A potent SARS-CoV-2 neutralizing human monoclonal antibody that reduces viral burden and disease severity in Syrian hamsters

The emergence of COVID-19 has led to a pandemic that has caused millions of cases of disease, variable morbidity and hundreds of thousands of deaths. Currently, only remdesivir and dexamethasone have demonstrated limited efficacy, only slightly reducing disease burden, thus novel approaches for clinical management of COVID-19 are needed. We identified a panel of human monoclonal antibody clones from a yeast display library with specificity to the SARS-CoV-2 spike protein receptor binding domain that neutralized the virus in vitro. Administration of the lead antibody clone to Syrian hamsters challenged with SARS-CoV-2 significantly reduced viral load and histopathology score in the lungs. Moreover, the antibody interrupted monocyte infiltration into the lungs, which may have contributed to the reduction of disease severity by limiting immunopathological exacerbation. The use of this antibody could provide an important therapy for treatment of COVID-19 patients.


Introduction
The emergence of a novel coronavirus disease , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was first described in December 2019 in Wuhan, China 1 .
Symptoms include cough, dyspnea, chest tightness and fever, plus occasional adverse gastrointestinal (GI) disturbances. In a vulnerable subset of patients, the disease often progresses to an atypical pneumonia with high morbidity and mortality rates [2][3][4][5][6] . Mortality based on case fatality rates have ranged from early estimates of 3.9% in Hubei province, China, to 0.9% in populations from countries with widespread testing. Recent estimates of infection fatality rates (IFRs) are between 0.6% to 1% and overall indicate that COVID-19 has a 10-fold greater mortality than seasonal influenza [6][7][8][9][10] . Alarmingly, the IFR is considerably higher (5.6%) in individuals ≥65 years of age where the prognosis for those hospitalized patients ranges from guarded to poor; they frequently require long periods of breathing support, with mortality rates of 20-30% 6,11 .
Consequently, the effect on healthcare resources is of major concern and there is an increasingly unmet medical need for effective therapies, particularly for older patients with comorbidities, including atherosclerosis, hypertension and diabetes mellitus, to prevent lethal pneumonia.
Cellular infection by coronaviruses is mediated by the viral homotrimeric spike glycoprotein binding to specific host cell receptors. Each protomer consists of an S1 domain, which mediates receptor binding, and an S2 domain that mediates membrane fusion and cell infection following conformational changes induced by host cell receptor binding [12][13][14] . The receptor for SARS-CoV, SARS-CoV-2 and HCoV-NL63 is angiotensin converting enzyme 2 (ACE2), which is expressed on mucosal epithelia of the upper respiratory tract, bronchioles, lungs and the GI tract [15][16][17][18][19] .
Convalescent serum and purified antibodies from recovering patients have yielded promising results in past viral outbreaks 20-26 , and this approach may also be promising for SARS-CoV-2.
One key challenge in the development of these antibodies is how to prevent viral escape whereby variant mutations nullify antibody binding to the RBD epitope. Multiple variants of SARS-CoV-2 have emerged, some of which have mutations in the RBD of the S protein and are predicted to have increased binding to ACE2 [27][28][29][30] . The best neutralizing antibodies are likely to recognize epitopes with high sequence conservation due to structural or functional constraints. The RBM binding site to ACE2 is such an example because sequence divergence within this region is constrained to the prerequisite that binding to ACE2 must be maintained.
An analysis of the crystal structures of the ACE2 -RBD SARS-CoV complex 31 and the ACE2-SARS-CoV-2 RBD complex 29,32 confirms that eight of the key RBM contact residues are conserved between these respiratory coronaviruses, although the SARS-CoV-2 RBM forms a more compact interface with ACE2 than that formed by SARS-CoV RBD and has a higher affinity for ACE2 29,32 . Many of the previously identified neutralizing antibody clones against SARS-CoV were shown to block S1 binding to its ACE2 receptor by binding to epitopes located within the RBM [21][22][23][24]33,34 . However, not all monoclonal antibodies isolated previously against the SARS-CoV RBD recognize the SARS-CoV-2 spike protein in the context of the full-length spike homotrimer 35,36 . Similarly, recent studies evaluating anti-SARS-CoV-2 antibodies isolated from convalescent patients indicate only a minor subset of the clones have the requisite viral neutralizing activity 35,36,46 .
Together, these studies indicate that there is an urgent need for SARS-CoV-2-specific human neutralizing antibody therapeutics that have proven ability to both neutralize SARS-CoV-2 viral host cell infection in vitro as well as the appropriate activity in vivo. The Syrian hamster has been shown to be a suitable model for SARS-CoV-2 infection given the similarity of clinical signs and lung pathology to human disease in the first few days of infection [37][38][39] . However, to date, there has been only a gross histological analysis of the lung pathological changes following infection and the impact of SARS-CoV-2 neutralizing antibody clones on lung immune infiltrates has yet to be fully assessed. In this study, we describe an anti-SARS-CoV-2 spike RBD clone isolated from a rationally designed, fully human antibody library that bound to native spike protein. This potent antibody could block the interaction of the spike protein with ACE2 and could also block SARS-CoV-2 infection of Vero E6 cells and the resultant cytopathic effect. This clone was then tested in a pilot therapeutic study in Syrian golden hamsters (Mesocricetus auratus) and was shown to reduce viral load and ameliorated the severity of bronchointerstitial pneumonia. Detailed histological and image analysis suggests that macrophages play a key role in the lung response to SARS-CoV-2 infection in hamsters.

Materials and Methods
Production and purification of antibody clones. The DNA inserts encoding the light and heavy chains of clone AvGn-B and its variants were separately cloned into different expression vectors carrying the constant regions of human IgG1 heavy chain and the kappa chain. The heavy and light chain plasmids were co-transfected into Expi293 suspension cells. After 5-7 days, secreted antibody was then purified from the culture supernatants by protein A chromatography, bufferexchanged into PBS pH 7.4 and its concentration determined by Nanodrop A280 assay. The quality of the IgG was assessed by SDS-PAGE and by HPLC. Endotoxin levels were tested using a Limulus amoebocyte lysate (LAL) assay (ThermoFisher Scientific, cat # A39552).

Determining affinity and receptor binding inhibitory activity of clone AvGn-B and its
variants. The wells of Immulon 2 HB ELISA 96-well plates were coated with AvGn-B or its variant IgGs at 4° C overnight. The wells were blocked, washed, then serial dilutions of the monomeric biotinylated RBD-antigen added to the wells and incubated for 1 h at room temperature. Bound antigen was then detected with HRP-labeled streptavidin. The OD readings were plotted against concentration using Prism software (Graphpad CA) for curve fitting and determination of the apparent KD value.
The IC50 values for inhibition of RBD binding to the ACE2 receptor were determined by pre-mixing a serial dilution of antibody with 2 nM of biotinylated SARS-COV-2-RBD and incubating for 30 min before adding to wells coated with recombinant ACE2-Fc fusion protein (residue Gln18-Ser760 fused to human Fc [Kactus BioSystems, Cat# ACE-HM401]). After 1 h, bound RBD was detected using HRP-labeled streptavidin and the OD readings plotted against antibody concentration using Prism software.
Real time kinetics were measured by bilayer interferometry (BLI) at a temperature of 30° C using

Determining the viral neutralization activity of antibody clones in vitro
Vero E6 cells were plated overnight in 96-well plates at 20,000 cells per well. Antibodies were diluted in Complete DMEM and serially diluted 1:3 resulting in a 12-point dose response dilution series run in 4 or 8 replicates. The dilutions of antibody were incubated with 100 TCID50 per 50 µL of SARS-CoV-2 (strain 2019-nCoV/USA-WA1/2020) for 1 h and added to the assay plates.
The plates were incubated for 3 days at 37° C, 5% CO2 and 95% relative humidity and the inhibitory effects of the antibodies were assessed.

Study design for animal infections
Approval of the study protocol was obtained from the Colorado State University Institutional Animal Care and Use Committee (protocol 993). Male Syrian hamsters (n=20, 10 weeks of age, obtained from Charles River Laboratory). The animals were held in the CSU animal facility and provided access to standard pelleted feed and water ad libitum prior to being moved into the biosafety level 3 facility for experimental challenge. Of the 20 hamsters, 18 were intranasally infected with 2.5x10 4 TCID50/mL equivalents of SARS-CoV-2 (strain 2019-nCoV/USA-WA1/2020) and divided into treatment groups as follows: "AvGn-B High" (2.5 mg AvGn-B) (n=5), "AvGn-B Low" (1 mg AvGn-B) (n=5), "Untreated" (no antibody) (n=6), and "Ab Control" (2.5 mg isotype control IgG) (n=2). Two uninfected hamsters received 2.5 mg AvGn-B (termed "Uninfected"). Each animal was dosed intraperitoneally with corresponding treatment at 24 and 72 hours post dosing (hpi). At 24, 48, 72, 96, and 120 hpi, each hamster was weighed and assessed for presence of clinical signs (lethargy, ruffled fur, hunched back posture, nasolacrimal discharge, and rapid breathing). At 120 hpi (5 dpi), each hamster was anesthetized with isoflurane and then euthanized via cardiac exsanguination and blood was collected.
Weight loss (calculated as percentage decrease from 0 dpi weight) and viral RNA load in lung were compared between treatment groups in Prism using multiple t-tests and Mann-Whitney U tests, respectively. p<0.05 was considered significant.

RNA extraction and qRT-PCR
Swabs in viral transport medium were vortexed thoroughly and centrifuged to pellet cellular debris.

Histopathology and immunohistochemistry
Lungs, tracheobronchial and hilar lymph nodes, thymus, esophagus, heart and liver from 20 hamsters labeled with ear tags 26-45 were extirpated en bloc and fixed whole in 10% neutral buffered formalin for at least 3 days to ensure virus inactivation prior to transfer to the CSU Veterinary Diagnostic Laboratory for trimming.
Four transverse whole-lung sections were stained with H&E or processed for IHC. Sections, 5 µm thick, were subjected to heat-induced epitope retrieval performed online on a Leica Bond-III IHC automated stainer using bond-epitope retrieval solution. Antibodies to SARS-CoV-2 nucleocapsid protein (mouse, 1:500), pancytokeratin, factor-VIII and ionized calcium binding adaptor molecule (IBA-1) (Leica Biosystems) or negative control slides primary antibody was replaced by a rabbit non-specific IgG isotype negative control antibody for 20 minutes. Labeling was performed on an automated staining platform. Fast red was used a chromogen and slides were counterstained with hematoxylin. Immunoreactions were visualized and blindly scored by a single pathologist. In all treated categories, reactive lung sections incubated with primary antibodies were used as positive immunohistochemical control. Negative control sections were incubated in diluent composed of Tris-buffered saline with a carrier protein and homologous nonimmune serum. All sequential steps of the immunostaining procedure were performed on negative controls following incubation.

Quantification of total nuclei, inflammatory cells and consolidation
The precise digital quantification of the total affected pulmonary parenchyma as well as counting of inflammatory cells per area (ROI, 1mm 2 ) in histological section was determined. A digital montage was compiled at 100´ magnification to include the four tissue classes that were differentially characterized (Table 1) to establish algorithm classifier. Affected ROIs were subsequently automatically identified using Olympus cellSens software by quantifying whole-lung mounts scanned from each hamster for total number of nuclei or nucleated cells (to exclude erythrocytes) stained with hematoxylin. Sections labeled by immunofluorescence for IBA-1 and SARS-CoV-2 were analyzed using the Count and Measure Module of cellSens. The algorithm predominantly extracted multispectral information from images with additional DIA processing using spatial, logical and threshold separators after manual annotations. All four classes were accurately identified as expected by a board-certified pathologist who was blinded to experimental groups of hamsters.

Isolation of an anti-SARS-CoV-2 RBD antibody clone that blocks receptor binding and can neutralize SARS-CoV-2 activity.
A rationally designed fully human antibody library displayed on yeast was screened by magnetic Individual antibody clones were tested for their abilities to block SARS-CoV-2-RBD binding to ACE2 using a competition ELISA and tested for their ability to bind to native SARS-CoV-2 spike protein expressed as a GFP fusion protein by transfected 293 cells compared to binding activity to non-transfected 293 cell controls. Those antibody clones that blocked the interaction of the RBD with ACE2 and bound to native spike protein were then tested for neutralization of SARS-CoV-2 in a cytopathic effect (CPE) assay with Vero E6 cells.
Clone AvGn-B was identified as the most potent in these assays. It exhibited an apparent affinity by ELISA for SARS-CoV-2 RBD of 0.17 nM (Fig. 1A), an apparent IC50 value for blocking RBD binding to ACE2-Fc of 2.2 nM (Fig. 1B), and an apparent KD for binding to native spike protein expressed by 293 cells of 1.02 nM compared to 10.9 nM for ACE2-Fc (Fig. 1D). Real time binding kinetics of AvGn-B to the isolated RBD recombinant protein measured by BLI indicated a KD of 0.37 nM (Fig. 1C). When tested for its ability to neutralize SARS-CoV-2 infectivity, Clone AvGn-B exhibited 100% protection from SARS-CoV-2 induced cell death down to 0.017 µg/mL. Using a colorimetric assay for quantitation of cell death, AvGn-B exhibited an IC50 value of that ranged from 0.008 µg/mL (experiment with 4 replicates) to 0.0054 µg/mL (experiment with 8 replicates) in the CPE assay (Fig. 1E).

Evidence for SARS-CoV-2 neutralization in vivo:
At 2 dpi, infected hamsters appeared quiet and began to progressively lose weight over the course of the study (up to ~13% reduction in untreated animals). There was no significant reduction in weight loss associated with AvGn-B treatment (p>0.05) ( Fig. 2A). None of the hamsters in the study died or met euthanasia criteria prior to study termination at 5 dpi. There was a significant reduction in viral RNA in the lungs of hamsters treated with AvGn-B (both 2.5 mg and 1 mg doses) compared to those that were untreated (p = 0.0173 and 0.0303, respectively) (Fig. 2B).

Pathology and immunohistochemistry results:
Lungs from the 2 uninfected and the 2 Ab Control hamsters (treated with 2.5 mg control isotype

Analysis of macrophage infiltration in lung tissue of animals exposed to SARS-CoV-2
Lungs were examined for the extent of macrophage infiltration and SARS-CoV-2 viral load at 5 days post infection by histopathological examination and by immunofluorescence imaging (Fig.   5). Whole mount sections of paraffin-embedded lung tissue were stained with hematoxylin and eosin and brightfield grayscale images were collected using a microscope equipped with a scanning motorized stage (Fig. 5A-E). Hematoxylin-positive macrophage soma were rendered as focal points within the regions of interest to calculate the percent hypercellularity of tissue following infection with SARS-CoV-2. By 5 dpi, lung tissue showed extensive infiltration of macrophages (Fig. 5A) that was decreased in dose-dependent fashion by treatment with AvGn-B (Fig. 5B, 5C). Treatment of uninfected hamsters with AvGn-B alone did not increase macrophage infiltration, whereas co-treatment with IgG control Ab during infection with SARS-CoV-2 (Ab Control group) still resulted in marked infiltration of macrophages (Fig. 5E). The percent of total lung area displaying macrophage hypercellularity was quantified in Fig. 5N, and Consequently, in parallel to the in vivo studies described above, antibody engineering of clone AvGn-B has been performed and a panel of more potent variants has now been isolated. As shown in Table 2 Future studies will explore the efficacy of the AvGn-B and/or its variants in non-human primate models, several of which have been described for use in SARS-CoV-2 pathogenesis and countermeasure development studies 40,60,61 . Larger studies will also be conducted to examine whether AvGn-B and/or its variants by reducing viral load and accumulation of macrophages within the lung will prevent downstream inflammatory and coagulation sequalae of SARS-CoV-2 infection within other parenchymatous organs, particularly heart, kidneys and liver. Should AvGn-B and/or its variants advance to clinical trials in human patients, its effect on Kawasaki-like disease (KD) in SARS-CoV-2 infected children merits clinical investigation. There is accumulating evidence that the monocyte/macrophage system releases cytokines that directly lead to vascular endothelial damage during acute KD 62,63 . Investigating the role of AvGn-B and/or its variants in suppressing the cytokine storm by suppressing a pivotal player, monocyte-macrophage system will be important.