Fc-engineered antibody therapeutics with improved efficacy against COVID-19

Monoclonal antibodies (mAbs) with neutralizing activity against SARS-CoV-2 have demonstrated clinical benefit in cases of mild to moderate SARS-CoV-2 infection, substantially reducing the risk for hospitalization and severe disease1–4. Treatment generally requires the administration of high doses of these mAbs with limited efficacy in preventing disease complications or mortality among hospitalized COVID-19 patients5. Here we report the development and evaluation of Fc-optimized anti-SARS-CoV-2 mAbs with superior potency to prevent or treat COVID-19 disease. In several animal models of COVID-19 disease6,7, we demonstrate that selective engagement of activating FcγRs results in improved efficacy in both preventing and treating disease-induced weight loss and mortality, significantly reducing the dose required to confer full protection upon SARS-CoV-2 challenge and treatment of pre-infected animals. Our results highlight the importance of FcγR pathways in driving antibody-mediated antiviral immunity, while excluding any pathogenic or disease-enhancing effects of FcγR engagement of anti-SARS-CoV-2 antibodies upon infection. These findings have important implications for the development of Fc-engineered mAbs with optimal Fc effector function and improved clinical efficacy against COVID-19 disease.

3 other mAbs that are currently awaiting regulatory approval offer a clear therapeutic benefit in patients with mild to moderate SARS-CoV-2 infection, reducing the risk for hospitalization and progression to severe COVID-19 disease by over 80% [1][2][3] . These results are in stark contrast to the findings from several phase III trials (e.g. ACTIV-3 protocols, NCT04501978) that assessed the therapeutic activity of these mAbs in hospitalized COVID-19 patients. In all cases, such trials failed to reach the primary study endpoints and were terminated prematurely, as none of the neutralizing anti-SARS-CoV-2 mAbs tested offered any therapeutic benefit over standard-ofcare, even when administered at exceedingly high doses or in combination with antivirals like remdesivir 5 .
For several acute and chronic viral infections, including influenza virus, RSV, HIV, and Ebola virus, the antiviral activity of IgG antibodies is the outcome of Fab-mediated virus neutralization coupled with the capacity of the Fc domain to mediate effector functions through interactions with Fcγ receptors (FcγRs) expressed on the surface of effector leukocytes 8 . Indeed, FcγR engagement mediates pleiotropic antiviral immune functions, including the clearance of viral particles 9,10 , the cytotoxic elimination of virus-infected cells 11 , as well as the induction of protective T cell responses that contribute to antiviral immunity [12][13][14] . In the context of SARS-CoV-2 infection, a growing body of experimental evidence from various animal disease models supports that Fc-FcγR interactions are essential for the in vivo antiviral activity of anti-SARS-CoV-2 mAbs, as loss of the capacity of the Fc domain of these mAbs to engage FcγRs is associated with reduced antiviral activity in vivo [15][16][17][18] .
Given the importance of Fc-FcγR interactions in the mAb-mediated protection, maximizing the capacity of clinical neutralizing anti-SARS-CoV-2 mAbs to engage and activate the appropriate FcγR pathways is expected to lower the mAb dose required for the treatment of mild to moderate SARS-CoV-2 infection, as well as improve their therapeutic activity in hospitalized patients with severe COVID-19 disease. Currently, most mAbs in clinical use or development are expressed as human IgG1, which despite its affinity for activating FcγRs also exhibits considerable binding to the inhibitory FcγRIIb, thereby limiting protective Fc effector activities 9 . Additionally, several clinical mAbs (etesevimab, AZD8895, and AZD1061) lack Fc effector function, as they are expressed as Fc domain variants with no FcγR binding activity. This was due to presumptive safety concerns over the capacity of antibodies to exacerbate disease through ADE (antibodydependent enhancement) mechanisms 8 . However, numerous in vivo studies in animal disease models have failed to provide evidence for ADE [15][16][17][18][19] , and therapeutic administration of high doses of convalescent plasma or neutralizing anti-SARS-CoV-2 mAbs in COVID-19 patients has not been associated with worse disease outcomes [1][2][3]5,20 . Likewise, comparable safety profiles were evident in clinical trials of neutralizing mAbs with intact or diminished Fc effector function.
To assess the role of FcγRs in the mAb-mediated protection and develop mAbs with superior therapeutic potency against COVID-19 disease, we selected well-established small animal SARS-CoV-2 infection models that recapitulate the clinical and pathological features of human COVID-19 disease 6,7,21 . One of these models involves the use of Syrian golden hamsters (Mesocricetus auratus), a species that not only sustains productive virus replication with SARS-CoV-2 clinical isolates, but also exhibits evidence of severe disease upon challenge 6 . However, a major obstacle in the in vivo study of Fc effector activity of human IgG antibodies is the substantial interspecies differences in the affinity of human IgG antibodies for FcγRs expressed by rodent species, such as hamsters 22 . We therefore cloned and expressed the four classes of hamster FcγRs and characterized their binding affinity for human, hamster, and mouse IgG 5 subclasses and Fc variants ( Fig. 1a- To assess the contribution of Fc-FcγR interactions to mAb-mediated protection, we selected neutralizing mAbs in clinical use or development, including casirivimab and imdevimab (REGN cocktail 23 ) and S309/VIR-7831 (Vir 24 ) and expressed them as human IgG1 or as Fc domain variants with defined affinity for hamster FcγRs. In agreement with recent reports 16 , we observed that when mAbs are administered prophylactically, Fc effector function has minimal contribution to the mAb antiviral activity in this model (Fig. 1c). By contrast, in the therapy setting (d+1 treatment), wild-type, but not FcR null (GRLR) variants are able to suppress lung viremia and prevent weight loss (Fig. 1d). Since previous studies in mouse models of influenza virus and HIV-1 infection support a key role for FcγRIV in mediating protection by antiviral mAbs 11,25,26 , we compared the in vivo therapeutic activity of two Fc domain variants -GAALIE and V11-that exhibit differential hamster FcγRIV binding activity, but comparable affinity for the other hamster FcγRs (Fig. 1b). Consistent with a protective role for FcγRIV, the FcγRIV-enhanced variant (GAALIE) demonstrates potent antiviral activity, whereas no therapeutic activity is evident for V11, which exhibits minimal affinity for hamster FcγRIV (Fig. 1e-g).
Although these findings support the importance of Fc-FcγR interactions in the mAb-mediated protection against SARS-CoV-2 infection, their translational relevance is rather limited, given the diversity of FcγR expression on immune cells, the structural complexity of the FcγR family and the divergence of these receptors between humans and other mammalian species 22 . To address this problem, we have previously developed a mouse strain in which only human FcγRs 6 are expressed in a pattern that recapitulates as faithfully as possible the expression pattern seen in human tissues 27 . Human FcγR expression among the various effector leukocyte populations is stable throughout mouse life span and does not differ between young and old mice (Extended Data Fig. 2). Infection of old (>15 weeks old), but not young (7 weeks), FcγR humanized mice with the mouse adapted SARS-CoV-2 strain MA10 (characterized in 7 ) results in rapid and often lethal weight loss, which is dependent upon the virus challenge dose ( Fig. 2a-b). Histological In a model of mAb-mediated therapy of SARS-CoV-2 infection, we observed that the REGN mAb cocktail (expressed as wild-type human IgG1) confers full protection of FcγR humanized mice when administered at 5 mg/kg one day after lethal challenge with SARS-CoV-2 (MA10, 10 4 pfu i.n.) ( Fig. 2c-f). By contrast, no therapeutic activity is evident in mice lacking FcγRs (FcγRnull) or when mAbs are expressed as variants (GRLR) with minimal affinity for human FcγRs, highlighting the importance of Fc effector function in the therapeutic activity of neutralizing mAbs (Fig. 2c-f). To determine the mechanisms by which human FcγRs contribute to the mAb-mediated protection, REGN cocktail mAbs were expressed as human IgG1 Fc variants that have been characterized extensively in previous studies 12,28 and exhibit differential affinities for the various classes of human FcγRs (Fig. 3a). Following the experimental strategy outlined in Figure 3b, we assessed the therapeutic activity of Fc variants of the REGN mAb cocktail at a dose (1 mg/kg) which wild-type human IgG1 confers minimal protection (Fig. 3c).
Consistent with a protective role for activating FcγRs, Fc variants enhanced for either FcγRIIa (GA) or FcγRIII (ALIE) show a trend for improved therapeutic potency over wild-type IgG1, whereas maximal therapeutic activity was evident only for the GAALIE variant, which is enhanced for both FcγRIIa and FcγRIII and has reduced affinity for the inhibitory FcγRIIb ( Fig.   3d-e). These findings suggest that synergy between the two activating FcγRs, FcγRIIa and FcγRIII likely accounts for the therapeutic activity of the GAALIE variant, which achieves the same degree of protection as wild-type IgG1, but at a 5-times lower dose. Importantly, treatment of SARS-CoV-2-infected mice with the GAALIE variant was not associated with enhanced disease not only when administered at a low dose (1 mg/kg), but also at a much higher dose (40 mg/kg) (Extended Data Fig. 4). At this dose, which is typically used in the clinical setting, FcR null variants (GRLR) also exhibit full protective activity comparable to the GAALIE variants, suggesting that Fc-independent protection could be achieved once neutralizing mAbs are administered at sufficiently high doses, as has been documented previously for other viral pathogens 25,26 . Additionally, the observed differences in the in vivo therapeutic activity among Fc variants could not be attributed to differences in Fab-mediated functions, as none of these Fc modifications impact the in vitro neutralization activity, antigen binding specificity, or in vivo half-life (Extended Data Fig. 5-6).
Similar results were obtained when we assessed the in vivo therapeutic activity of another mAb cocktail (C135+C144; BMS/RU 29,30 ) that is currently in clinical development and targets the SARS-CoV-2 Spike and mediates potent neutralizing activity 29 . In in vivo titration experiments, we observed that therapeutic administration of the BMS/RU mAb cocktail expressed as wild-type human IgG1 protects FcγR humanized from lethal SARS-CoV-2 challenge in a dosedependent manner (Fig. 3f). When BMS-RU mAbs are administered to mice at 1 mg/kg, only GAALIE variants, but not wild-type human IgG1 confer protective activity and prevent diseaseinduced weight loss, confirming our findings on the REGN mAb cocktail, which demonstrated improved therapeutic activity when engineered for selective engagement of activating FcγRs (Fig. 3g).
Our findings in hamsters suggest that when neutralizing mAbs are administered prophylactically, Fc-FcγR interactions are not critical for their antiviral activity (Fig. 1c). However, given the substantial interspecies differences in FcγR biology between hamsters and humans, we assessed the contribution of activating FcγR engagement to the mAb-mediated prophylaxis of SARS-CoV-2-challenged FcγR humanized mice (Fig. 4a). When administered at a dose where wildtype human IgG1 exhibits no protective activity (0.5 mg/kg) (Fig. 4b) demonstrated that enhanced FcγR engagement is associated with increased pathology 34 . By

Viruses, Cell Lines, and Animals
A P1 stock of the SARS-CoV-2 MA10 strain 7 was amplified in VeroE6 cells obtained from the ATCC that were engineered to stably express TMPRSS2 (VeroE6TMPRSS2). To generate a P2 μg/ml streptomycin (ThermoFisher). All cell lines were maintained at 37 o C at 5% CO2.
Expi293F cells (ThermoFisher) were maintained at 37 o C, 8% CO2 in Expi293 expression medium (ThermoFisher) supplemented with 10 U/ml penicillin and 10 μg/ml streptomycin. All cell lines have been tested negative for mycoplasma contamination.
In vivo experiments were approved by the Rockefeller University Institutional Animal Care and Use Committee in compliance with federal laws and institutional guidelines. Hamsters and mice were maintained at the Comparative Bioscience Center at the Rockefeller University at a controlled ambient temperature environment with 12-h dark/light cycle. Golden Syrian hamsters were purchased from Charles River laboratories (strain code 049) and maintained in compliance with USDA regulations. FcγR knockout (mFcγRα −/− ; Fcgr1 −/− ) and FcγR humanized mice (mFcγRα −/− , Fcgr1 −/− , hFcγRI + , hFcγRIIaR131 + , hFcγRIIb + , hFcγRIIIaF158 + , and hFcγRIIIb + ) were generated in the C57BL/6 background and characterized in previous studies 12,27 .

Cloning, Expression, and Purification of Recombinant Proteins
Human IgG1 Fc domain variants were generated by site-directed mutagenesis using specific primers as previously described 10 and recombinant IgG antibodies were expressed and purified using previously described protocols 12 . Purity was assessed by SDS-PAGE followed by SafeStain blue staining (ThermoFisher). All antibody preparations were more than 90% pure and endotoxin levels were less than 0.05 EU/mg, as measured by the limulus amebocyte lysate assay.
The two plasmid-based HIV/NanoLuc-SARS-CoV-2 pseudovirus system was kindly provided by Dr. Paul Bieniasz (described in 36 ). The S gene was modified by side-directed mutagenesis to introduce the amino acid changes present in the MA10 strain 7 . SARS-CoV-2MA10 pseudovirus particles were generated by transfection of the two plasmid-based system to 293T cells using X-tremeGENE HP DNA transfection reagent (Sigma).

Surface Plasmon Resonance
All experiments were performed with a Biacore

Neutralization Assay
Neutralization activity of IgG1 Fc domain variants was measured as previously described 36 .
Briefly, HT1080ACE2 cells were seeded in 96 U-well plates 24 h prior to infection with SARS-CoV-2MA10 pseudoviruses. Pseudovirus particles were pre-incubated with mAbs (four-fold serially diluted starting at 10 μg/ml) for 1 h at 37 o C and then added to a monolayer of
Plates were developed using the TMB two-component peroxidase substrate kit (KPL) and reactions were stopped with the addition of 1 M phosphoric acid. Absorbance at 450 nm was immediately recorded using a SpectraMax Plus spectrophotometer (Molecular Devices) and background absorbance from negative control samples was subtracted. Data were collected and analysed using SoftMax Pro v.7.0.2 software (Molecular Devices).

Quantification of Serum IgG Levels
Blood was collected into microvette serum gel tubes (Sarstedt) and serum was fractionated by centrifugation (10,000g, 5 min). IgG levels were determined by ELISA following previously described protocols 12 .
After infection, animals were monitored daily and humanely euthanized by CO2 asphyxiation at endpoints authorized by the Rockefeller University Institutional Animal Care and Use Committee, including any of the following: marked lethargy or inactivity, severe respiratory distress or labored breathing, inability to ambulate, and weight loss of greater than 20% of baseline. Animals were randomized based on age, gender, and weight. Before treatment, we ensured that the mean weight, gender, and age were comparable among the various treatment groups. For antibody-mediated prophylaxis, antibodies were administered intravenously one day before virus challenge, whereas for antibody-mediated therapy, antibodies were administered one day after infection. Antibody dose was calculated as mg/kg.

Histological Analysis
Lungs from euthanized mice were instilled with 10% neutral buffered formalin and fixed overnight by submersion in 10% formalin. Fixed tissues were embedded in paraffin, sectioned at 4 μm thickness, and stained with hematoxylin and eosin. Sections of lung were microscopically evaluated by a board-certified veterinary anatomic pathologist and representative images were captured with an Olympus BX45 light microscope using an SC30 camera with the cellSens Dimension software.

Determination of Lung Viral Titers
Hamsters were euthanized at the indicated timepoints following infection and lung weights were recorded. Lungs were lysed in Trizol (ThermoFisher) and dissociated in gentle MACS M tubes using the gentleMACS Octo Dissociator (Miltenyi Biotec). Samples were transferred into Phasemaker tubes (ThermoFisher) and chloroform was added (200 μl chloroform/1 ml TRIzol).
After vigorous shake, tubes were rested for 5 min and then centrifuged for 15 min at 12,000g at Samples that were stained with isotype control antibodies were also blocked with unlabeled anti- anti-human FcγRIIb (clone 2B6), and anti-human FcγRIIIa/b (clone 3G8) (used at 10 μg/ml and incubated for 5 min prior to staining with fluorescently labelled antibodies). Samples were analyzed on an Attune NxT flow cytometer (ThermoFisher) using Attune NxT software v3.1.2 and data were analyzed using FlowJo (v10.7) software.

Statistical Analysis
Results from multiple experiments are presented as mean ± s.e.m. One-or two-way ANOVA was used to test for differences in the mean values of quantitative variables, and where statistically significant effects were found, post hoc analysis using Bonferroni (adjusted for multiple comparisons) test was performed. Statistical differences between survival rates were analyzed by comparing Kaplan-Meier curves using the log-rank (Mantel-Cox) test. Data were analyzed with GraphPad Prism v.9.1 software (GraphPad) and p < 0.05 were considered to be statistically significant.