Novel bispecific human antibody platform specifically targeting a fully open spike conformation potently neutralizes multiple SARS-CoV-2 variants

Abstract

Rapid emergence of new variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has prompted an urgent need for the development of broadly applicable and potently neutralizing antibody platform against the SARS-CoV-2, which can be used for combatting the coronavirus disease 2019 (COVID-19). In this study, based on a noncompeting pair of phage display-derived human monoclonal antibodies (mAbs) specific to the receptor-binding domain (RBD) of SARS-CoV-2 isolated from human synthetic antibody library, we generated K202.B, a novel engineered bispecific antibody with an immunoglobulin G4-single-chain variable fragment design, with sub- or low nanomolar antigen-binding avidity. Compared with the parental mAbs or mAb cocktail, the K202.B antibody showed superior neutralizing potential against a variety of SARS-CoV-2 variants in vitro. Furthermore, structural analysis of bispecific antibody-antigen complexes using cryo-electron microscopy revealed the mode of action of K202.B complexed with a fully open three-RBD-up conformation of SARS-CoV-2 trimeric spike proteins by simultaneously interconnecting two independent epitopes of the SARS-CoV-2 RBD via inter-protomer interactions. Intravenous monotherapy using K202.B exhibited potent neutralizing activity in SARS-CoV-2 wild-type- and B.1.617.2 variant-infected mouse models, without significant toxicity in vivo. The results indicate that this novel approach of development of immunoglobulin G4-based bispecific antibody from an established human recombinant antibody library is likely to be an effective strategy for the rapid development of bispecific antibodies, and timely management against fast-evolving SARS-CoV-2 variants.

Abbreviations

BSA

Bovine serum albumin

BsAb

Bispecific antibody

BUN

Blood urea nitrogen

COVID-19

Coronavirus disease 2019

CRE

Creatinine

Dpi

Days post infection

E

Envelope

ELISA

Enzyme-linked immunosorbent assay

EMA

European Medicines Agency

EM

Electron microscopy

GOT

Glutamic oxaloacetic transaminase

GPT

Glutamic pyruvic transaminase

HA

Hemagglutinin

HRP

Horseradish peroxidase

hACE2

Human angiotensin-converting enzyme 2

HUVEC

Human umbilical vein endothelial cell

IC50

Half-maximum inhibitory concentration

IgG

Immunoglobulin G

mAb

Monoclonal antibody

PBS-T

PBS containing 0.05% (v/v) Tween 20

RBD

Receptor-binding domain

RdRp

RNA-dependent RNA polymerase

RT

Room temperature

RT-qPCR

Reverse transcription-quantitative polymerase chain reaction

S

Spike

SARS-CoV-2

Severe acute respiratory syndrome coronavirus 2

scFv

Single-chain variable fragment

SEC

Size exclusion chromatography

SPR

Surface plasmon resonance

TG

Transgenic

TMB

3,3′,5,5′-tetramethylbenzidine;

US FDA

United States Food and Drug Administration

WT-RBD-His

His-tagged wild-type SARS-CoV-2 RBD

1. Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a novel, highly transmissible, and pathogenic beta-coronavirus that first emerged in late 2019 (Hu et al., 2021). SARS-CoV-2 caused outbreak of an acute respiratory illness, coronavirus disease 2019 (COVID-19), which resulted in a global pandemic (Baloch et al., 2020). As of February 2023, around 754 million people have been infected with COVID-19, with a confirmed fatality of 6.8 million individuals worldwide (WHO, 2023). The COVID-19 pandemic has imposed an unprecedented level of stress on global healthcare systems, along with socioeconomic crisis (Nicola et al., 2020). Although multiple vaccines and therapeutic interventions have been developed, the rapid emergence of SARS-CoV-2 variants represents an ongoing challenge in COVID-19 management (Tregoning et al., 2021; Yin et al., 2022). Thus, the potential severe outcomes of this infection highlight the need for development of a novel therapeutic platform for the treatment of COVID-19.

Neutralizing monoclonal antibodies (mAbs) provide a promising therapeutic approach against infectious viral diseases, including human immunodeficiency, Ebola, and human respiratory syncytial viruses (Levine, 2019; Pelegrin et al., 2015). SARS-CoV-2 infection is mediated by spike (S) proteins, which form homotrimers protruding from the viral surface. The S protein comprises two functional subunits: S1 and S2. The receptor-binding domain (RBD) within the S1 subunit is responsible for direct interaction with human angiotensin-converting enzyme 2 (hACE2), a critical receptor for SARS-CoV-2 expressed on the surface of host cells (Bangaru et al., 2020; Huang et al., 2020). Several mAbs and mAb cocktails specifically targeting this RBD have received approval by the United States Food and Drug Administration (US FDA), or European Medicines Agency (EMA) for emergency use against COVID-19 or have been validated in clinical trials (Kumar et al., 2021). However, rapidly emerging SARS-CoV-2 variants including Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2), and Kappa (B.1.617.1) exhibit S protein mutations that can evade neutralization by many single and cocktails of mAbs with overlapping epitopes (Baum et al., 2020; Biswas et al., 2022; Planas et al., 2021). These variants may pose an undetermined resurgence risk. Furthermore, a new health threat in the forms of the Omicron (BA.1–BA.5) variants of SARS-CoV-2, has emerged globally (Khandia et al., 2022). Therefore, rapid development and acquisition of a broadly applicable and potent neutralizing antibody platform technology will be critical as a timely measure against the rapid emergence of new or resurgent SARS-CoV-2 viruses.

As a first, in this study, we aimed to develop a novel immunoglobulin G4 (IgG4)-[single-chain variable fragment (scFv)]₂ form of human bispecific antibodies (bsAbs) based on a noncompeting pair of parental mAbs specific to the SARS-CoV-2 RBD isolated from human synthetic antibody library. Using intensive biochemical, molecular, and virological testing, we demonstrate the therapeutic potential of this bsAb, K202.B, that exhibits broad and potent neutralizing activity against multiple SARS-CoV-2 variants with high avidity and specificity. Moreover, our comprehensive structural analyses provide clear insights into the mechanism of action of the K202.B for the better understanding of its potent neutralizing effects. Therefore, the present study suggests that this novel approach for the development of phage display-derived bsAb from an established human recombinant antibody library may be widely useful for the rapid and effective management of the constantly changing SARS-CoV-2.

2. Materials and methods

2.1. Cell culture

The cell lines 293T, K562, and THP-1 were purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA). VeroE6 cells were obtained from the Korea Microbial Resource Center (Korean Collection for Type Cultures (KCTC), Daejeon, Republic of Korea), Expi293 cells from Thermo Fisher Scientific (Waltham, MA, USA), and human umbilical vein endothelial cells (HUVECs) from Lonza (Basel, Switzerland). The 293T and VeroE6 cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Thermo Fisher Scientific), whereas K562 and THP-1 cells were cultured in Roswell Park Memorial Institute (RPMI, Thermo Fisher Scientific) media supplemented with 10% (v/v) fetal bovine serum (FBS, Thermo Fisher Scientific) and 100 U/mL penicillin–streptomycin (Thermo Fisher Scientific) at 37 °C and 5% CO₂. HUVECs were maintained in endothelial growth medium-2 (EGM-2; Lonza). Expi293™ cells were cultured in Expi293™ Expression Media in shaking incubators at 37 °C, 125 rpm, and 8% CO₂.

2.2. Isolation of human antibodies specific to SARS-CoV-2 RBD

Biopanning was performed to select human antibodies specific to SARS-CoV-2 RBD from a human scFv antibody library (Cho et al., 2022). Four rounds of biopanning were conducted with M-270 epoxy Dynabeads™ (Invitrogen, Carlsbad, CA, USA) coated with 4 μg recombinant His-tagged wild-type SARS-CoV-2 RBD (WT-RBD-His) (Sino Biological, Beijing, China). Ninety-six phage clones, randomly selected from colonies grown on output plates, were tested for reactivity to WT-RBD-His using a phage enzyme-linked immunosorbent assay (ELISA). DNAs from these clones were sequenced, and four scFv clones with different complementarity-determining region sequences were finally selected.

2.3. Phage ELISA

Phage ELISA was performed as described previously, with minor modifications (Lee et al., 2022). In brief, 0.1 μg of SARS-CoV-2 WT-RBD-His was incubated overnight in 96-well high binding plates (Corning, Lowell, MA, USA) at 4 °C. Then, the plates were blocked with 3% (w/v) bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 2 h at 37 °C. Subsequently, the plates were incubated with 96 randomly selected scFv-displayed phage clones from the fourth round of biopanning for 2 h at 37 °C. After washing with PBS containing 0.05% (v/v) Tween 20 (PBS-T), the samples were incubated with horseradish peroxide (HRP)-conjugated anti-hemagglutinin (HA) antibody (1:3000; Bethyl Laboratories, Montgomery, TX, USA) for 1 h at 37 °C. The colorimetric reaction was started with 100 μl of 3,3′,5,5′-tetramethylbenzidine (TMB; Thermo Fisher Scientific) substrate solution, and quenched with 2 N H₂SO₄ solution. The optical density was measured at 450 nm using a microplate reader (Synergy H1, BioTek, Winooski, VT, USA).

2.4. Construction, expression, and purification of mono- and bispecific antibodies

To generate an IgG4 antibody, each variable heavy or light chain gene of the selected scFv clones was individually cloned into a bicistronic mammalian expression vector encoding IgG4 containing S228P mutation. The constructed IgG4 mAbs were designated as K102.1, K102.2, K102.3, and K102.4. Two different formats of IgG-scFv bsAbs were generated as follows: Format A was constructed by fusing the scFv of K102.2 to the C-terminus of a K102.1 heavy chain with a short G₃S linker; format B was constructed by fusing the scFv of K102.2 to the N-terminus of K102.1 light chain with a long (G₄S)₃ linker. The constructed mAbs and bsAbs were transiently expressed using the Expi293 Expression System (Ecker et al., 2020). Briefly, recombinant vectors encoding mAbs and bsAbs were transfected into Expi293F cells according to the manufacturer's recommendations. The antibodies were overproduced and purified from culture media using affinity column chromatography with protein A sepharose (Repligen, Waltham, MA, USA).

2.5. Competition ELISA

WT-RBD-His (0.1 μg) was coated in a 96-well microplate and incubated overnight at 4 °C. After washing with PBS-T, the plate was blocked for 2 h with blocking buffer (3% BSA in PBS), and incubated with or without 1.5 μg of K102.1, K102.2, K102.3, or K102.4 for 1 h at room temperature (RT). The plates were then washed with PBS-T and incubated with 0.15 μg of HRP-conjugated K102.1 for 1 h at RT. After washing with PBS-T, anti-human Fc-HRP was incubated for 1 h at RT. The plates were washed thrice with PBS-T, and colorimetric detection was performed using the TMB substrate. The optical density was measured at 450 nm using a microplate reader (Synergy H1).

2.6. Surface plasmon resonance (SPR)

The binding kinetics of antibodies to SARS-CoV-2 RBDs were analyzed at RT on an iMSPR-mini instrument (iCLUEBIO, Seongnam, Republic of Korea) using 10 mM HEPES pH 7.4, 700 mM NaCl, 2 mM CaCl₂, 1 mM MnCl₂, and 0.005% (v/v) Tween-20 as a running buffer. The recombinant SARS-CoV-2 RBDs (wild-type and variants including B.1.1.7, B.1.351, P.1, B.1.617.2, and B.1.617.1) were covalently immobilized on the surface of a COOH–Au chip (iCLUEBIO) up to 500 response units through standard amine coupling. Increasing concentrations of K202.A or K202.B (8, 16, 32, 64, and 128 nM, respectively) were injected onto the surface of a sensor chip at a flow rate of 50 μL/min. Kinetics evaluation data was obtained using a 1:1 binding model.

To evaluate the ability of K202.B to bind to different regions of the RBD, competition experiments were performed; following the immobilization of 5 nM WT-RBD-His on the surface of a COOH–Au chip, a high concentration (512 nM) of K102.1 or K102.2 antibody was added to saturate the corresponding binding sites on the RBD. Then, 128 nM K202.B was added. Conversely, following the addition of 512 nM K202.B to the surface of recombinant WT-RBD-His-immobilized sensor chip, 256 nM K102.1 or K102.2 was subsequently added. Curve fitting and data analysis were performed using the iMSPR analysis software (Tracedrawer; iCLUEBIO).

2.7. SARS-CoV-2 RBD-hACE2 interaction inhibition assay

The ability of antibodies to inhibit the interaction of the SARS-CoV-2 RBD with hACE2 was investigated using ELISA; 50 ng of purified Fc-tagged hACE2 (hACE2-Fc) (R&D Systems, Minneapolis, MN, USA) was coated in each well of a 96-well plate and incubated for 2 h at RT. After washing with immunobuffer (BPS Bioscience, San Diego, CA, USA), the plates were blocked with blocking buffer (BPS bioscience) for 1 h at RT. Simultaneously, 25 nM of purified SARS-CoV-2 WT- or variant-RBD-His (B.1.1.7, B.1.351, P.1, B.1.617.2, or B.1.617.1) (Sino Biological) was pre-incubated in the presence or absence of K202.B [(wild-type, B.1.1.7, or P.1: 100, 25, 6.25, 1.56, 0.39, 0.098, 0.024, 0.0061 nM) and (B.1.351, B.1.617.2, or B.1.617.2: 100, 25, 6.25, 1.56, 0.39, 0.098, 0.024, 0.0061, 0.0015, 0.00038, 0.000095 nM)] for 1 h at RT. For RBD variants that were observed during early 2020, 25 nM of purified variant-RBD-His (V341I, F342L, R408I, A435S, G476S, V483A, W436R, V376F, and N354D + D364Y) (Sino Biological) was pre-incubated in the presence or absence of 1 or 10 nM of K202.B for 1 h at RT. After washing with immunobuffer thrice, the preincubated mixtures were added into wells for 1 h at RT. After washing with immunobuffer thrice, HRP-conjugated anti-His secondary antibody (BPS Bioscience) was added and incubated for 1 h at RT. The neutralizing activity was detected using an ELISA enhanced chemiluminescence substrate (BPS Bioscience). Chemiluminescence intensity was measured using a Synergy H1 microplate reader (BioTek). Dose-response curves for half-maximum inhibitory concentration (IC₅₀) values were determined with GraphPad Prism 8.0 software (La Jolla, CA, USA) using the equation log(inhibitor) versus normalized response-variable slope.

2.8. Establishment of hACE2-overexpressing 293T stable cell lines (293T/hACE2 cell lines)

To generate stable 293T/hACE2 cell lines, a pUNO1-hACE2 plasmid (InvivoGen, San Diego, CA, USA) was transfected into 293T cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Forty-eight hours after transfection, the cells were cultured in a medium containing 20 μg/mL blasticidin (InvivoGen) to select positive cell populations. The expression of hACE2 was determined using immunoblot and immunocytochemical analysis. For immunoblot analysis, 293T and 293T/hACE2 cell lines were lysed with sodium dodecyl sulfate sample buffer, and subjected to immunoblot analysis using a polyclonal anti-hACE2 antibody (R&D Systems). The distribution of hACE2 in the cell membrane was studied through immunocytochemical analysis. Briefly, 293T and 293T/hACE2 cells were plated on Nunc® Lab-Tek® II chamber slides (Thermo Fisher Scientific) coated with poly-L-lysine (0.1 mg/mL) (Sigma-Aldrich, St. Louis, MO, USA). After 24 h, the cells were fixed with 4% formaldehyde for 10 min and washed twice with PBS. The cells were blocked using PBS containing 1% (w/v) BSA and incubated with polyclonal anti-hACE2 antibody overnight at 4 °C. After washing thrice with PBS, the cells were subsequently incubated with Alexa Fluor 488-labeled anti-goat secondary antibody (Invitrogen) at RT for 1 h, then mounted with mounting solution (Dako North America, Carpinteria, CA, USA). The stained cells were imaged using confocal microscopy (LSM510; Carl Zeiss, Oberkochen, Germany).

2.9. SARS-CoV-2 pseudotyped virus neutralization assay

Pseudotyped replication-deficient lentiviral particles carrying the SARS-CoV-2 S protein of the wild-type or D614G (B.1) variant, and a firefly luciferase reporter gene were prepared using Lenti-X™ SARS-CoV-2 packaging mix according to the manufacturer's instruction (Takara Bio, Kusatsu, Japan). Briefly, the packaging mix was transiently transfected into Expi293™ cells with ExpiFectamine™ 293 reagent. After culturing for 72 h, the supernatants containing the pseudotyped viruses were collected and centrifuged briefly (500×g for 10 min) to remove cellular debris. Virus titration was measured using Lenti-X GoStix™ Plus (Takara Bio) according to the manufacturer's instructions. The pseudotyped replication-deficient Moloney murine leukemia virus particles carrying the SARS-CoV-2 S protein of B.1.1.7, B.1.351, P.1, B.1.617.2, B.1.617.1, or BA.1 variants, and a firefly luciferase reporter gene were obtained from eEnzyme (Gaithersburg, MD, USA).

To determine the neutralization activity of mAbs or bsAbs on pseudotyped virus infection, 1 × 10⁴ 293T/hACE2 cells in 50 μL culture medium were seeded in 96-well tissue culture plates overnight. Serial dilutions of the antibodies were pre-incubated at RT for 10 min with 50 μL of each pseudotyped virus [1 × 10⁷ plaque-forming unit (PFU)/mL], and the mixture was subsequently incubated with the cells for 24 h. The firefly luciferase reporter gene expression (which is indicative of viral presence) was measured using ONE-Glo™ luciferase substrate (Promega, Madison, WI, USA). Next, the culture medium was removed and incubated with 100 μL of ONE-Glo™ substrate. After 5 min, 70 μL supernatant was transferred to white flat-bottom 96-well assay plates (Corning) and the luminescence signal was measured using the Synergy H1 microplate reader. The recorded relative luminescence units were normalized to those derived from cells infected with each SARS-CoV-2 pseudotyped virus in the absence of antibodies. Dose-response curves for IC₅₀ values were determined with GraphPad Prism 8.0 software using the equation log(inhibitor) versus normalized response-variable slope.

2.10. SARS-CoV-2 live virus neutralization assays

SARS-CoV-2 (BetaCoV/Korea/KCDC03/2020, NCCP no. 43326) was provided by the Korea Disease Control and Prevention Agency (KDCA, Osong, Republic of Korea). The viruses were prepared by propagation in Vero E6 cells, titered using a plaque assay, and stored at −80 °C for in vitro infection assessment. Next, 1.5 × 10⁴ VeroE6 cells in 100 μL culture medium were seeded on 96-well tissue culture plates overnight. For 1 h at RT, 50 μL of serial dilutions of the antibodies were preincubated with 50 μL of SARS-CoV-2 (4 × 10² TCID₅₀/mL). Media was then removed from the seeded VeroE6 cells and replaced with the virus/antibody mixture. Three days post-inoculation, neutralization potency was determined via viral RNA quantification using reverse transcription-quantitative polymerase chain reaction (RT-qPCR). RT-qPCR was performed with the QuantiFast SYBR® Green RT-PCR kit (QIAGEN, Hilden, Germany) using forward primer (5′-CCCTGTGGGTTTTACACTTAA), reverse primer (5′-ACGATTGTGCATCAGCTGA), and a probe (5′-FAM-CCGTCTGCGGTATGTGGAAAGG TTATGG-BHQ1-3′). The heating cycle was conducted as follows: 2 min at 50 °C and 10 min at 92 °C, followed by 30 cycles of 15 s at 92 °C and 1 min at 60 °C.

2.11. Expression and purification of SARS-CoV-2 S protein

Hexa-Pro constructs (F817P, A892P, A899P, A942P, K986P, V987P, and 682-GSAS-685 substitutions at the furin cleavage site) were introduced into cDNA expressing wild-type SARS-CoV-2 S protein using PCR-based site-directed mutagenesis (Hsieh et al., 2020). These constructs were then cloned into a baculovirus expression vector (pFASTBac1), which was kindly provided by Professor Ji-Joon Song (KAIST, Daejeon, Republic of Korea) and contained a C-terminal HRV3C cleavage site, 6 × His-tag, and Strep-tag. Hexa-Pro S protein was expressed by Spodoptera frugiperda (Sf9) cells in a secreted form using the Bac-to-Bac® baculovirus system (Invitrogen). Media containing the Hexa-Pro S protein were loaded onto a HisTrap excel column (Cytiva, Marlborough, MA, USA), washed with washing buffer (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, and 20 mM imidazole), and eluted with elution buffer (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, and 500 mM imidazole). Next, the Hexa-Pro S protein was incubated with HRV3C protease (Takara Bio) for 16 h at 4 °C to cleave the tag-peptides. The tag-peptides were removed using a second Ni-NTA column (Takara Bio), and the buffer was replaced with 20 mM Tris-Cl (pH 7.5) containing 150 mM NaCl to exclude imidazole. Finally, the Hexa-Pro S protein was purified via size exclusion chromatography (SEC) using a HiLoad 26/600 Superdex 200 pg (Cytiva) with SEC buffer (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, and 1 mM dithiothreitol), and fractions containing Hexa-Pro S protein were collected.

2.12. Grid preparation, cryo-electron microscopy (EM) data collection and processing

Purified Hexa-Pro S protein was complexed with K202.B in a 2:3 ratio (final concentration: 0.45 mg/mL) at 4 °C. After 30 min, the sample (4 μL) was applied to a glow-discharged 200-mesh Quantifoil R 1.2/1.3 holey carbon copper grid, and blotted for 5 s with a blot force of 0, prior to plunge-freezing with liquid ethane cooled by liquid nitrogen. Grid vitrification was performed using the Vitrobot Mark IV (Thermo Fisher Scientific) system at 100% humidity at 4 °C. Cryo-EM data were collected on a Titan Krios G4 Cryo-Transmission Electron Microscopy (Cryo-TEM; Thermo Fisher Scientific) with a Falcon 4 direct electron detector operated in energy-filtered TEM mode, with an energy filter of 20 eV slit width. A total of 9,797 micrographs were acquired automatically with a pixel size of 0.9013 Å/pix (130,000 × magnification) for a total dose of 40 e⁻/Ų using EPU software (Thermo Fisher Scientific) operated in electron-counting mode.

The image processing, 3D classifications, and refinements were carried out using cryoSPARC (v.3.3.1) software. Movie fractions were aligned via patch-based motion correction, and contrast transfer function parameters were measured using a patch CTF estimation; 1,086,347 initial particles were auto-picked using Blob-picker, and 275,077 particles were selected using 2D classification with a box size of 400 pixels. The particles with a good 2D class average were combined and run through the next round of 2D classification. Ab-initio 3D initial model reconstitution with two classes using these 220,653 particles was performed, and the results were subjected to local motion correction, global contrast transfer function estimation, and local estimation for heterogeneous refinement. Using 3D variability analysis, four classes with large variance were excluded among the 20 classes. After non-uniform refinement with 194,852 particles, a Coulomb potential map was obtained at an average resolution of 3.1 Å with C1 symmetry. To reveal the epitope of S protein, focused refinement using local refinement was performed with a soft mask in UCSF Chimera, resulting in 3.3 Å resolution for the RBD and K202.B. All reported resolutions were based on the gold-standard Fourier Shell Correlation (FSC) = 0.143 criteria. The overall workflow of cryo-EM data processing is shown in Fig. S1.

2.13. Model building and refinement

The SARS-CoV-2 S trimer (PDB ID: 7VXM), K202.B^Fab (PDB ID: 7D85), and K202.B^scFv (PDB ID: 2A9N) structures were used for initial model building. All models were fitted into the EM potential map using UCSF Chimera 1.15 (Goddard et al., 2018). The amino acid chain was refined manually in Coot-0.9.8.1 (Emsley and Cowtan, 2004). For further adjustment, the structure was refined in PHENIX-1.19.2 (Adams et al., 2010). The statistics were validated using PHENIX-1.19.2 and are summarized in Table S5. Visualization and evaluation of the 3D volume map were performed in USCF Chimera with or without the PyMOL Molecular Graphics System, version 2.2 (Schrödinger, LLC).

2.14. In vivo mouse study

For in vivo efficacy studies, 8-week-old female B6.Cg-Tg(K18-ACE2)2Prlmn/J (hACE2) mice (The Jackson Laboratory, CA, USA), were housed in a certified A/BSL3 facility (Korea Zoonosis Research Institute, Iksan, Republic of Korea). All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at KNOTUS (No. 22-KE-0076), and all experimental protocols requiring biosafety were approved by the Institutional Biosafety Committee of Jeonbuk National University (approval number: JBNU 2020-11-003-003) and performed in a biosafety cabinet at the BL3 and ABL3 facilities of Korea Zoonosis Research Institute at Jeonbuk National University. The hACE2-transgenic (hACE2-TG) mice (n = 7) were intranasally inoculated with 30 μL of wild-type or B.1.617.2 variant virus (1 × 10⁴ PFU) under anesthesia. Three hours after infection, PBS, mAbs, or bsAbs were injected intravenously. The mice were monitored daily for weight change and clinical severity based on the criteria listed in Table S1.

The SARS-CoV-2 burden in lung tissues was determined via RT-qPCR. Lung tissues were harvested from hACE2-TG mice 6 days after SARS-CoV-2 wild-type or B.1.617.2 variant infection, and total RNAs were extracted from the collected tissues using Wizol™ Reagent (Wizbiosolutions, Seongnam, Republic of Korea). Following the reverse transcription of total RNA using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster, CA, USA), the samples were subjected to RT-qPCR using a CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA). The reaction mixture (20 μL total) contained 2 μL of template cDNA, 10 μL of 2 × Premix Ex Taq, 200 nM primer, and a probe (E gene: forward primer 5′-ACAGGTACGTTAATAGTTAATAGCGT-3′, reverse primer 5′-ATATTGCAGCAGTACGCACACA-3′, probe 5′-FAM-ACACTAGCCATCCTTACTGCG CTTCG-BHQ1-3′; RdRp gene: forward primer 5′-ATGAGCTTAGTCCTGTTG-3′, reverse primer 5′-CTCCCTTTGTTGTGTTGT-3′, probe 5′-HEX-AGATGTCTTGTGCTGCCGGTA-BHQ1-3′) (Corman et al., 2020). These reactions were denatured at 95 °C for 30 s, and then subjected to 45 cycles of 95 °C for 5 s and 60 °C for 20 s. After completion of the reaction cycles, the temperature was increased from 65 to 95 °C at a rate of 0.2 °C/15 s and fluorescence was measured every 5 s to construct a melting curve. A control sample lacking template DNA was run with each assay. All measurements were performed in duplicate to ensure reproducibility. The authenticity of the amplified product was determined using melting curve analysis (Cho et al., 2008). All data were analyzed using Bio-Rad CFX Manager analysis software version 2.1 (Bio-Rad Laboratories). The viral burden was expressed by the copy number of viral RNA per nanogram of total RNA after calculating the absolute copy number of viral RNA in comparison with the standard cDNA template.

2.15. Histology

Excised mouse lung tissues were fixed with 4% (v/v) paraformaldehyde (PFA) in PBS and processed for paraffin embedding. The paraffin blocks were sliced into 3 μm sections using a microtome (HistoCore MULTICUT R; Leica, Germany) and mounted on silane-coated glass slides (5116-20F; Muto, Tokyo, Japan). Hematoxylin and eosin, periodic acid–Schiff, and modified Masson's trichrome stains were used to identify histopathological changes in all the organs. The histopathology of the lung tissue was observed using light microscopy (Axio Scope A1; Carl Zeiss). Pathological scores were determined based on the percentage of inflammation area for each section in each group using the following scoring system: 0, no pathological change; 1, affected area (≤10%); 2, affected area (10–50%); 3, affected area (≥50%); an additional 0.5 point was added when pulmonary edema and/or alveolar hemorrhage was observed.

2.16. In vitro antibody-dependent enhancement assay

Each SARS-CoV-2 pseudotyped virus (50 μL; 1 × 10⁷ PFU/mL) was preincubated with different concentrations of K202.B (0.044, 0.138, 0.42, 1.24, 3.7, 11.1, 33.3, and 100 nM) in culture medium. After 30 min of incubation at RT, the mixture was added to 293T, 293T/hACE2, K562, or THP-1 cells (1 × 10⁴ cells in a 96-well plate). The cells were cultured for 24 h, and the luciferase activity of infected cells was measured as described in “pseudotyped virus neutralization assay.”

2.17. Endothelial cell viability assay

Approximately 5 × 10³ HUVECs were plated in 96-well plates and incubated in the presence or absence of 20 μg/mL K202.B or 36 μg/mL 5-fluorouracil for 24 h at 37 °C. Cell viability was determined using the Cell Counting Kit-8 (Sigma) according to the manufacturer's instructions. The final absorbance was measured at 450 nm using a spectrophotometer (BioTek).

2.18. Flow cytometry

The effects of K202.B on endothelial cell activation were evaluated by incubating 2 × 10⁵ HUVECs with or without 20 ng/mL of human tumor necrosis factor-α (hTNFα; Millipore), 20 μg/mL of K202.B, or control IgG for 24 h. The cells were fixed with 4% (v/v) PFA in PBS and incubated with 10 μg/well of intercellular cell adhesion molecule-1 (ICAM-1; Abcam, Cambridge, MA, USA) or vascular cell adhesion molecule-1 (VCAM-1; Abcam) antibody for 1 h at 25 °C. Then, Alexa Fluor 647-conjugated anti-mouse IgG or antirabbit IgG (1:1000; Invitrogen) was incubated for 1 h at 25 °C. All samples were analyzed using flow cytometry with the aid of FlowJo software (TreeStar, Ashland, OR, USA).

2.19. In vivo toxicity and serum pharmacokinetic analysis

In vivo toxicity and serum pharmacokinetic studies using animals were approved by the IACUC (Approval No. NCC-21-693) of the National Cancer Center, Republic of Korea. Eight-week-old female Institute of Cancer Research (ICR) mice (Orient Bio Inc., Seongnam, Republic of Korea) were intravenously injected with 5 or 30 mg/kg of K202.B (n = 3 per group). At 4, 8, 24, 72, 120, 168, 264, 384, and 504 h post-inoculation, blood samples (50 μL) were collected from each mouse and centrifuged at 5000×g for 20 min at 4 °C. The serum was stored at −80 °C for evaluation of biochemical parameters. Serum levels of glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), creatinine (CRE), and blood urea nitrogen (BUN) were measured using a Fuji Dri-Chem 3500 Biochemistry Analyzer (Fujifilm, Tokyo, Japan). Serum levels of K202.B were determined using a human IgG ELISA kit (Abcam) according to the manufacturer's instructions. Optical density was measured using a Synergy H1 microplate reader, and values were compared to those from a concurrently analyzed standard curve.

2.20. Statistical analysis

Data were analyzed with GraphPad Prism 8.0 software using two-tailed Student's t-test for comparisons between two groups, and one-way analysis of variance (ANOVA) with Bonferroni's correction for multiple comparisons. All data represent the mean ± standard deviation (S.D.). A P-value less than 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001).

3. Results

3.1. Design, generation, and characterization of bsAbs

Biopanning was performed using phage-display technology to isolate four SARS-CoV-2 RBD-specific human scFvs from the human synthetic scFv library. To prevent Fab arm exchange that results in an unwanted heterogeneous mixture of antibodies by half molecule exchanged with endogenous IgG4, we created IgG4-based mAbs with S228P mutations [IgG4 (S228P)] (Silva et al., 2015). The selected IgG4 mAbs were designated as K102.1, K102.2, K102.3, and K102.4 (Fig. 1 A and B). Among these, we identified a noncompeting pair of mAbs, K102.1 and K102.2, that recognized independent epitopes of the SARS-CoV-2 RBD using a competition ELISA (Fig. 1C). Based on these parental mAbs, we generated two forms of IgG4 (S228P)-(scFv)₂ bsAb, K202.A and K202.B (Fig. 1D and E).

Fig. 1.

Design and generation of engineered bsAbs. (A) Schematic representation of a SARS-CoV-2 specific scFv selection method using phage display biopanning. (B) After four rounds of biopanning, 96 scFv clones were selected and tested for reactivity analysis using phage ELISA. Open bars: BSA binding, Black bars: SARS-CoV-2 RBD binding. (C) Screening of noncompeting monoclonal antibody in K102.1-RBD interaction using competition ELISA. Results are represented as mean ± S.D. of duplicates and represent one of two independent experiments. Schematic diagram of engineered antibodies with bicistronic mammalian expression vector design of K202.A (D) and K202.B (E). Using K102.1 IgG as a template, K102.2 scFv was conjugated to the C-terminal of the heavy chain (K202.A) or the N-terminal of the light chain (K202.B). bsAbs, bispecific antibodies; scFv, single-chain variable fragment; ELISA, enzyme-linked immunosorbent; RBD, receptor-binding domain.

Then, we conducted SPR analysis to measure the binding kinetics of bsAbs for the purified RBDs of wild-type, and B.1.1.7, B.1.351, P.1, B.1.617.2, and B.1.617.1 variants of SARS-CoV-2. K202.B showed strong binding to the RBD of SARS-CoV-2 wild-type and the variants in the sub- or low nanomolar range (Fig. 2 A and Table S2). To further confirm whether K202.B could recognize two independent epitopes of SARS-CoV-2 RBD, we performed competition assays using SPR. The results showed that K202.B could bind to the RBD after saturation with K102.1 or K102.2 (Fig. 2B). In contrast, neither K102.1 nor K102.2 bound to the RBD after saturation with K202.B (Fig. 2C), suggesting that the bsAb K202.B specifically recognized two independent binding sites.

Fig. 2.

Biochemical characterization of bsAbs for SARS-CoV-2 RBDs. (A) Characterization of the binding kinetics of bsAbs K202.A (green) and K202.B (red) to SARS-CoV-2 wild-type and variant RBDs using SPR. (B) and (C) Cross-reactive properties of mAbs [K102.1 (blue) and K102.2 (purple)] and K202.B (red) to the wild-type SARS-CoV-2 RBD. (B) K202.B was injected onto an RBD-immobilized sensor chip saturated with K102.1 (left) or K102.2 (right). (C) Inversely, K102.1 (left) or K102.2 (right) was injected on the chip saturated with K202.B. Paler colors denote controls with the second injected antibody alone. Results are representative of at least two independent experiments. bsAbs, bispecific antibodies; RBD, receptor-binding domain; SPR, surface plasmon resonance.

3.2. Inhibitory activity of bsAb in hACE2-RBD interaction and SARS-CoV-2 pseudotyped and live virus infection in vitro

To assess the inhibitory activity of K202.B in hACE2-RBD interactions, we performed ELISA-based inhibition assays with recombinant hACE2 and RBD proteins in the presence or absence of K202.B. In the case of hACE2 binding to all the tested RBDs of SARS-CoV-2 wild-type and variants, the antibody exhibited potent inhibitory effects in sub- or low nanomolar range (Fig. S2A and Table S3). Furthermore, K202.B also exhibited a strong inhibitory effect on hACE2 binding to RBDs with N354D/D364Y, V367F, W436R, R408I, G476S, V483A, V341I, F342L, or A435S mutations, which were observed during the early transmission phase (Fig. S2B).

To investigate the neutralizing potency of K202.B against live viral infections, we performed neutralization assays with SARS-CoV-2 wild-type and B.1.617.2 variant in the presence or absence of K102.1 and K202.B. To test the impact of K202.B on the expression of viral envelope (E) gene, we performed RT-qPCR (Fig. 3 A and B). The result shows that K202.B exhibited more potent inhibitory effects on the expression of viral E gene compared with the parental mAb, K102.1.

Fig. 3.

Neutralizing properties of bsAbs against SARS-CoV-2 live and pseudotyped virus in vitro. Measurement of neutralization efficacy via viral gene expression of K102.1 and K202.B against live SARS-CoV-2 wild-type (A) and B.1.617.2 variant (B) using RT-qPCR. Data are represented as mean ± S.D. of duplicates and represent one of two independent experiments. (C) Neutralizing activity of parental mAbs, mAb cocktail, or bsAbs against pseudotyped virus infections with SARS-CoV-2 wild-type and variants in 293T/hACE2 cells. Data are represented as mean ± S.D. of duplicates and represent one of two independent experiments. bsAbs, bispecific antibodies; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; mAbs, monoclonal antibodies, hACE2, human angiotensin-converting enzyme 2; SPR, surface plasmon resonance; S.D., Standard Deviation.

To further evaluate the neutralizing ability of K202.B in the infection of a variety of SARS-CoV-2 variants, we conducted SARS-CoV-2 pseudotyped virus neutralization assays using hACE2-overexpressing 293T stable cell lines (293T/hACE2 cells) in the presence or absence of parental mAbs, mAb cocktail, and K202.B (Figs. S3A and B). K202.B exhibited stronger inhibitory effects on the infection of almost all the tested pseudotyped viruses than parental mAbs or the mAb cocktail with the IC₅₀ value of mostly subnanomolar or nanomolar concentration (Fig. 3C and Table S4). However, all antibodies had no effect on that of BA.1 variant.

3.3. Structural analysis of K202.B complexes with RBDs of SARS-CoV-2 S proteins

To understand the structural basis of K202.B in SARS-CoV-2 neutralization, we determined the cryo-EM structure of K202.B in a complex with the trimeric SARS-CoV-2 S proteins at 3.1 Å resolution (Fig. S1 and Table S5). Intriguingly, unlike the previously reported mAbs specifically targeting RBD, K202.B was observed to bind only to a fully open, three-RBD-up conformation of the trimeric S proteins (Fig. 4 A and B). Furthermore, the cryo-EM data also revealed that K202.B independently interconnects the two different RBD subsets in a SARS-CoV-2 S homotrimer via inter-protomer interactions. More specifically, K202.B^Fab (K102.1 Fab part of K202.B) interacted with S protomer 1, whereas K202.B^scFv (K102.2 scFv part of K202.B) was bound to S protomer 3 in a counterclockwise direction (Fig. 4C).

Fig. 4.

Cryo-electron microscopy (EM) structure and mechanism of action of K202.B in a complex with SARS-CoV-2 spike proteins (A) Side-view of the complex structure. The trimeric complex of SARS-CoV-2 spike protein (green and grey) reveals a three-open state conformation (EMD-33734, PDB: 7YC5) when K202.B, comprising K202.B^Fab (blue and light blue) and K202.B^scFv (purple and light purple), individually binds to all RBDs of the trimeric spike protein. The complex structure has C1 symmetry. (B) Side-view of the complex structure as a cartoon. (C) K202.B protomer 1 comprises K202.B^Fab (blue) and K202.B^scFv (purple). K202.B^Fab of protomer 1 interacts with spike protomer 1 (green), while K202.B^scFv of protomer 1 simultaneously interacts with spike protomer 3 (grey). K202.B^scFv C-terminus and light chain N-terminus of K202.B^Fab are covalently linked with G4S linker (red dash line) at a distance of 15.1 Å. (D) Local refinement results of SARS-CoV-2 RBD (green), K202.B^Fab (blue), and K202.B^scFv (purple) (EMD-33642, PDB: 7Y6K). (E) Overlap mapping of K202.B epitope and hACE2 binding sites. Overlapped sites (red) between the K202.B^Fab epitope and hACE2 binding sites. (F) The epitopes of K202.B^Fab (blue) and K202.B^scFv (purple) are independently shown in SARS-CoV-2 RBD (grey). RBD, receptor-binding domain; hACE2, human angiotensin-converting enzyme 2.

To determine the exact binding mode of the antibody, we conducted focused refinement on the interface region of SARS-CoV-2 RBD and K202.B (Fig. 4D). The recognition sites for K202.B^Fab were identified as Arg403, Thr415, Lys417, Tyr421, G446, Tyr449, Tyr453, Leu455-Phe456, Ala475, Val483-Asn487, Tyr489-Phe490, Gln493-Ser494, and Gln498 of the RBD, indicating that the major binding sites are located in the receptor-binding motif (RBM; residues 438–506), a pivotal region of the RBD for interaction with hACE2 (Fig. 4E and F) (Lan et al., 2020). The binding sites for K202.B^scFv were also determined as Tyr369-Ser371, Phe374-Tyr380, Val382, Pro384-Thr385, Gly404-Asp405, Val407-Arg408, Val503-Gly504, and Tyr508 revealing that most binding sites are in the RBD core region (residues 331–437, 507–524) far from the RBM (Fig. 4E and F). The buried surface area of RBD by K202.B was about 1882 Å², whereas those for K202.B^Fab and K202.B^scFv were 989 and 893 Å², respectively, showing that the RBD was more broadly masked by K202.B than K202.B^Fab or K202.B^scFv alone.

3.4. In vivo efficacy of K202.B in wild-type SARS-CoV-2-infected animal models

To evaluate the in vivo efficacy of K202.B against wild-type SARS-CoV-2, we first investigated the pharmacokinetics of K202.B, and found that K202.B exhibited an in vivo half-life of approximately 78 h in mice (Fig. S4). Next, the viruses were intranasally administered to the K18-hACE2 transgenic (TG) mice. After 3 h, the mice received intravenous injections of two doses (5 and 30 mg/kg) of K202.B, or a single dose of K102.1 (30 mg/kg) (Fig. 5 A). At 6 days post-infection (dpi), each K202.B-treated group exhibited a clinical severity score of $\le$ 1 (mostly score 0), whereas PBS- or K102.1-treated groups displayed scores greater than 3, characterized by a very ruffled coat, slightly closed eyes, and/or a moribund state (Fig. 5B). Lung samples from all mice sacrificed at 6 dpi were subjected to RT-qPCR to determine the relative expression of viral E and RNA-dependent RNA polymerase (RdRp) genes. The expression of both viral genes was significantly reduced in a dose-dependent manner in each K202.B-treated group when compared with that in the PBS-treated group (Fig. 5C and D).

Fig. 5.

In vivo efficacy of K202.B and K102.1 in hACE2 transgenic mice infected with wild-type SARS-CoV-2. (A) Schematic description of the experimental design for in vivo efficacy evaluation of K102.1 and K202.B on wild-type SARS-CoV-2-infected K18-hACE2 transgenic mice (n = 7 per group). (B) Clinical severity at 6 dpi. The level of the viral E (C) or RdRp (D) gene expression in lung tissue was measured using RT-qPCR at 6 dpi. (E) Two representative histopathological images of the lung tissue in each group at 6 dpi. Scale bar (40 × ): 500 μm; scale bar (400 × ): 50 μm. All values represent the mean ± S.D. of seven biological replicates. Data were statistically analyzed using two-tailed Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001). hACE2, human angiotensin-converting enzyme 2; RdRp, RNA-dependent RNA polymerase; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; S.D., Standard Deviation.

Histopathological examination of the lungs from infected mice at 6 dpi showed that PBS- and K102.1-treated mice scored $\ge$ 1 or 2 due to significant pulmonary lesions. In comparison, a high proportion of the K202.B-treated group showed a score of 0 at both, 5 and 30 mg/kg doses (Table S6). Further histopathological analyses revealed normal features in the K202.B-treated lungs, whereas PBS- and K102.1-treated mice exhibited severe pulmonary edema or alveolar hemorrhage (Fig. 5E).

3.5. In vivo efficacy of K202.B in SARS-CoV-2 B.1.617.2 variant-infected animal models

To evaluate the in vivo efficacy of K202.B against the SARS-CoV-2 B.1.617.2 variant, 5 or 30 mg/kg K202.B was intravenously injected into K18-hACE2 TG mice 3 h after an intranasal viral challenge (Fig. 6 A). K202.B-treated mice showed an average score of ≤1 for clinical severity (most mice scored 0) (Fig. 6B). Additionally, K202.B effectively reduced the expression of both viral E and RdRp genes. In the 30 mg/kg K202.B-treated group, viral gene expression was barely detected at 6 dpi (Fig. 6C and D). Histopathological examination on 6th dpi revealed that the pathological score of 0 was increased in a dose-dependent manner in K202.B-treated group, whereas scores of 1–2 (mostly score 2) with pulmonary lesion was observed in PBS-treated group (Table S7). Histopathological analyses revealed normal features in the K202.B-treated lungs, whereas PBS-treated mice showed distinct pulmonary edema or alveolar hemorrhage (Fig. 6E).

Fig. 6.

In vivo efficacy of K202.B in SARS-CoV-2 B.1.617.2 variant–infected hACE2 transgenic mice. (A) Experimental design to test in vivo efficacy of K202.B in hACE2 transgenic mouse models against the SARS-CoV-2 B.1.617.2 variant. (B) Clinical severity at 6 dpi. The level of the viral E (C) or RdRp (D) gene expression in lung tissue was measured using RT-qPCR at 6 dpi. (E) Two representative histopathological images of the lung tissue in each group on 6 dpi. Scale bar (40 $\times$): 500 μm; scale bar (400 $\times$): 50 μm. All of the values represent the mean ± S.D. of seven biological replicates. Data were statistically analyzed using two-tailed Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001). hACE2, human angiotensin-converting enzyme 2; RdRp, RNA-dependent RNA polymerase; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; S.D., Standard Deviation.

To assess the effect of K202.B on the antibody-dependent enhancement (ADE), we used permissive (293T/hACE2 cells) and Fc gamma receptor-bearing cells (293T, K562, and THP-1 cells). No significant changes were observed in any pseudotyped virus infection, indicating that K202.B may not induce ADE in vivo (Fig. S5A). Next, our in vitro toxicity assays revealed no significant effect of K202.B on endothelial cell viability and activation in the K202.B-treated group compared with the control group, suggesting no severe endothelial toxicity (Figs. S5B and C). Furthermore, we administered 5 or 30 mg/kg K202.B to mice via a single intravenous injection and evaluated body weight change and hepatic and renal toxicity by biochemically measuring enzyme activity; no significant changes were observed in the K202.B-treated groups, indicating no in vivo toxicity (Figs. S5D–F).

4. Discussion

The rapid emergence and continuous spread of SARS-CoV-2 variants have presented a significant challenge to public healthcare, and novel mutations threaten the effectiveness of current mAb-based therapy against COVID-19 (Pinto et al., 2020; Singh et al., 2022). In this regard, the rapid development of a novel antibody platform with therapeutic potential is necessary for protection against new or resurgent SARS-CoV-2 variants. In the present study, based on a noncompeting pair of phage display-derived mAbs isolated from an established human recombinant antibody library, our engineered K202.B bsAb with IgG4 (S228P)-(scFv)₂ showed therapeutic potential with superior neutralizing activity against a variety of SARS-CoV-2 variants. Furthermore, we elucidated the structural basis of the bsAb in SARS-CoV-2 neutralization using cryo-EM analysis. Therefore, our study not only provides clear structural insights for the better understanding of the mode of action of the novel engineered bsAb, but also suggests that compared with most of the existing SARS-CoV-2-specific neutralizing antibodies derived from B cells of convalescent whole blood or immunized transgenic mice, which is highly time consuming and labor intensive, the rapid development of phage display-derived neutralizing bsAb may be useful as an alternative method for implementing timely measures against a life-threatening COVID-19 (Brouwer et al., 2020; Cao et al., 2020; Jones et al., 2021; Kim et al., 2021; Pinto et al., 2020; Shan et al., 2021).

BsAbs harness the specificities of two mAbs and combine them to simultaneously recognize two independent epitopes or antigens (Pantaleo et al., 2022). Here, we propose that the IgG4-based K202.B developed by us in this study may have therapeutic potential with broad and potent neutralizing activity against a variety of SARS-CoV-2 variants. Several lines of evidence support our notion. K202.B is a fully human bsAb that may have a low immunogenicity risk. Similar to modern IgG4-based therapeutic antibodies including pembrolizumab, an anti-PD-1 therapeutic antibody approved by the US FDA, IgG4-based K202.B possesses the S228P point mutation, converting the hinge to that of an IgG1-like molecule, thus preventing the formation of bispecific molecules between the drug and endogenous human IgGs (Levine et al., 2019). K202.B exhibited a high avidity to SARS-CoV-2 variant RBDs in sub- or low nanomolar ranges, which is possible for in vivo SARS-CoV-2 specific targeting. Furthermore, it showed superior and broad-spectrum of inhibitory activity in almost all SARS-CoV-2 wild-type and variants in pseudotyped and live virus infection tests than parental mAbs or mAb cocktails without ADE. Through intravenous injections, the most widely used clinical route of antibody drug administration for systemic circulation (Keizer et al., 2010), monotherapy with a low dose (5 mg/kg) of K202.B compared with K102.1 (30 mg/kg) showed potent neutralizing ability in two different SARS-CoV-2, wild-type and B.1.617.2 variant, infected-K18-hACE2 TG mice that represent the features of human COVID-19 disease progress and pathogenesis, without significant toxicity in vivo. Thus far, most virus-neutralizing antibodies have been developed as an IgG1 subtype exhibiting strong effector functions, such as antibody-dependent cell cytotoxicity and cell phagocytosis (Hansen et al., 2020; Pinto et al., 2020; Tay et al., 2019). However, they may also show unwanted in vivo toxicity. Here, we present for the first time a novel neutralizing bsAb with an IgG4 subtype that may also be effective against COVID-19 without distinct in vivo toxicity. In addition, although we verified that K202.B had no effect on BA.1 neutralization, our findings lead us to speculate that the novel IgG4-based bsAb may be used as a promising antibody platform technology to generate multi-specific antibodies as a timely measure against the rapid emergence of new or resurgent SARS-CoV-2 viruses.

To date, several neutralizing mAbs specifically targeting the RBD have been approved by the US FDA or EMA for emergency use, and their epitopes have been identified using cryo-EM or X-ray crystallography. Most mAbs including imdevimab, casirivimab, regdanvimab, bamlanivimab, etesevimab, tixagevimab, and cilgavimab recognize epitopes located in and/or overlapping with the RBM, whereas sotrovimab binds to an RBD core region outside the RBM, indicating that both regions are critical for SARS-CoV-2 neutralization (Table S8) (Hansen et al., 2020; Jones et al., 2021; Kim et al., 2021; Pinto et al., 2020; Shi et al., 2020; Zost et al., 2020). In this study, our cryo-EM analysis revealed that the major binding sites for K202.B^Fab are located in the RBM, whereas the binding sites for K202.B^scFv are residues of RBD core region located at a distance from the RBM. Furthermore, K202.B^Fab and K202.B^scFv interconnects two independent epitopes of the SARS-CoV-2 RBD via inter-protomer interactions, proposing that RBD may be more broadly masked by the simultaneous binding of K202.B than K202.B^Fab or K202.B^scFv alone. Our results in Fig. 3 revealed that K202.B exerted superior neutralizing activity against SARS-CoV-2 variants, compared with that of each parental mAb or mAb cocktail. Thus, we suggest that the superior neutralizing activity of K202.B to mAbs or mAb cocktail may be due to not only the increased numbers of the paratopes in the K202.B but also the simultaneous recognition of two independent epitopes, allowing the bsAb to block the RBD-hACE2 interaction more efficiently than single-epitope mAbs. To our knowledge, this is the first study directly showing the structural basis of a bsAb complexed with the RBDs of SARS-CoV-2 S proteins for the better understanding of the superior neutralizing activity of bsAbs.

The up state of the SARS-CoV-2 S trimer is important for ensuring stability in an open state, which favors virulence of new variants (Giron et al., 2021). Previous structural and functional analyses have revealed that emerging SARS-CoV-2 variants with escape mutations exhibit a high infectivity, correlating with an increased proportion of the S trimer in the open conformation state (Wang et al., 2021a, 2021b, 2022b; Xu et al., 2021). Therefore, antibodies specific to the open conformation of SARS-CoV-2 S trimer may be critical for the effective neutralization of SARS-CoV-2 variants. Intriguingly, our intensive cryo-EM analysis revealed that 149,542 of 194,852 particles showed a K202.B–S trimer complex bound only to a fully open three-RBD-up conformation of the S trimer, whereas the rest of the particles were observed in the fully closed three-RBD-down conformation without K202.B binding. This indicates that the mode of action of the K202.B is to bind specifically to the fully open three-RBD-up conformation of the S trimer different from most other existing SARS-CoV-2 neutralizing antibodies (Barnes et al., 2020; Planas et al., 2021; Raybould et al., 2021). According to Wang et al., antibody binding to up conformations of the RBD may facilitate the destabilization and dissociation of S trimers, and result in decreased infectivity of SARS-CoV-2 (Wang et al., 2022a). Therefore, K202.B specific to a fully open three-RBD-up conformation of the trimeric S proteins may allow for more potent neutralization of a variety of SARS-CoV-2 variants than that yielded by mAb-based therapy.

5. Conclusions

This study is the first to present the broad and potent neutralizing activity of phage display-derived bsAb with an IgG4 (S228P)-(scFv)₂ form against a variety of SARS-CoV-2 variants and its mode of action based on structural analysis. On the basis of currently available evidence, we propose an underlying mechanism, whereby the K202.B specifically binds to, and interconnects two independent RBD epitopes in a fully open three-RBD-up conformation of the S trimer in the proximal region of hACE2-expressing host cells through inter-protomer interactions. K202.B bsAb effectively interferes with the RBD-hACE2 interactions by specifically targeting a fully open spike conformation that is critical for the viral infection, ultimately inducing conformational changes in the S trimer and eventual destabilization that reduces the infectivity of SARS-CoV-2 variants. We believe that the novel approach for the development of the IgG4-based bsAb represents an effective antibody development strategy for timely and successful management of constantly changing SARS-CoV-2, resulting in COVID-19 pandemic. In the near future, we plan to generate IgG4-based bsAbs based on a noncompeting pair of Omicron-specific and phage display-derived mAbs isolated from an established human recombinant antibody library, showing utility of the antibody platform for combating the current SARS-CoV-2 variants of concern.

Funding sources

This research was supported by the Bio & Medical Technology Development Program of the National Research Foundation of Korea (grant number: NRF-2020M3A9I2107093, NRF-2019M3E5D6063903, and NRF-2016R1A5A1010764) and the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (grant number: HI22C0360).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.