key: cord-1008031-4ohz92vo
authors: Mullen, Tracey E.; Abdullah, Rashed; Boucher, Jacqueline; Brousseau, Anna Susi; Dasuri, Narayan K.; Ditto, Noah T.; Doucette, Andrew M.; Emery, Chloe; Gabriel, Justin; Greamo, Brendan; Patil, Ketan S.; Rothenberger, Kelly; Stolte, Justin; Souders, Colby A.
title: Accelerated Antibody Discovery Targeting the SARS-CoV-2 Spike Protein for COVID-19 Therapeutic Potential
date: 2021-05-31
journal: bioRxiv
DOI: 10.1101/2021.05.31.446421
sha: b5d064080a6f4b0a584169c64b30c67b99d8df13
doc_id: 1008031
cord_uid: 4ohz92vo

Rapid deployment of technologies capable of high-throughput and high-resolution screening is imperative for timely response to viral outbreaks. Risk mitigation in the form of leveraging multiple advanced technologies further increases the likelihood of identifying efficacious treatments in an aggressive timeline. In this study, we describe two parallel, yet distinct, in vivo approaches for accelerated discovery of antibodies targeting the SARS-CoV-2 spike protein. Working with human transgenic Alloy-GK mice, we detail a single B-cell discovery workflow to directly interrogate antibodies secreted from plasma cells for binding specificity and ACE2 receptor blocking activity. Additionally, we describe a concurrent accelerated hybridoma-based workflow utilizing a DiversimAb™ mouse model for increased diversity. The panel of antibodies isolated from both workflows revealed binding to distinct epitopes with both blocking and non-blocking profiles. Sequence analysis of the resulting lead candidates uncovered additional diversity with the opportunity for straightforward engineering and affinity maturation. By combining in vivo models with advanced integration of screening and selection platforms, lead antibody candidates can be sequenced and fully characterized within one to three months.

The global pandemic caused by severe acute respiratory Ebola Virus in 2014, 3, 4 and Zika Virus in 2015. 5 Lessons learned from these public health threats helped guide the strategy for the accelerated response to COVID-19.

In particular, the understanding that neutralizing antibody function is fundamental to combating disease progression 6 helped streamline early antibody-based drug therapy discovery strategies.

serum titer to the S and S1 proteins at serum dilution factors of at least 1:70,000 or higher (data not shown).

Following a high-efficiency electrofusion to generate hybridoma lines, resulting colonies were initially screened by ELISA to identify S protein-reactive clones. Preliminary clones of interest were subsequently screened via high-throughput biolayer interferometry (BLI) kinetic screening on the ForteBio Octet® system to select candidates for scale up and antibody purification from hybridoma cultures.

Simultaneously, sequencing of immunoglobulin genes was performed following a high-throughput hybridoma sequencing procedure and sequence analysis was performed using the Geneious Biologics software. Screening results from a subset of representative candidates are shown in Table 1 .

Concurrently, a single B cell discovery approach was employed whereby plasma cells were enriched from primary tissues prior to loading onto OptoSelect™ chips with the Berkeley Lights Beacon. Following single cell Table 1 . Binding characteristics of anti-S protein candidates, including ELISA binding, single point kinetic measurements to the trimeric S protein, affinity characterization to the monomeric S1 protein and effect on ACE2 binding to S1 protein (S1:ACE2 interaction blocking activity).

* Values <1E-05 (s-1) are beyond the limit of detection in this experiment and are estimated ** Values <1nM are beyond the limit of detection in this experiment and are estimated *** Calculated as % increase or decrease in signal upon exposure of Antibody:S1 complex to 100nM ACE2 (as compared to Antibody:S1 signal) two separate mouse strains were immunized with the S1 subunit (which contains the receptor binding domain): a humanized strain to facilitate the discovery of fully human antibodies (Alloy GK mice), and an engineered mouse strain designed to elicit greater epitopic diversity and overall immune response (Abveris DiversimAb™ mice). Furthermore, two distinct upstream discovery methods were applied: a hybridoma discovery platform optimized for high-content deposition into NanoPens, assay mixtures containing capture beads and fluorescently labeled target proteins were imported into the channels above the NanoPens.

Throughout the course of the assay, antibody secreted from the plasma B cells diffused from the NanoPen chambers into the channel above. Upon bead binding, fluorescence from either directly labeled protein(s) or secondary detection antibodies was concentrated on the surface of the bead, resulting in the time-dependent development of fluorescent halos in the channels above the pens containing antigen-specific plasma cells ( Figure 1 ).

were exported for immunoglobulin sequence capture and analysis with the Geneious Biologics software. The resulting naturally paired heavy and light chains were cloned into expression plasmids and recombinantly expressed.

Purified antibodies were screened similarly to the strategy used for hybridoma candidates via ELISA and BLI with a subset of representative candidates displayed in Table 1 .

High-throughput and high-content screening on lead candidates was performed on the Carterra LSA to elucidate kinetic profiles to the monovalent S1 protein ( Table 1 and Figure 2 ). Full kinetic profiles were assessed in triplicate under regenerative Assay 1 Assay 2 Mouse IgG capture beads + 200nM S1 Protein-AF647 + 500nM ACE2

Mouse IgG capture beads + anti-mouse IgG-AF488 + 200nM S1 Protein-AF647 Figure 1 . Example Beacon screening data for candidate D59047-11955. Anti-mouse IgG capture beads (brightfield image) were imported into the channel above the pens. Antibody secretion from a single B cell contained within a pen bound to capture beads at the mouth of the pen. In assay 1, antibody secretion was assessed by detection of total IgG in the FITC detection channel via binding of an anti-mouse IgG-AF488 conjugated secondary and simultaneously the specificity for S1 protein was determined in the CY5 detection channel with AF647 conjugated S1 protein at 200nM. In assay 2, binding competition between secreted antibody from the B cell and ACE2 receptor was assessed by precomplexing AF647-conjugated S1 protein with a molar excess of recombinant ACE2. A lack of antibody binding to S1 protein under these conditions demonstrated binding to a similar epitope as ACE2, thus indicative of a potential blocking candidate.

D70678-13531-S1 . Example Carterra sensogram for antibody blocking of the S1:ACE2 interaction. Antibody:S1 protein complex was captured on the chip surface and ACE2 protein binding was assessed. A non-blocking candidate, D70678-12637-S1, complexed with S1 protein does not inhibit interaction with ACE2 (green). Conversely, a blocking candidate, D59047-11955, prevents ACE2 binding when S1 is complexed with the antibody (yellow).

D70678-12637-S1 Figure 4 . Select antibodies were characterized for competitive binding to the S1 protein to elucidate epitopic coverage. (a) All antibodies capable of binding S1 protein and preventing the S1:ACE2 interaction (ID highlighted in yellow) focused on a similar epitope (competitive binding indicated by red squares in grid). Antibodies that did not function as ACE2 blocking candidates (ID highlighted in light green) were distributed across two distinct core epitopes, with some antibodies binding at the interface of these epitopes, and an additional two clones appeared to bind distinct epitopes. Interestingly, an intermediate S1:ACE2 blocking candidate (1E5; ID highlighted in beige) bound at the interface between the ACE2 blocking epitope and a separate non-blocking epitope, thus supporting the partial blocking observation. No direct correlation was observed between affinity and binding epitope when assessed in either the monovalent binding format to the monomeric S1 protein or avid binding to the trimeric S protein.

(b) A community network plot illustrates the bin clustering and distinct binding regions for each group of candidates.

and non-regenerative conditions with both purified antibodies and crude supernatant samples. Target S1 protein was used as an analyte in an ascending concentration series ranging from 0.49nM to 500nM with four-fold dilutions ( Figure 2 ).

All conditions (regeneration vs. non-regeneration and purified vs. crude antibody samples) yielded similar affinity values. The average resulting values from all asays are reported in Table 1 .

Additionally, candidates were assayed on the Carterra LSA for the ability to block the S1:ACE2 binding interaction (Table 1 and Figure 3 ). ACE2 receptor blocking activity was interrogated by forming an antibody:S1 protein complex on the chip surface in a sequential format. Following complex formation, ACE2 was introduced as an analyte at 100nM ( Figure 3 ) and the percent increase in RU value as a result of ACE2 binding was quantified using the RU signal from the antibody:S1 complex formation as the baseline. a b

The average of triplicate measurements is reported in Table 1 . Lastly, candidate antibodies were assayed in a classical binning competition format ( Figure 4 ) to identify the relative binding epitopes on the S1 protein. Epitope binning was performed on the Carterra LSA in a sequential format and Table 2 . Heavy and light chain family and full sequence information for select characterized candidate antibodies.

binding of each antibody combination was assessed simultaneously. Non-competitive binding pairs and competitive binding pairs are highlighted by green or red squares, respectively, in Figure 4a (self interactions shown in dark red squares). The data is also presented as a community network plot in Figure 4b to better visualize the relative binding epitopes among the candidates assessed.

Sequence information for each candidate is presented in Table 2, including full paired heavy and light chain variable regions along with V-region and mutation rate analysis. Table 3 highlights common in silico liability assessments for each candidate based on published motifs 10,11 for antibodies. overall diversity. 17 Alternative immunization workflows involving rapid schedules and/or immunogen manipulation are effective risk mitigation strategies to circumvent these challenges. 18 In this case, the DiversimAb mice were immunized following an accelerated strategy to maximize epitopic diversity, and although the overall average affinity of these candidates was lower as a result, a subset of high affinity blocking clones were discovered. It is also important to note that recent studies have identified non-blocking neutralizing epitopes, 19 indicating further testing of non- 

Prior to Beacon workflows, recombinant S1 (Acro) and ACE2

(Acro) proteins were labeled at primary amine sites with AF488 or AF647 using Alex Fluor 488 TFP Ester or Alexa Fluor 647 The chips were imaged using a FITC and/or Cy5 filter cube. For sequential assays, the bead assay mixture was flushed from the chip before loading a second assay mixture. All assays were automatically scored and confirmed with human verification. Both non-regenerative (single cycle) and regenerative (multicycle) kinetics analyses were performed using monomeric S1

protein analyte with six sequential injections at different concentrations ranging from 0.46nM to 500nM in HBSTE + 0.1% BSA running buffer. Analyte was injected for 5 minutes to measure association followed by 15 minutes of buffer injection to assess dissociation. For multi-cycle kinetics, standard regeneration conditions were used with alternating injections of 10mM glycine pH2.5 and running buffer.

To assess ACE2 blocking profiles, a single injection of 100nM ACE2 in running buffer was performed following S1 binding at 500nM to the captured antibodies on the chip surface.

Using the RU value of antibody:S1 complex as a baseline, the increase in RU response due to ACE2 binding was calculated as a percentage of the baseline. Kinetic and blocking data were analyzed using Kinetics software v1.6 and association and dissociation rate curve fits were calculated using a 1:1 Langmuir binding model.

An epitope competition assay was performed in the classical sandwich format whereby the captured and crosslinked array of experimental and control antibodies was sequentially exposed to the following series of analyte injections: 200 µg/ml polyclonal mIgG for 5 minutes to block additional surface sites, 100nM S1 protein in running buffer for 5 minutes, 10 µg/ml (purified) or 1:1 dilution (supernatant) in running buffer of an individual experimental or control antibody. Each injection series was preceded by and ended with a buffer injection.

Standard regeneration conditions were used between each full cycle. Analyte injection series were repeated until all experimental and control antibodies were assessed as analytes. Crude supernatant and purified antibodies were assessed in bi-directional format as both ligands and analytes. Binning data was analyzed using Epitope software v1.6 with manual verification to generate clustered heat map and community network diagrams. Competition for S1 binding was based on normalized signal to the S1 binding step for each interaction and self-self-competition was used as a baseline for determining positive binding.

Samples with sufficient S1 protein capture and adequate regeneration profiles were reported with bi-directional data, otherwise a unidirectional interaction (analyte only) was reported.

Proof of principle for epitope-focused vaccine design

Recent advances in therapeutic applications of neutralizing antibodies for virus infections: an overview

Deconstructing the Antiviral Neutralizing-Antibody Response: Implications for Vaccine Development and Immunity

Deamidation and isomerization liability analysis of 131 clinical-stage antibodies

Susceptibility of Antibody CDR Residues to Chemical Modifications Can Be Revealed Prior to Antibody Humanization and Aid in the Lead Selection Process

Recent insights for the emerging COVID-19: Drug discovery, therapeutic options and vaccine development

Pandemic Preparedness: Developing Vaccines and Therapeutic Antibodies For COVID-19

Understanding and Manipulating Viral Immunity: Antibody Immunodominance Enters Center Stage

designed and executed experiments, performed data analysis and reporting, and participated in manuscript review

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Pandemic preparedness and response--lessons from the H1N1 influenza of 2009

Discovery of an antibody for pan-ebolavirus therapy

Isolation of potent neutralizing antibodies from a survivor of the 2014 Ebola virus outbreak

The Zika Virus Epidemic in Brazil: From Discovery to Future Implications

Neutralization of Virus Infectivity by Antibodies: Old Problems in New Perspectives

Immunodominant antibody germlines in COVID-19

Identification of immunodominant linear epitopes from SARS-CoV-2 patient plasma

Restricted Clonality and Limited Germinal Center Reentry Characterize Memory B Cell Reactivation by Boosting

Antibody repertoire analysis of mouse immunization protocols using microfluidics and molecular genomics

Receptor-binding domain-specific human neutralizing monoclonal antibodies against SARS-CoV and SARS-CoV-2. Signal Transduct Target Ther

Genetic background and immunological status influence B cell repertoire diversity in mice. Sci Rep

Systems Analysis Reveals High Genetic and Antigen-Driven Predetermination of Antibody Repertoires throughout B Cell Development

Development of therapeutic antibodies for the treatment of diseases

Electrofusion of cells: hybridoma production by electrofusion and polyethylene glycol

We would like to thank Alloy Therapeutics for supplying the GK mice, Carterra for assistance and use of the LSA and Gary Ng of Abveris for preparation of the manuscript.

T.E.M. and C.A.S. conceived of and designed experiments, performed data analysis, and authored the manuscript. R.A., J.B.,