Screening Antibodies Raised against the Spike Glycoprotein of SARS-CoV-2 to Support the Development of Rapid Antigen Assays

Severe acute respiratory coronavirus-2 (SARS-CoV-2) is a novel viral pathogen and therefore a challenge to accurately diagnose infection. Asymptomatic cases are common and so it is difficult to accurately identify infected cases to support surveillance and case detection. Diagnostic test developers are working to meet the global demand for accurate and rapid diagnostic tests to support disease management. However, the focus of many of these has been on molecular diagnostic tests, and more recently serologic tests, for use in primarily high-income countries. Low- and middle-income countries typically have very limited access to molecular diagnostic testing due to fewer resources. Serologic testing is an inappropriate surrogate as the early stages of infection are not detected and misdiagnosis will promote continued transmission. Detection of infection via direct antigen testing may allow for earlier diagnosis provided such a method is sensitive. Leading SARS-CoV-2 biomarkers include spike protein, nucleocapsid protein, envelope protein, and membrane protein. This research focuses on antibodies to SARS-CoV-2 spike protein due to the number of monoclonal antibodies that have been developed for therapeutic research but also have potential diagnostic value. In this study, we assessed the performance of antibodies to the spike glycoprotein, acquired from both commercial and private groups in multiplexed liquid immunoassays, with concurrent testing via a half-strip lateral flow assays (LFA) to indicate antibodies with potential in LFA development. These processes allow for the selection of pairs of high-affinity antispike antibodies that are suitable for liquid immunoassays and LFA, some of which with sensitivity into the low picogram range with the liquid immunoassay formats with no cross-reactivity to other coronavirus S antigens. Discrepancies in optimal ranking were observed with the top pairs used in the liquid and LFA formats. These findings can support the development of SARS-CoV-2 LFAs and diagnostic tools.


Introduction
The appearance of a novel coronavirus disease 2019 (COVID- 19) was first reported in the city of Wuhan, Hubei Province, China in 2019 1 . Since then COVID-19 has progressed to pandemic levels with over 23 million reported cases including at least 800,000 associated deaths reported globally 2 . The pathogen responsible is the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a novel betacoronavirus. The coronaviruses are enveloped positive-stranded RNA viruses that are 70 -90 nm in size and characterized by a crown like morphology associated with the display of spike (S) glycoproteins on the host membrane-derived and lipid bilayered viral envelope 3;4 . The structure of the S glycoprotein of SARS-CoV-2 has been resolved and is known to be essential for the viral infection of the host cell via its binding to the cellular receptor angiotensin-converting enzyme 2 to promote fusion and entry into the cell 5 . The S glycoprotein is poorly conserved across coronaviruses, with 85.3% of the antibody epitopes found in SARS CoV-2 S protein considered unique 6;7 . Conversely, high conservation is noted across SARS CoV-2 isolates from Europe, Asia, and the US, resulting in an antigen that offers greater specificity over more conserved targets like the N antigen.
The rapid spread of COVID-19 has resulted in an urgent need for effective diagnostic tests to support disease management, monitoring, surveillance and pandemic control against SARS-CoV-2 8 . In high income countries molecular testing, typically using real time reverse transcription PCR (RT PCR), has been the primary test method implemented to diagnose SARS-CoV-2 in both symptomatic and asymptomatic cases but the accurate detection of early infection remains challenging giving false negative results 9;10 . As of August 24 th , 141 commercial or clinical laboratory derived molecular tests have been granted emergency use authorization (EUA) in the USA by the Food and Drug Administration (FDA) 11 . The vast majority of the tests are predominantly unsuitable for use at the point of care as many are in open assay format with which significant engagement are needed from skilled operators to prepare the samples for testing, prepare test reagents, operate complex equipment and finally to process the data and interpret the test results. Automated high throughput molecular platforms are available and are capable of processing large numbers of samples with significantly reduced operator input [12][13][14][15] . However, acquiring and operating such equipment comes with high capital costs and a need for appropriate infrastructure, not only for housing the equipment and reagents but also requiring effective specimen collection and transport and the reporting of test data to patients, clinicians, and health care programs after processing. In the current pandemic, global demand for has affected all countries and so sufficient access to reagents, consumables and more other materials such as personnel protective equipment, swabs and transport media is necessary ensure consistent testing 16 .
Lack of access to key reagents and consumables has highlighted that there is a market for SARS-CoV-2 diagnostic immunoassay-based lateral flow assays (LFAs) in high income countries. Low-and middle-income countries (LMICs) already faced serious constraints in diagnostic capacity and accessibility before the COVID-19 pandemic stuck. SARS-CoV-2 will have an amplified effect in these countries that have limited access to care with and already greater burden of infectious diseases 17 . LMICs lack time and finances for the swift uptake of new diagnostic technologies. Furthermore, a lack of resources and skilled laboratorians limits the number of test facilities, the ability to scale testing, while access to critical reagents is limited as high-income countries dominate procurement, culminating in inability to perform molecular tests at the scale required 18 . Without access to expanded molecular test capacity and capability, other diagnostic tools must be developed to support COVID-19 infection control. Therefore, LFAs serve as a best alternative in regions lacking sufficient access to widespread molecular testing for SARS-coV-2.
For detection and control of COVID-19 in LMICs, an antigen LFA format makes a more viable option to the serologic LFAs that currently dominate the market due to their ability to detect SARS-CoV-2 directly and earlier in the infection process. Serology-based assays are insensitive in early infection requiring individuals to be diseased for at least a week before the antibody response can first be detected (IgA, IgM, and/or IgG) 19 , which is enough time for infected individual to unknowingly spread the disease 20 . In terms of operation and cost, LFAs can be manufactured at a very large scale and at a relatively low cost per unit in comparison to molecular tests. While LFAs typically require some limited training of users, they are easy to use, give a test result in minutes, most do not require associated equipment and their use is broadly disseminated from hospitals to clinics to community-and home-based testing (e.g. malaria, HIV-1, and pregnancy testing).
The performance of antigen detection LFAs is variable depending on the performance of the antibodies used in the test and while visually read LFA reach the level of sensitivity that molecular assays offer, the use of readers can further increase test senstivity. The recent FDA EUA to Lumira Diagnostics (Stirling, UK) for their SARS-CoV-2 Ag assay has claims of a sensitivity of 97.6% as compared to RT PCR testing. Therefore, rapid antigen assays using high performance antibodies can offer may offer sufficient clinical sensitivity to detect infectious patients in decentralized settings where molecular testing is not readily available today. Furthermore, the LFA format can be manufactured at extremely high volumes and very low costs, and can offer increased testing capacity in LMICs where molecular testing is not readily or sufficiently. Other markets where LFAs can play a key role is in disseminated testing models such as employed in community-and home-based testing, and self-testing [21][22][23] .
The WHO's recently released target product profile for a point of care test for suspected COVID-19 cases (e.g. a rapid antigen assay) has listed the acceptable characteristics for sensitivity and specificity at ≥70% and ≥97%, respectively 24 . A current challenge to antigen test development is understanding the performance of the SARS-CoV-2 antibodies that are on or entering the market, with the screening of large numbers of unqualified antibodies a resource sink for developers aiming to develop direct antigen tests. Abundant targets include the four major structural proteins: the spike (S), membrane (M), envelope (E) and the nucleocapsid (N) proteins. The S glycoprotein represented an attractive candidate due to the unique structural changes relative to SARS-COV1 and other seasonal coronaviruses, offering the potential of high specificity for SARS-COV2 6 . Scientific), and the incorporation ratio for each label was measured. Briefly, the concentration of biotinylated antibodies after desalting was measured at 280 nm via spectrophotometer (Nanodrop ND-1000, ThermoFisher Scientific); biotin incorporation was measured using a Biotin quantification kit (Pierce™, ThermoFisher Scientific). For measuring the incorporation of the SULFO-TAG, the protein concentration was estimated using the bicinchoninic acid (BCA) protein assay (ThermoFisher Scientific), and the SULFO-TAG label spectrophotometrically measured at 455 nm.

Preparation of U-plex plates
The biotinylated capture antibodies were coupled via biotin-streptavidin binding to U-PLEX linkers. To prepare the capture antibody arrays, up to 10 antibody-linker conjugates were pooled together in U-PLEX stop buffer at a concentration of 0.29 µg/mL per antibody, and 50 µL of this mixture was added to individual wells of the U-PLEX plates. The plates were incubated for 1 hour with shaking (500 rpm) to allow the antibody array to selfassemble to the complimentary antibody linker binding sites and unbound material then removed by washing 3 times with 150 µL/well of phosphate buffered saline + 0.05% Tween 20 (PBS-T, pH 7.5) using a BioTek ELX405R microplate washer (BioTek Instruments Inc., Winooski, VT, USA).

Processing U-plex plates
Appropriate serial dilutions of the trimeric S glycoprotein in Diluent 100 (MSD) were prepared. Clinical specimens and cell lysates were prepared by adding 25 µL into 25 µL of Diluent 100. The 50 µL of each prepared sample was added to each antibody array well in the U-PLEX plate, and incubated with shaking for 1 hour at room temperature. Plates were washed 3 times in 1X PBST and then 25 µL of 2 µg/mL SULFO-TAG-labeled detection antibody in Diluent 3 (MSD) was added to each well with incubation for an hour with shaking. Plates were then washed 3 times to remove excess detection reagent and the wells filled with 150 µL of 2X read buffer T (MSD). The plates were inserted into the MESO QuickPlex SQ 120 plate reader (MSD) and the electrochemiluminescence (ECL) from each individual array spot was subsequently measured. In the absence of a control, the array spot that gave the highest signal to noise in each plate was expressed as 100% and each of the array spots in each plate expressed as percentile of this value. When serial dilutions of the S glycoprotein were used to generate a calibration curve, the relationship of ECL signal to S glycoprotein concentration was then fitted to a four-parameter logistic (4-PL) function in the Discovery Workbench v4 program. S glycoprotein concentrations for gamma-irradiated SARS-CoV-2 were calculated by back-fitting ECL signals to the 4-PL fit.

Antibody Evaluation
The identification of the optimal antibody pairs for capture and detection of the S glycoprotein was determined via a three-stage process using the MSD immunoassay platform. MSD U-plex plates with a 10-plex array/well format were prepared for capture antibody binding as above. Antibodies were screened in a matrix format, acting both as capture and detector antibody. Round 1. All 41 AbCellera antibodies were screened together in a matrix format using 10 ng/mL of trimeric S glycoprotein antigen (Acro Biosystems) in triplicate. The capture and detection antibody pairs that recorded 25% or greater ECL per plate were further evaluated over a greater range of S antigen concentration (1000, 100 and 10 ng/mL) to verify the initial results. The highest ECL readings across each concentration ranges were then used to rank antibodies for round 2 screening. Round 2. Six antibody candidates from round 1 were evaluated further in a matrixed format alongside 3 antibodies from Sino Biological using 7-point dilutions of the S glycoprotein antigen in diluent 100 (ranging from 1250 to 0.016 pg/mL) in duplicate. Antibody pairs were ranked in terms of the limit of detection (LOD). Specificity was evaluated by challenging the pairs with irradiated viral cultures of SARS-CoV-2 and other human coronavirus species at concentrations equivalent to 10 4 TCID50/mL or PFU/mL in Diluent 100.
Round 3: An additional 4 antibodies from Leinco were evaluated in a matrix format with the 4 best performing antibodies from round 2 and 2 from round 1. Their analytical sensitivity was evaluated by challenging the antibody pairs with a 7-point calibration curve of the S antigen, and a dilution series of the irradiated SARS-CoV-2. Specificity was evaluated by challenging the pairs with irradiated viral cultures and supernates of other human coronavirus species (OC43,229E, MERS and SARS) at concentrations equivalent to 10 4 TCID50/mL or PFU/mL in Diluent 100. Antibody pairs were ranked in terms of LOD and ECL signal with the best performing pair further evaluated for clinical sensitivity and specificity with 53 clinical specimens.

Antibody/antigen evaluation by SDS-PAGE
Antigens were evaluated for purity and size using SDS-PAGE. Concentration was measured for all proteins using BCA assay (ThermoFisher Scientific ). Samples were premixed NuPAGE™ LDS 4x Sample Buffer (ThermoFisher Scientific) and heated at 70 o C for 10 minutes. Gels with a 4-12% Bis-Tris gradient were used to achieve separation. Novex Sharp Pre-stained protein standard (ThermoFisher Scientific) was used as a molecular weight marker Coomassie Imperial™ Protein Stain (ThermoFisher Scientific) was used to stain each gel and visualize protein bands.

Latex bead conjugation
For both test and control line detection conjugates, 400 nm carboxylic blue latex beads (Magsphere, Pasadena CA, USA) were washed three times with 0.1 M MES buffer (pH 6). Then, latex beads were activated using 1ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride / N-hydroxysuccinimide (ThermoFisher Scientific) coupling reagents at 0.15 and 10 mg/mL respectively for 30 minutes. Afterwards, the blue latex particles were conjugated in 1× PBS (pH 7.2) to various anti-spike antibodies at a w/w ratio of 20:1 and 10:1 (bead: antibody) for test and control line antibodies, respectively, for three hours. Finally, latex conjugates were quenched using 0.1 M ethanolamine before being washed and blocked with 6% (w/v) casein in H2O (preparation method is proprietary), final concentration 1.2%, overnight. The latex conjugates were stored in buffer containing 50 mM borate (pH 8.5) and 1% casein. The latex conjugates were quantified using the spectrophotometer by measuring absorbance at 660 nm and comparing to absorbance of unconjugated beads.

LFA reagent deposition and strip assembly
Unlabeled capture antibodies were diluted to 1 mg/mL in 1× PBS (pH 7.4) with 2.5% (w/v) sucrose, and were striped at 1 µL/cm (ZX1010, BioDot, Irvine, CA, USA) on nitrocellulose CN95 (20 mm wide, CN95, Sartorius Lab Instruments GmbH & Co. KG, Otto-Brenner-Straße 20, Göttingen, Germany) and dried at 25°C for 30 min. The control line was 0.75 mg/mL Donkey anti-Chicken IgY (Jackson ImmunoResearch, West Grove, PA, USA), striped at 1 µL/cm. The test and control lines were located at 8 mm and 13 mm from the upstream edge of the nitrocellulose membrane. For antibody screening, the nitrocellulose was left unblocked.
Card assembly was performed on a clamshell laminator (Matrix 2210, Kinematic Automation, Sonora, CA, USA). Pads were placed on the backing card in the following order: nitrocellulose, cover tape, conjugate pad, sample pad, wicking pad. Individual strips (3.3 mm wide) were cut with a Matrix 2360 sheet cutter (Kinematic Automation, Mono Vista, CA, USA) and assembled in cassettes (proprietary design) using an assembly roller (YK725, Kinbio Tech Co., Shanghai, China).

Hamilton screening procedure for LFA screening of antibodies
Antibody pairs were screened on an integrated robotic system we have previously used to test antibody performance directly on nitrocellulose 25;27 . In this system, the Hamilton STAR automated liquid handling robot (Hamilton Company, Reno, NV, USA), camera (IDS UI-1460SE-C-H detector with a Tamron M118FM16 lens) custom LFA holders, and custom control software developed in-house were combined to allow rapid screening of antibody pairs directly in LFA format. The robot used 8-channel pipetting for parallel application to LFAs and the camera for imaging. The custom LFA framework held a maximum of 96 LFA cassettes per robot run. The custom control software applied 1 µL of latex bead conjugate mix (0.15% anti-spike -latex bead, 0.1% or 0.05% Chicken IgY latex bead in 50mM borate [pH 8.5]) to the conjugate pad in the LFA. After a 10-minute delay to let the conjugate mix dry, 75 µL of sample diluted in 2.5% BSA in PBST, spike glycoprotein or buffer (2.5% BSA in PBST or 2.5% BSA and 1% IGEPAL in 1× PBS) was added to the sample pad. Images were acquired 20 minutes after sample addition. Four technical replicates were run for each antibody pair per sample type.

Screening recombinant antigens on LFAs
LFAs were screened across two rounds using a recombinant spike glycoprotein as the as the antigen target. The first, with the best-available at the time spike antigen (from Sino Biological), at 80 ng/mL. The second round used a different recombinant antigen produced in house was subsequently determined preferable, was also used at a concentration of 80 ng/mL,. A complete list of all pairs screened from all rounds is in Table 1si (suppl. info).

Liquid immunoassay screening
All of the data generated from screening antibodies using the liquid platform in the following section is publicly accessible. 29 A total of 48 human monoclonal antibodies (AbCellera, 41; Sino Biological, 3; Leinco, 4) were assessed for their performance as capture and detection antibodies for the SARS CoV-2 S glycoprotein using the MSD U-PLEX immunoassay format across 3 rounds of testing. Each well in a 96 well U-PLEX plate can host 10 different capture antibodies in a geometric planar array by assessing ten capture antibodies per well (960 per plate) enabled rapid screening of multiple combinations to identify the most promising candidate pairs that would enable sensitive and specific capture and detection of SARS CoV-2.
In the preliminary evaluations, a recombinant S glycoprotein antigen expressed from insect cells (BEI) was used to screen the AbCellera antibodies however, this particular antigen resulted in the generation of very low ECL signals, at the concentration used. We postulate that as the post-translational modifications that can arise during antigen production will differ between insect cells and mammalian cells, the antigen initially used may have had or lack modifications that made it unsuitable for our study 30 . To identify an antigen most suitable for this work we evaluated 3 recombinant S glycoproteins across a range of dilutions (1000 to 0.24 pg/mL) using AbC525 and AbC397 as capture and detector respectively; this pair had generated the strongest ECL in the preliminary screen. The signal intensities and LOD varied with respect to each of the three antigens used. The mammalian cell-derived recombinant S glycoprotein from Acro Biosystems produced the strongest and more consistent signal as compared to the baculovirus expressed antigens, and the lowest LOD ( Figure 1). Thus, this antigen was selected for use as the standard in all antibody screens.
In round 1, 41 antibodies from AbCellera were assessed in both capture and detector format (1681 unique antibody pairings in total) using a low S glycoprotein antigen concentration of 10 ng/mL to allow for more stringent down-selection. Table 1 summarizes the Round 1 screening results in a matrixed array for each antibody combination. In the absence of a positive control assay, the ECL values from each array spot in each well were normalized based on the percentile of signal-noise (S-N) in each plate versus the spot with the maximum S-N produced in each plate. A total of 117 (7.0%) antibody pairs produced at least 25% of the maximum signal (marked in blue). These pairs, that consisted of 20 capture and 23 detection antibodies, were then further screened in a total of 460 combinations with S antigen in a range of 10, 100 and-1000 ng/mL to confirm the initial results ( Figure 2). The ten antibody pairs that generated highest ECL signals were selected for evaluation in round 2, and included two capture antibodies (AbC447 and AbC525) and five detector antibodies (AbC513, AbC518, AbC459, AbC447 and AbC511). No self-pairing antibodies were identified presumably due to the presence of only single epitope on the recombinant antigen that would limit binding to only one form of the respective labeled antibody.  In round 2, the 10 AbCellera optimal antibody pairs were assessed further in a matrix format alongside three antibodies from Sino Biological (MM443, MM57 and D003). Screening with an 8-point standard curve indicated that the Sino Biological antibodies resulted in higher ECL signals and lower LODs than the best AbCellera pair (AbC447/AbC513) ( Table 2). Notably the Sino 447/MM43 and 447/D003 pairs exhibited similarly low LODs at 43 and 45 pg/mL respectively, in addition to the highest ECL signals at when challenged with ≥ 625 ng/mL of trimeric S antigen. Antibody pairs AbC447/MM43 and AbC447/D003 were then further challenged with a range of concentrations of SARS-CoV-2 virions, both  In round 3, four antibodies procured from Leinco (L2215, L2355, L2381 and L2838) screened with the 6 top antibody candidates identified from round 2 (D003, MM42, AbC447, and AbC513) and round 1 (AbC353 and AbC525). When used either as a capture or detector, the Leinco antibodies typically generated higher ECL signal and lower LODs than previously observed (Table 2), many with the LOD generally 5-10 times lower than for the best performing antibody in round 2. The L2381/MM43 and L2355/L2215 combinations had near identical LODs at 3 and 4 pg/mL respectively, with L2355/L2215 were selected for further study due to greater affinity to the target as indicated by significantly higher ECL signal when challenged with S antigen at 625 ng/mL ( Table 2). The antibody pair L2355/L2215 was challenged with a titered SARS-CoV-2 (BEI), resulting in the generation of a dose-dependent curve ( Figure 3) with an estimated LOD of 2 TCID50/mL virions or 7.4 x10 3 genome equivalents/mL. To demonstrate assay performance of the L2355/L2215 antibody pair with clinical samples, a panel of fifty-three de-identified clinical samples, comprising of 20 COVID19-negatives and 33 COVID19-positives were used to challenge the assay. Of these, 44 of the 53 samples were correctly identified as either positive or negative ( Table 5). The viral load of the specimen was important as nine positive samples, each with a cycle threshold of > 29.5, were incorrectly scored as negative. This was likely in part due to dilution of the sample as each nasopharyngeal swab was collected in 3 mLs of viral transport medium. Overall the assay had a sensitivity and specificity of 73% and 100% respectively (

Specificity screening of top candidate pairs with other species of coronavirus in liquid immunoassays
The cross-reactivity of the antibody pairs from rounds 2 and 3 ( Table 2) were also evaluated by challenging them with alpha-and betacoronavirus isolates including inactivated MERS and SARS virions and human coronaviruses OC43 and 229E cell culture lysates at concentrations equivalent to 10 4 TCID50/mL or 10 4 PFU/mL. None of the ten antibody pairs showed any cross-reactivity with other human CoV indicating a high specificity towards SARS-CoV-2.
Candidate screening via lateral flow assays.
The candidate antibodies were also evaluated in the LFA format in two rounds of screens, to assess if the performance of the candidate antibodies varied between the liquid and LFA test formats. A total of 8 antibodies (AbC131 from AbCellera, D003 from Sino Biological and 6 other antibodies from Sino Biological and Creative Diagnostics) were evaluated in Round 1 on LFAs in an 8 × 8 matrix (64 unique pairs, see Table  1si). For each pair, one antibody was striped on nitrocellulose as a test line (the "capture" antibody) and the other was coupled to latex nanoparticles using EDC/NHS chemistry (the "detector" antibody). The results of the first round are given in Figure 5(A). The positive control used round 1 was 80 ng/mL S glycoprotein from Sino Biological, selected due the presence of both the S1 and S2 domains of the native spike trimer. The negative control was 2.5% BSA in PBST.  Table 2si). "Sino Bio" antigen was sourced from Sino Biological and "in-house spike" recombinant antigen was produced in and purified at Global Health Labs.
After the first round, the best five pairs were D003/D002, D004/D002, D001/D004, D004/D001 and D003/D001 (index 564, 589, 511, 568, and 563, Table 3). Each of the top pairs from round 1 consisted exclusively of antibodies from Sino Biological, which was unsurprising considering recombinant antigen choice and the fact that most antibodies screened were from Sino Biological. As with the liquid immunoassay screen, selfpairs did not perform well, as expected, a consequence of the monomeric recombinant antigen likely containing a single copy of the target sequence. However, we would expect self-pairs to do better against the native antigen in clinical samples because it is trimeric. After round 1, 57 anti-S pairs were eliminated and the top seven pairs carried to round 2, along with 22 new antibodies. These new antibodies included the 12-top performing AbCellera antibodies from round 1 liquid immunoassay screen, MM43 from Sino Biological, and 9 antibodies from Leinco Technologies, including the 4 antibodies already screened with liquid immunoassay (Figure 1si). The grid for round 2 was larger at 26 × 26 (616 pairs), however limited access to material meant 60 pairs were ultimately excluded ( Figure 1si). Results from round 2 are shown in a scatterplot in Figure 5(B). The positive control used here was a trimeric spike glycoprotein produced inhouse, considered superior to the recombinant form due to its ability to better mimic the protein folding seen in native structures. The negative control used was 2.5% BSA in PBST. Based on S/N and S-N metrics, the five best performing antibody indices from round 2 were 259, 262, 440, 284, 523, and 610 (Table 3).

Discussion
In this paper, we present the screening of a panel of antibodies targeting the S glycoprotein of SARS Cov-2 to identify candidate capture and detector pairs that may be suitable for development of LFA antigen detection assays. We gained access to a large private collection but with limited access to sufficient materials resulting in some antibodies being screened in one assay and not the other. Commercially available antibodies were typically screened on both formats. A key to this work is the availability of a good native antigen proxy, and as antigen sources can vary considerably, it is important to assess them prior to commencing work. Using the highly sensitive MSD immunoassay platform we are able to achieve an analytical sensitivity in the range of to 7.4 x 10 3 genomic copies/mL and a specificity of 100% when using a limited specimen panel. The TPP for a test for diagnosis or confirmation of acute or subacute SARS-CoV-2 infection, suitable for low or high-volume needs notes a sensitivity of under 1000 copies which this test does not currently meet. However, the intent of this project was to screen antibodies that have optimal potential for implementation in LFAs, and not to develop a diagnostic assay. If necessary, the platform can use a further enhance signal format not used here, the S-PLEX which MSD claim can further improve sensitivity by 10 -100X or into the lower femtogram range. While the S glycoprotein is less abundant that the N protein, there may be utility for combining S as a target to create highly sensitive multiplex immunoassays, with its additional distinct epitopes enabling improved accuracy, especially at lower limits of detection 31;32 .
The liquid assays identified pairs that gave an analytical sensitivity to the S antigen into the low picogram range, a tenfold improvement over previous N immunoassays reported for SARS but the ECL detection feature of the MSD device does also offer greater sensitivity over traditional colorimetric detection employed by most enzyme immunoassay methods 33;34 . Interestingly, the assay format had a distinct effect on the optimal candidate pairs identified. The L2135 clone was the best antibody in either format and as both capture and detector. In contrast, no AbCellera antibodies showed good performance in the liquid assay, though in the LFA format AbC459 was present as capture or detector in 4/5 top pairs. The use of a different source of recombinant antigen may have played a role in this as we did observe some difference in binding using mammalian recombinant sources of antigen. This finding serves as an insight to LFA developers wherein screening of all antibodies should be performed on nitrocellulose rather than using traditional liquid immunoassays. The best antibodies candidates screened in the liquid format appeared to be highly specific to SARS-CoV-2 as they were not reactive with SARS, MERS and OC43 HCoVs that are in the same genus as SARS-CoV-2 35;36 . While we did not have access to HKU1, another beta-CoV species associated with respiratory illness, we do expect it is unlikely to be reactive as the other more closely related beta-CoVs screened were non-reactive.
On the LFA platform, the best pairs, as measured by S/N and S-N, were from a combination of vendors (e.g. AbCellera, Leinco, and Sino Biological), likely because these high-affinity antibodies were raised via unique processes and therefore recognize different epitopes on the antigen.
Interestingly, the liquid and LFA formats did identify very different optimal pairs for the detection of the S antigen. Restricted resources meant that entire antibody sets could not be fully evaluated on both platforms but it was evident that some pairs were better suited to one format over the other. In the liquid format, none of the AbCellera antibodies were in the top candidates as either capture or detector by round 3 but with the LFA, AbC459 and AbC525 were represented in several optimal pairings (Table 2). With the Sino Biological antibodies a similar trend was noted wherein no candidates shone with the liquid immunoassay format while the LFA had two, D001 and D002 (Tables 2 and 3). Antibodies from Leinco were highly represented in the optimal liquid assay design with each of the top five pairs having at least one Leinco antibody in the pairing. By contrast, with the top five candidates in the LFA format, three pairs used a single Leinco antibody, L2355, either as capture or detector, though in combination with differing antibodies to the liquid format.
Our goal is to qualify reagents and methods that are publicly available to any developer who sees value in their use, removing the need for them to invest time and resources on antibodies with little or no potential. Further work is ongoing with our groups to develop a POC LFA with the potential for manufacturing at scale. An advantage of using recombinant antibodies like those from AbCellera and Leinco is that the variable antibody region of single antigen-specific memory B cells derived from convalescent patients is cloned into an expression vector enabling cost efficient scaled production of antibodies. In addition, this work uses recombinant IgG antibodies which are monomeric; with the possibility of manipulating the same variable region sequences to create recombinant IgM type antibodies, decameric forms of which may improve capture and/or detector efficiency leading to more effective rapid antigen assays for COVID-19 diagnosis.

Disclaimers
Other than funding the PATH authors salaries, the BMGF did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. This research was, in part, funded by the U.S. Government. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the U.S. Government.