key: cord-0970609-661f4ktg
authors: Deshpande, Ashlesha; Harris, Bethany D.; Martinez-Sobrido, Luis; Kobie, James J.; Walter, Mark R.
title: Epitope classification and RBD binding properties of neutralizing antibodies against SARS-CoV-2 variants of concern
date: 2021-04-13
journal: bioRxiv
DOI: 10.1101/2021.04.13.439681
sha: a60ffefc1a0bee9edef7e548e10942b45cbdae5f
doc_id: 970609
cord_uid: 661f4ktg

Severe acute respiratory syndrome coronavirus-2 (SAR-CoV-2) causes coronavirus disease 2019 (COVID19) that is responsible for short and long-term disease, as well as death, in susceptible hosts. The receptor binding domain (RBD) of the SARS-CoV-2 Spike (S) protein binds to cell surface angiotensin converting enzyme type-II (ACE2) to initiate viral attachment and ultimately viral pathogenesis. The SARS-CoV-2 S RBD is a major target of neutralizing antibodies (NAbs) that block RBD - ACE2 interactions. In this report, NAb-RBD binding epitopes in the protein databank were classified as C1, C1D, C2, C3, or C4, using a RBD binding profile (BP), based on NAb-specific RBD buried surface area and used to predict the binding epitopes of a series of uncharacterized NAbs. Naturally occurring SARS-CoV-2 RBD sequence variation was also quantified to predict NAb binding sensitivities to the RBD-variants. NAb and ACE2 binding studies confirmed the NAb classifications and determined whether the RBD variants enhanced ACE2 binding to promote viral infectivity, and/or disrupted NAb binding to evade the host immune response. Of 9 single RBD mutants evaluated, K417T, E484K, and N501Y disrupted binding of 65% of the NAbs evaluated, consistent with the assignment of the SARS-CoV-2 P.1 Japan/Brazil strain as a variant of concern (VoC). RBD variants E484K and N501Y exhibited ACE2 binding equivalent to a Wuhan-1 reference SARS-CoV-2 RBD. While slightly less disruptive to NAb binding, L452R enhanced ACE2 binding affinity. Thus, the L452R mutant, associated with the SARS-CoV-2 California VoC (B.1.427/B.1.429-California), has evolved to enhance ACE2 binding, while simultaneously disrupting C1 and C2 NAb classes. The analysis also identified a non-overlapping antibody pair (1213H7 and 1215D1) that bound to all SARS-CoV-2 RBD variants evaluated, representing an excellent therapeutic option for treatment of SARS-CoV-2 WT and VoC strains.

side of RBD that forms the ACE2 binding site (e.g. C1 NAbs), to the surface opposite the ACE2 269 binding site, where C2 NAbs bind without hindrance to the RBD in either up or down 270 conformations. Thus, the accessibility of the C2 NAb epitopes are greater than for C1 and C1D 271

NAbs. As previously reported, the orientations of several C2 NAbs, including Ab2-4, C121, and 272 C144, shows they can block adjacent RBDs in the pre-fusion trimer from adopting the "up" 273 ACE2-binding-competent conformations (9, 27) . 274

The C3 NAb (n=4) binding epitope is localized to the RBD knob and adjacent residues 333-347. 276 C3 NAbs adopt distinct orientations, relative to the C1-C1D-C2 NAbs (Fig. 4) . A unique feature 277 of C3 NAb epitope is its location near RBD-RBD trimer interfaces. As a result, some C3 NAbs 278 bind across adjacent RBDs in the trimeric SARS-CoV-2 S and sterically block the RBD up 279 conformation of adjacent RBDs. Some C3 NAbs only block the up conformation of the RBD, 280 while others sterically block ACE2 binding, and some have both neutralization mechanisms. 281

Like C2 NAbs, C3 NAb epitopes are accessible whether SARS-CoV-2 S RBD is in the up or 282 down conformation. C4 NAbs (n= 6) bind to an RBD surface comprising a helix-strand-helix 283 motif (residues 366-389) that is buried in the trimeric S, even when the RBD is in the up 284 conformation. Thus, C4 NAbs, required SARS-CoV-2 S to undergo significant conformational 285 changes, beyond RBD up-down movements for binding (11) . For many C4 NAbs, their epitope 286 and orientation prevent them from sterically blocking ACE2 binding. This may explain why 287 many C4 NAbs, such as CR3022, do not potently neutralize SARS-CoV-2 compared to NAbs in 288 the other classes. However, this is not universal, as the orientations C4 NAbs COVA1-16 and 289 H014 allow them to block ACE2 binding and they are reported to potently neutralize SARS-CoV-290 2 in cell culture assays and in animal models (28, 29) . 291 292

We previously isolated a series of potent SARS-CoV-2 NAbs from a single convalescent patient 295 and characterized their efficacy in vitro, and for one NAb (1212C2), its efficacy in golden Syrian 296 hamsters, delivered by inhalation (16). The classified NAb/RBD structures (Table 1, Fig. 4 ) 297 were used to predict the binding epitopes of these NAbs and the resulting assignments were 298 experimentally tested using RBD variant binding and epitope mapping studies. Thus, unknown 299

NAb heavy chains were aligned to the NAb database sequences classified as C1, C1D, C2, C3, 300 or C4 (Figs. 3 and 4) . Of 20 NAbs evaluated, 9 (45%) were classified as C1, 10 (50%) were 301 classified as C2, and one (5%) was classified as C4. The analysis suggested that none of the unknown NAbs were C3 NAbs, and one of the 20 NAbs (1213H7) was misclassified by 303 sequence analysis as a C2 NAb, but subsequently reassigned to the C1D classification based 304 on epitope mapping and RBD variant binding analysis (Figs. 6 and 7) . Alignments performed 305 with the light chains did not improve the unknown NAb assignments. 306 307

The predicted impact of RBD variant residues (Fig. 2) on NAb binding was evaluated using the 310 BPs, derived from the NAb/RBD complex structures ( Table 1) . Variable RBD residues (freq. 311 >0.01) found in at least 75% of the epitopes for each NAb class were identified ( Table 2) . This 312 analysis identified 11 residues exhibiting increased variability in C1 and C1D NAb epitopes, 7 313 residues in the C2 epitope, and 2 in C3 NAb epitope. Variable residues that consistently bury 314 surface area in at the C1 NAb class include K417, Y453, L455, A475, F486, and N501. Thus, 315 C1 NAb epitopes are predicted to be sensitive to K417N/T, Y453F, L455F, A475V, F486L, and 316 N501Y variants. Consistent with the structural relationships between C1, C1D, and C2 NAb 317 epitopes (Fig. 4) , conserved residues within the C1D epitope include residues shared with C1 318

NAbs (e.g. L455, A475, and F486), as well unique residues E484, and G485. Thus, RBD 319 variants that could disrupt C1D NAbs include L455F, A475V, F486L, as well as E484K, and 320 G485R. Contacts conserved across C2 NAbs include variable RBD residues G446, L452, 321 V483, E484, F490, and S494. This is consistent with many C2 NAbs being sensitive to the 322 E484K variant (9) and Fig. 5 ). Additional conserved RBD contacts within the C2 NAb epitope 323 suggests RBD variants G446V, L452R, and S494P can also impact C2 interactions. Based on 324

only four examples, the C3 NAb epitope is diverse and the conserved C3 contact region exhibits 325 limited RBD residue variation. Only two RBD residues in the C3 epitope (N440 and K444) 326 exhibit residue variability greater than 0.01%. Five additional RBD amino acids are contacted 327 by 75% of the C3 NAbs, but these amino acid sidechains are highly conserved in the SARS-328

CoV-2 RBD sequences evaluated. In contrast, amino acid variation within the C1, C1D, and C2 329 epitopes is much greater (maximal variation 0.12%-2.5%) and extends over more residues. 330

Finally, there are several common contact residues in the C4 NAb epitope, suggesting that 331 N370S, T376I, V382L, P384L, T385I, and R408I could disrupt C4 NAbs. 

Based on the in silico NAb epitope analysis, nine RBD variants that localize to the C1 (K417T, 338 A475V, N501Y, Y505W), C1D/ C2 (G446V, L452R and E484K) and C3 epitopes (T345I, 339 N440Y, and G446V) were produced for binding studies with the unknown NAbs. C4 NAbs were 340 excluded from the analysis. Our structural analysis, and recent literature (9), confirm that the 341 loss of NAb binding to certain SARS-CoV-2 RBD variants can be used to identify NAb classes. 342

Furthermore, the assays can be used to identify NAbs that are not sensitive to RBD variants 343 and thus potentially useful as therapeutics. To address these questions, the unknown NAbs 344 and three controls (CB6 (30), RGN10933 and RGN10987(4)) were tested for their ability to bind 345 to the SARS-CoV-2 RBD variants (Fig. 5) . NAb-RBD interactions were defined as "significantly 346 disrupted" if NAb binding to an RBD variant was reduced by 50% or more, relative to the 347 reference WT RBD. Based on this definition, C1 RBD variants K417T (25%) and N501Y (25%), 348 and C2 variant E484K (30%), most frequently disrupted NAb binding (Fig. 5) . A second group 349 of RBD variants, L452R (15%), G446V (10%), and N440Y (5%) disrupted fewer NAb 350 interactions, while all NAbs bound efficiently (>50%, relative to WT reference) to RBD variants 351 T345I, A475V, and Y505W. Efficient binding of all NAbs to T345I (freq. = 0.0003%) and Y505W 352 (freq. = 0.004%) is consistent with their high conservation within the SARS-CoV-2 RBD (Table  353 2). Furthermore, for those NAbs exhibiting >50% reduction in binding levels, relative to WT 354 RBD, E484K and N501Y exclusively disrupted NAbs classified as C2 and C1, respectively (Fig.  355   5c) . 356 357

In addition to disrupting NAb binding, SARS-CoV-2 RBD mutations can impact RBD-ACE2 360 binding affinity, which could impact virus infectivity. Thus, ACE2 binding affinity was measured 361

for each RBD variant and the reference Wuhan-1 RBD protein (Fig. 6) . The two RBD variants 362 that disrupted the largest number of NAb interactions (E484K, and N501Y) bound to ACE2 within 15% (e.g. error of the measurements) of the WT RBD-ACE2 interaction. In contrast, 364

K417T, which disrupts a salt bridge with ACE2 ( Fig. 2b) , exhibited two-fold lower binding affinity 365 for ACE2, relative to WT RBD (Fig. 6b) . This data suggests 2 of the 3 most disruptive RBD 366 mutations are NAb evading mutations that do not alter interactions with the viral receptor, ACE2. 367 T345I and N440Y also exhibited WT RBD-ACE2 binding affinity, while G446V and A475V 368 exhibited disrupted ACE2 binding affinities, like K417T. Two RBD mutations (L452R and 369 Y505W) exhibited higher affinity for ACE2 than WT RBD. Thus, L452R disrupts several NAbs, 370 but also binds to ACE2 with higher affinity. Thus, the L452R mutant, associated with a recently 12 NAbs were subjected to epitope mapping analysis (Fig. 7) . Overall, there was excellent 381 correlation between the NAb assignments ( Table 1 ) and the epitope analysis. For example, C1 382 vs. C1 NAbs, or other "like-like" classifications prevented binding of the second NAb. In 383 contrast, some C1 NAbs allowed simultaneous binding of a second C2 NAb, but this was not 384 universal. For example, C1 NAbs 1212D5, 1212F2, 1213F2, 1212D4, and 1212C8, allowed 385 simultaneous binding of C2 NAbs 1207B4, 1215D1, and 1207F10 (Fig. 7) . In contrast, the C2 386

NAb 1212C2 blocked all C1 and C2 NAbs that were assayed (Fig. 7a) . To address these 387 observations, structures of the NAb-RBD complexes assigned to the different C2 NAbs were 388 compared (Fig. 7d) . Specifically, 1215D1, 1207B4, and 1207F10 NAbs were assigned to the 389 C386-2/RBD complex, while 1212C2 was assigned to the C121/RBD complex. As shown in 390 As a second method of epitope analysis, the NAbs were tested for their ability to bind to an 397 RBD-ACE2 fusion protein (FP). The RBD-ACE2FP was designed to provide a stable RBD-398 ACE2 complex to characterize the overlap of NAb epitopes with the RBD-ACE2 binding site. As 399 expected, most NAbs bound very poorly to the RBD-ACE2FP, consistent with their epitope 400 being blocked by ACE2 (Fig. 8) . However, three C2-classified NAbs (1215D1, 1207B4, and 401 1207F10) exhibited high levels of binding to the RBD-ACE2FP (Fig. 8) . Based on the location 402 of the RBD ACE2 binding site (Fig. 4) , this data further validates the assignment of these NAbs 403 to the C2 class (Fig. 7d) . It also suggests that this subclass of C2 NAbs may disrupt RBD-404 ACE2 interactions even after the complex is formed, presumably by a conformational change, 405 providing a novel mechanism of neutralization for this group of C2 NAbs. The goal of this study was to use SARS-CoV-2 NAb/RBD structural information found in the 411 protein databank and SARS-CoV-2 S RBD sequence variation data from the GSIAD, to 412 characterize a series of NAbs for potential use as therapeutics against SARS-CoV-2. Optimal 413

NAbs against SARS-CoV-2 have always required that they exhibit high potency neutralizing 414 capabilities, but as the pandemic has continued, they now must also exhibit broad specificity Ab-mediated neutralization and/or exhibit increased disease severity and transmissibility 424 (cdc.gov). To further evaluate these single mutations at the molecular level, we tested a panel 425

of SARS-CoV-2 NAbs, isolated from a convalescent patient, for their ability to bind to a series of 426 SARS-CoV-2 RBD variants. The analysis found that E484K is the most disruptive mutation and 427

impacted the greatest number of NAbs tested in our study (30%), with N501Y and K417T being slightly less disruptive, but still each impacting 25% of the NAbs evaluated. Combined, the 3 429 SARS-CoV-2 RBD variants (K417T, E484K, and N501Y) present in the P. providing advantages in immune evasion against NAbs, and increased cell infectivity via 454 enhanced ACE2 binding affinity, ultimately leading to increased disease severity. A weakness 455 in our studies is that our RBD-ACE2 binding analysis studies are based on a 1:1 monomeric 456 ACE2-RBD interaction. It is possible that the conformation and function of the RBD variants are 457 distinct in the context of the trimeric SARS-CoV-2 S. Recent data suggests that the D614G 458 mutation stabilizes the S trimer and is necessary to observe the functional impact of the N501Y 459 and L452R mutations on cell infectivity (31,32). It is also possible that large changes in RBD-460 ACE2 affinity, higher or lower, disrupts optimal viral infectivity, while more subtle changes 461 around the reference WT RBD-ACE2 binding affinity can cause large changes in cell infectivity. 462

From an overall analysis of SARS-CoV-2 anti-RBD NAbs, C3 NAbs appear to have an 464 advantage by targeting an RBD epitope that has undergone less variation than NAb epitopes 465 associated with C1, C1D, and C2 NAbs. However, variability in CDR length, structure, and 466 amino acid composition has thus far been sufficient for at least some C1, C1D, and C2 NAbs to 467 maintain their neutralization efficacy, despite RBD variability. Furthermore, we find the C3 468 RGN10987 NAb is highly sensitive to the G446V variant, suggesting the virus is able to identify 469 residues to disrupt NAbs in all NAb classes. For the unknown NAbs evaluated in this report, the 470 NAb/RBD database, RBD variant binding studies, and epitope mapping allowed their 471 assignment into specific NAb classes. Thus, our study provides a strategy for characterizing 472 NAb/RBD epitopes at a larger scale to further define NAb sensitivity to RBD variants in 473 individuals recovered from natural infections and changes in the anti-RBD SARS-CoV-2 474 structural repertoire induced by vaccination. Finally, our analysis identified a non-overlapping 475 antibody pair, 1213H7/1215D1, which both bind to all RBD variants found in current SARS-CoV-476 2 VoC, suggesting they represent a potential therapeutic treatment for SARS-CoV-2 WT and 477 newly identified VoC. hydrogen bonds = 2, monovalent sidechain hydrogen bonds =1, and mainchains hydrogen 543 bonds are denoted by "mc". Frequency that variant RBD amino acids (variant) observed in 544 SARS-CoV-2 GISAID SARS-CoV-2 S sequence database, relative to the Wuhan-1 reference 545 sequence. The color code (CC) for Figs. 2b, 2c, and 2d is also shown. 546 547 Fig. 3 . Binding Profile (BP) correlations to define RBD-binding NAb classes. 548

BPs were defined for NAb/RBD complexes ( Table 1 ) that were extracted from the pdb (n=45). 549

For some structures, more than one NAb-RBD complex was evaluated leading to the A, B, C 550 designations for some NAbs. Buried surface area was not calculated for some complexes, due 551 to almost complete lack of sidechains for the NAbs. The BPs were subjected to hierarchical 552 clustering and then similarity analysis (correlation coefficient calculation) to define five NAb 553 classes C1 (n=17), C1D (n=5), C2 (n=12), C3 (n=4), and C4 (n=6). Sensitivity of NAb classes to the different RBD variants. The number of NAbs that display less 571 than 50% of WT binding in Fig. 5a (red and yellow boxes) were counted (y axis) and plotted 572 according to their NAb classification (Table 1) (Table 1) . Figures 7a, 7b , 579 and 7c, represent 3 distinct mapping experiments performed with common and distinct NAbs. 580 (d) Graphical summary, showing NAb/RBD complexes that were assigned to the unknown 581

NAbs and used to explain the epitope mapping data. P2C-1F11/RBD was used to model C1 582

NAbs, C2 NAbs were modeled by C368-2/RBD (1215D1, 1207B4 and 1207F10) and C121/RBD 583 (1212C2) structures, while the C4 NAb, 1215D11, was modeled by the RBD/CR3022 structure. 584

The assigned structures explain how 1212C2 epitope blocks all other NAbs from binding to 585 RBD, while 1212D4 (or 1213H7) and 1215D1 bind to non-overlapping RBD epitopes and also 586 bind to all RBD variants they were tested against (Fig. 5) . Two NAbs evaluated in the mapping 587 data (1213H7 and ctl RGN10933) are not listed on the graphical summary. 1213H7 and 588 RGN10933 were both assigned as C1D NAbs, which is consistent with their common epitope 589 mapping profiles, and their ability to bind simultaneously with 1215D1. captured on anti-IgG biacore chips and sensorgrams were recorded for an RBD-ACE2 fusion 593 protein (20nM) injected over the surface of each NAb. Variability in fusion protein binding levels 594

Detection of SARS-CoV-2-Specific Humoral and 600 Cellular Immunity in COVID-19 Convalescent Individuals

Heterogeneous antibodies against SARS-CoV-2 spike 605 receptor binding domain and nucleocapsid with implications for COVID-19 immunity

Safety and Immunogenicity of Two RNA-Based Covid-19 Vaccine Candidates. The New 612 England journal of medicine

Studies in humanized mice and convalescent humans yield a SARS-CoV-2 antibody 621 cocktail

Isolation of potent SARS-CoV-2 neutralizing antibodies and protection from disease in a 628 small animal model

REGN-COV2 antibody cocktail prevents and treats SARS-CoV-2 infection 635 in rhesus macaques and hamsters. bioRxiv

Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Cryo-EM structure of the 2019-nCoV spike in the 641 prefusion conformation

SARS-CoV-2 645 neutralizing antibody structures inform therapeutic strategies

A highly conserved cryptic epitope in the receptor binding domains of SARS-648

653 (2020) Neutralization of SARS-CoV-2 by Destruction of the Prefusion Spike

2021) mRNA vaccine-661 elicited antibodies to SARS-CoV-2 and circulating variants

CoV-2 spike mutations that attenuate monoclonal and serum antibody neutralization

Antibody cocktail to SARS-CoV-2 spike 670 protein prevents rapid mutational escape seen with individual antibodies

Complete Mapping of Mutations to the SARS-CoV-2 676

Spike Receptor-Binding Domain that Escape Antibody Recognition

Therapeutic activity of an inhaled potent SARS-CoV-2 neutralizing human monoclonal 682 antibody in hamsters. bioRxiv

2017) Data, disease and diplomacy: GISAID's 684 innovative contribution to global health

Inference of macromolecular assemblies from 686 crystalline state

The PyMOL Molecular Graphics System DeLano Scientific

The EMBL-EBI search 691 and sequence analysis tools APIs in 2019

A new coronavirus associated with 695 human respiratory disease in China

Kinetic analysis of 697 cytokine-mediated receptor assembly using engineered FC heterodimers

Structure of the SARS-CoV-2 spike receptor-binding domain 701 bound to the ACE2 receptor

Structure of SARS coronavirus 703 spike receptor-binding domain complexed with receptor

Recognition of the SARS-CoV-2 706 receptor binding domain by neutralizing antibodies

715 (2020) A Therapeutic Non-self-reactive SARS-CoV-2 Antibody Protects from Lung 716 Pathology in a COVID-19 Hamster Model

Potent neutralizing antibodies against multiple epitopes on SARS-CoV-2 721 spike

Structural basis for neutralization of 725 SARS-CoV-2 and SARS-CoV by a potent therapeutic antibody

Cross-Neutralization of a SARS-CoV-2 Antibody to a 730 Functionally Conserved Site Is Mediated by Avidity

A 734 human neutralizing antibody targets the receptor-binding site of SARS-CoV-2

2021) Transmission, infectivity, and 743 antibody neutralization of an emerging SARS-CoV-2 variant in California carrying a 744 L452R spike protein mutation. medRxiv : the preprint server for health sciences

The D614G mutation in the SARS-CoV-2 spike protein reduces S1 747 shedding and increases infectivity. bioRxiv

Structural basis of a public antibody response to SARS-CoV-2

Structural basis for potent neutralization of SARS-CoV-2 and role of 753 antibody affinity maturation

A noncompeting pair of human 757 neutralizing antibodies block COVID-19 virus binding to its receptor ACE2

Human neutralizing antibodies 762 elicited by SARS-CoV-2 infection

765 (2020) An alternative binding mode of IGHV3-53 antibodies to the SARS-CoV-2 receptor 766 binding domain

Molecular basis for a germline-biased 769 neutralizing antibody response to SARS-CoV-2

Structurally Resolved 774 SARS-CoV-2 Antibody Shows High Efficacy in Severely Infected Hamsters and Provides 775 a Potent Cocktail Pairing Strategy

Mapping Neutralizing 784 and Immunodominant Sites on the SARS-CoV-2 Spike Receptor-Binding Domain by 785 Structure-Guided High-Resolution Serology

Potent Neutralizing Antibodies against 790 SARS-CoV-2 Identified by High-Throughput Single-Cell Sequencing of Convalescent 791 Patients' B Cells

Cross-796 neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV antibody

Structural basis for the neutralization of SARS-805

CoV-2 by an antibody from a convalescent patient