key: cord-1056564-u8m7dnl7
authors: Andreano, Emanuele; Nicastri, Emanuele; Paciello, Ida; Pileri, Piero; Manganaro, Noemi; Piccini, Giulia; Manenti, Alessandro; Pantano, Elisa; Kabanova, Anna; Troisi, Marco; Vacca, Fabiola; Cardamone, Dario; De Santi, Concetta; Torres, Jonathan L.; Ozorowski, Gabriel; Benincasa, Linda; Jang, Hyesun; Di Genova, Cecilia; Depau, Lorenzo; Brunetti, Jlenia; Agrati, Chiara; Capobianchi, Maria Rosaria; Castilletti, Concetta; Emiliozzi, Arianna; Fabbiani, Massimiliano; Montagnani, Francesca; Bracci, Luisa; Sautto, Giuseppe; Ross, Ted M.; Montomoli, Emanuele; Temperton, Nigel; Ward, Andrew B.; Sala, Claudia; Ippolito, Giuseppe; Rappuoli, Rino
title: Extremely potent human monoclonal antibodies from COVID-19 convalescent patients
date: 2021-02-23
journal: Cell
DOI: 10.1016/j.cell.2021.02.035
sha: 98856090d999ab258e7e38521afc130e554dc78a
doc_id: 1056564
cord_uid: u8m7dnl7

Human monoclonal antibodies are safe, preventive and therapeutic tools, that can be rapidly developed to help restore the massive health and economic disruption caused by the coronavirus disease 2019 (COVID-19) pandemic. By single cell sorting 4,277 SARS-CoV-2 spike protein specific memory B cells from 14 COVID-19 survivors, 453 neutralizing antibodies were identified. The most potent neutralizing antibodies recognized the spike protein receptor binding domain, followed in potency by antibodies that recognize the S1 domain, the spike protein trimer and the S2 subunit. Only 1.4% of them neutralized the authentic virus with a potency of 1-10 ng/mL. The most potent monoclonal antibody, engineered to reduce the risk of antibody dependent enhancement and prolong half-life, neutralized the authentic wild type virus and emerging variants containing D614G, E484K and N501Y substitutions. Prophylactic and therapeutic efficacy in the hamster model was observed at 0.25 and 4 mg/kg respectively in absence of Fc-functions.

Isolation Isolation Isolation Isolation and characterization and characterization and characterization and characterization of S of S of S of S----protein specific antibodies from protein specific antibodies from protein specific antibodies from protein specific antibodies from SARS SARS SARS SARS----CoV CoV CoV CoV----2 con 2 con 2 con 2 convalescent valescent valescent valescent 118 patients patients patients patients 119

To retrieve mAbs specific for SARS-CoV-2 S-protein, peripheral blood mononuclear cells (PBMCs) 120 from fourteen COVID-19 convalescent patients enrolled in this study were collected and stained 121 with fluorescently labelled S-protein trimer to identify antigen specific memory B cells (MBCs). 122 Figure 1 summarizes the overall experimental strategy. The gating strategy described in Figure  123 S1A was used to single cell sort, into 384-well plates, IgG + and IgA + MBCs binding to the SARS-124

CoV-2 S-protein trimer in its prefusion conformation. The sorting strategy aimed to specifically 125 identify class-switched MBCs (CD19 + CD27 + IgD -IgM -) to identify only memory B lymphocytes that 126 underwent maturation processes. A total of 4,277 S-protein-binding MBCs were successfully 127 retrieved with frequencies ranging from 0.17% to 1.41% (Table S1 ). Following the sorting 128 procedure, S-protein + MBCs were incubated over a layer of 3T3-CD40L feeder cells in the 129 presence of IL-2 and IL-21 stimuli for two weeks to allow natural production of immunoglobulins 130 (Huang et al., 2013) . Subsequently, MBC supernatants containing IgG or IgA were tested for their 131 ability to bind either the SARS-CoV-2 S-protein trimer in its prefusion conformation or the S-protein 132 S1 + S2 subunits (Figure 2A Table S1 ). 135 136 Identification of S Identification of S Identification of S Identification of S----protein specific mAbs able to neutralize SARS protein specific mAbs able to neutralize SARS protein specific mAbs able to neutralize SARS protein specific mAbs able to neutralize SARS----CoV CoV CoV CoV----2 2 2 2 137

The 1,731 supernatants containing S-protein specific mAbs, were screened in vitro for their ability 138 to block the binding of the streptavidin-labelled S-protein to Vero E6 cell receptors and for their 139 ability to neutralize authentic SARS-CoV-2 virus by in vitro microneutralization assay. In the 140 neutralization of binding (NoB) assay, 339 of the 1,731 tested (19.6%) S-protein specific mAbs 141 were able to neutralize the antigen/receptor binding showing a broad array of neutralization 142 potency ranging from 50% to 100% (Table S1 and Figure S2C ). 143 J o u r n a l P r e -p r o o f As for the authentic virus neutralization assay, supernatants containing naturally produced IgG or 144

IgA were tested for their ability to protect the layer of Vero E6 cells from the cytopathic effect 145 triggered by SARS-CoV-2 infection. To increase the throughput of our approach, supernatants 146 were tested at a single point dilution and to increase the sensitivity of our first screening a viral titer 147 of 25 TCID 50 was used. For this screening mAbs were classified as neutralizing, partially 148 neutralizing and non-neutralizing based on their complete, partial or absent ability to prevent the 149 infection of Vero E6 cells respectively. Out of 1,731 mAbs tested in this study, a panel of 453 150 (26.2%) mAbs neutralized the authentic virus and prevented infection of Vero E6 cells (Table S1) . 151

The percentage of partially neutralizing mAbs and neutralizing mAbs (nAbs) identified in each 152 donor was extremely variable ranging from 2.6 -29.7% and 2.8 -26.4% respectively ( Figure 2B  153 and Table S2 ). The majority of nAbs were able to specifically recognize the S-protein S1 domain 154 (57.5%; N=244) while 7.3% (N=53) of nAbs were specific for the S2 domain and 35.2% (N=156) 155 did not recognize single domains but only the S-protein in its trimeric conformation ( Figure S2A ; 156 Table S3 ). From the panel of 453 nAbs, we recovered the heavy and light chain variable regions of 157 220 nAbs which were expressed as full length IgG1 using the transcriptionally active PCR (TAP) 158 approach to characterize their neutralization potency against the live virus at 100 TCID 50 . The vast 159 majority of nAbs identified (65.9%; N=145) had a low neutralizing potency and required more than 160 500 ng/mL to achieve 100% inhibitory concentration (IC 100 ). A smaller fraction of the antibodies had 161 an intermediate neutralizing potency (23.6%; N=52) requiring between 100 and 500 ng/mL to 162 achieve the IC 100 , while 9.1% (N=20) required between 10 and 100 ng/mL. Finally, only 1.4% (N=3) 163 of the expressed nAbs were classified as extremely potent nAbs, showing an IC 100 lower than 10 164 ng/mL ( Figure 2C and Figure S2B ; Table S4 ). 165

CoV----2 2 2 2 neutralizing antibodies neutralizing antibodies neutralizing antibodies neutralizing antibodies can be can be can be can be classified classified classified classified in in in into to to to four groups four groups four groups four groups 167

Based on the first round of screening, 14 nAbs were selected for further characterization. All nAbs 168 were able to bind the SARS-CoV-2 S-protein in its trimeric conformation ( Figure 3A ). The mAbs 169 named J08, I14, F05, G12, C14, B07, I21, J13 and D14 were also able to specifically bind the S1 170 domain ( Figure 3B ). The nAbs named H20, I15, F10 and F20 were not able to bind single S1 or S2 171 The fourteen selected nAbs were further characterized by a competition assay that allowed 210 speculation on the S-protein regions recognized by these antibodies. Briefly, beads were coated 211 with SARS-CoV-2 trimeric S-protein and incubated with a primary unlabeled antibody in order to 212 saturate the binding site on the antigen surface. Following the first incubation step a secondary 213

Alexa-647 labeled antibody was incubated with the antigen/unlabeled-mAb complex. If the 214 secondary labeled-antibody did not recognize the same epitope as the primary unlabeled-mAb a 215 fluorescent signal would be detected when tested by flow cytometry. Through this assay, we 216 observed that all Group I nAbs competed amongst themselves for binding to the S-protein RBD, 217

indicating that these antibodies possibly clash against each other and recognize a similar epitope 218 region. All Group II nAbs, showed a different competition profiles and competed with Group II and 219

Group III nAbs. These results confirmed that Group III antibodies can recognize various regions on 220 the S-protein surface as they compete with themselves as well as with antibodies belonging to 221

Group II. Interestingly, nAbs belonging to Group II also competed with the B07 RBD-directed 222 antibody and thereby suggesting that this latter nAb may have a different binding orientation 223 compared to other nAbs included in the Group I. Finally, the Group IV nAb L19 did not compete 224 with any of the other groups identified in this study suggesting that this class of nAbs recognize a 225 distant epitope region as compared to Group I -II and III nAbs ( Figure 4A Genetic characterization of SARS----CoV CoV CoV  CoV----2 nAbs  2 nAbs  2 nAbs  2 

The genes encoding the heavy and light chains of the 14 selected nAbs, were sequenced and their 229 IGHV and IGKV genes compared with publicly available SARS-CoV-2 neutralizing antibody 230 sequences ( Figure 5A -B) . Four nAbs used one of the most predominant heavy chain V genes for 231 SARS-CoV-2 nAbs (IGHV1-69), while three nAbs used one of the least representative heavy chain 232 V genes (IGHV1-24). Other two nAbs, employed the most common germline observed for SARS-233

CoV-2 nAbs which is IGHV3-53 ( Figure 5A ) . Interestingly, while IGHV1-69 and 234 IGHV1-24 accommodate IGHJ diversity, nAbs belonging to the IGHV3-53 gene family only showed 235 recombination with the IGHJ6 gene (Table S6 ). The heavy chain V genes somatic hypermutation 236 level and complementary determining region 3 (H-CDR3) length were also evaluated. Our selected 237 nAbs displayed a low level of somatic mutations when compared to the inferred germlines with 238 sequence identities ranging from 95.6 to 99.3% (Figure 5C left panel; Table S6 ) confirming what 239 was observed in previous publications (Pinto et al., 2020 , Zost et al., 2020b , Rogers et al., 2020 , 240 Griffin et al., 2020 . The H-CDR3 length spanned from 7 to 21 amino acids (aa) with the majority of 241 the antibodies (N=6; 42.0%) having a length of 14 to 16 aa that is slightly bigger than previously 242 observed (Figure 5C right panel; Table S6 ). All of our nAbs used the κ-chain and the majority of 243 them used the common genes IGKV1-9 and IGKV3-11 (N=6; 42.0%) ( Figure 5B ; Table S6 ). The 244 level of IGKV somatic hypermutation was extremely low for light chains showing a percentage of 245 sequence identities ranging from 94.3 to 98.9% ( Figure 5D left panel; Table S6 ). The light chain 246 CDR3 (L-CDR3) length were ranging from 5 to 10 aa which is in line with what was previously 247 observed for SARS-CoV-2 nAbs (Figure 5D right panel; Table S6 ). When paired heavy and light 248 chain gene analysis was performed, IGHV1-69 derived nAbs were found to rearrange exclusively 249 with IGKV3 gene family while IGHV1-24 derived nAbs accommodate light chain diversity (Table  250 S6). Of note, some of our candidates showed unique heavy and light chain pairing when compared 251 to the public SARS-CoV-2 nAb repertoire. Particularly, five different heavy and light chain 252 rearrangements not previously described for nAbs against SARS-CoV-2 were identified. These 253 included the IGHV1-24;IGKV1-9, IGHV1-24;IGKV3-15, IGHV1-46;IGKV1-16, IGHV3-30;IGKV1-9,  254 IGHV3-53;IGKV1-17 ( Figure 5E ). 255 J o u r n a l P r e -p r o o f Fc c c  c receptor binding  receptor binding  receptor binding  receptor binding and extend half  and extend half  and extend half  and extend half----life life life  life  256 Antibody-dependent enhancement (ADE) of disease, is a potential clinical risk following 257 coronavirus infection (Lee et al., 2020) . Therefore, to optimize the suitability for clinical 258 development and reduce the risk of ADE, five different point mutations were introduced in the 259 constant region (Fc) of the three most potent nAbs (J08, I14 and F05) which were renamed J08- FcγRs and cell-based activities (Schlothauer et al., 2016) . 265

To confirm the lack of FcγR binding as well as the extended half-life, a beads-based Luminex 266 assay was performed. Briefly the beads were coated with SARS-CoV-2 S-protein RBD. Antibodies 267 were tested at eight-point dilutions and the binding was detected with FcγR2A and neonatal Fc 268 receptor (FcRn) at pH6.2 and 7.4. The FcγR2A was selected as it is predominantly expressed on 269 the surface of phagocytic cells (such as monocytes, macrophages and neutrophils) and is 270 associated with phagocytosis of immune complexes and antibody-opsonized targets (Ackerman et 271 al., 2013) . On the other hand, FcRn, which is highly expressed on endothelial cells and circulating 272 monocytes, was selected as it is responsible for the recycling and serum half-life of IgG in the 273 circulation (Mackness et al., 2019). This latter receptor was shown to possess a tighter binding at 274 lower pH (e.g. pH 6.2) compared to a physiological pH (e.g. pH 7.4) (Booth et al., 2018) . Results 275 shown in Figure S6 demonstrate that binding to the FcγR2A was completely abrogated for the 276 mutated version of candidate nAbs (J08-MUT, I14-MUT and F05-MUT) compared to their 277 respective wild type (WT) versions (J08, I14 and F05) and controls (CR3022 and unrelated protein) 278 ( Figure S6A ). Furthermore, Fc-engineered antibodies showed increased binding activity to the 279 FcRn at both pH 6.2 and 7.4 compared to their WT counterpart ( Figure S6B -C) . Finally, to 280 evaluate the lack of Fc-mediated cellular activities by our three candidate nAbs, the antibody- . For the ADNP assay, primary human neutrophils were used to detect antibody binding to 284 SARS-CoV-2 S-protein RBD coated beads, while ADNK activity was evaluated by using primary 285 human NK cells and detecting the release of the proinflammatory cytokine interferon gamma (IFN-286 γ). Complete abrogation of both ADNP and ADNK was observed for all three Fc-engineered 287

candidate nAbs compared to their WT versions and control antibody (CR3022), thus confirming the 288 lack of Fc-mediated cellular activities ( Figure S6D -E) . The three engineered antibodies were tested to confirm their binding specificity and neutralization 292 potency against both the WT, the widespread SARS-CoV-2 D614G mutant and the emerging 293 variant B.1.1.7 (Korber et al., 2020 , CDC, 2021 ) to evaluate their cross-neutralization ability. The 294 three engineered nAbs maintained their S1-domain binding specificity and extremely high 295 neutralization potency with J08-MUT and F05-MUT being able to neutralize both the WT and the 296 D614G variant with an IC 100 lower than 10 ng/mL (both at 3.9 ng/mL for the WT and the D614G 297 strains) (Figure S6F -K; Table S5 ). The antibody J08-MUT also showed extreme neutralization 298 potency against emerging variants as it was able to neutralize the B.1.1.7 with an identical IC 100 299 compared to the WT virus ( Figure S6K ; Table S5 ) and has also showed to neutralize variants that 300 include the E484K mutation (Andreano et al., 2020) . 301

Since it has been reported that SARS-CoV-2 elicited antibodies that can cross-react with human 302 tissues, cytokines, phospholipids and phospholipid-binding proteins (Zuo et al., 2020 , Bastard et 303 al., 2020 , Kreer et al., 2020 , the three candidate mAbs in both their WT and MUT versions were 304 tested through an indirect immunofluorescent assay against human epithelial type 2 (HEp-2) cells 305 which expose clinically relevant proteins to detect autoantibody activities ( Figure S7A ). As reported 306

in Figure S7B , the positive control presents a different range of detectable signals based on the 307 initial dilution steps (from bright-green at 1:1 to very dim-green at 1:100). Among all samples 308 tested, only F05 showed moderate level of autoreactivity to human cells while no signal could be 309 measured for the other antibodies ( Figure S7B) . uctural analyses ctural analyses ctural analyses ctural analyses of of of of candidate nAbs candidate nAbs candidate nAbs candidate nAbs 312

Single particle negative stain electron microscopy (nsEM) was used to visualize a stabilized SARS-313 2-CoV-6P-Mut7 spike protein in complex with three separate Fabs: J08, I14 and F05. This 314 recombinant, soluble spike protein primarily exhibits 3 RBD's "down" but can switch to RBD "up" 315 conformation with antibody bound. Inspection of the 2D class averages revealed a mixed 316 stoichiometry of unbound spike protein, 1 Fab bound, and 2 Fab bound classes, which allowed for 317 3D refinements of each ( Figure 6A ). The three different Fabs bind to the RBD in the "up" 318 conformation, although at different angles and rotations, likely due to the flexibility of the RBD. 319

Model docking of PDB 7BYR (one RBD "up" bound to antibody) shows that the fabs overlap with 320 the receptor binding motif (RBM), and therefore are positioned to sterically block receptor hACE2 321 engagement ( Figure 6B ). To determine the epitope, heavy chain (HC) and light chain (LC) 322 sequences of Fabs J08, I14, and F05 were used to create synthetic models for docking into the 323 nsEM maps. Based on the docking, we predicted that a loop containing residues 477 to 489 324 (STPCNGVEGFNCY) appeared to be involved in the binding specifically with residue F486 325 extending into a cavity that is in the middle of the HC and LC of each antibody. IgG1 isotype control groups were included in the study which received a saline solution and an 336 anti-influenza antibody at the concentration of 4 mg/kg respectively. The J08-MUT at 4 mg/kg 337 group and the 1 and 0.25 mg/kg groups were tested in two independent experiments. The IgG1 338 isotype control group was tested in parallel with the J08-MUT 4 mg/kg group while the placebo is 339 an average of the two experiments. Animals were challenged with 100 µL of SARS-CoV-2 solution 340

(5 x 10 5 PFU) via intranasal distillation twenty four hours post-administration of the antibody. Three 341 hamsters per group were sacrificed at three days post-infection while the remaining animals were 342 culled at day 8 ( Figure 7A ). Body weight change was daily evaluated and considered as a proxy for 343 disease severity. Animals in the control group and those that received the IgG1 isotype antibody 344 lost more than 5% of their original body weight from day 1 to day 6 and then stabilized. These data 345 are in line with previously published data of SARS-CoV-2 infection in a golden Syrian hamster 346 model (Kreye et al., 2020 , Liu et al., 2020 . In marked contrast, in the prophylactic study, all 347 animals that received J08-MUT were significantly protected from weight loss. Protection was 348 present at all J08-MUT concentrations and was dose dependent ( Figure 7B ). When J08-MUT was 349 administered at 4 mg/kg we observed protection from SARS-CoV-2 infection and only a minimal 350 weight loss (average -1.8% of body weight) was noticed one day post viral challenge. A higher 351 body weight loss was observed 1 day post infection in hamsters that received J08-MUT at 1 mg/kg 352 (from -1.8% to -3.3%) and 0.25 mg/kg (from -1.8% to -4.7%). In the J08-MUT 4 mg/kg group all 353 animals quickly recovered and reached their initial weight by day 3. From day 4 on all hamsters 354 gained weight increasing up to 5% from their initial body weight. Hamsters that received the 1 and 355 0.25 mg/kg dosages completely recovered their initial body weight at day 6 and 8 respectively. 356

Hamsters in the control groups did not recover their initial body weight and at day 8 still showed 357 around 5% of weight loss ( Figure 7B ). The prophylactic activity of J08-MUT was also reflected in 358 the complete absence of viral titer in the lung tissue at three days post-infection in all hamsters that 359 received J08-MUT at 4 and 1 mg/kg and also in two out of three hamsters that received J08-MUT 360 at 0.25 mg/kg. On the other hand, hamsters that received the IgG1 isotype control or in the placebo 361 group showed a significantly higher viral titer ( Figure 7D ). 362

Finally, we performed an ELISA assay to detect the presence of human IgG in hamster sera. All 363 samples that received J08-MUT or the IgG1 isotype control showed detectable human IgGs in the 364 sera in a dose-dependent fashion, while no human IgGs were detected in the placebo group 365 For the therapeutic study, 3 groups of 6 animals each were used to evaluate the ability of J08-MUT 370 to treat SARS-CoV-2 infection in the golden Syrian hamster model. One group received J08-MUT 371 via intraperitoneal injection at 4 mg/kg and the other two groups received placebo and 4mg/kg IgG1 372 isotype control respectively. The experiment was performed in parallel with the initial prophylactic, 373 study where J08-MUT was administered at 4 mg/kg, and the two control groups. Animals were 374 challenged with 100 µL of SARS-CoV-2 solution (5 x 10 5 PFU) via intranasal distillation twenty four 375 hours prior to the administration of the antibody. Three hamsters per group were sacrificed at three 376 days post-infection while the remaining animals were culled at day 12 ( Figure 7A ). Despite J08-377 MUT and control groups showed a similar trend in weight loss in the first four days post-infection, 378

the treatment group showed a significantly quicker weight recovery ( Figure 7C ). At day 12, only 379 hamsters that received J08-MUT recovered the initial body weight ( Figure 7C ). When we analyzed 380 the viral titer in lung tissues we observed complete absence of the virus at day 3 in all the hamsters 381 treated with J08-MUT at 4 mg/kg while animals that received the IgG1 isotype control or in the 382 placebo group showed a significantly higher viral titer ( Figure 7G ). To evaluate the presence of 383 human antibodies in hamster sera, we performed an ELISA assay. All samples that received J08-384 MUT or the IgG1 isotype control showed detectable human IgGs in the sera in a dose-dependent 385 fashion, while no human IgGs were detected in the placebo group ( Figure In the search for potent antibodies, we found that approximately 10% of the total B cells against the 397 S-protein isolated produce neutralizing antibodies and these can be divided into 4 different groups 398

recognizing the S1-RBD, S1-domain, S2-domain and the S-protein trimer. Most potently 399 neutralizing antibodies are extremely rare and recognize the RBD, followed in potency by 400 antibodies recognizing the S1 domain, the trimeric structure and the S2 subunit. From these data 401 we can conclude that in COVID-19 convalescent patients most of the observed neutralization titers 402 are likely mediated by antibodies with medium-high neutralizing potency. Indeed, the extremely 403 potent antibodies and the antibodies against the S2 subunit are unlikely to contribute to the overall 404 neutralizing titers because they are respectively too rare and too poor neutralizers to be able to 405 make a difference. We and others found that the antibody repertoire of convalescent patients is 406 mostly germline-like. This may be a consequence of the loss of Bcl-6-expressing follicular helper T 407 cells and the loss of germinal centers in COVID-19 patients which may limit and constrain the B cell 408 affinity maturation (Kaneko et al., 2020) . It will be therefore important to perform similar studies 409 following vaccination as it is likely that the repertoire of neutralizing antibodies induced by 410 vaccination may be different from the one described here. 411

Out of the 453 neutralizing antibodies that were tested and characterized, one antibody (J08) 412

showed extremely high neutralization potency against both the wild type SARS-CoV-2 virus 413 isolated in Wuhan and emerging variants containing the D614G, E484K and N501Y variants. 414

During the last few months several groups reported the identification, three-dimensional structure 415 and passive protection in animal models of neutralizing antibodies against SARS-CoV-2. Most of 416 these studies, with few exceptions, reported antibodies which require from 20 to several hundred 417 ng/mL to neutralize 50% of the virus in vitro. While these antibodies are potentially good for 418 therapy, they will require a high dosage which is associated with elevated cost of goods, low 419 production capacity and delivery by intravenous infusion. 420

The extremely potent mAb described in our study is likely to allow to use lower quantities of 421 antibodies to reach prophylactic and therapeutic efficacy and as a consequence decrease the cost 422 of goods and enable sustainable development and manufacturability. This solution may increase 423 the number of doses produced annually and therefore increase antibodies availability in high 424 income countries as well as low-and middle-income countries (LMICs). Therefore our antibodies 425

have the potential to meet the expectations of the call to action to expand access to monoclonal 426 antibody-based products, recently published by the Wellcome Trust, and supported by the WHO 427 and the Coalition for Epidemic Preparedness Innovations (CEPI) (Wellcome, 2020). 428

A potential issue associated with the use of human monoclonal antibodies against viral pathogens 429 is the potential selection of escape mutants. This is usually addressed by using a combination of 430 antibodies directed against non-overlapping epitopes. While this is an ultimate clear solution, it 431 increases the complexity of development, costs of production, drug availability and affordability. In 432 our case we believe that selection of escape mutants upon treatment with a single monoclonal 433

antibody may be quite difficult as the SARS-CoV-2 RNA-dependent polymerase possesses a 434 proofreading machinery (Romano et al., 2020) and the epitope recognized by the antibodies herein 435 described overlaps with the region necessary to bind the hACE2 receptor. In this regard, it took 436 more than 70 days of continuous co-culture of the virus in presence of the antibodies before we 437 were able to detect the first emergence of escape mutants of the wild-type SARS-CoV-2 (data not 438

Finally, a peculiar part of our approach consisted in depleting possible antibody Fc-mediated 440 functions of the antibodies to avoid the risk of ADE. While there is no evidence of ADE in SARS-441

CoV-2, and most vaccines and mAbs tested so far seem to be safe, it is too early to make definitive 442

conclusions. In addition, two recently published reports suggested that we need to continue to 443 monitor the potential risk of ADE. The first report showed that severe SARS-CoV-2 patients are 444 characterized by an increased proinflammatory signature mediated by the Fcγ receptors triggered 445 by afucosylated IgG1 antibodies (Chakraborty et al., 2020) . The second report described that one 446 antibody was associated with worse clinical outcomes when administered to hospitalized patients 447 requiring high flow oxygen or mechanical ventilation (Lilly, 2020). Therefore, we believe it is While we believe that our antibodies are extremely potent when compared to most of those 457 described in literature, we acknowledge that in most cases direct comparison was not performed 458 and we rely on published data. 459

The second limitation of the study is that in vitro neutralization and in vivo protection in the SARS- S1 S1 S1. . . . Gating strategy for single cell sorting and m Gating strategy for single cell sorting and m Gating strategy for single cell sorting and m Gating strategy for single cell sorting and monoclonal antibodies screening for S onoclonal antibodies screening for S onoclonal antibodies screening for S onoclonal antibodies screening for S----protein protein protein protein 603 S1 + S2 subunits S1 + S2 subunits S1 + S2 subunits S1 + S2 subunits binding binding binding binding and neutralization of binding (NoB) and neutralization of binding (NoB) and neutralization of binding (NoB) and neutralization of binding (NoB) activity activity activity activity, related to harvested three and six days after transfection. Cells were separated from the medium by 722 centrifugation (1,100 g for 10 min at 24°C). Collected supernatants were then pooled and clarified 723 by centrifugation (3,000 g for 15 min at 4°C) followed by filtration through a 0.45 µm filter. 724

Chromatography was conducted at room temperature using the ÄKTA go purification system from 725 GE Healthcare Life Sciences. Expressed proteins were purified by using an immobilized metal 726 affinity chromatography (FF Crude) followed by dialysis into final buffer. Specifically, the filtrated 727 culture supernatant was purified with a 5 mL HisTrap FF Crude column (GE Healthcare Life 728 Sciences) previously equilibrated in Buffer A (20 mM NaH 2 PO 4 , 500 mM NaCl + 30 mM imidazol 729 pH 7.4). 730

The flow rate for all steps of the HisTrap FF Crude column was 5 mL/min. The culture supernatant 731 of spike and RBD cell culture was applied to a single 5 mL HisTrap FF Crude column. The column 732 was washed in Buffer A for 4 column volumes (CV) with the all 4 CV collected as the column wash. The final protein concentration was determined by measuring the A520 using the Pierce™ BCA 741 protein assay kit (Thermo Scientific™ ). Final protein was dispensed in aliquots of 0.5 ml each and 742 stored at -80°C. 743 744 ELISA assay with S1 and S2 subunits of SARS ELISA assay with S1 and S2 subunits of SARS ELISA assay with S1 and S2 subunits of SARS ELISA assay with S1 and S2 subunits of SARS----CoV CoV CoV CoV----2 S 2 S 2 S 2 S----protein protein protein protein 745

The presence of S1-and S2-binding antibodies in culture supernatants of monoclonal S-protein- CoV----2 S 2 S 2 S 2 S----protein prefusion trimer protein prefusion trimer protein prefusion trimer protein prefusion trimer and S1 and S1 and S1 and S1 ----S2 subunits S2 subunits S2 subunits S2 subunits 768 ELISA assay was used to detect SARS-CoV-2 S-protein specific mAbs and to screen plasma from 769 SARS-CoV-2 convalescent donors. 384-well plates (384 well plates, microplate clear; Greiner Bio-770 one) were coated with 3 µg/mL of streptavidin (Thermo Fisher) diluted in coating buffer (0.05 M 771 carbonate-bicarbonate solution, pH 9.6) and incubated at RT overnight. Plates were then coated 772 with SARS-CoV-2 S-protein, S1 or S2 domains at 3 µg/mL and incubated for 1h at RT. 50 µL/well 773 of saturation buffer (PBS/BSA 1%) was used to saturate unspecific binding and plates were 774 incubated at 37°C for 1h without CO 2 . For the first round of screening, supernatants were diluted 775 1:5 in PBS/BSA 1%/Tween20 0.05% in 25 µL/well final volume and incubated for 1h at 37°C 776 without CO 2 . For purified antibodies, and to assess EC 50 , mAbs were tested at a starting 777 concentration of 5 µg/mL and diluted step 1:2 in PBS/BSA 1%/Tween20 0.05% in 25 µL/well final 778 volume for 1h at 37°C without CO 2 . 25 µL/well of alkaline phosphatase-conjugated goat anti-human 779

IgG (Sigma-Aldrich) and IgA (Southern Biotech) were used as secondary antibodies. Wells were 780 washed three times between each step with PBS/BSA 1%/Tween20 0.05%. pNPP (p-nitrophenyl 781 phosphate) (Sigma-Aldrich) was used as soluble substrate to detect SARS-CoV-2 S-protein, S1 or 782 S2 specific mAbs and the final reaction was measured by using the Varioskan Lux Reader (Thermo 783 Expi293F™ cells (Thermo Fisher) were transiently transfected with plasmids carrying the antibody 859 heavy chain and the light chains with a 1:2 ratio. Cells were grown for six days at 37 °C with 8% 860 CO 2 shaking at 125 rpm according to the manufacturer's protocol (Thermo Fisher); 861

ExpiFectamine™ 293 transfection enhancers 1 and 2 were added 16 to 18 hours post-transfection 862 to boost cell viability and protein expression. Cell cultures were harvested three and six days after 863 transfection. Cells were separated from the medium by centrifugation (1,100 g for 10 min at 24°C). 864

Supernatants collected were then pooled and clarified by centrifugation (3000 g for 15 min, 4°C) 865 followed by filtration through a 0.45 µm filter. Chromatography was conducted at room temperature 866 using the ÄKTA go purification system from GE Healthcare Life Sciences. Affinity chromatography 867 was used to purify expressed monoclonal antibodies using an immobilized protein G column able 868 to bind to Fc region. Specifically, filtrated culture supernatants were purified with a 1 mL HiTrap 869 stock. Viral titers were determined in confluent monolayers of Vero E6 cells seeded in 96-well 894 plates using a 50% tissue culture infectious dose assay (TCID 50 ). Cells were infected with serial 895 1:10 dilutions (from 10 -1 to 10 -11 ) of the virus and incubated at 37°C, in a humidified atmosphere 896 with 5% CO 2 . Plates were monitored daily for the presence of SARS-CoV-2 induced CPE for 4 897 days using an inverted optical microscope. The virus titer was estimated according to Spearman-898

Karber formula (Kundi, 1999) and defined as the reciprocal of the highest viral dilution leading to at 899 least 50% CPE in inoculated wells. After 1 hour incubation at 37°C, 5% CO 2 , 25 µL of each virus-supernatant mixture was added to the 911 wells of a 96-well plate containing a sub-confluent Vero E6 cell monolayer. Following a 2-hours 912 incubation at 37°C, the virus-serum mixture was removed and 100 µl of DMEM 2% FBS were 913 added to each well. Plates were incubated for 3 days at 37°C in a humidified environment with 5% 914 CO 2 , then examined for CPE by means of an inverted optical microscope. Absence or presence of 915 CPE was defined by comparison of each well with the positive control (plasma sample showing 916 high neutralizing activity of SARS-CoV-2 in infected Vero E6 cells (Andreano et al., 2020) and 917 negative control (human serum sample negative for SARS-CoV-2 in ELISA and neutralization 918 assays). Following expression as full-length IgG1 recombinant antibodies were quantitatively 919 tested for their neutralization potency against both the wild type, D614G variant and the B.1.1.7 920 emerging variants. The assay was performed as previously described but using a viral titer of 100 921 TCID 50 . Antibodies were prepared at a starting concentration of 20 µg/mL and diluted step 1:2. 922

Technical triplicates were performed for each experiment. Aliquots were stored at -80°C. Titration assays were performed by transduction of HEK293T/17 931 cells pre-transfected with ACE2 and TMPRRS2 plasmids to calculate the viral titer and infectious 932 dose (PV input) for neutralization assays. SARS-CoV-2 D614G pseudotype was produced using 933 the same procedure as described above. SARS-1 pseudotype was produced in a 1:0.5:0.8 ratio. 934

MERS-pseudotype was produced as previously described . 935

CoV----2 pseudotyped lentivirus neutrali 2 pseudotyped lentivirus neutrali 2 pseudotyped lentivirus neutrali 2 pseudotyped lentivirus neutraliz z z zation assay ation assay ation assay ation assay 937

The potency of the neutralizing mAbs was assessed using lentiviral particles expressing SARS-938

CoV-2 spike protein on their surface and containing firefly luciferase as marker gene for detection 939 of infection. Briefly, 2 x 10 6 HEK293T cells were pre-transfected in a 10 cm dish the day before the 940 neutralization assay with ACE2 and TMPRSS2 plasmids in order to be used as optimal target cells 941

for SARS-CoV-2 PV entry. In a 96-well plate mAbs were 2-fold serially diluted in duplicate in culture 942 medium (DMEM supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin) 943 starting at 20 µg/mL in a total volume of 100 µL. 1x10 6 RLU of SARS-CoV-2 pseudotyped lentiviral 944 particles were added to each well and incubated at 37ºC for 1h. Each plate included PV plus cell 945 only (virus control) and cells only (background control). 1 x 10 4 pre-transfected HEK293T cells 946 suspended in 50 µL complete media were added per well and incubated for 48h at 37°C and 5% 947 CO2. Firefly luciferase activity (luminescence) was measured using the Bright-Glo assay system 948 with a GloMax luminometer (Promega, UK). The raw Relative Luminescence Unit (RLU) data 949 points were converted to a percentage neutralization value, whereby 100% neutralization equals 950 the mean cell only RLU value control and 0% neutralization equals the mean PV only RLU value 951 control. The normalized data was then plotted using Prism 8 (GraphPad) on a neutralization 952 percentage scale and a NT50 value calculated, using the non-linear regression analysis. Plasma Antigen Antigen Antigen Antigen----specific F specific F specific F specific Fcγ cγ cγ cγR binding R binding R binding R binding 982

Fluorescently coded microspheres were used to profile the ability of selected antibodies to interact 983 with Fc receptors (Boudreau et al., 2020) . The antigen of interest (SARS -CoV-2 S-protein RBD) 984 was covalently coupled to different bead sets via primary amine conjugation. The beads were 985 incubated with diluted antibody (diluted in PBS), allowing "on bead" affinity purification of antigen-986 specific antibodies. The bound antibodies were subsequently probed with tetramerized 987 recombinant human FcγR2A and FcRN and analyzed using Luminex. The data is reported as the 988 median fluorescence intensity of PE for a specific bead channel. phagocytosis was allowed to proceed for 1 hour. The cells were then washed and fixed, and the 999 extent of phagocytosis was measured by flow cytometry. The data is reported as a phagocytic 1000 score, which considers the proportion of effector cells that phagocytosed and the degree of 1001 phagocytosis. Each sample is run in biological duplicate using neutrophils isolated from two distinct 1002

donors. The mAb were tested for ADNP activity at a range of 30 µg/mL to 137.17 ng/mL. 1003 1004 Antibody Antibody Antibody Antibody----dependent NK cell activation dependent NK cell activation dependent NK cell activation dependent NK cell activation 1005

Antibody-dependent NK cell activation (ADNKA) assesses antigen-specific antibody-mediated NK 1006 cell activation against protein-coated plates. This assay was performed as previously described 1007 J o u r n a l P r e -p r o o f were then washed and blocked. Diluted antibody (diluted in PBS) was added to the antigen coated

Enhanced phagocytic activity of HIV-specific antibodies correlates with natural production of 1120 immunoglobulins with skewed affinity for FcγR2a and FcγR2b

Polyfunctional HIV-Specific Antibody Responses Are Associated 1123 with Spontaneous HIV Control

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

SARS-CoV-2 escape <em>in 1132 vitro</em> from a highly neutralizing COVID-19 convalescent plasma

Jobless America: the coronavirus unemployment crisis in figures. The Guardian

A perspective on potential antibody-1136 dependent enhancement of SARS-CoV-2

Autoantibodies against type I IFNs in patients with life-threatening COVID-1153 19

REGN-1159 COV2 antibodies prevent and treat SARS-CoV-2 infection in rhesus macaques and hamsters

Extending human IgG half-life using structure-guided design

Selective 1165 induction of antibody effector functional responses using MF59-adjuvanted vaccination. The

A Sample-Sparing Multiplexed ADCP Assay

Pseudotype-based 1169 neutralization assays for influenza: a systematic analysis

An Optimized Method for 1171 the Production Using PEI, Titration and Neutralization of SARS-CoV Spike Luciferase Pseudotypes

Proinflammatory IgG Fc structures in patients with severe COVID-19

Infection Depends on Cellular Heparan Sulfate and ACE2. bioRxiv : the preprint server for biology

The COVID-19 Pandemic and the $16 Trillion Virus

Broadly neutralizing hemagglutinin stalk-specific 1190 antibodies require FcγR interactions for protection against influenza virus in vivo

Safety and pharmacokinetics of the Fc-1200 modified HIV-1 human monoclonal antibody VRC01LS: A Phase 1 open-label clinical trial in healthy 1201 adults

An optimised method for the production of MERS-CoV 1203 spike expressing viral pseudotypes

Single-Dose 1206 Nirsevimab for Prevention of RSV in Preterm Infants

Studies in humanized mice and convalescent 1214 humans yield a SARS-CoV-2 antibody cocktail

Reassessing therapeutic antibodies for neglected and tropical diseases. PLoS 1217 neglected tropical diseases

Structure-based design of prefusion-1222 stabilized SARS-CoV-2 spikes

Isolation of human monoclonal antibodies 1225 from peripheral blood B cells

Syrian 1231 hamsters as a small animal model for SARS-CoV-2 infection and countermeasure development

Dried SARS-CoV-2 virus maintains infectivity to Vero E6 cells for up to 48 h

& 1242 MASSACHUSETTS CONSORTIUM ON PATHOGEN READINESS SPECIMEN WORKING, G. 2020. The 1243 Loss of Bcl-6 Expressing T Follicular Helper Cells and the Absence of Germinal Centers in COVID-19. 1244 SSRN

A versatile high-throughput assay to characterize 1247 antibody-mediated neutrophil phagocytosis

Antibody-dependent enhancement of severe dengue disease in humans

Developing therapeutic monoclonal antibodies at pandemic pace

Tracking Changes in SARS-CoV-2 Spike: Evidence that D614G Increases Infectivity of the 1257 COVID-19 Virus

1262 Longitudinal Isolation of Potent Near-Germline SARS-CoV-2-Neutralizing Antibodies from COVID-19 1263 Patients

SARS-CoV-2 Antibody Protects from Lung Pathology in a COVID-19 Hamster Model

One-hit models for virus inactivation studies

Successful Ebola treatments promise to tame outbreak

Appion: an integrated, 1277 database-driven pipeline to facilitate EM image processing

Lilly's neutralizing antibody bamlanivimab (LY-CoV555) receives FDA emergency use 1282 authorization for the treatment of recently diagnosed COVID-19

Potent 1286 neutralizing antibodies against multiple epitopes on SARS-CoV-2 spike

Antibody Fc engineering for 1289 enhanced neonatal Fc receptor binding and prolonged circulation half-life

Evaluation of SARS-CoV-2 neutralizing antibodies using a CPE-based 1292 colorimetric live virus micro-neutralization assay in human serum samples

FDA approves antibody cocktail for Ebola virus

Use of broadly neutralizing antibodies 1295 for HIV-1 prevention

UCSF Chimera--a visualization system for exploratory research and analysis

Cross-neutralization of SARS-CoV-2 by a 1303 human monoclonal SARS-CoV antibody

Isolation of potent SARS-CoV-2 neutralizing antibodies and 1311 protection from disease in a small animal model

A Structural View of SARS-CoV-2 1313 RNA Replication Machinery: RNA Synthesis, Proofreading and Final Capping

Antibody potency, effector function and 1317 combinations in protection from SARS-CoV-2 infection in vivo. bioRxiv : the preprint server for 1318 biology

RELION: implementation of a Bayesian approach to cryo-EM structure determination

Novel human IgG1 and IgG4 Fc-engineered antibodies 1323 with completely abolished immune effector functions

A human neutralizing antibody targets the 1327 receptor-binding site of SARS-CoV-2

Pathogenesis and transmission of 1330 SARS-CoV-2 in golden hamsters

Therapeutic antibodies for infectious 1332 diseases

Automated molecular microscopy: the new Leginon system

The trinity of COVID-19: immunity, 1337 inflammation and intervention

Efficient 1339 generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression 1340 vector cloning

DoG Picker and 1342 TiltPicker: software tools to facilitate particle selection in single particle electron microscopy

Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Structural and Functional Basis of SARS-CoV-2 Entry by Using 1348 Human ACE2

Expanding access to monoclonal antibody-based products: a global call to action

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

Structural basis of a 1357 public antibody response to SARS-CoV-2. bioRxiv : the preprint server for biology

Enhanced antibody half-life improves in vivo activity

Potently neutralizing and protective human antibodies against SARS-1369 CoV-2

Rapid isolation and profiling of a 1375 diverse panel of human monoclonal antibodies targeting the SARS-CoV-2 spike protein

Single-cell RNA-seq data analysis on the 1378 receptor ACE2 expression reveals the potential risk of different human organs vulnerable to 2019-1379 nCoV infection

Prothrombotic autoantibodies in serum from 1383 patients hospitalized with COVID-19

memory B cells encoding extremely potent neutralizing antibodies are rare

Most potent antibodies recognize the tip of the spike receptor binding domain

Selected neutralizing antibody neutralizes SARS-CoV-2 emerging variants

plates, and unbound antibodies were washed away. NK cells, purified from healthy blood donor 1010 leukopaks using commercially available negative selection kits (StemCell EasySep Human NK Cell 1011Isolation Kit) were added, and the levels of IFN-γ was measured after 5 hours using flow cytometry. 1012The data is reported as the percent of cells positive for IFN-γ. Each sample is tested with at least 1013 two different NK cell donors, with all samples tested with each donor. The monoclonal antibodies 1014were tested for ADNKA activity at a range of 20 µg/mL to 9.1449 ng/mL. Na acetate pH 5.0 at the concentration of 25 µg/mL was injected for 360 sec over the dextran 1020 matrix, which had been previously activated with a mixture of 0.1M 1-ethyl-3(3-1021 dimethylaminopropyl)-carbodiimide (EDC) and 0.4 M N-hydroxyl succinimide (NHS) for 420 sec. 1022After injection of the antibody, Ethanolamine 1M was injected to neutralize activated group. 10 1023 µL/min flow rate was used during the whole procedure. Anti-SPIKE protein human mAbs were 1024 diluted in HBS-EP+ (Hepes 10 mM, NaCl 150 mM, EDTA 3.4 mM, 0.05% p20, pH 7.4) and injected 1025 for 120 sec at 10 µL/min flow rate over one of the two flow cells containing the immobilized Anti-1026Human IgG Antibody, while running buffer (HBS-EP+) was injected over the other flow cell to be 1027 taken as blank. Dilution of each mAb was adjusted in order to have comparable levels of RU for 1028 each capture mAb. Following the capture of each mAb by the immobilized anti-human IgG 1029 antibody, different concentrations of SPIKE protein (20 µg/mL, 10 µg/mL, 5 µg/mL, 2.5 µg/mL and 1 1030 µg/mL in HBS-EP+) were injected over both the blank flow cell and the flow cell containing the 1031 captured mAb for 180 sec at a flow rate of 80 µL/min. Dissociation was followed for 800 sec, 1032 regeneration was achieved with a pulse (60 sec) of Glycine pH 1.5. Kinetic rates and affinity 1033 constant of SPIKE protein binding to each mAb were calculated applying a 1:1 binding as fitting 1034 model using the Bia T200 evaluation software 3. The NOVA Lite HEp-2 ANA Kit (Inova Diagnostics) was used in accordance to the manufacturer's 1038 instructions to test antibodies the autoreactivity of selected antibodies which were tested at a 1039 concentration of 100 µg/mL. Kit positive and negative controls were used at three different dilutions 1040(1:1, 1:10 and 1:100). Images were acquired using a DMI3000 B microscope (Leica) and an Lung tissues were homogenized in 1 mL of DMEM containing 1% fetal bovine serum (FBS) and 1% 1100 penicillin/streptomycin. The lung homogenate supernatant was diluted 10-fold (10 0 to 10 6 ) and 1101 used to determine median tissue culture infection dose (TCID 50 ) in Vero E6 cells as previously 1102described (Jang and Ross, 2020) . 1103

Human IgG detection in hamster sera Human IgG detection in hamster sera Human IgG detection in hamster sera Human IgG detection in hamster sera 1105 ELISA assay was used to detect the human IgG J08-MUT in hamster sera. 384-well plates (384 1106 well plates, Microplate Clear; Greiner Bio-one) were coated with 2 µg/mL of unlabled goat anti-1107 human IgG (SouthernBiotech) diluted in sterile PBS and incubated at 4°C overnight. 50 µL/well of 1108 saturation buffer (PBS/BSA 1%) was used to saturate unspecific binding and plates were incubated 1109 at 37°C for 1h without CO 2 . Hamster sera were diluted in PBS/BSA 1%/Tween20 0.05% at a 1110 starting dilution of 1:10. Fourteen reciprocal dilutions were performed. Alkaline phosphatase-1111 conjugated goat anti-human IgG (Sigma-Aldrich) was used as secondary antibody and pNPP (p-1112 nitrophenyl phosphate) (Sigma-Aldrich) was used as soluble substrate. Wells were washed three 1113 times between each step with PBS/BSA 1%/Tween20 0.05%. The final reaction was measured by 1114 using the Varioskan Lux Reader (Thermo Fisher Scientific) at a wavelength of 405 nm. Samples 1115 were considered as positive if OD at 405 nm (OD 405 ) was twice the blank. 1116 1117