key: cord-0779177-wjkyo3qn
authors: Westendorf, Kathryn; Žentelis, Stefanie; Foster, Denisa; Vaillancourt, Peter; Wiggin, Matthew; Lovett, Erica; van der Lee, Robin; Hendle, Jörg; Pustilnik, Anna; Sauder, J. Michael; Kraft, Lucas; Hwang, Yuri; Siegel, Robert W.; Chen, Jinbiao; Heinz, Beverly A.; Higgs, Richard E.; Kalleward, Nicole; Jepson, Kevin; Goya, Rodrigo; Smith, Maia A.; Collins, David W.; Pellacani, Davide; Xiang, Ping; de Puyraimond, Valentine; Ricicova, Marketa; Devorkin, Lindsay; Pritchard, Caitlin; O’Neill, Aoise; Dalal, Kush; Panwar, Pankaj; Dhupar, Harveer; Garces, Fabian A.; Cohen, Courtney; Dye, John; Huie, Kathleen E.; Badger, Catherine V.; Kobasa, Darwyn; Audet, Jonathan; Freitas, Joshua J.; Hassanali, Saleema; Hughes, Ina; Munoz, Luis; Palma, Holly C.; Ramamurthy, Bharathi; Cross, Robert W.; Geisbert, Thomas W.; Menacherry, Vineet; Lokugamage, Kumari; Borisevich, Viktoriya; Lanz, Iliana; Anderson, Lisa; Sipahimalani, Payal; Corbett, Kizzmekia S.; Wang, Lingshu; Yang, Eun Sung; Zhang, Yi; Shi, Wei; Graham, Barney S.; Mascola, John R.; Fernandez, Tara L.; Hansen, Carl L.; Falconer, Ester; Jones, Bryan E.; Barnhart, Bryan C.
title: LY-CoV1404 (bebtelovimab) potently neutralizes SARS-CoV-2 variants
date: 2021-09-09
journal: bioRxiv
DOI: 10.1101/2021.04.30.442182
sha: 702dd0041dc0fc6d2e3cceec4826f2a2f6e73c19
doc_id: 779177
cord_uid: wjkyo3qn

LY-CoV1404 (also known as bebtelovimab) is a highly potent, neutralizing, SARS-CoV-2 spike glycoprotein receptor binding domain (RBD)-specific antibody identified from a convalescent COVID-19 patient sample, obtained approximately 60 days after symptom onset. In pseudovirus studies, LY-CoV1404 retains potent neutralizing activity against numerous variants including B.1.617.2, B.1.1.7, B.1.351, B.1.427/B.1.429, P.1, and B.1.526, binding to these variants in the presence of their underlying RBD mutations (which include K417N, L452R, E484K, and N501Y). LY-CoV1404 also neutralizes multiple isolates of the authentic SARS-CoV-2 virus in two different assays. The RBD positions comprising the LY-CoV1404 epitope are highly conserved, with the exception of N439 and N501; notably the binding and neutralizing activity of LY-CoV1404 is unaffected by the most common mutations at these positions (N439K and N501Y). New variant-resistant treatments such as LY-CoV1404 are desperately needed, given that some of the existing therapeutic antibodies are less effective or ineffective against certain variants and the impact of variants on vaccine efficacy is still poorly understood. The breadth of variant binding, potent neutralizing activity and the relatively conserved epitope suggest that LY-CoV1404 is one in a panel of well-characterized, clinically developable antibodies that could be deployed as potentially long-term solutions to address current and emerging variants. In Brief LY-CoV1404 is a potent SARS-CoV-2-binding antibody that neutralizes all known variants of concern and whose epitope is rarely mutated. Highlights LY-CoV1404 potently neutralizes SARS-CoV-2 authentic virus and all known variants of concern including the B.1.617.2 (Delta) variant No loss of potency against currently circulating variants Binding epitope on RBD of SARS-CoV-2 is rarely mutated in GISAID database Breadth of neutralizing activity and potency supports clinical development

Variants of SARS-CoV-2 continue to alter the trajectory of the COVID-19 pandemic, which at the time of writing has infected over 190 million people world-wide and is responsible for more than 4 million deaths (https://covid19.who.int/). As predicted, SARS-CoV-2 has continued to evolve as the pandemic has progressed (Mercatelli and Giorgi, 2020; Pachetti et al., 2020) .

Selective pressures and viral adaptation during prolonged, suboptimally-treated infections are thought to have generated numerous variants (McCormick et al., 2021) , some significantly diminishing the effectiveness of COVID-19 clinical countermeasures (Altmann et al., 2021; Plante et al., 2020) . Variants of concern (VOC) represent a closely monitored subset of the many detected SARS-CoV-2 variants, due to their potential for increased transmissibility, and ability to evade immunity produced by infection or vaccines while reducing the efficacies of antibody-based treatments (Hoffmann et al., 2021; Kuzmina et al., 2021; McCormick et al., 2021; Thomson et al., 2021; Wang et al., 2021b) . As of April 2021, the number of countries reporting VOC continues to increase (World Health Organization, 2021) , with emerging variants threatening to slow the pace and success of global vaccination efforts and limit the effectiveness of existing COVID-19 treatments Munnink et al., 2021; Plante et al., 2020; Tegally et al., 2021) .

Many studies have demonstrated the clinical safety and efficacy of antibody-based COVID-19 therapies and their potential for easing the strain on economies and healthcare infrastructures during the pandemic (Chen et al., 2020; Jiang et al., 2020; Jones et al., 2020) . Key target populations for such monoclonal antibody treatments include immunocompromised individuals, patients that are particularly susceptible due to comorbidities (cardiovascular disease, diabetes, chronic respiratory conditions), and those aged over 65 (Jiang et al., 2020) . To date, several antibody therapies targeting the spike (S) protein of SARS-CoV-2 have been successfully used in the fight against the virus (Chen et al., 2020; Gottlieb et al., 2021) . One example is Eli Lilly and Companies' bamlanivimab (LY-CoV555), which was the first anti-SARS-CoV-2 monoclonal antibody tested clinically in June 2020, three months after initiating discovery efforts (Jones et al., 2021) . Lilly's combination of bamlanivimab with another antibody, etesevimab, has received emergency authorization in several countries. Bamlanivimab is also being assessed with the monoclonal antibody VIR-7831 (GSK4182136) in a clinical setting (Medicine, 2020) . Clinical testing has shown that antibodies from Regeneron, GSK/Vir, AstraZeneca, and others are also safe and effective (AstraZeneca, 2020; GlaxoSmithKline, 2021; Regeneron, 2021) . As variants continue to emerge and spread, antibodies can provide highly effective, immediately available therapeutic options to vulnerable populations against COVID-19. However, for antibody therapies to be successful, they must retain potent neutralizing breadth against emerging SARS-CoV-2 lineages carrying mutations. A high neutralization potency may also provide the opportunity to explore lower clinical doses (Andreano et al., 2021) and alternate routes of administration.

Multiple reports have consistently shown that mutations in SARS-CoV-2 can substantially reduce the binding affinity and neutralization of antibodies Yao et al., 2021) . Variants such as B.1.1.7, B.1.351, P.1, B.1.526, B.1.427 and B.1.429 affect the in vitro binding of antibodies being tested clinically or on those already authorized for emergency use to varying degrees (Kuzmina et al., 2021; Liu et al., 2021; Rees-Spear et al., 2021) . Mutations at amino acid positions 417, 439, 452, 484, and 501 were found to have the greatest consequences on the functionality of both antibodies and vaccines (Thomson et al., 2021) . Therefore, antibodies that retain binding and potent neutralization activity in the presence of these mutations could be extremely valuable tools to rapidly respond to these variants.

In this report, we describe LY-CoV1404 (also known as bebtelovimab), a fully human IgG1 monoclonal SARS-CoV-2 antibody targeting the receptor binding domain (RBD), identified in our ongoing pandemic-response efforts. As of the writing of this paper (August 2021), LY-CoV1404 binds and potently neutralizes all currently known VOC of SARS-CoV-2. LY-CoV1404 binds to an epitope that is distinct from the mutations identified to be widely circulating within the newly emerged variants, including those mutations that reduce the effectiveness of vaccines.

Importantly, the LY-CoV1404 interaction with the S protein is driven by amino acids that are rarely mutated in the global GISAID EpiCoV database, indicating LY-CoV1404 has the potential to be a long-term, solution for reducing COVID-19-related illness and death.

Variants of SARS-CoV-2 continue to be identified globally ( Figure 1A) and can abrogate the binding of monoclonal antibody therapies or reduce vaccine effectiveness (Hoffmann et al., 2021; Wang et al., 2021b) . To identify potent neutralizing antibodies to SARS-CoV-2 variants, a high-throughput screening approach (Jones et al., 2021) was conducted on peripheral blood mononuclear cells (PBMC) that were isolated from a COVID-19 convalescent donor. The sample was obtained approximately 60 days after the onset of symptoms.

Three different screening strategies were employed to identify SARS-CoV-2-binding antibodies from the PBMCs in this donor sample. We used a soluble antigen assay with fluorescently labeled prefusion-stabilized full-length trimeric SARS-CoV-2 S protein as well as the SARS-CoV-2 RBD subunit, a live-cell-based assay using mammalian cells that transiently expressed the SARS-CoV-2 S protein, and a multiplexed bead assay using S proteins from multiple coronaviruses ( Figure 1B) . A total of 740,000 cells were screened in this discovery effort and machine learning (ML)-based analyses were used to automatically select and rank 1,692 antibody "hits" (0.22% frequency). Of these, 1,062 single antibody-secreting cells were selected for recovery. Libraries of antibody genes from recovered single B cells were generated and sequenced using next-generation sequencing, and a custom bioinformatics pipeline with MLbased sequence curation was used to identify paired-chain antibody sequences, with 290 unique high-confidence paired heavy and light chain sequences identified ( Figure 1B) . The antibody sequences corresponded to 263 clonal families and used a diverse set of 41 VH genes. The mean sequence identity to germline was 96.8% from this 60-days post-symptomonset blood draw in this donor.

High-throughput SPR experiments on a selected subset of 69 recombinantly-expressed antibodies (including benchmark controls) were performed to assess S protein epitope coverage. These studies included epitope binning, isolated subdomain binding, and binding competition with ACE2 (Figures 1C/S1 ). We employed 21 benchmark antibodies including those known to bind to S protein subunits S1, N-terminal domain (NTD), RBD, S2 subunit, and several internally-identified antibodies with known binding domains. The antibodies (including benchmarks) were clustered into 46 bins using Carterra epitope analysis software based on the competition heatmap ( Figure 1C) , and were delineated into bins among known subunits of the S protein. Several of the antibodies discovered are cross-blocked with those that are currently in clinical testing or authorized for use in COVID-19 patients including bamlanivimab. We therefore focused analyses on antibodies outside of these established epitopes and identified LY-CoV1404, an ACE2-blocking antibody, which is found in a unique bin shared with one other antibody. LY-CoV1404 was observed to compete directly for S protein RBD binding with the S309 (Pinto et al., 2020) and REGN10987 (Hansen et al., 2020) antibodies, suggesting a binding epitope somewhat distinct from other ACE2-blocking antibodies such as bamlanivimab or etesevimab.

Characterization of LY-CoV1404's binding kinetics revealed that the Fab fragment of the antibody bound to the S protein of D614G with high affinity, characterized by binding constant (K D ) values ranging between 790 pM and 4nM, depending on the assay ( Figure 1D , Table 1 ).

We further analyzed the binding affinity of LY-CoV1404 as a human IgG1 antibody and determined the affinity to be between 75 pM and 220 pM (depending on the assay design) for LY-CoV1404 binding to the S protein ( Figure 1E , Figure 1E ). These data suggest that LY-CoV1404 binds to a region that is not significantly affected by the prevalent VOC mutations. We further tested the binding of LY-CoV1404 to a wider range of S mutants and VOC using an assay system with S protein expressed on the surface of mammalian cells. In these assays, we compared binding of LY-CoV1404 to isotype control antibodies. Consistent with the affinity data, LY-CoV1404 maintained binding to all variants and mutants tested ( Figure 1F ). These data indicate that SARS-CoV-2 S protein variants identified and reported to be of significant concern in this pandemic are still potently bound by LY-CoV1404.

The in vitro viral neutralization activity LY-CoV1404 was assessed against an authentic viral strain (SARS-CoV-2/MT020880.1) using an immunofluorescence assay (IFA) readout. In these studies, LY-CoV1404 potently neutralized the authentic virus with half maximal inhibitory concentration (IC50) values ranging from 9 ng/mL to 22.1 ng/mL. Bamlanivimab (LY-CoV555), another potent anti-RBD binding antibody, consistently had a 2-3-fold higher (less potent) IC50 compared to LY-CoV1404 (Figure 2A) .

We also tested the VIR-S309 antibody that was recently described to have a unique binding epitope, outside of many of the mutated positions in current VOC (Pinto et al., 2020) . This antibody is cross-reactive to SARS-CoV-1 S protein and retains neutralization capacity against B.1.351, B.1.1.7, and P.1 (Pinto et al., 2020) . Comparing separate assay runs, we observed VIR-S309 had an IC50 of 118 ng/mL (highly consistent with a recent report on the clinically active derivative VIR-7831 (Cathcart et al., 2021) ) compared to bamlanivimab with a range of 19.2 to 29.7 ng/mL IC50 in multiple independent assays ( Figure 2B ). Amongst other antibodies tested, LY-CoV1404 exhibited the most potent neutralization of authentic virus in this assay ( Figure 2C ). We also tested authentic virus neutralization in a plaque reduction neutralization test (PRNT) against a Canadian strain of SARS-CoV-2 (hCoV-19/Canada/ON_ON-VIDO-01-2/2020) with multiple antibodies (Figure 2C) , and against other natural SARS-CoV-2 isolates tested with LY-CoV1404 alone and in combination with other antibodies ( (Table 3B) .

These studies were performed using two different systems with highly comparable results.

These data demonstrate that LY-CoV1404 is extremely potent against VOCs, including those most rapidly spreading worldwide (for example, B.1.617.2).

In order to fully characterize the binding epitope of LY-CoV1404 we determined the X-ray crystal structure of LY-CoV1404 Fab bound to S protein RBD and CryoEM structure of intact S protein bound to LY-CoV1404 (Figure 3) . The Fab of LY-CoV1404 binds to a region overlapping the ACE2-interacting site of the S protein that is accessible in both the "up" and "down" conformers of the RBD on the S protein ( Figure S2 ). While this property would suggest that LY-CoV1404 is a Class 2 binder (Barnes et al., 2020), the structural location of the epitope is closer to the canonical Class 3 binder, VIR-S309 (Wang et al., 2021b) . Interestingly, the binding epitope of LY-CoV1404 is very similar to previously described antibodies imdevimab (REGN10987) (Hansen et al., 2020) and Fab 2-7 (Cerutti et al., 2021; Liu et al., 2020) . This similarity is apparent from the structural superposition of the Fabs as well as which RBD residues interact with the antibodies ( Figure S3 ). Furthermore, though independently discovered, LY-CoV1404

and Fab 2-7 (Cerutti et al., 2021) share 92% amino acid sequence identity in the variable regions of both their heavy and light chains, and they both engage the RBD similarly through their heavy and light chains. By contrast, REGN10987 is more sequence divergent and nearly all of its interactions with the RBD are through its heavy chain. LY-CoV1404 has a contact surface area on the RBD of 584 Å 2 , compared to 496 Å 2 for Fab 2-7, and only 343 Å 2 for REGN10987 (Table S2) .

To determine the frequency of mutations of the amino acid residues of the S protein that are close contacts with LY-CoV1404 as determined by structural analysis, we used the publicly deposited mutations observed at these sites in the GISAID EpiCoV database ( Table 4 ). There were very few changes to these identified amino acids, with the only two residues below the >99% unchanged threshold: N439 and N501 (reported as having 97.832% and 86.966%

conservation at these positions respectively), demonstrating that the LY-CoV1404 epitope has remained relatively unchanged during the pandemic. We identified the most prevalent mutations at epitope residues in the GISAID EpiCoV database and characterized their impact on the functional activity of LY-CoV1404. Interestingly, the observed N439K and N501Y mutations had no impact on LY-CoV1404 binding or neutralization ( Table 5 ). The lack of effect by the N493K mutation is interesting in the context of the loss of antibody binding previously observed for REGN10987 (Thomson et al., 2021) demonstrating that beyond the epitope, the specific interactions are critical to determining whether resistant mutations will arise for any particular antibody. A loss of binding, ACE2 competition and neutralization activity was observed with mutations at positions 444 and 445 (which could arise by a single nucleotide change) ( Table 5) . Importantly however, these changes are extremely rare in the general population as reported in the GISAID database ( Table 4 ). The ability of LY-CoV1404 to maintain neutralization potency despite a variety of mutations that have been shown to nullify activity of several other potent neutralizing antibodies (Rappazzo et al., 2021) indicates that this antibody may be uniquely suited to combat the current VOC. Critically, potent activity against all of these variants suggests that LY-CoV1404 binds to an epitope that has acquired few mutations.

With the exception of N439 and N501, the residues with which LY-CoV1404 interacts are not prevalent in current VOC (e.g. 417, 439, 452, 484) . Importantly, while the LY-CoV1404 binding epitope includes residues both N439 and N501, LY-CoV1404 still binds the N501Y mutant ( Figure 1F) , and variants B. Figure 3B , Figure   S2 ). However, unlike imdevimab (Thomson et al., 2021) , LY-CoV1404 retains full functional neutralization against pseudovirus with the N439K mutant ( activity against all of the tested variants suggests LY-CoV1404 binds uniquely to an epitope that has acquired few mutations and is not sensitive to the few mutations that have arisen.

We explored the potential for resistant mutations to arise by yeast-display of the S protein RBD.

A library was constructed in which all amino acid residues were sampled at positions 331 to 362

and 403 to 515 of the spike glycoprotein containing the epitope of LY-Cov1404. Cells containing RBD substitutions that were still able to bind soluble ACE2 after antibody treatment were isolated and sequenced to determine the location and identity of the change. Further functional characterizations of isolates were limited to those containing a change of a single amino acid residue which can be generated by a single nucleotide change based on the codon used in the Wuhan isolate sequence.

Consistent with the structure of the RBD complex, selection using LY-CoV1404 identified potential susceptibility to some, but not all, substitutions at residues K444, V445, G446, and P499. None of the observed G446 changes identified from the selection could arise from a single nucleotide substitution but nevertheless suggested a potentially liable residue with LY-CoV1404. The presence of the G446V variant in the GISAID database, as well as the ability of this variant to be resistant to imdevimab treatment prompted further analysis. The ability of LY-CoV1404 to inhibit ACE2 binding to RBD substitutions that have been identified in common circulating viral variants and/or have been shown to impart resistance to the authorized antibodies bamlanivimab, etesevimab, casirivimab, and imdevimab was also analyzed ( Table   6) .

We further characterized LY-CoV1404 activity in the presence of mutations known to weaken the effectiveness of other antibodies in clinical testing. Of the mutations tested, only K444Q, V445A, and G446V impacted LY-CoV1404 binding (Table 5) . Interestingly, while the K444Q and V445A mutations also resulted in reduced ACE2 competition and neutralization activity, the G446V mutation only appeared to significantly impact the monomeric binding affinity, along with modest changes in ACE2 competition and neutralization. These residues are very rarely mutated according to GISAID reports (Tables 4,5), thus suggesting a low risk of mutations emerging at these residues and restriction in functional activity of LY-CoV1404. We examined the frequency of all variants observed at residues 444 and 445 within the GISAID-EpiCov database (Shu and McCauley, 2017) , focusing on recent data (from December 16, 2020 to March 16, 2021; see Materials and Methods). These two sites were rarely mutated, with 0.027% of samples collected within the past 3 months reported to carry mutations. It is not known whether all variants at these locations would confer resistance. Of the specific mutations tested (Table 5) , we find that only 0.0045% of samples contained the known resistance V445A, while no occurrences of K444Q were observed. This suggests LY-CoV1404 may remain effective even as new strains emerge. These data also indicate that LY-CoV1404 is not only likely to be highly effective against current variants, but is less likely to be impacted by future mutations given the low level of changes observed to date in its binding epitope (McCormick et al., 2021) .

We report the discovery of LY-CoV1404, a highly potent SARS-CoV-2 S protein RBD-binding antibody that maintains binding and neutralizing activity across currently known and reported VOC. This antibody binds to or neutralizes variants including those first identified in India has been determined to stabilize the S protein, leading to more efficient ACE2 interactions and infections (Santos and Passos, 2021) . More recently, the rapid spread of the B. Regeneron, 2021). With the proven clinical effectiveness of antibody therapies at preventing hospitalization, reducing severity of disease symptoms and death (Darby and Hiscox, 2021) , the importance of access to effective antibody treatments targeting conserved neutralizing epitopes for response to viral variants must be recognized.

It remains unknown whether vaccines will significantly alter the mutation profile of the virus (Choi et al., 2020; Kemp et al., 2021) . If, as vaccination continues world-wide, the virus responds to this selection pressure with particular mutations to avoid the effects of vaccination, it will be important to maintain alternative therapies to help overcome these variants. It has been hypothesized that certain variants may have arisen in immunocompromised individuals, which provided the virus with an extended timeframe for continued viral replication and the accumulation of many mutations, seeding these novel VOC (Aschwanden, 2021). A potent neutralizing monoclonal antibody, such as LY-CoV1404, would be a potential solution for these immunocompromised patients, with rapid neutralization of the virus providing both protection for the patient, and less opportunity for viral mutation and evolution. Finally, immunocompromised individuals are less likely to respond well to vaccines, further highlighting the utility of potent monoclonal antibody therapy for protecting this population. Monoclonal antibodies with potent neutralizing effects could provide a safe and robust medical countermeasure against variants or a necessary, complementary alternative form of protection for those individuals who cannot, or do not receive vaccines. For the reasons listed above, it has been proposed that achieving "herd immunity" is unlikely in the near term for SARS-CoV- 2 (Cathcart et al., 2021; Pinto et al., 2020) , further emphasizing the need for medical alternatives to combat viral resistance.

Recently, an antibody that binds to an epitope of the SARS-CoV-2 S protein that is distinct from current VOCs has been described. This antibody, VIR-7831 ( Though it remains to be tested, this increased potency has the potential to enable lower doses and subcutaneous administration for either treatment or

prophylaxis. Interestingly, VIR-7831 was derived from the antibody S309, which was identified from a SARS-CoV-1 convalescent patient (Cathcart et al., 2021) . This antibody is cross-reactive to SARS and binds to one of the few epitopes shared between the viruses, which includes an Nlinked glycan. By comparison, LY-CoV1404 was discovered from a SARS-CoV-2 convalescent patient and selected based on its cross-reactivity to SARS-CoV-2 variants, thereby allowing testing and identification of much more potent neutralizing epitopes. A more recently identified antibody from a COVID-19-recovered patient was described with broad cross-reactivity to variants (including B.1.1.7, B.1.351, B.1.429, and P.1) and several zoonotic strains (Tortorici et al., 2021) . This antibody, while able to bind widely to coronaviruses, had neutralizing capacity against SARS-CoV-2 in authentic virus-neutralization assays that was substantially less potent than LY-CoV1404. Finally, an engineered antibody with broad cross-reactivity against coronaviruses and potent neutralization was recently described (Rappazzo et al., 2021) , which was optimized for binding following identification from a SARS convalescent patient.

Mutations within the S protein and in particular in the RBD are inevitable, particularly as SARS-

CoV-2 has been present in the human population for only 18 months (Jones et al., 2020) . Since the pandemic began, we have continued to screen patient samples and have identified thousands of human antibodies with associated functional data. We analyzed this antibody database in response to the increase in VOC and identified many potential solutions based on these variants. LY-CoV1404 is one example of this screening and analysis. In order to fully prepare for the inevitability of additional mutants and reduction in treatment effectiveness, we into clinical testing concurrent with this report represents a significant milestone in response to the evolving pandemic.

Recombinant antibodies and other proteins were produced as previously described (Jones et al., 2020) . The antigen binding fragment (Fab) portion of LY-CoV1404 was generated by proteolytic digestion using immobilized papain (ThermoFisher Scientific), followed by removal of Fab:RBD complex was prepared by mixing the components with 20% molar excess of the RBD, and then the complex purified from the excess RBD by size-exclusion chromatography. Beads or cells were flowed onto microfluidic screening devices and incubated with single antibody-secreting cells, and mAb binding to cognate antigens was detected via a fluorescently labeled anti-human IgG secondary antibody, or soluble antigen labeled with fluorophore.

Positive hits were identified using machine vision and recovered using automated roboticsbased protocols.

Single cell polymerase chain reaction (PCR) and custom molecular biology protocols generated NGS sequencing libraries (MiSeq, Illumina) using automated workstations (Bravo, Agilent).

Sequencing data were analyzed using a custom bioinformatics pipeline to yield paired heavy and light chain sequences for each recovered antibody-secreting cell (Jones et al., 2020) . Each sequence was annotated with the closest germline (V(D)J) genes, degree of somatic hypermutation, and potential sequence liabilities. Antibodies were considered members of the same clonal family if they shared the same inferred heavy and light V and J genes and had the same CDR3 length. The variable (V(D)J) region of each antibody chain was PCR amplified and inserted into expression plasmids using a custom, automated high-throughput cloning pipeline.

Plasmids were verified by Sanger sequencing to confirm the original sequence previously identified by NGS. Antibodies were recombinantly produced by transient transfection in either human-embryonic kidney (HEK293) or CHO cells as described in Supplemental Methods.

All epitope binning and ACE-2 blocking experiments were performed on a Carterra® LSA™ instrument equipped with an HC-30M chip type (Carterra-bio), using a 384-ligand array format as previously described (Shu and McCauley, 2017) . For epitope binning experiments, antibodies coupled to the chip surface were exposed to various antibody:antigen complexes. Samples were prepared by mixing each antibody in 10-fold molar excess with antigen (1:1 freshly prepared mix of 400 nM antibody and 40 nM antigen, both diluted in 1X HBSTE + 0.05% BSA running buffer). Each antigen-antibody premix was injected sequentially over the chip surface for 5 minutes (association phase to ligand printed onto chip previously), followed by a running buffer injection for 15 minutes (dissociation phase). Two regeneration cycles of 30 seconds were performed between each premix sample by injecting 10 mM glycine pH 2.0 onto the chip surface. An antigen-only injection (20 nM concentration in the running buffer) was performed every 8 cycles to assess maximum binding to S protein and in-order to accurately determine the binning relationship.

The data were analyzed using the Carterra Epitope analysis software for heat map and competition network generation. Analyte binding signals were normalized to the antigen-only binding signal, such that the antigen-only signal average is equivalent to one RU (response unit). A threshold window ranging from 0.5 RU to 0.9 RU was used to classify analytes into 3 categories: blockers (binding signal under the lower limit threshold), sandwichers (binding signal over the higher limit threshold) and ambiguous (binding signal between limit thresholds).

Antibodies with low coupling to the chip, poor regeneration or with absence of self-blocking were excluded from the binning analysis. Like-behaved antibodies were automatically clustered to form a heat map and competition plot.

To test the antibodies' ability to block ACE2, antibodies coupled to the HC-30M chip as described above were exposed to SARS-CoV-2 S protein:ACE2 complex. A freshly prepared 1:1 mix of 40 nM of SARS-CoV-2 S protein and 400 nM of untagged ACE2, both diluted in HBSTE + 0.05% BSA, were tested for binding to the immobilized mAbs on the prepared HC-30M chips, with association for 5 minutes and dissociation for 5 minutes. A SARS-CoV-2 S protein injection at 20 nM was included to assess for maximum binding, as well as a ACE2

injection at 200 nM to assess for non-specific binding. Regeneration was performed in 20 mM glycine pH 2.0 with 1M NaCl for 30 seconds twice.

Surface plasmon resonance (SPR) capture experiments were performed on a Biacore 8K To investigate binding of LY-CoV1404 mAb to SARS-CoV-2 S protein, D614G S protein variant and RBD protein, a HC30M chip was used, and the array preparation was performed as described above, coupling antibody diluted to 3 µg/mL in 10 mM acetate, pH 4.0, for 10 minutes, and deactivation for 7 minutes in 1 M ethanolamine, pH 8.5.

To measure binding kinetics and affinity, the mAb-coupled HC30M chip surface was exposed to injections of the proteins, with an association period of 5 minutes and dissociation period of 15 minutes. The tested concentrations of the trimeric S proteins were 300, 100, 33.3, 11.1, 3.70, 1.23, and 0.41 nM in HBS-EP+ containing 0.1 mg/mL BSA. The tested concentrations of the RBD protein were 400, 100, 25, 6.25 and 1.56 nM in HBS-EP+ containing 0.1mg/mL BSA.

Regeneration of the chip surface between the different concentrations was performed using 20 mM glycine, pH 2.0, for 30 seconds twice. Kinetic data was analyzed using Carterra KIT™ software using a 1:1 Langmuir binding model.

To investigate binding of LY-CoV1404 Fab to reference and mutant RBDs a HC30M chip was used, and the array preparation was performed as described above, coupling Fc-fused RBD proteins diluted 5 µg/mL and 10 µg/mL in 10 mM acetate, pH 4.0, for 10 minutes, and deactivation for 7 minutes in 1 M ethanolamine, pH 8.5.

To measure binding kinetics and affinity, the RBD-coupled HC30M chip surface was exposed to injections of LY-CoV1404 Fab, with an association period of 5 minutes and dissociation period 

The capture molecule, an anti-human IgG (Fc) antibody, was immobilized on a Biacore CM5 chip by direct coupling. The chip surface was first activated by flowing a freshly prepared 1:1 activation mixture of 100 mM S-NHS, 400 mM EDC for 10 min at a flow rate of 10 uL/min. Antihuman IgG (Fc) antibody was diluted to 25 ug/mL in 10 mM Sodium Acetate buffer pH 4.5, then injected using all 8 channels, at a flow rate of 10 uL/min for 10 min. The chip was washed with HBS-EP+ (10 mM HEPES, 150 mM NaCl, 3 mM EDTA and 0.05% v/v Surfactant P20) for 1 min, at a flow rate of 30 uL/min. Finally, excess reactive esters were quenched by flowing 1 M ethanolamine for 10 min at a flow rate of 10 uL/min, followed by 3 conditioning steps of 30 s each, in 10 mM NaOH buffer, at a flow rate of 10 uL/min.

Multi cycle kinetics of LY-CoV1404 was performed using the previously prepared CM5 chip on the Biacore instrument as described herein. LY-CoV1404 was diluted to 5 nM in HBS-EP+ buffer (as above), then flowed over the CM5 chip for 30 sec at a flow rate of 10 uL/min. Each antigen of interest was diluted in HBS-EP+ buffer to various concentrations (100 nM, 33.3 nM, 11.1 nM, 3.7 nM, 1.2 nM). Each concentration was then serially flown over, from lowest concentration to highest, starting with a buffer (blank) injection, for 120 sec at a flow rate of 30 uL/min. After each antigen injection, HBS-EP+ buffer was injected for 600 sec at a flow rate of 30 uL/min. Then the chip was regenerated with a single injection of 3M MgCl2 for 30 s at a flow rate of 30 uL/min before the next antigen injection.

The data were analyzed using the Biacore Insight Evaluation Software. The curves were referenced and blank subtracted and then fit to a 1:1 Langmuir binding model to generate apparent association (ka) and dissociation (kd) kinetic rate constants and binding affinity constants (KD).

Unique antibody sequences were confirmed to bind the screening target ( 

Neutralization assays were carried out as described (Nie et al., 2020 

SARS-CoV-2 spike pseudotyped lentiviruses that harbor a luciferase reporter gene were produced and neutralization assay was performed as described previously Naldini et al., 1996) . Pseudovirus was produced by co-transfection of 293T cells with plasmids encoding the lentiviral packaging and luciferase reporter, a human transmembrane protease serine 2 (TMPRSS2), and SARS-CoV-2 S (Wuhan-1, Genbank #: MN908947.3) or S variant genes (Wang et al., 2021a) . Forty-eight hours after transfection, supernatants were harvested, filtered and frozen. For neutralization assay serial dilutions (2 dilutions at 10 and 1 µg/ml for the initial screen assay or 8 dilutions for the full curve at 10-0.0006 µg/ml) of monoclonal antibodies 

The potential for SARS-CoV-2 to develop resistance under selective pressure of LY-CoV1404, was studied using in vitro directed evolution. The RBD comprised of residues 319 to 541 of performed at room temperature for 20 minutes followed by transfer of reactions to ice for another 10 minutes to quench dissociation. Fluorescence was measured after addition of streptavidin-phycoerythrin for 20 minutes on ice followed by multiple washes with cold buffer.

Mean fluorescence intensity (MFI) was normalized for each concentration response curve (CRC) using the maximum MFI at no mAb inhibitor (MAX) and the background MFI (Bgnd).

Percent inhibition of S protein binding to ACE2 was defined as follows:

where x was the MFI at the tested concentration of mAb. CRCs were fit with a four-parameter logistic function. All four parameters were estimated from the fitting. Absolute IC50 (50% absolute inhibition) values were reported. Standard error and 95% confidence intervals for IC50

estimates are reported and, in some cases, a fixed top (Top = 100) was used to stabilize the standard error estimate. IC50 ratios between mutant S protein and wild type (WT) were reported. For mutant S proteins that exhibited no mAb inhibition, fold-change was reported at 1x of the highest concentration of mAb tested. For any single experimental batch, inhibition data were obtained in duplicate or triplicate. WT RBD was analyzed with every experiment and the consensus estimate as of the date indicated is reported. If select mutants of interest were analyzed in more than one experimental batch, then the geometric mean is reported.

Work with authentic SARS-CoV-2 at USAMRIID was completed in BSL-3 laboratories in accordance with federal and institutional biosafety standards and regulations. Vero-76 cells were inoculated with SARS-CoV-2 (GenBank MT020880.1) at a MOI = 0.01 and incubated at 37°C with 5% CO2 and 80% humidity. At 50 h post-infection, cells were frozen at -80°C for 1 h, allowed to thaw at room temperature, and supernatants were collected and clarified by centrifugation at ~2,500 x g for 10 min. Clarified supernatant was aliquoted and stored at -80°C. Sequencing data from this virus stock indicated a single mutation in the S glycoprotein (H655Y) relative to Washington state isolate MT020880.1.

Authentic SARS-CoV-2 plaque reduction assays were also conducted in BSL3 laboratories at UTMB. Three natural isolates were used to measure mAb neutralization: USA/WA/1/2020 (BEI isolates (in the case of the rWA1, rWA1 E484K, and rWA1 E484Q viral isolates) were constructed using reverse genetics as previously described (Xie et al., 2021b) . Virus stocks were grown by inoculating cultured Vero E6 cells, followed by incubation at 37ºC until cytopathic effects (CPE) were evident (typically 48 to 72 hours). Expansion was limited to 1 to 2 passages in cell culture to retain integrity of the original viral sequence. The virus stock was quantified by standard plaque assay, and aliquots were stored at -80ºC. A freshly thawed aliquot was used for each neutralization experiment.

A pre-titrated amount of authentic SARS-CoV-2/MT020880.1, at final multiplicity of infection of 0.2, was incubated with serial dilutions of monoclonal antibodies for 1 h at 37°C. The antibodyvirus mixture was applied to monolayers of Vero-E6 cells in a 96-well plate and incubated for 1 hour at 37°C in a humidified incubator. Infection media was then removed, and cells were washed once with 1X PBS, followed by addition of fresh cell culture media. Culture media was removed 24 hours post infection and cells were washed once with 1X PBS. PBS was removed and plates were submerged in formalin fixing solution, then permeabilized with 0.2% Triton-X for 10 minutes at room temperature and treated with blocking solution. Infected cells were detected using a primary detection antibody recognizing SARS-CoV-2 nucleocapsid protein (Sino Biological) following staining with secondary detection antibody (goat α -rabbit) conjugated to AlexaFluor 488. Infected cells were enumerated using the Operetta high content imaging instrument and data analysis was performed using the Harmony software (Perkin Elmer).

Vero-E6 cells were seeded in a 24-well plate 48 hours before the assay. Seventy-five plaque forming units (pfu) of infectious clone hCoV-19/Canada/ON_ON-VIDO-01-2/2020 were mixed with serial dilutions of monoclonal antibodies and incubated at 37°C for 60 minutes. Virus and antibody mix was added to each well and incubated for 1 h in a 37°C + 5% CO2 incubator with rocking every 10-15 min. Plaque assay media (complete MEM media with 1% BGS + 1% low melting point agarose) was overlaid on top of the inoculum and incubated at 37°C + 5% CO2

incubator for 48 hours. For plaque visualization, an MEM-Neutral Red overlay was added on day 2 and plaques counted manually on day 3 or day 4.

PRNT assay (Table 2) .

Plaque Reduction Neutralization test assays were performed in 6-well plates. Vero E6 cells were seeded at a concentration of approximately 106 cells/well and grown overnight at 37°C in 5% CO2 to reach 95% confluency. The next day, serial three-fold dilutions of LY-CoV1404 or bamlanivimab were prepared in Eagle's minimal essential medium, mixed with approximately 100 pfu of SARS-CoV-2, and incubated for 1 to 2 hours on ice or at 37°C. The mAb/virus mixture was inoculated directly onto the cells (in duplicate wells) and allowed to adsorb for 1 hour at 37°C with 5% CO2, with rocking at 15-minute intervals. An overlay media composed of 1.25% Avicel RC 581 (FMC BioPolymer) in Eagles minimum essentials medium (MEM) with 5% FBS was added, and plates were incubated for 48 hours at 37°C with 5% CO2 for virus plaques to develop. After incubation, overlays were removed by aspiration and the cells were fixed with 10% buffered formalin containing crystal violet stain for 1 hour. Plaques were counted manually, and plaque forming units were determined by averaging technical replicates per sample.

Percent neutralization was determined relative to virus-only control-treated samples.

A 10 mg/mL solution of LY-CoV1404 Fab (with CrystalKappa mutations (Lieu et al., 2020b) ) in complex with RBD was set up in vapor diffusion sitting drops at a ratio of 1:1 with a well solution of 100 mM Tris HCl pH 6.5, 20 % PEG MME 2K and 200 mM Trimethylamine N-oxide. Crystals appeared within two days, grew to their full size and were harvested on the fifth day after the set up. Crystals were flash-frozen in liquid nitrogen following 1-minute incubation in cryoprotectant solution containing mother liquor supplemented with additional 5% PEG MME 2K.

Diffraction data were collected at Lilly Research Laboratories Collaborative Access Team (LRL-CAT) and beamline at Sector 31 of the Advanced Photon Source at Argonne National Laboratory, Chicago, Illinois. Crystals stored in liquid nitrogen were mounted on a goniometer equipped with an Oxford Cryosystems cryostream maintained at a temperature of 100 K. The wavelength used was 0.9793 Å collecting 900 diffraction images at a 0.2 degree oscillation angle and 0.1 seconds exposure time on a Pilatus3 S 6M detector at a distance of 385 mm. The diffraction data were indexed and integrated using autoPROC (Vonrhein et al., 2011) /XDS (Kabsch, 2010) and merged and scaled in AIMLESS (Evans and Murshudov, 2013) from the CCP4 suite (Winn et al., 2011) . Non-isomorphous data readily yielded initial structures by molecular replacement using the Fab portion of crystal structures from the proprietary Eli Lilly structure database and the SARS-CoV-2 S protein RBD from the public domain structure with the access code 6yla (Huo et al., 2020) . The initial structure coordinates were further refined using Buster (Smart et al., 2012) CONTACT (Winn et al., 2011) and custom shell/Perl scripts.

A stabilized version of the SARS-CoV-2 S protein was employed for all studies (Hsieh 2020) .

The Fab generation was initiated with plasmids encoding the truncated heavy chain and light chain of LY-CoV1404 (1 ug/mL each) that were co-transformed into expi293 cells (expifectamine, Gibco) for protein expression (5 days). Cell supernatants were collected, clarified and incubated with Ni-NTA beads to capture the Fab, followed by washing in PBS buffer (pH 7.4) and elution in the same buffer supplemented with 250 mM Imidazole. Falcon IV detector at a 130,000x nominal magnification corresponding to a pixel size of 0.98A . The RELION program was used for particle picking, 2D and 3D class averaging (Scheres 2016) .

To estimate the incidence of SARS-CoV-2 variants with potential resistance to LY-CoV1404, we analyzed data from GISAID EpiCoV database (Shu and McCauley, 2017) (Li and Durbin, 2010) , and performed variant calling and annotation using SAMTools (Danecek et al., 2021) . All further analysis (i.e.

slicing by time or geographic region) was performed using custom-written Python scripts. Abbreviations: D = aspartate; del = deletion; E = glutamate; G = glycine; H = histidine; I = isoleucine; IC 50 , IC90 = concentration inhibiting maximal activity by 50% and 90%, respectively; K = lysine; L = leucine; N = asparagine; NA = not applicable; P = proline; Q = glutamine; R = arginine; S = serine; T = threonine; Y = tyrosine. a Absolute IC estimates presented are a result of a meta-analysis of multiple replicate experiments. * Viral constructs for the lentivural platform include D614G ** Viral constructs for the VSV platform include D614G 

SARS-CoV-2 mRNA vaccine design enabled by prototype pathogen preparedness

Twelve years of SAMtools and BCFtools. Gigascience 10

Covid-19: variants and vaccination

Estimated transmissibility and impact of SARS-CoV

Antibody evasion by the P.1 strain of SARS-CoV-2. Cell

How good are my data and what is the resolution?

Effect of Bamlanivimab as Monotherapy or in Combination With Etesevimab on Viral Load in Patients With Mild to Moderate COVID-19

Vir Biotechnology and GSK announce VIR-7831 reduces hospitalisation and risk of death in early treatment of adults with COVID-19

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

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

SARS-CoV-2 variants B.1.351 and P.1 escape from neutralizing antibodies

Paper from the New and Emerging Respiratory Virus Threats Advisory Group (NERVTAG) on new coronavirus

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

Neutralizing nanobodies bind SARS-CoV-2 spike RBD and block interaction with ACE2

Neutralizing antibodies for the treatment of COVID-19

LY-CoV555, a rapidly isolated potent neutralizing antibody, provides protection in a non-human primate model of SARS-CoV-2 infection

The neutralizing antibody, LY-CoV555, protects against SARS-CoV-2 infection in non-human primates

SARS-CoV-2 evolution during treatment of chronic infection

SARS CoV-2 spike variants exhibit differential infectivity and neutralization resistance to convalescent or post-vaccination sera

Fast and accurate long-read alignment with Burrows-Wheeler transform

Rapid and robust antibody Fab fragment crystallization utilizing edge-to-edge betasheet packing

Rapid and robust antibody Fab fragment crystallization utilizing edge-to-edge betasheet packing

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

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

The SARS-CoV-2 Spike Variant D614G Favors an Open Conformational State

The emerging plasticity of SARS-CoV

A Study of Immune System Proteins in Participants With Mild to Moderate COVID-19 Illness (BLAZE-4)

Geographic and Genomic Distribution of SARS-CoV-2 Mutations

Transmission of SARS-CoV-2 on mink farms between humans and mink and back to humans

Efficient transfer, integration, and sustained long-term expression of the transgene in adult rat brains injected with a lentiviral vector

Establishment and validation of a pseudovirus neutralization assay for SARS-CoV-2

Emerging SARS-CoV-2 mutation hot spots include a novel RNA-dependent-RNA polymerase variant

Spike mutation D614G alters SARS-CoV-2 fitness

Broad and potent activity against SARS-like viruses by an engineered human monoclonal antibody

The impact of Spike mutations on SARS-CoV-2 neutralization

The high infectivity of SARS-CoV-2 B.1.1.7 is associated with increased interaction force between Spike-ACE2 caused by the viral N501Y mutation

Chapter Six Processing of Structurally Heterogeneous Cryo-EM Data in RELION

GISAID: Global initiative on sharing all influenza data -from vision to reality

Exploiting structure similarity in refinement: automated NCS and target-structure restraints in BUSTER

Sixteen novel lineages of SARS-CoV-2 in South Africa

Circulating SARS-CoV-2 spike N439K variants maintain fitness while evading antibody-mediated immunity

Structural basis for broad sarbecovirus neutralization by a human monoclonal antibody

Data processing and analysis with the autoPROC toolbox

Antibodies with potent and broad neutralizing activity against antigenically diverse and highly transmissible SARS-CoV-2 variants

Antibody Resistance of SARS-CoV-2 Variants B.1.351 and B.1.1.7

Generation of VSV pseudotypes using recombinant Δ G-VSV for studies on virus entry, identification of entry inhibitors, and immune responses to vaccines

COVID-19 Weekly Epidemiological Update (World Health Organization)

Overview of the CCP4 suite and current developments

Neutralization of SARS-CoV-2 spike 69/70 deletion, E484K and N501Y variants by BNT162b2 vaccine-elicited sera

Engineering SARS-CoV-2 using a reverse genetic system

Spike mutations decrease SARS-CoV-2 sensitivity to neutralizing antibodies but not ACE2-Ig in vitro

Structural and Functional Analysis of the D614G SARS-CoV-2 Spike Protein Variant

We would like to thank the following: Kristi Huntington (Eli Lilly and Company) and Emma McCarren (AbCellera Biologics Inc.) for project leadership and coordination; Franz Trianna, Douglas Burtrum, Prabakaran Narayanasamy, Xiaomin Yang, Ricky Lieu, Dongmei He, Henry Koo, of Eli Lilly and Company for reagent and antibody cloning, expression and purification; Craig Dickinson, Kristina Coleman and Jeffrey Boyles of Eli Lilly and Company for initiation and reagent generation for crystallography experiments; David Driver of Eli Lilly & Company for assistance with pseudovirus genomic titering; Natalie Thornburg, Kenneth Plante, and the University of Texas Medical Branch (UTMB) World Reference Collection for Emerging Viruses and Arboviruses (WRCEVA) for the WA-1 isolate (deposited by N. Thornburg), San Diego isolate (deposited by N. Thornburg), the Maryland isolate (Mehul Suthar), and the Delta isolate; the SARS-CoV-2/INMI-1-Isolate/2020/Italy used in this publication was kindly provided by the European Virus Archive Goes Global project that has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement no. 653316 Steven Widen of the NextGen Sequencing Core at UTMB for genomic sequencing of the variants used in this study; the SARS-CoV-2/INMI-1-Isolate/2020/Italy used in this publication was kindly provided by the European Virus Archive Goes Global project that has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement no. 653316; Sherie Duncan, Anders Klaus, Keith Mewis, Karine Herve, Amanda Moreira, and Emilie Lameignere of AbCellera Biologics Inc. for technical support; Chad Thiessen of AbCellera Biologics Inc. for development of features for Celium TM required for antibody selection; Clara Ng-Cummings of AbCellera Biologics Inc. for figure generation; Wolfgang Glaesner of Eli Lilly and Company for management and personnel resources. We gratefully acknowledge the authors from the originating laboratories and the submitting laboratories, who generated and shared via GISAID genetic sequence data, on which this research is based.