Guided by molecular modeling, we generated a set of VHH72 variants with point mutations in the residues that line the cavity that accommodates Y369 of SARS-CoV-2 RBD. These variants were rapidly prototyped in parallel as Fc fusions in the yeast Pichia pastoris to ensure that we would only prioritize mutations that enhanced affinity in our intended bivalent Fc-fusion drug context. Introduction of S56A resulted in slower dissociation in this bivalent context and enhanced binding to SARS-CoV-2 spike expressed on the surface of 293T cells (fig. S4). Based on this prototyping in Pichia pastoris, we selected VHH72 with a S56A mutation in the CDR2 for further analysis. All VHH72-Fc fusion constructs mentioned below were produced in and purified from transiently transfected CHO cells.
All the VHH72-Fc variants were efficiently produced in transiently transfected ExpiCHO cells with yields as high as 1.2 g per liter of cell culture after purification (table S4). The resulting humVHH_S56A/LALAPG-Fc/Gen2 construct displayed similar RBD-binding kinetics and SARS-CoV-2 neutralizing activity as humVHH_S56A/LALAPG-Fc (table S1, and fig. S7E). Combining our two potency-enhancing modalities, we also introduced the S56A mutation in each of the four VHH72 moieties in the tetravalent format in construct (humVHH_S56A)2/LALAPG-Fc/Gen2, which further increased the in vitro antiviral potency, reaching a PRNT50 value of 0.02 μg/ml (a 50-fold improved over the pre-lead parental construct, fig. S7E and table S1). Finally, we also generated hIgG1 LALA variants of the final bivalent sequence design, with or without the S56A VHH mutation (table S1).
There remains a need for safe and effective anti-SARS-CoV-2 drugs that can prevent or treat COVID-19. Here, we report the protein engineering-based drug development of a potent cross-neutralizing, VOC-resistant anti-COVID-19 biologic. This biologic was based on a humanized, in silico affinity-enhanced VHH72 variant, which, in bivalent and tetravalent VHH-Fc format, demonstrates strong anti-viral efficacy in a hamster challenge model. Compared with conventional human monoclonal antibodies, the VHH-Fc fusion construct is smaller (80 kDa versus 150 kDa) and encoded by a single gene, giving advantages with respect to dosing and manufacturability. For example, the almost two times smaller size of a VHH-Fc is a considerable advantage for subcutaneous formulations, where a very high molarity of the antibody drug needs to be provided in a small (1 to 2 ml) volume that is feasible for injection via that route. Furthermore, the single-gene encoded nature of the VHH-Fc format, as well as its simpler homodimer assembly pathway, allow for manufacturing in alternative microbial host systems such as yeast. Such a microbial expression system comes with considerably lower cost of goods and rapid pandemic response re-manufacturing opportunities.
Our study has several limitations. Although the binding region of VHH72 in SARS-CoV-2 RBD is highly conserved, suggesting a functional constraint for this part of the RBD, in vitro escape selection experiments are needed to estimate the likelihood of the emergence and fitness of mutant viruses with reduced susceptibility to XVR011. Second, although we demonstrated in vivo protection against challenge with three different SARS-CoV-2 virus strains, these were all derived from clinical isolates that circulated early in the pandemic. Finally, the safety and effectiveness of XVR011 in COVID-19 patients has not yet been demonstrated. Healthy volunteer Phase 1a trial and a Phase 1b/2 trial in patients with moderate disease severity are presently ongoing to address this question.
The pharmaceutically fully developed VHH72-based biologic named XVR011, that combines potent neutralizing activity with high stability, broad coverage and silenced Fc effector functionality for enhanced safety, has currently completed cGMP-manufacturing and formal preclinical development. Clinical studies are now being started to evaluate safety and efficacy of rapid administration upon hospitalization of patients within the first week of COVID-19 symptoms (NCT04884295). In conclusion, XVR011 represents a promising antibody-based countermeasure against disease caused by SARS-CoV-2 and potential future zoonotic outbreaks with related Sarbecoviruses.
MATERIALS AND METHODS
Study design. This study was designed to generate a candidate VHH72-Fc-based biologic, that is ready for clinical development. We used structure-guided molecular modelling combined with a rapid yeast-based screening approach to identify VHH72-Fc derivatives with enhanced affinity for SARS-CoV-2 RBD. Tandem repeats of VHH72 fused to human IgG1 Fc were generated as an alternative strategy to increase the affinity for SARS-CoV-2 RBD. VHH72 frame work regions were humanized, linkers between the VHH and human IgG1 Fc were optimized, and we opted for an Fc domain with mutations that reduce Fc effector functions and CHO cells for recombinant protein expression. Mass-spectrometry, dynamic light scattering, a polyethylene glycol protein solubility assay and hydrophobic interaction chromatography analysis were performed to select the VHH-Fc candidates with optimal physico-chemical properties. Lack of non-specific antigen binding was demonstrated using a human membrane protein microarray assay and surface plasmon resonance was used to show reduced FcγR binding of the mutated Fc. Candidate VHH72-Fc derivatives for further development were also compared in in vitro virus neutralization assays using VSV pseudotyped with SARS-CoV-1 and -2 spike and authentic SARS-CoV-2 virus, including VOCs.
Prophylactic and therapeutic activity of selected VHH72-Fc constructs were assessed in SARS-CoV-2 challenged K18-hACE2 transgenic mice and hamsters. The sample size of the K18-hACE2 transgenic mice was estimated based on experience with other respiratory viruses to give statistical power while minimizing animal use. Experimentalists involved in the mouse studies were not blinded, because objective measurements were used. To evaluate antiviral activity in the hamster model, we wanted to detect at least 1 log reduction of viral RNA in treated compared to untreated infected control animals. The group size was calculated based on the independent t test with an effect size of 2.0 and a power of 80% (effect size = delta mean/SD = 1 log decrease in viral RNA/0.5 log), resulting in 6 animals/group. The sample sizes were maximized considering the limits in the BSL3 housing capacity and the numbers of animals that can be handled under BSL3 conditions. All caretakers and technicians involved in the hamster studies were blinded to group allocation in the animal facility, and to sample numbers for analysis (qPCR, titration, and histology).
Protein expression and purification. P. pastoris cultures were grown in liquid YPD (1% yeast extract, 2% peptone, 2% D-glucose) or on solid YPD-agar (1% yeast extract, 2% peptone, 2% D-glucose, 2% agar) and selected with 100 μg/ml Zeocin® (InvivoGen). For small scale expression screening, 2-3 single colonies of P. pastoris OCH1 transformed with pX-VHH72-xxx-hIgGhinge-hIgGFc were inoculated in 2 ml BMDY (1% yeast extract, 2% peptone, 100 mM KH2PO4/K2HPO4, 1.34% YNB, 2% D-glucose, pH 6) or BMGY (same composition but with 1% glycerol replacing the 2% D-glucose) in a 24 deep well block. After 50 hours of expression in a shaking incubator (28°C, 225 rpm), the medium was collected by centrifugation at 1.500 g, 4°C for 5 min. Protein expression was evaluated on Coomassie-stained SDS-PAGE of crude supernatant. Crude supernatant was used immediately for analytics purposes (biolayer interferometry and mass spectrometry, see below) or stored at -20°C.
VHH72-Fc constructs were produced in ExpiCHO cells (Thermo Fisher Scientific) by transient transfection of the respective expression plasmids. 25 ml cultures with 6 x106 cells/ml were grown at 37°C and 8% CO2 and transfected with 20 μg of pcDNA3.3-based plasmid DNA using ExpiFectamine CHO reagent. Twenty four hours after transfection, 150 μL of ExpiCHO enhancer and 4 mL of ExpiCHO feed was added to the cells, and the cells were further incubated at 32°C and 5% CO2. Transfected cells were fed a second time 5 days post-transfection. As soon as cell viability had dropped below 75%, the culture medium was harvested. The VHH72-Fc constructs were purified from the clarified culture medium using a 5 ml MabSelect SuRe column (GE Healthcare). After a wash step with McIlvaine buffer pH 7.2, bound proteins were eluted using McIlvaine buffer pH 3. The eluted protein-containing fractions were neutralized with a saturated Na3PO4 buffer. These neutralized fractions were then pooled, and loaded onto a HiPrep Desalting column for buffer exchange into storage buffer (25 mM L Histidine, 125 mM NaCl pH 6) or onto a HiLoad 16/600 Superdex 200 pg size-exclusion column (GE-Healthcare) calibrated with phosphate-buffered saline (PBS) or storage buffer. Where data are labeled as ‘XVR011’, protein material manufactured from a stable CHO cell line has been used in the experiments, of identical sequence as humVHH_S56A/LALA-Fc/Gen2 (which is the naming we use for the protein produced in ExpiCHO transient transfection).
Open reading frames corresponding to the light and heavy chains of the hIgG1 anti-SARS-CoV-2 antibody S309 were ordered synthetically at IDT. For both, an optimized Kozak sequence was added upstream of the start codon. The secretion signal of a Mus musculus Igκ chain was used to direct secretion of the S309 light chain and of a Mus musculus IgG heavy chain to direct secretion of the S309 heavy chain. The carboxy-terminal lysine residue of the S309 heavy chain was omitted. The synthetic DNA fragments were solubilized in ultraclean water at a concentration of 20 ng/μl, A-tailed using the NEBNext-dA-tailing module (NEB), purified using CleanPCR magnetic beads (CleanNA), and inserted in pcDNA3.3-TOPO vector (ThermoFisher). The ORF of positive clones was sequence verified and pDNA of selected clones was prepared using the NucleoBond Xtra Midi kit (Machery-Nagel). Expression in ExpiCHO cells and purification of S309 was performed as described above for the VHH72-Fc constructs except that for S309 the heavy chain and light chain encoding plasmids were mixed in a ratio of 1:2. CB6 (heavy chain GenBank MT470197, light chain GenBank MT470196) was custom produced in a mammalian cell system by Genscript.
Dose-dependent neutralization of distinct VHH-Fc constructs was assessed by mixing the VHH-Fc constructs at different concentrations (three-fold serial dilutions starting from a concentration of 20 μg/ml), with 100 PFU SARS-CoV-2 in DMEM supplemented with 2% FBS and incubating the mixture at 37°C for 1h. VHH-Fc-virus complexes were then added to Vero E6 cell monolayers in 12-well plates and incubated at 37°C for 1h. Subsequently, the inoculum mixture was replaced with 0.8% (w/v) methylcellulose in DMEM supplemented with 2% FBS. After 3 days incubation at 37°C, the overlays were removed, the cells were fixed with 3.7% PFA, and stained with 0.5% crystal violet. Half-maximum neutralization titers (PRNT50) were defined as the VHH-Fc concentration that resulted in a plaque reduction of 50% across 2 or 3 independent plates.
Mouse challenge experiments. Transgenic (K18-hACE2)2Prlmn mice were originally purchased from The Jackson Laboratory. Locally bred hemizygous 10-14-week-old animals of both sexes were used for experiments. Infection experiments were performed in accordance with the guidelines of the Federation for Laboratory Animal Science Associations and the national animal welfare body. All experiments were in compliance with the German animal protection law and approved by the animal welfare committee of the Regierungspräsidium Freiburg (permit G-20/91).
WT-VHH/12GS-WT-Fc treatments were performed either by injecting approximately 500 μl of a 250 μg/ml solution of WT-VHH/12GS-WT-Fc into the peritoneum of non-anesthetized mice to reach a dose of 5 mg/kg or by applying 40 μl of the WT-VHH/12GS-WT-Fc at 1 mg/ml to the nostrils of isoflurane-anesthetized mice. Infection of isoflurane-anesthetized mice with 8 × 103 PFU of a clinical SARS-CoV-2 isolate (Muc-IMB-1/2020) was done by applying 40 μl samples to the nostrils. Infected mice were monitored for weight loss and clinical signs of disease for 14 days. Mice were euthanized by cervical dislocation on day 3 post infection to determine infectious virus in lungs by plaque assay using Vero E6 cells. Infected mice were euthanized and scored dead if they lost 25% of their initial body weight or if showing severe signs of disease. All infection experiments were performed under BSL3 conditions.
The hamster model performed at Viroclinics was as follows. Male 14-15 weeks old Syrian hamsters (Mesocricetus auratus), weighing 106 to 158 g were obtained from Janvier. The animals were housed for 7 days in individually ventilated cages (2 hamsters per cage) under BSL-2 conditions prior to transfer to a BSL-3 animal house. Six animals were used per group. Palivizumab (20 mg/kg), humVHH_S56A/LALAPG-Fc/Gen2 (20, 7, and 2 mg/kg) and (humVHH_S56A)2/LALAPG-Fc/Gen2 (20, 7 and 2 mg/kg) were administered by intraperitoneal injection in a volume of 1ml per 100 g of hamster body mass 4 hours after the challenge. One group of hamsters received 20 mg/kg of humVHH_S56A/LALAPG-Fc/Gen2 by intraperitoneal injection 24 hours before challenge infection. All animals were infected intranasally with 104 TCID50 of SARS-CoV-2 BetaCoV/Munich/BavPat1/2020 (Passage 3, grown on Vero E6 cells) in a total volume of 0.1 ml. On day 4 after infection all animals were euthanized by exsanguination under anesthesia.
Statistical analysis. Normal distribution of the data was tested by a Shapiro-Wilk test and differences in variance among groups with normally distributed data was tested by an F test. For statistical analyses in which two groups with normally distributed data were compared an unpaired student’s t test was used. Welsh’s correction was applied for groups with significant difference in variance. When multiple groups with normally distributed data and no significant differences in variance were compared to the control group a 1-way ANOVA test and a Dunnet’s multiple comparisons test was used for post hoc analysis to correct for multiple comparisons. When multiple groups with not normally distributed data were compared to the control group a Kruskal-Wallis test and a Dunn’s multiple comparisons test was used for post hoc analysis to correct for multiple comparisons. Differences in bodyweight between two groups was tested by 2-way ANOVA with Sidak’s multiple comparison. Differences in survival were tested with a log-rank (Mantel-Cox) test. All statistical analysis was done using GraphPad Prism 9.10.
We are grateful to Gert Zimmer, Stefan Pöhlmann and Markus Hoffmann for providing reagents to generate VSV pseudotype particles, Jason McLellan and Daniel Wrapp for providing SARS-CoV-2 spike and RBD proteins and expression plasmids for these proteins, Jesse Bloom for providing the S. cerevisiae RBD display library, and Vincent Munster and Michael Letko for providing chimeric SARS-CoV-1 spike expression constructs. We thank Nadja van Boxel and Joana Reis Pedro for excellent technical assistance in the generation of humanized VHHs, stability and functional characterization, and pharmacokinetic assessments. We thank Leon de Waal from Viroclinics for coordinating the hamster challenge study. We thank Carolien De Keyzer, Lindsey Bervoets, Elke Maas, Thibault Francken, Tina Van Buyten, Jasper Rymenants and Kathleen Van den Eynde for excellent technical assistance in the hamster challenge studies. We thank Valentijn Vergote and Elisabeth Heylen for facilitating the animal studies. We are grateful to Hanna Hailu and Sarfaraj Topia for material generation, purification and distribution. We thank Oliver Zaccheo and Bruce Carrington for helpful advice in the conducting of the FcγR binding studies. We thank the staff of the VIB Flow Core Ghent for providing access to flow cytometry equipment and their support with flow cytometry experiments. We thank the VIB Protein Service Facility and Savvas Savvides for making multi-angle light scattering (MALS) and biolayer interferometry (BLI) equipment available. We acknowledge all Global initiative on sharing all influenza data (GISAID) contributors for their sharing of sequencing data. We thank the entire teams at VIB-Center for Medical Biotechnology, VIB Innovation and Business, VIB Discovery Sciences and UCB for support.
Funding: This work was supported by a PhD fellowship of the Fund for Scientific Research Flanders (FWO) to SDC, CL, HE, DVH, SC, GDM, WW, and SJ; a junior postdoctoral fellowship of the FWO to IRo; a China Scholarship Council grant (grant No.201906170033) to XZ; by FWO projects VIREOS (EOS ID: 30981113) and G0B1917N to XS; FWO project G0G4920N to NC; UGent GOA and BOF projects to XS and NC; VUB grant OZR3571 to ND; and the Bundesministerium fuer Bildung und Forschung (BMBF) through the Deutsches Zentrum für Luft- und Raumfahrt (DLR, grant number 01KI2077) to MS and PS.
Author contributions: BS, LvS, WN, KR, FP, PS, MS, ND, DT, MG, JH, AH, AP, ME, KB, AT, BD, CS, JN, NC, and XS conceptualized the study. BS, LvS, WN, KR, DF, SD, DS, CSF, RA, SJFK, JR-P, DJ, LD, RB, HJT, KD, FP, PS, ND, DT, AH, AP, KB, ME, AT, BD, CS, JN, NC, and XS conceptualized the methodology. BS, LvS, WN, KR, DF, SD, WW, JBo, AFO, LS, GVV, BG, JBe, DS, IRe, CSF, RA, SJFK, JR-P, DJ, LD, PS, ND, DT, MG, AH, AP, AT, BD, CS, JN, NC, and XS performed the formal analysis. BS performed VSV pseudotype and XZ and RB SARS-CoV-2 virus neutralization assays. LsV and JBo performed affinity and protein stability assays. WN and FP performed molecular modeling. KR, SV, and KS produced and purified recombinant proteins from CHO cells. SD, WW, and AVH performed experiments using mass spectrometry and N-glycan analysis. WVB, SDC, DVH, JPC, SC, DDV, GM, JCZM, and GDM performed molecular cloning experiments. ME designed and generated final expression constructs (molecular cloning). DF analyzed SARS-CoV-2 RBD sequence diversity. JH performed polyethylene glycol solubility and hydrophobic interaction chromatography assays. B.S., LvS, CL, HE, and IRo performed yeast expression and RBD-display experiments. JBe, DS, PS, and AO performed experiments with K18-hACE2 transgenic mice. LLa, SJ, StH, LLi, RB, HJT, CSF, RA, and SJFK performed hamster challenge experiments. PM isolated, cultured, characterized and provided SARS-CoV-2 viruses. BS, LvS, WN, KR, SD, WW, and DS performed data analysis and presentation. DS, NC, and XS. BS, LvS, WN, DF, NC, and XS wrote the original draft. KR, SD, JBo, KD, SJFK, LD, PS, AH, AP, BD, and CS participated in manuscript writing and editing. BS, LvS, WN, DF, SD, DS, JG, AT, and XS made graphs to visualize the data.
Competing interests: BS, LvS, WN, KR, WVB, DF, HE, DDV, CL, SDC, SJFK, JR-P, LD, BD, CS, JN, NC, and XS are named as inventors on patent application Coronavirus Binders (WO 2021/156490 A2), published on August 12 2021. NC and XS are scientific founders of and consultants for ExeVir Bio, and are in receipt of ExeVir Bio share options. DT is employed by ExeVir Bio and is in receipt of ExeVir Bio share options. MG, AH, AP, KB, and AT are in receipt of UCB shares and share options.
Data and materials availability: All data associated with this study are in the paper or supplementary materials. VHH-Fc fusion proteins described in this work will be made available on request to the corresponding authors under a Material Transfer Agreement with VIB.