementation of the Patch CTF job. Processing for all datasets was carried out in a similar fashion. Briefly, an initial set of particles was picked based on templates from 2D classification of blob picked particles on a small sub-set of images. This set was pared down through several rounds of 3D classification. An ab initio job on a small good subset of these particles revealed distinct states and junk particles. Full set of particles was heterogeneously refined against distinct states, as well as a junk class acting as a trap for bad particles. Particles from each were separately refined using non-uniform refinement in C1 symmetry. Particles from distinct states were re-extracted without binning and were further refined separately in several rounds of 3D classification. Particles were subdivided into groups J o u r n a l P r e -p r o o f 53 based on beam-tilt, refined separately for CTF parameters and aberration correction. For all states, a soft mask (3-pixel extension, 6-pixel soft edge) was generated to exclude Fab constant domains for local non-uniform refinements. To improve the density for the BG1-22 -RBD, BG7-20 -RBD, and BG1-24 -RBD interfaces, a soft mask was generated around the RBD and Fab variable domain and particles were subjected to non-uniform local refinement in cryoSPARC.

The resolution at the Fab-RBD interface was modestly improved and used to model coordinates in the overall Fab -S trimer complex structures. Overall reported resolutions are based on gold standard FSC calculations.

Coordinates for initial complexes were generated by docking individual chains from reference structures into cryo-EM density using UCSF Chimera (Goddard et al., 2018 ) (see Table S5 for PDB coordinates). Models were then refined into cryo-EM maps rigid body and real space refinement with morphing in Phenix (Terwilliger et al., 2018) . Sequence-updated models were built manually in Coot (Emsley et al., 2010) and then refined using iterative rounds of real-space refinement in Phenix and Coot. Glycans were modeled at potential N-linked glycosylation sites (PNGSs) in Coot using 'blurred' maps processed with a variety of B-factors generated in cryoSPARC v2.15. Validation of model coordinates was performed using MolProbity (Table S5) .

Crystallization trials for a stoichiometric complex of BG4-25 -SARS-CoV2 RBD -CR3022 were carried out at room temperature using the sitting drop vapor diffusion method by mixing equal volumes of the Fab-RBD complex and reservoir using a TTP LabTech Mosquito robot and commercially-available screens (Hampton Research). Crystals were obtained in 0.05 M citric acid, 0.05 M BIS-TRIS propane pH 5.0 and 16% polyethylene glycol 3350 and quickly cryo-protected in a solution matching the reservoir + 20% glycerol. X-ray diffraction data were collected for Fab-RBD complex at the Stanford Synchrotron Radiation Lightsource (SSRL) beamline 12-2 on a Pilatus 6M pixel detector (Dectris). Data from a single crystal were indexed and integrated in XDS (Kabsch, 2010) and merged using AIMLESS in CCP4 (Winn et al., 2011) ( 

CDR lengths and Kabot numbering were calculated based on IMGT definitions (Lefranc et al., 2015) . Structure figures were made with UCSF ChimeraX. Local resolution maps were calculated using cryoSPARC v 2.15. Buried surface areas were calculated using PDBePISA (Krissinel and Henrick, 2007 ) and a 1.4 Å probe. Potential hydrogen bonds were assigned as interactions that were < 4.0Å and with A-D-H angle > 90˚. Potential van der Waals interactions between atoms were assigned as interactions that were < 4.0Å. Hydrogen bond and van der Waals interaction assignments are tentative due to resolution limitations.

SPR experiments were performed using a Biacore T200 instrument (GE Healthcare). ACE2 microbody (Tada et al., 2020) was immobilized on a CM5 chip by primary amine chemistry at pH 4.5 to a final response level of ~1000 resonance units (RUs). Fabs were complexed with 100 nM SARS2 S 6P or SARS2 RBD at a 10:1 molar ratio and incubated for a minimum of 1 hour.

Antigen or Fab-antigen complex was injected over immobilized ACE2 microbody surface at a J o u r n a l P r e -p r o o f 

scRNA-seq analysis mRNA and VDJ sequence reads were mapped to the reference human genome GRCh38-3.0.0 with the cloud-based Cumulus workflows , using the CellRanger 3.0.2 software pipeline. Cells with both high-quality VDJ sequence and transcriptome information were kept for the downstream analysis by filtering out the cells which had less than 300 detected genes or which had poor quality VDJ contig information defined by i) being non-productive by 10x standards, ii) having more than four productive VDJ contigs, iii) having less than three filtered

UMIs. For the transcriptome analysis, batch effects were removed with the ComBat algorithm (Johnson et al., 2007) implemented in SVA R Package version 3.38.0.. For the transcriptome mRNA count normalization, dimensionality reduction, clustering, cell cycle scoring, cluster marker genes detection and differential gene expression analysis steps Seurat R package (Butler et al., 2018) (Stuart et al., 2019) was employed. For the normalization step gene expression counts for each cell were divided by the total counts for that cell and multiplied by 1e6, which was then log-transformed using log1p. Dimensionality reduction was done by PCA with selecting 50 first principal components.

For clustering of the cells into transcriptome clusters, first the k-nearest neighbor (kNN) graph of the cells was constructed. Second, this kNN graph was used to generate the shared nearest neighbor (sNN) graph by calculating Jaccard index between every cell and its k nearest neighbors. Third, the leiden algorithm (Traag et al., 2019) was used to find the clusters of the cells based on the generated sNN graph.

Cell cycle scoring was done by calculating the module scores of the cell cycle genes defined in (Tirosh et al., 2016) . Positive cluster marker genes and differentially expressed genes were detected with a log-fold change threshold of 0.25, where only the genes that were detected in a minimum fraction of 20% in either of the six transcriptome populations were considered.

Expression levels of Ig genes were discarded during the clustering and differential gene expression analysis steps. Source codes for scRNA-seq analyses were deposited in Github (https://github.com/EraslanBas/Sars_Cov2_Antibodies).

10x V(D)J contig assembly algorithm takes many forms of noise specific to scRNA-seq data into account while generating the assembled V(D)J sequences (https://support.10xgenomics.com/single-cell-vdj/software/pipelines/latest/algorithms/assembly).

Nevertheless, only the cells with high-quality V(D)J contig sequences were selected and V(D)J gene annotations were assigned by using IGBLAST (version 1.14.0) software with the Change-J o u r n a l P r e -p r o o f 59 O R package (Gupta et al., 2015) . Donor specific B cell clones were identified by the Change-O R package (Gupta et al., 2015) , where the appropriate threshold for trimming the hierarchical clustering into B cell clones was found by inspecting the bimodal distribution of the distance between each sequence in the data and its nearest-neighbor.

Mutation inference based on the scRNA-seq VDJ sequences of the donor and the control cells (Rubelt et al., 2012) was performed by the Shazam R Package (Yaari et al., 2012) where the region definition parameter was set to be "IMGT_V_BY_SEGMENTS" which provides no subdivisons and treats the entire V segment as a single region.

CDRH3 length was defined based on IMGT definition (Lefranc et al., 2015) with the addition of two conserved amino acid residues that were added to assist in clonal analysis (Nouri and Kleinstein, 2018) . This addition was corrected for all analyses involving specific CDRH3 length, such as selection of VH3-53/3-66 mAbs with CDRH3 shorter than 14 amino acid length.

CDRH3 amino acid charges were calculated by the Alakazam R package (Gupta et al., 2015) using the method of Moore (Moore, 1985) excluding the N-terminus and C-terminus charges, and normalizing by the number of informative positions. Hydrophobicity scores were calculated with the Alakazam R package using the method of Kyte and Doolittle (Kyte and Doolittle, 1982) .

Shannon entropy values were calculated using Alakazam R package. For each donor the transcriptome cluster specific Hill diversity index, proposed in (Hill, 1973) , improved by (Chao et al., 2014; Chao et al., 2015) was calculated by setting the diversity order equal to 1 with Alakazam R package. For each run the number of bootstrap realizations is set to be 400, and the minimum number of observations to sample is set to be 10. Source codes for Ab repertoire analyses were deposited in Github (https://github.com/EraslanBas/Sars_Cov2_Antibodies). 10 -1 10 0 10 1 10 2 10 3 10 4 10 -1 10 0 10 1 10 2 10 3 10 4 10 -1 10 0 10 1 10 2 10 3 10 4 0.00 VSV-S neutralization for selection resistance mutations against BG10-19 2E1-WT 2E1 L441P 2E1 G339R (2) 2E1 G339R (10) 10 3 10 2 10 1 10 0 10 -1 10 -2 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2

SARS-CoV-2 mutant RBD and sarbecovirus RBD ELISA binding assay Binding of Fabs to SARS-CoV-2 RBDs containing single mutations or to sarbecovirus RBDs was evaluated by ELISA. RBD antigens (mutant or wild-type) were adsorbed to 384-well Nunc MaxiSorp plates (Sigma) at a concentration of 2 µg/mL overnight

05% Tween20) for 1 h at room temperature, then 5-fold serial dilutions starting at 10 µg/mL of Fab were added. Plates were washed with TBS-T and bound Fab was detected using an HRP-conjugated secondary Ab (Genscript) and SuperSignal ELISA Femto Substrate (Thermo Scientific). AUC for each Fab-antigen pair was calculated using

Fold decrease in AUC was calculated relative to SARS-CoV2 RBD AUC for the same Fab. Data shown are representative of two independent experiments

SARS-CoV-2 mutant pseudotyped reporter virus and mutant pseudotyped virus neutralization assay

Briefly, 293T cells were transfected with pNL4-3∆Env-nanoluc and pSARS-CoV-2-S ∆19 . For generation of RBD-mutant pseudoviruses, pSARS-CoV-2-S ∆19 carrying indicated spike mutations was used instead

Particles were harvested 48 h post-transfection

PHENIX: a comprehensive Python-based system for macromolecular structure solution

SARS-CoV-2 neutralizing antibody structures inform therapeutic strategies

Structures of Human Antibodies Bound to SARS-CoV-2 Spike Reveal Common Epitopes and Recurrent Features of Antibodies

The role of IgG Fc receptors in antibodydependent enhancement

Integrating single-cell transcriptomic data across different conditions, technologies, and species

Potent neutralizing antibodies against SARS-CoV-2 identified by high-throughput single-cell sequencing of convalescent patients' B cells

Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies

Unveiling the species-rank abundance distribution by generalizing the Good-Turing sample coverage theory

VH1-69 antiviral broadly neutralizing antibodies: genetics, structures, and relevance to rational vaccine design

MolProbity: all-atom structure validation for macromolecular crystallography

Mosaic RBD nanoparticles elicit neutralizing antibodies against SARS-CoV-2 and zoonotic coronaviruses. bioRxiv

Broadly neutralizing antiviral antibodies

Prophylactic and postexposure efficacy of a potent human monoclonal antibody against MERS coronavirus

Dynamics of neutralizing antibody titers in the months after SARS-CoV-2 infection

Protocol and Reagents for Pseudotyping Lentiviral Particles with SARS-CoV-2 Spike Protein for Neutralization Assays

Estimated transmissibility and severity of novel SARS-CoV-2 Variant of Concern 202012/01 in England. medRxiv

SARS and MERS: recent insights into emerging coronaviruses

Restricting HIV-1 pathways for escape using rationally designed anti-HIV-1 antibodies

SAbDab: the structural antibody database

Data, disease and diplomacy: GISAID's innovative contribution to global health

Defining antigen-specific plasmablast and memory B cell subsets in human blood after viral infection or vaccination

Features and development of Coot

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

Complete Mapping of Mutations to the SARS-CoV-2 Spike Receptor-Binding Domain that Escape Antibody Recognition

Natively glycosylated HIV-1 Env structure reveals new mode for antibody recognition of the CD4-binding site

Change-O: a toolkit for analyzing large-scale B cell immunoglobulin repertoire sequencing data

Polyreactive Broadly Neutralizing B cells Are Selected to Provide Defense against Pandemic Threat Influenza Viruses

Studies in humanized mice and convalescent humans yield a SARS-CoV-2 antibody cocktail

Diversity and Evenness: A Unifying Notation and Its Consequences

Memory B Cell Activation, Broad Antiinfluenza Antibodies, and Bystander Activation Revealed by Single-Cell Transcriptomics

HIV-1 suppression and durable control by combining single broadly neutralizing antibodies and antiretroviral drugs in humanized mice

A strategy for risk mitigation of antibodies with fast clearance

Structure-based design of prefusion-stabilized SARS-CoV-2 spikes

Structural basis for potent neutralization of SARS-CoV-2 and role of antibody affinity maturation

Inherited CD70 deficiency in humans reveals a critical role for the CD70-CD27 pathway in immunity to Epstein-Barr virus infection

Adjusting batch effects in microarray expression data using empirical Bayes methods

MAFFT online service: multiple sequence alignment, interactive sequence choice and visualization

Changes in CD44 expression during B cell differentiation in the human tonsil

Inference of macromolecular assemblies from crystalline state

Establishing Correlates of Protection for Vaccine Development: Considerations for the Respiratory Syncytial Virus Vaccine Field

A simple method for displaying the hydropathic character of a protein

The Multifaceted B Cell Response to Influenza Virus

IMGT(R), the international ImMunoGeneTics information system(R) 25 years on

Correlation analysis between frequencies of circulating antigen-specific IgG-bearing memory B cells and serum titers of antigen-specific IgG

Cumulus provides cloud-based data analysis for large-scale single-cell and single-nucleus RNA-seq

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus

Neutralizing Activity of BNT162b2-Elicited Serum -Preliminary Report

Structural basis for neutralization of SARS-CoV-2 and SARS-CoV by a potent therapeutic antibody

Automated electron microscope tomography using robust prediction of specimen movements

Single cell BCR and transcriptome analysis after respiratory virus infection reveals spatiotemporal dynamics of antigen-specific B cell responses. bioRxiv

Phaser crystallographic software

Evidence for HIV-associated B cell exhaustion in a dysfunctional memory B cell compartment in HIV-infected viremic individuals

Amino acid and peptide net charges: A simple calculational procedure

Integrative transcriptome and chromatin landscape analysis reveals distinct epigenetic regulations in human memory B cells

Polyreactivity increases the apparent affinity of anti-HIV antibodies by heteroligation

Development of potency, breadth and resilience to viral escape mutations in SARS-CoV-2 neutralizing antibodies. bioRxiv

Regulation of CXCR3 and CXCR4 expression during terminal differentiation of memory B cells into plasma cells

Molecular cloning and characterization of a murine pre-B-cell growth-stimulating factor/stromal cell-derived factor 1 receptor, a murine homolog of the human immunodeficiency virus 1 entry coreceptor fusin

Spec-seq unveils transcriptional subpopulations of antibody-secreting cells following influenza vaccination

The role of CXCR4 in maintaining peripheral B cell compartments and humoral immunity

A spectral clustering-based method for identifying clones from high-throughput B cell repertoire sequencing data

Mapping neutralizing and immunodominant sites on the SARS-CoV-2 spike receptor-binding domain by structure-guided high-resolution serology

Cross-neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV antibody

Structure of severe acute respiratory syndrome coronavirus receptor-binding domain complexed with neutralizing antibody

Conformational Plasticity in Broadly Neutralizing HIV-1

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

Convergent antibody responses to SARS-CoV-2 in convalescent individuals

Structural basis for potent cross-neutralizing human monoclonal antibody protection against lethal human and zoonotic severe acute respiratory syndrome coronavirus challenge

SARS-CoV-2-specific ELISA development

Onset of immune senescence defined by unbiased pyrosequencing of human immunoglobulin mRNA repertoires

Broad diversity of neutralizing antibodies isolated from memory B cells in HIV-infected individuals

Sequence and structural convergence of broad and potent HIV antibodies that mimic CD4 binding

Measuring SARS-CoV-2 neutralizing antibody activity using pseudotyped and chimeric viruses

Passive transfer of modest titers of potent and broadly neutralizing anti-HIV monoclonal antibodies block SHIV infection in macaques

Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega

Comprehensive Integration of Single-Cell Data

Potent neutralization of severe acute respiratory syndrome (SARS) coronavirus by a human mAb to S1 protein that blocks receptor association

Decreased neutralization of SARS-CoV-2 global variants by therapeutic anti-spike protein monoclonal antibodies. bioRxiv

An ACE2 Microbody Containing a Single Immunoglobulin Fc Domain Is a Potent Inhibitor of SARS-CoV-2

Identification of human neutralizing antibodies against MERS-CoV and their role in virus adaptive evolution

Emergence and rapid spread of a new severe acute respiratory syndrome-related coronavirus 2 (SARS-CoV-2) lineage with multiple spike mutations in South Africa. medRxiv

A fully automatic method yielding initial models from high-resolution cryo-electron microscopy maps

Potent binding of 2019 novel coronavirus spike protein by a SARS coronavirus-specific human monoclonal antibody

Efficient generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression vector cloning

Autoreactivity in human IgG+ memory B cells

Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq

Ultrapotent human antibodies protect against SARS-CoV-2 challenge via multiple mechanisms

From Louvain to Leiden: guaranteeing well-connected communities

An efficient method to make human monoclonal antibodies from memory B cells: potent neutralization of SARS coronavirus

The costimulatory molecule CD27 maintains clonally diverse CD8(+) T cell responses of low antigen affinity to protect against viral variants

Microneutralization assay titres correlate with protection against seasonal influenza H1N1 and H3N2 in children

Germinal centers

Transcriptional and clonal characterization of B cell plasmablast diversity following primary and secondary natural DENV infection

Enhanced SARS-CoV-2 Neutralization by Secretory IgA in vitro

mRNA vaccine-elicited antibodies to SARS-CoV-2 and circulating variants

Predominant autoantibody production by early human B cell precursors

Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants

Evaluation of CD4-CD4i antibody architectures yields potent, broadly cross-reactive anti-human immunodeficiency virus reagents

Computational analysis of anti-HIV-1 antibody neutralization panel data to identify potential functional epitope residues

SARS-CoV-2 501Y.V2 escapes neutralization by South African COVID-19 donor plasma

Overview of the CCP4 suite and current developments

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Serum Neutralizing Activity Elicited by mRNA-1273 Vaccine -Preliminary Report

An Alternative Binding Mode

Antibodies to the SARS-CoV-2 Receptor Binding Domain

A noncompeting pair of human neutralizing antibodies block COVID-19 virus binding to its receptor ACE2

Quantifying selection in high-throughput Immunoglobulin sequencing data sets

Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2

CD80 and PD-L2 define functionally distinct memory B cell subsets that are independent of antibody isotype

We thank all study participants who devoted time to our research and the clinical staff. We thank Heather Kang for editorial assistance with the manuscript and figures. We thank members of the Bjorkman, Xavier, and Bieniasz laboratories for helpful discussions. We thank Liat Amir-Zilberstein and Novalia Pishesha for support with sample processing and Christy Lavine