key: cord-0835635-b36vq2pj
authors: Clark, Sarah A.; Clark, Lars E.; Pan, Junhua; Coscia, Adrian; McKay, Lindsay G.A.; Shankar, Sundaresh; Johnson, Rebecca I.; Brusic, Vesna; Choudhary, Manish C.; Regan, James; Li, Jonathan Z.; Griffiths, Anthony; Abraham, Jonathan
title: SARS-CoV-2 evolution in an immunocompromised host reveals shared neutralization escape mechanisms
date: 2021-03-16
journal: Cell
DOI: 10.1016/j.cell.2021.03.027
sha: 6459022b7023d5587a1263cd2b65a217865b82c0
doc_id: 835635
cord_uid: b36vq2pj

Many individuals mount nearly identical antibody responses to SARS-CoV-2. To gain insight into how the viral spike (S) protein receptor-binding domain (RBD) might evolve in response to common antibody responses, we studied mutations occurring during virus evolution in a persistently infected immunocompromised individual. We use antibody Fab/RBD structures to predict, and pseudotypes to confirm, that mutations found in late-stage evolved S variants confer resistance to a common class of SARS-CoV-2 neutralizing antibodies we isolated from a healthy COVID-19 convalescent donor. Resistance extends to the polyclonal serum immunoglobulins of four out of four healthy convalescent donors we tested and to monoclonal antibodies in clinical use. We further show that affinity maturation is unimportant for wildtype virus neutralization but is critical to neutralization breadth. As the mutations we studied foreshadowed emerging variants that are now circulating across the globe, our results have implications to the long-term efficacy of S-directed countermeasures.

SARS-CoV-2 has infected over 110 million individuals worldwide, resulting in over 2.4 78 million deaths to date. The SARS-CoV-2 spike protein (S) is a central target for vaccine and 79 drug design efforts (Abraham, 2020; Krammer, 2020) . S is heavily glycosylated and forms 80 trimers of heterodimers on the virion surface. Each S protomer has two functional subunits; S1, 81 which contains a receptor-binding domain (RBD) that binds the cellular receptor, ACE2 82 Coronaviruses encode a viral exonuclease that increases replication fidelity (Denison et 93 al., 2011) , which probably makes antigenic drift in SARS-CoV-2 less significant than in other 94 enveloped RNA viruses. Changes in SARS-CoV-2 S have nonetheless occurred over time and 95 become fixed among circulating variants; the D614G S mutation is a prime example 96 (Yurkovetskiy et al., 2020) . This mutation, however, does not seem to impact the activity of 97 RBD-targeting neutralizing antibodies (Yurkovetskiy et al., 2020) . Ultimately, evolution of S 98 antibody escape mutations could impact the long-term effectiveness of vaccines and 99 monoclonal antibody-based therapeutics that target S. 100

In our efforts to study SARS-CoV-2 antibody neutralization and to predict escape 101 mutations, we examined sequences of S variants that evolved in a persistently infected 102 J o u r n a l P r e -p r o o f

To study neutralizing antibody responses to SARS-CoV-2, we obtained a peripheral 118 blood sample from a healthy adult male individual ("C1") who had been infected by 119 SARS-CoV-2 five weeks prior to sampling. Polyclonal IgG purified from the blood of this 120 individual neutralized SARS-CoV-2 lentivirus pseudotype ( Figure S1A ). We generated a soluble 121 SARS-CoV-2 S construct that is stabilized through mutations and the addition of a trimerization 122 tag to remain in the S "pre-fusion" conformation ("S2P") (Wrapp et al., 2020) and used it as an 123 antigen to isolate 116 memory B cells (CD19 + , IgG + ) by FACS ( Figure S1B ). We could produce 124 48 recombinant monoclonal antibodies in sufficient amount for further characterization; forty-125 three of these bound S2P by ELISA, and 18 also bound the RBD ( Figure S1C and Table S1 ). 126

Most antibodies were derived from the V H 3 heavy chain subgroup and had kappa light chains 127 ( Figure S1D ). Antibody CDR H3 and CDR L3 loops had an average length of 15 and 9 amino 128 al., 2020). The Q493K RBD change or a similar mutation at the same position (Q493R RBD ) have 233 been recently described in other human-derived SARS-CoV-2 sequences ( Figure 3B and Table  234 S3). To determine the role of the Q493K/R RBD mutations in resistance to C1A-V H 3-53 235 antibodies, we generated pseudotypes containing either mutation in addition to the D614G S 236 change. We also included an N439K RBD variant, a recently described antibody neutralization 237 escape mutant (Thomson et al., 2021) . The Q493K RBD mutation caused substantial resistance to 238 the C1A-V H 3-53 antibodies that bind the most weakly to the RBD ( Figures 5A and 5B) . We 239 observed similar findings with the Q493R RBD pseudotypes, although the decrease in 240 neutralization sensitivity was more severe. The only exception was C1A-gl, which neutralized 241 Q493K/R RBD pseudotypes better that C1A-gl*, likely because a serine instead of an arginine at 242 CDR H3 position 100a would better accommodate these RBD mutations ( Figure 4A ). The 243 N439K RBD mutation had no effect on pseudotype neutralization by C1A-V H 3-53 antibodies 244 ( Figure 5A ), which was expected because this mutation falls outside of the V H 3-53 antibody 245 epitope on the RBD. 246

To determine if resistance extends to V H 3-53-derived antibodies isolated from different 247 COVID-19 convalescent donors, we also tested antibodies B38 The monoclonal antibody cocktail REGN-COV2 comprises two antibodies that bind 255 non-overlapping sites on the RBD to suppress the emergence of antibody neutralization escape 256 mutations (Baum et al., 2020; Hansen et al., 2020) . REGN10933 binds a region of the RBD that 257 J o u r n a l P r e -p r o o f overlaps significantly with the ACE2-binding site, while REGN10987 binds a region that has little 258 to no overlap ( Figure 5D ). Of the S mutations that evolved during persistent SARS-CoV-2 259 infection in the immunocompromised individual, the Q493K RBD change, found in day 146 260 sequencing, was previously detected in tissue cell culture passaging experiments using 261 REGN10933 and rVSV-S (Baum et al., 2020) . In our experiments, the Q493K RBD mutation 262 decreased REGN10933 pseudotype neutralization potency by fifteenfold ( Figures 5C and S2D The N440D RBD mutation, which was only detected on day 146 sequencing ( Figure 3B ), 268 falls on the REGN10987 RBD-binding site. It is adjacent to a N439K RBD mutation that is found in 269 circulating variants with reported REGN10987 resistance (Thomson et al., 2021) ( Figure 5D ). 270

The day 146* variant had a fourfold decrease in REGN10987 neutralization sensitivity, while the 271 N439K RBD mutation caused fourteenfold decrease in sensitivity ( Figures 5C and S2D) . 272 REGN10987, therefore, is the only antibody we tested that had demonstrable activity against 273 the day 146* S pseudotypes, but a single substitution at an adjacent position (N439K RBD ) found 274 in circulating variants (Thomson et al., 2021) All of the potent SARS-CoV-2 neutralizing antibodies we isolated from the healthy 278 COVID-19 convalescent donor (C1) were clonotypes of a single V H 3-53/V K 1-9 antibody, 279 suggesting that this individual's memory B cell response was narrowly focused on this class of 280 neutralizing antibodies ( Figure S1 ; Table S1 ). While purified C1 IgG could neutralize WT 281 (D614G S ) pseudotypes, day 146* and 152* S pseudotypes were resistant to C1 serum IgG 282 neutralization ( Figures 6A and 6B ). The Q493K/R RBD mutations also conferred near complete 283 J o u r n a l P r e -p r o o f escape mutation that falls outside of the RBD epitope for V H 3-53-derived neutralizing antibodies, 285 had no effect on C1 polyclonal IgG neutralization ( Figures 6A and 6B ). To determine whether 286 our findings extended to other COVID-19 convalescent donors that may have less epitope 287 biased antibody responses, we performed similar experiments with purified IgG from three 288 additional donors ("C2", "C3", and "C4"). The neutralizing activity of purified IgG from these 289 donors was mostly unaffected by the single mutations (Q493K/R RBD or N439K RBD ), but day 146* 290 and 152* S pseudotypes were resistant to neutralization ( Figures 6A and 6B) . 291 292 V H 3-53 antibody affinity maturation partially overcomes neutralization escape 293

Although the benefits of antibody RBD affinity are limited in SARS-CoV-2 neutralization 294 ( Figure 2F ), affinity gains, in principle, could compensate for losses of contacts or potential 295 clashes that are caused by escape mutations. Indeed, the highest affinity binding antibodies 296 were seemingly the least impacted by neutralization escape mutations (Figures 5A, 5B, and 297 S2C). We selected C1A-B12, our most potent neutralizing antibody against infectious 298 SARS-CoV-2 (Table 1) , to directly test whether additional affinity enhancing mutations could 299 overcome neutralization escape. To generate affinity enhanced versions of C1A-B12, we 300 introduced into its sequence somatic hypermutation changes found in other V H 3-53/3-66 301 antibodies, including antibodies described elsewhere (Hurlburt et bound to the RBD with a six-to-ten-fold increase in affinity as compared to the parent C1A-B12 304 Fab (Table 1; Figure S3 ). Affinity enhanced variants potently neutralized D614G S pseudotypes 305 and infectious SARS-CoV-2 ( Figures 6A and 6B ). The Q493K/R RBD mutations, however, had less of an effect on the serum 319 IgG of three additional COVID-19 convalescent donors ( Figures 6A and 6B) , suggesting that the 320 C1 donor may be a rare example of an overly focused antibody response. 321

The immunocompromised individual we studied received REGN-COV2 (REGN10933 322 and REGN10987) on day 145 of their illness, so the S mutations detected on days 146 and 152 323 could have been influenced by selective pressure from this therapeutic antibody cocktail, as 324 described in a recent report (Starr et al., 2021) . Nonetheless, several of the RBD mutations we 325 studied were detected well before day 146 ( Figure 3B Figure 3B ; Table S3 ). The only unique mutation is F486I RBD , 331 although a nearly identical mutation, F486L RBD , has been observed in another human derived 332 SARS-CoV-2 S sequence (Table S3) . Importantly, Q493K RBD /N501Y RBD , Q493R RBD /N501Y RBD , 333 albeit with very low frequency for the time being ( Figure 3B and Table S3 ). Additional mutations 335 that we did not study directly but that could also substantially impact neutralization by 336 The S mutations we studied are only from one immunocompromised individual (n = 1) 409 (Choi et al., 2020). While some of the S mutations we mention in our work are also now found in 410 other human-derived SARS-CoV-2 sequences available in public data bases (e.g., B.1.1.7 411 variants containing the Q493K/R RBD mutations; see Table S3 ), these variants for the time being 412 are rare and the context in which they arose is also not defined based on public information 413 (e.g., whether they occurred in a healthy person or an immunocompromised individual, or 414 whether the individual received treatment with convalescent plasma or therapeutic antibodies 415 prior to sampling, etc.). While we used infectious SARS-CoV-2 in some assays, we used 416 replication defective pseudotyped lentiviruses as a surrogate system for studying the effects of 417 S mutations. The replication fitness of infectious SARS-CoV-2 carrying the S mutations we 418 studied remains to be determined. We used multiple V H 3-53-derived monoclonal antibodies 419 isolated from a healthy donor (n=1), and single antibodies from two additional donors (B38 from 420 (Wu et al., 2020) and CC12.1 (Rogers et al., 2020)), but did not test all V H 3-53-derived 421 monoclonal antibodies identified to date. Other V H 3-53-derived antibodies may be differently 422 impacted by specific mutations because of differences in their light chain genes and CDR H3 423 loops. Lastly, while we predicted that residue R100a VH was a serine in the germline C1A-V H 3-53 424 antibody based on our analysis using the IMGT/V-QUEST database (Brochet et al., 2008) , this 425 database is likely missing alleles. To prove that our D gene assignment was accurate, we would 426 have had to sequence D gene segments in the PBMC donor C1, and we did not perform this 427 analysis. There is, therefore, the possibility that an arginine or lysine would be found at position 428 100a VH in a germline C1A-V H 3-53 antibody. We worked to ensure gender balance in the recruitment of human subjects. We worked to 464 ensure ethnic or other types of diversity in the recruitment of human subjects. We worked to 465 ensure that the study questionnaires were prepared in an inclusive way. One or more of the 466 authors of this paper self-identifies as an underrepresented ethnic minority in science. One or 467 more of the authors of this paper received support from a program designed to increase minority 468 representation in science. While citing references scientifically relevant for this work, we also 469 actively worked to promote gender balance in our reference list. The author list of this paper 470 includes contributors from the location where the research was conducted who participated in 471 the data collection, design, analysis, and/or interpretation of the work. Figure 4 . Unique reagents generated in this study are available from the Lead Contact upon request with 737 completed material transfer agreements (MTA). GISAID accession numbers for the sequences 738 analyzed in Figure 3 are provided in Figure S6 and Table S3 . substitutions at residues 986 and 987 (Wrapp et al., 2020) , and a C-terminal foldon trimerization 819 motif followed by a BirA ligase site, a Tobacco Etch Virus (TEV) protease site, a FLAG tag, and 820 a His 6 -tag into a pHLsec vector (Aricescu et al., 2006) , which contains its own secretion signal 821 sequence. We note that two N-terminal S residues (residues 14 and 15) downstream of the 822 native S signal peptide were inadvertently omitted from the S2P construct during subcloning. 823

We transfected Expi293F TM cells using an ExpiFectamine TM transfection kit (Thermo Fisher 824 Scientific Cat# A14525) according to the manufacturer's protocol. We purified the protein using 825 anti-FLAG M2 Affinity Gel (Sigma-Aldrich Cat# A2220; RRID: AB_10063035) according to 826 manufacturer's protocol and removed the FLAG tag and His 6 -tag with TEV digestion followed by 827 reverse nickel affinity purification and size-exclusion chromatography on a Superose 6 Increase 828 previously described (Clark et al., 2018) . 830

To obtain recombinant S2P for ELISAs, we used Ni Sepharose ® Excel (GE Healthcare Life 832

Sciences Cat# 17-3712-02) to purify His 6 -tagged SARS-CoV-2 S2P from the supernatant of 833

Expi293F cells stably expressing this protein (a gift of Bing Chen). We further purified the 834 protein using size exclusion chromatography on a Superpose 6 Increase column. His 6 -tag, a TEV protease site and a short linker (amino acids SGSG). For BLI-binding assays, 852 the construct includes an N-terminal His 6 -tag, followed by a TEV protease site, a BirA ligase 853 site, and a 7-residue linker (amino acids GTGSGTG). We produced protein for ELISA and BLI-854 transfect HEK293T cells grown in suspension and purified by nickel affinity purification. For BLI-856 binding assays the protein was digested with TEV protease to remove the His 6 -tag followed by 857 reverse nickel affinity purification. We biotinylated proteins with BirA ligase as previously 858 described (Mahmutovic et al., 2015) , followed by a reverse nickel affinity purification step to 859 remove BirA ligase, which contains a His 6 -tag and cannot be separated by size exclusion 860 chromatography from the SARS-CoV-2 RBD due to its similar size. For crystallography, we 861 produced RBDs by PEI MAX transfection of GnTI -/-HEK293S cells grown in suspension or 862 HEK293T cells grown in suspension and also in the presence of kifunensine (Sigma-Aldrich 863

Cat# K1140) at 5 µM, purified these by nickel affinity, and removed the His 6 -tag by TEV 864 digestion followed by reverse nickel affinity purification. As a final step, we used size exclusion 865 on a Superdex 200 Increase column, in which each recombinant RBD protein ran as a single 866 peak at the expected retention volume. 867

We subcloned the ectodomain of human ACE2 (GenBank ID: BAB40370.1) residues 18-740, 869 with cDNA obtained as a gift from Michael Farzan, with a C-terminal Fc tag into a pVRC8400 870 vector containing human IgG1 Fc (a gift from Aaron Schmidt). We expressed the protein in 871 Expi293F TM cells using an ExpiFectamine TM transfection kit according to the manufacturer's 872 protocol and purified the protein with MabSelect SuRE Resin using the manufacturer's protocol, 873 followed by size exclusion chromatography on a Superose 6 Increase column, with the protein 874 eluting at the expected retention volume. 875 876

We prepared each Fab:SARS-CoV-2 RBD complex by mixing RBD with 1.5 molar excess of 878

Fab. The mixtures were incubated at 4°C for 1 h prior to purification on a Superdex 200 879

Increase column in buffer containing 150mM NaCl, 25 mM Tris-HCl, pH 7.5. Each complex co-880 each complex to 13 mg ml -1 and screened for crystallization conditions in hanging drops 882 containing 0.1 µl of protein and 0.1 µl of mother liquor using a Mosquito protein crystallization 883 robot (SPT Labtech) with commercially available screens (Hampton Research) (see Key 884

Resources Table) . chains where density was apparent. During refinements, we updated TLS groups calculated 906

horseradish peroxidase (HRP)-coupled anti-human (Fc) antibody (Sigma-Aldrich Cat# A0170)

Biolayer interferometry assays

We performed BLI experiments with an Octet RED96e (Sartorius). For affinity measurements

SA) sensor (ForteBio) at 1.5 µg 991 ml -1 in kinetic buffer (PBS containing 0.02% Tween and 0.1% BSA) for 100 s. After a baseline 992 measurement for 60 s in kinetic buffer, antibody Fabs were associated for 300 s followed by a 993 300 s dissociation step. We used ForteBio data analysis software to determine kinetics

For ACE2-Fc competition experiments, we loaded biotinylated SARS-CoV-2 RBD onto

SA sensors (ForteBio) at 1.5 µg ml -1 for 80 s. We associated C1A-B12 Fab or CR3022-Fab at 997 250 nM or buffer for 180 s followed by an association with ACE2-Fc or CR3022 Fab at a 998 concentration of 250 nM for 180 s

All statistical analysis was performed using Prism v8

Statistical details for experiments are found in figure 1003 legends, including the statistical tests used, the exact value of n, and what n represents

Passive antibody therapy in COVID-19

Python-based system for macromolecular structure solution

A time-and cost-efficient system for high-level 1014 protein production in mammalian cells

SARS-CoV-2 shedding from an asymptomatic immunocompromised individual with cancer

Antibody cocktail to SARS-CoV-2 spike protein prevents rapid 1021 mutational escape seen with individual antibodies

BUSTER version 2.10.3. (Cambridge, United 1024 Kingdom: Global Phasing Ltd

IMGT/V-QUEST: the highly customized 1026 and integrated system for IG and TR standardized V-J and V-D-J sequence analysis

Genome-wide CRISPR screen in a mouse model of tumor 1030 growth and metastasis

A neutralizing human antibody binds to the N-terminal domain of the spike 1033 protein of SARS-CoV-2

Persistence and evolution of SARS-CoV-2 in an 1036 immunocompromised host

Vaccine-elicited receptor-binding site 1039 antibodies neutralize two New World hemorrhagic fever arenaviruses

Coronaviruses: an RNA proofreading machine regulates replication fidelity and diversity

Structurally resolved SARS-CoV-2 antibody shows high efficacy in severely infected 1045 hamsters and provides a potent cocktail pairing strategy

Data, disease and diplomacy: GISAID's innovative 1047 contribution to global health

How good are my data and what is the resolution? 1049

Adaptation of SARS-CoV-2 in BALB/c mice for testing vaccine efficacy

Studies in humanized mice and convalescent humans yield a 1055 SARS-CoV-2 antibody cocktail

Severe Acute Respiratory Syndrome Coronavirus 2 from 1058 patient with coronavirus disease, United States

SARS-CoV-2 cell entry 1061 depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

Structural basis for potent neutralization 1065 of SARS-CoV-2 and role of antibody affinity maturation

Improved methods for building 1067 protein models in electron density maps and the location of errors in these models

Xds

SARS-CoV-2 evolution during treatment of 1072 chronic infection

SARS-CoV-2 vaccines in development

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 1076 receptor

The Immunoglobulin FactsBook

A mouse-adapted SARS-CoV-2 induces acute lung 1080 injury and mortality in standard laboratory mice

Structure of SARS coronavirus spike 1082 receptor-binding domain complexed with receptor

Receptor and viral determinants of SARS-coronavirus adaptation to 1085 human ACE2

Potent neutralizing antibodies against multiple epitopes on SARS-1088

CoV-2 spike

Molecular basis for antibody-mediated neutralization of New World Hemorrhagic fever 1091 mammarenaviruses

Phaser crystallographic software

Transferrin receptor 1 is a cellular 1096 receptor for New World haemorrhagic fever arenaviruses

Convergent antibody responses

CoV-2 in convalescent individuals

Deciphering key features in protein structures with the new 1101

ENDscript server

Isolation of potent SARS-CoV-2 neutralizing antibodies and 1104 protection from disease in a small animal model

A method for identification of HIV 1107 gp140 binding memory B cells in human blood

Sequence and structural convergence of broad 1110 and potent HIV antibodies that mimic CD4 binding

Analysis of a SARS-CoV-2-infected 1113

individual reveals development of potent neutralizing antibodies with limited somatic mutation

Structural basis of receptor recognition by SARS-CoV-2

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

Prospective mapping of viral mutations that escape antibodies used to 1122 treat COVID-19

Broadening of neutralization activity to directly block a dominant antibody-1125 driven SARS-coronavirus evolution pathway

The circulating SARS-CoV-2 1128 spike variant N439K maintains fitness while evading antibody-mediated immunity

Receptor recognition by the 1131 novel coronavirus from Wuhan: an analysis based on decade-long structural studies of SARS 1132 coronavirus

Escape from neutralizing 1135 antibodies by SARS-CoV-2 spike protein variants

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

Mechanisms of host receptor 1140 adaptation by severe acute respiratory syndrome coronavirus

A noncompeting pair of human neutralizing antibodies block COVID-19 virus binding to 1143 its receptor ACE2. Science

Structural basis of a shared antibody response to SARS-CoV-2. Science

1148 (2020b). A highly conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 1149 and SARS-CoV

Structural and functional analysis of the 1152

D614G SARS-CoV-2 spike protein variant

Cellular Nanosponges Inhibit SARS-CoV-2 Infectivity

A pneumonia outbreak associated with a new coronavirus of probable bat 1158 origin

calculated in Buster. During model building, we also customized geometry restraints to prevent 908 large displacement of unambiguous contacts in poor regions; the restraints were released once 909 refinements became stable. Water molecules were automatically picked and updated in Buster, 910 followed by manual examination and adjustment till late-stage refinement. The structures of the 911 C1A-B3:RBD complex (space group P2 1 2 1 2 1 , three copies per ASU), C1A-C2:RBD complex 912 (space group C222 1 , one copy per ASU) and C1A-F10:RBD complex (space group C222 1 , one 913 copy per ASU) were determined using RBD and the C1A-B12 Fab variable and constant 914 domains as search ensembles with CDR and flexible loops truncated, with iterative model 915 building and refinement as described above. Data collection, processing and refinement 916 statistics are summarized in Table S2 . PDB validation reports are included as Supplemental 917 Data S1. We coated NUNC MaxiSorp plates (Thermo Fisher Scientific Cat# 44-2404) with His 6 -tagged 982 SARS-CoV-2 S2P, SARS-CoV-2 RBD, or LUJV GP1 in PBS overnight at 4 °C, followed by a 983