Molecular clamp stabilised Spike protein for protection against SARS-CoV-2

Efforts to develop and deploy effective vaccines against Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) continue at pace with more than 30 candidate vaccines now in clinical evaluation. Here we describe the preclinical development of an adjuvanted, prefusion-stabilised Spike (S) protein “Sclamp” subunit vaccine, from rational antigen design through to assessing manufacturability and vaccine ecacy. In mice, the vaccine candidate elicits high levels of neutralising antibodies to epitopes both within and outside the receptor binding domain (RBD) of S, as well as broadly reactive and polyfunctional S-specic CD4+ and cytotoxic CD8+ T cells. We also show protection in Syrian hamsters, which has emerged as a robust animal model for pulmonary SARS-CoV-2 infection. No evidence of vaccine enhanced disease was observed in animal challenge studies and pre-clinical safety was further demonstrated in a GLP toxicology study in rats. The Sclamp vaccine candidate is currently progressing rapidly through clinical evaluation in parallel with large-scale manufacture for pivotal ecacy trials and potential widespread distribution. industry standard mammalian cell bioprocessing facilities. We utilised several, patient derived, SARS-CoV-2 cross-reactive mAbs (CR3022 10 , S309 13 and CB6 14 to demonstrate high anity binding to the Sclamp protein hence conformational integrity (S309 kD = 0.08 nM; CB6 kD = 0.15 nM). We also conrmed that the trimeric conformation and antibody recognition of Sclamp are retained, even after prolonged exposure to heat stress (Fig. 1H and Extended Data Fig. indicating that a long shelf-life may be achievable without the need for low temperature, frozen storage in the optimised vaccine formulation. (Gibco™) and transfection was manufacturer’s protocols or 7 days to harvest of the supernatant and protein purication. Lonza GS Xceed® CHOK1SV GS-KO® cells were transfected via electroporation with linearised GS expression vector encoding SARS-CoV-2 Sclamp as per manufacturer’s instructions (GS Xceed® manual, Version 06 2019). Approximately 24 h later, enriched pools were selected using 50 µM L-Methionine sulfoximine (MSX) over a period of 3–4 weeks. Stable pool shaker ask expression was assessed over 12 days Lonza’s abridged fed-batch shake ask screen (v8.10) and clone selection was performed using the Beacon Optouidic platform (Berkeley Stable pools were loaded onto the OptoSelect™ 1750b Chip as single cells. Cells chip days before pens analysed for secretion of SARS-CoV-2 Sclamp using uorescently tagged anti-Clamp IgG. pens plate and scaled-up asks. Clones µm A Fc-tagged and Eluates following from supernatant respectively. Both isolates were passaged three times in Vero E6 cells and titrated by focus-forming assay on Vero E6 cells. Serial dilutions of mouse sera and BAL were incubated with ~ 6000 focus-forming units per ml of SARS-CoV-2 for 1 h at 37 °C. Serum-virus mixtures were added to Vero E6 cell monolayers (pre-seeded at 40,000 cells/well in 96-well plates and incubated overnight) and incubated at 37 °C for 30 min. Subsequently, cells were overlaid with 1% (w/v) medium viscosity carboxymethyl cellulose in M199 (Gibco) supplemented with 2% heat-inactivated fetal bovine serum supplemented with 1% Penicillin-Streptomycin (Sigma-Aldrich) and incubated at 37 o C in 5% CO 2 for 14 h. After incubation, overlay was removed and cells xed with 80% cold-acetone in PBS for 30 min at -20 °C. Plates were then dried, blocked with blocking buffer (1xKPL in 0.1% PBS-Tween 20) for 1 h and then incubated with 1 µg/ml of CR3022 anti-S mAb and 0.2 µg/ml IR-Dye800-conjugated goat anti-human IgG (LI-COR) in blocking buffer. Plates were washed 3 times after antibody incubations by submerging in PBS with 0.1% Tween-20. Plates were then dried prior to visualising using Odyssey (LI-COR). Immunoplaques were manually counted in blinded fashion.


Introduction
The scienti c community, including critical industry and academic partnerships, have embarked on an unprecedented race to develop and produce effective COVID-19 vaccine(s) for global use 1 . Several recent reports validate the SARS-CoV-2 Spike (S) protein as a promising target for COVID-19 vaccine development 2-5 . However, these approaches utilise platforms based on nucleic acids, viral vectors or insect cell production systems which could face challenges during large-scale manufacture and distribution.
Our candidate vaccine, SARS-CoV-2 Sclamp, consists of the recombinant viral S glycoprotein stabilised in its prefusion trimeric form, through the incorporation of a "molecular clamp" stabilisation domain, and produced in mammalian cells (Chinese Hamster Ovary) 6 . Stabilisation of the prefusion antigen is a major goal of next generation vaccines that target enveloped viruses. This approach ensures that vaccines elicit the production of speci c antibodies that recognise a wider array of conserved epitopes, including conformational epitopes, that are displayed on the virion surface 7,8 .
Herein we describe the rapid development of the Sclamp vaccine, including antigen design and characterisation, process optimisation, and animal immunogenicity, safety and e cacy studies.

Accelerated development of a highly expressed prefusion conformed Sclamp
The virus genomic sequence rst became available on the 12th of January, 2020 9 and we immediately commenced development of a candidate subunit vaccine utilising the molecular clamp platform (a timeline of research and development and an overview of the molecular clamp platform are included in Extended Data Figs. 1 and 2). Within 34 days, we had expressed and screened > 200 antigens incorporating truncations and modi cations at key sites within the coding sequence of the S glycoprotein to identify a lead candidate, with high expression and a nity to S-speci c monoclonal antibody (mAb), CR3022 10 Table 1). The optimal lead construct incorporated the S protein native signal peptide, replacement of aa680-690 with a Glycine linker (GSG), truncation at aa1204 and followed by the molecular clamp coding sequence. This construct was termed M1GSG (Supplementary Fig. 1 and Supplementary Table 1).
We solved the 5 Å resolution structure of this Sclamp construct in the closed trimeric prefusion conformation by Cryo-TEM (Fig. 1E). Using sizeexclusion high-performance liquid chromatography (SE-HPLC) and TEM, we demonstrated that Sclamp can also adopt the open prefusion conformation 11 , and that the antigen exists in an equilibrium between these two conformations, with exposure to varying pH or temperature driving transition between the open and closed forms (Extended Data Figs. 3 and 4).
Transient gene expression in CHO cells was initially utilised to allow rapid screening of antigen design panels with CHO stable cell lines being established following lead candidate selection. Transitioning to CHO stable pools and using a fed batch bioprocess resulted in increased expression yields in shaker asks of 100-150 mg/L and in bioreactors up to 500 mg/L (Fig. 1F). This level of mammalian cell-based protein expression is an order of magnitude higher than previously reported for an "optimised" S protein construct 12 and has the potential to generate many millions of doses per bioreactor run using industry standard mammalian cell bioprocessing facilities.
We utilised several, patient derived, SARS-CoV-2 cross-reactive mAbs (CR3022 10 , S309 13 and CB6 14 ), to demonstrate high a nity binding to the Sclamp protein and hence conformational integrity (S309 kD = 0.08 nM; CB6 kD = 0.15 nM). We also con rmed that the trimeric conformation and antibody recognition of Sclamp are retained, even after prolonged exposure to heat stress ( Fig. 1H and Extended Data Fig. 5), indicating that a long shelf-life may be achievable without the need for low temperature, frozen storage in the optimised vaccine formulation.
MF59C.1-adjuvanted Sclamp vaccination elicited effective neutralising antibodies against SARS-CoV-2 Puri ed protein-based vaccines require adjuvant co-administration to enhance immunogenicity and to reduce the total amount of antigen required to provide protective immunity 15 . The squalene-oil-in-water adjuvant MF59C.1® (Seqirus) was selected for use with our vaccine due to its well-established safety record and its ability to stimulate a balanced T helper (Th1/Th2) cell response 16 . Alhydrogel was included as a comparator adjuvant 17,18 . BALB/c mice received intramuscular (IM) injections of PBS (placebo) or two doses of Sclamp with or without Alhydrogel or MF59C.1 ( Fig. 2A). Vaccinated mice developed a robust antigen-speci c IgG response after a single dose with either adjuvant and this was boosted following the second dose (Fig. 2B).
All mice receiving two doses of Sclamp with either Alhydrogel or MF59C.1 produced a strong neutralising antibody response as assessed in a microneutralisation (MN) assay (Fig. 2C). Virus neutralisation, as assessed by a plaque reduction neutralisation test (PRNT 50 ) assay, was observed equally for both 614D and 614G SARS-CoV-2 variants in serum, and also in bronchoalveolar lavage (BAL) samples ( Fig. 2F and Supplementary Fig. 2). A subsequent assay was conducted that included a WHO recommended reference serum (NIBSC code 20:130) 19 to allow benchmarking of the neutralising titre to other platforms. In this assay, sera from mice vaccinated with MF59C.1-adjuvanted Sclamp neutralised SARS-CoV-2 with an ~ 2-fold higher geometric mean titre (GMT) compared to the reference serum (Extended Data Fig. 6A).
Using a depletion ELISA to block the clamp tag-, RBD-and Sclamp-speci c antibodies, the relative speci city of the antibody response to the clamp trimerisation tag, RBD subdomain and SARS-CoV-2 S ectodomain, was assessed in the initial immunogenicity study (Fig. 2D/G), and in a second immunogenicity study designed to assess post-formulation stability (Extended Data Fig. 6). In both studies, we found that on average 70-76% of antibodies were speci c to the ectodomain, of which roughly half were directed to the RBD and half to other epitopes within the ectodomain. Interestingly, while the relative percentage of the immune response to the clamp itself varied between individual animals, this 'off target' response was not associated with a decrease in either the overall response to the ectodomain or the MN 50 titres (Extended Data Figs. 6C, 7A and 7B). RBD-speci c antibodies contributed, on average, ~ 50% of MN activity, with a high degree of inter-animal variation (SD = 27.9%). In comparison, MN activity was reduced by nearly 100% by the Sclamp depletion control (Fig. 2E/H). This nding indicates that antibodies to the RBD contribute substantially to MN activity, but that antibodies speci c to S domains outside the RBD also make a signi cant contribution to overall neutralising activity. Notably, the S-speci c IgG titre was a better correlate of MN activity compared to the RBD-speci c IgG titre (R 2 t = 0.77 and 0.22, respectively; Extended Data Fig. 7C and 7D).
In this assay, up-regulation of CD69 on S peptide-pulsed B220 + cells recovered from challenged mice indicates the presence of antigen-speci c Th cell responses and the killing of cognate peptide-pulsed targets is a result of cytotoxic T lymphocyte (CTL) responses in vivo. Speci cally, our ndings revealed that Alhydrogel-adjuvanted vaccination was more effective in eliciting S-speci c Th cell and CTL responses compared to Sclamp only and placebo control groups (  Tables 4-5). Furthermore, these mice developed higher numbers of IFN-γ + , TNF-α + and/or IL-2 + T cells compared to IL-4 + and/or IL-13 + counterparts (Fig. 3E).

Sclamp vaccination was safe and induced protection against pathogenic SARS-CoV-2 challenge
Given that the molecular clamp platform has not previously been used in humans, we conducted a repeat dose toxicity study in Sprague Dawley rats (data not shown). Three doses of MF59C.1-adjuvanted Sclamp at the maximum feasible dose administered by IM injection at twoweek intervals, were well tolerated. Treatment-related observations were limited to the injection site and included mild in ammatory cell in ltration and minimal myodegeneration. No other adverse ndings were observed.
To evaluate Sclamp-mediated protection and to investigate the possibility of vaccine-enhanced respiratory disease, we completed two studies in which hamsters were challenged with intranasal inoculation following either a single dose or a prime/boost vaccination regimen ( Fig. 4 and Extended Data Fig. 8), and one in ferrets following a similar inoculation and challenge regimen (Extended Data Fig. 9). In the hamster model, a single dose of MF59C.1-adjuvanted Sclamp resulted in a substantially reduced virus load in the lungs at day 4 post infection (p < 0.001), with 5/10 animals having undetectable levels of virus in the lungs (Extended Data Figure Fig. 8D). In the prime/boost study, administration of two doses of MF59C.1-adjuvanted Sclamp resulted in complete protection in the lungs for all but one animal at 4 days post infection. The two-dose regimen reduced viral TCID 50 titres in the nasal turbinates from 2.9 × 10 7 to 3.8 × 10 6 , although this was not statistically signi cant (Fig. 4D).
However, peak viral loads in daily throat swabs collected in both the single dose and prime/boost studies were signi cantly reduced compared to placebo controls ( Fig. 4C and Extended Data Fig. 8C).
Histopathological assessment of placebo vaccinated hamsters at day 4 and at day 8 provides an interesting snapshot of disease, with in ammation primary con ned to the upper respiratory tract and larger airways (trachea and bronchus) at day 4 but then proceeding to the smaller airways (bronchioles and alveoli) at day 8 and impacting a greater percentage of lung tissue, despite the apparent clearance of virus from the lung by this timepoint (Fig. 4 and Extended Data Fig. 8). Importantly, two doses of MF59C.1-adjuvanted Sclamp was found to be signi cantly protective against rhinitis, tracheitis and bronchitis at day 4, as well as bronchiolitis and alveolitis at day 8 ( Fig. 4F and I and Extended Data Fig. 10). In addition two doses of MF59C.1-adjuvanted Sclamp signi cantly reduced the overall percentage of the lung affected at day 8 ( Fig. 4H) and provided protection against peribrochial/pervascular cu ng, alveolar oedema and alveolar haemorrhage (Supplementary table 6). A single dose of MF59 adjuvanted Sclamp was also shown to provide a signi cant level of protection against bronchitis, bronchiolitis and alveolitis (Extended Data Fig. 9F).
In the ferret model of SARS-CoV-2 infection, viral replication was primarily restricted to the upper respiratory tract, with no virus isolated from lung tissue at day 4 post infection (Extended Data Fig. 10). While there was a tendency for reduced viral loads in daily nasal and throat swabs from animals receiving a single dose of MF59C.1-adjuvanted Sclamp, there was substantial inter-animal variation and consequently the differences were not statistically signi cant. Nonetheless, histopathological assessment of rhinitis demonstrated a signi cant reduction of in ammatory cells following MF59C.1-adjuvanted Sclamp vaccination (p = 0.01).

Discussion
Vaccine development in response to the COVID-19 pandemic has been unprecedented both in terms of speed and the diversity of the underlying technology platforms. These approaches have included inactivated viruses, nucleic acid-based vaccines, viral vectors and protein subunits 22 .
While each of these platforms comes with its own strengths and limitations, this diversity should also increase the likelihood of success and allow for the selection of the optimal vaccine(s) to progress into wide-spread global use based on key parameters, including safety and e cacy 23 .
Driven by the rationale that any improvement in protein yield directly translates into a future increase in dose availability, we dedicated considerable effort to the selection of an optimal candidate and development of a high-yielding and scalable bioprocess. The ability to manufacture at large-scale, ensuring consistent product quality, together with supply chain stability, will be a critical differentiator in achieving global vaccine supply 23 .
In mice, vaccination with MF59C.1-adjuvanted Sclamp was found to elicit a robust humoral immune response that e ciently neutralised both the original SARS-CoV-2 strain and the D614G variant which has since become the dominant circulating strain. We con rmed that the vaccine elicited neutralising antibodies against a broad array of epitopes, including those both inside and outside the RBD. These ndings provide con dence that the breadth of vaccine e cacy should extend to 'drifted' strains, and minimise the potential impact of the emergence of escape mutants 24 .
A major issue in comparing results between vaccine platforms is that there is no standardised method for assessing virus neutralisation. To address this issue, WHO has recommended the use of a pooled convalescent patient reference serum (NIBSC code 20:130) until an international standard can be made available at the end of 2020 19 . Serum from mice that received two doses of MF59C.1-adjuvanted Sclamp was found to neutralise SARS-CoV-2 at twice the level of the reference serum. To our knowledge no other programs have reported neutralising titres relative to this reference serum. Therefore, this data provides a useful cross-platform comparator.
In addition to the humoral immune response, type 1 biased S-speci c T cell immunity likely contributes to protection in convalescent COVID-19 patients 25 . Broadly-reactive and/or type 1 cytokine producing T cell responses are also characteristic of DNA, Adenovirus serotype 26 and chimpanzee Adenovirus vaccines that protected non-human primates against SARS-CoV-2 challenge and are in clinical development 2,4,5 . In vivo and in vitro T cell analyses in our study demonstrated that MF59C.1-adjuvanted Sclamp vaccination elicited a robust, broad and type 1 biased polyfunctional T cell response in mice, consistent with features of a T cell response that could contribute to protection against SARS-CoV-2.
Although non-human primates represent an attractive model for SARS-CoV-2 vaccine development 2-5 , multiple studies 26-28 including our own suggest that the Syrian hamster provides an effective animal model to evaluate SARS-CoV-2 vaccine e cacy for the following reasons: 1) low dose (i.e. 10 2 -10 3 TCID 50 ) i.n. administration of SARS-CoV-2 results in replication of infectious virus to high titres in the respiratory tract (lungs and nasal turbinates), 2) hamsters are permissive to communicable transmission 26 , 3) hamsters can develop severe pneumonia and lung injury similar to COVID-19 patients 28 , and 4) protection can be reproducibly evaluated based on infectious viral load rather than viral genomic copy levels [26][27][28] .
A single dose of MF59C.1 adjuvanted Sclamp afforded a level of protection to hamsters from SARS-CoV-2 infection, with a signi cant decrease in viral replication, reduced lung in ammation and disease severity. Administration of two doses further improved lung protection in hamsters and suggested this protection may extend to the upper respiratory tract. These ndings are in line with our mouse immunogenicity studies which demonstrated that a prime/boost strategy maximised the neutralising antibody response to SARS-CoV-2. Moreover, we demonstrated some evidence of protection in the ferret model, although ferrets were markedly less permissive to SARS-CoV-2 infection. Importantly, no evidence of vaccine-enhanced respiratory disease was observed in either hamsters or ferrets.
Overall, this work demonstrates the ability of MF59C.1 adjuvanted Sclamp to provide protection against SARS-CoV-2 infection and supports continued development and progression through human clinical trials. This candidate vaccine has signi cant advantages that make it wellsuited to meet the global response to the COVID-19 pandemic, notably the well-established safety record for the use of subunit vaccines and the MF59C.1 adjuvant, and the compatibility of the manufacturing process with industry standard bioprocess facilities.

Declarations
All the authors contributed to the design and/or analysis of experimental data and edited the manuscript. K.J.C., D.W., D.K.W. and N.M. wrote the initial drafts of the manuscript, and constructed the gures and tables. K.J.C., D.W., and P.R.Y. conceived the project as a component of the

Constructs and plasmids
To express the prefusion S ectodomain, codon-optimised SARS-CoV-2 S (Genbank accession number: MN908947) gene with variations including (i) substitution at the furin cleavage site, (ii) substitution at signal peptide, (iii) truncation at C-terminal domain was generated with primers containing overlapping sequence by PCR mutagenesis using Phusion polymerase (New England Biolabs). These amplicons were introduced upstream of the clamp trimerisation motif. SARS-CoV-2 S RBD (aa residues 319-526) was expressed as a Fc fusion protein, which was cleaved and polished by size exclusion chromotography prior to use in downstream assays. Plasmid encoding variable domains of heavy and light chain of CR3022 10 , S309 13 , B38 13 , H4 29 , CB6 13 , G4 (anti-MERS S) 30 or anti-clamp HIV1281 31 was cloned into the mammalian expression vector, pNBF-Hv or pNBF-Lv in-frame with IgK signal peptide.

Recombinant protein expression
The ExpiCHO-S expression system (ThermoFisher Scienti c) was used for transient Spike protein and antibody expression. CHO cells were cultured in ExpiCHO-S Expression Medium (Gibco™) and transfection was conducted following the manufacturer's protocols for 5 or 7 days prior to harvest of the culture supernatant and protein puri cation. Stable cell lines were generated using the Lonza GS Xceed® System. CHOK1SV GS-KO® cells were transfected via electroporation with linearised GS expression vector encoding SARS-CoV-2 Sclamp as per manufacturer's instructions (GS Xceed® manual, Version 06 2019). Approximately 24 h later, enriched pools were selected using 50 µM L-Methionine sulfoximine (MSX) over a period of 3-4 weeks. Stable pool shaker ask expression was assessed over 12 days via Lonza's abridged fed-batch shake ask screen (v8.10) and clone selection was performed using the Beacon Opto uidic platform (Berkeley Lights). Stable pools were loaded onto the OptoSelect™ 1750b Chip as single cells. Cells were cultured on chip for 3-5 days before pens were analysed for secretion of SARS-CoV-2 Sclamp using uorescently tagged anti-Clamp IgG. Selected pens were then exported into a 96-well plate and scaled-up into shaker asks. Clones were further assessed via Lonza's abridged fed-batch shake ask screen (v8.10).
Recombinant protein puri cation SARS-CoV-2-stabilised spike protein was puri ed using immunoa nity chromatography on an ÄKTA pure protein puri cation system (Cytiva). This was achieved using an in-house made immunoa nity chromatography column -the anti-clamp mAb HIV1281 coupled to 1 or 5 mL HiTrap-NHS activated HP Columns (Cytiva). CHO expression cultures were centrifuged at 4000 × g for 10 min at 4 o C and resultant supernatant ltered through a lter unit (0.22 µm pore size). Supernatant was added to an anti-clamp protein a nity column to purify clamp-tagged proteins or Protein A HP column (Cytvia) to purify Fc-tagged proteins and antibodies. Eluates following puri cation from culture supernatant were neutralised immediately and buffer-exchanged into PBS using Merck Amicon Ultra-4 or Ultra-15 centrifugal lter units. Protein concentration was determined using the NanoDrop One (ThermoFisher Scienti c) or via the Pierce BCA protein assay kit (ThermoFisher Scienti c). Homogeneity of puri ed proteins was analysed using SDS-PAGE and visualised following staining with Coomassie blue.

SE-HPLC analysis of Sclamp
High resolution assessment of the oligomeric state of Sclamp was conducted using an Agilent 1200 HPLC. Duplicate 95 µl samples were injected onto a Waters X-Bridge 300 mm column pre-calibrated with PBS mobile phase and using a ow rate of 0.8 ml/min.
Negative staining electron microscopy of pre-fusion Sclamp protein SARS-CoV-2 proteins were diluted at ~ 10 µg/ml in PBS. Diluted proteins (4 µl) were adsorbed onto carbon-coated grids (ProSciTech) for 2 min and glow discharged for 5 sec in 25 mA. The grids were blotted and washed three times in water and stained twice with 1% Uranyl acetate with blotting in between. The grids were air dried and imaged using a Hitachi HT7700 microscope operated at 120 Kv.
Cryo-EM analysis of Sclamp SARS-CoV-2 proteins were diluted to 0.5 mg/ml in PBS. Diluted proteins (4 µl) were adsorbed onto glow discharged quantifoil grids (Q2/1) and plunge frozen using a EMGP2 system (Leica). Grids were imaged on a CRYO ARM 300 (JEOL) equipped with a K3 detector (Gatan). 50 frame, 5 second movies were collected using JADAS software at a magni cation of 50,000x, corresponding to a pixel size of 0.48 Å/pixel and a dose rate of 9e/pix/s. Movies were binned 2x during motion correction using MotionCor2 (v1.1.0) 32 , giving a nal pixel size of 0.96 Å. The contrast transfer function (CTF) parameters of each image were determined using CTFFIND (v. 4.1) 33 . Initial 2D references were generated using particles manually picked in RELION 3.1 34 . A 60 Å ltered map from EMD-21452 was used as an initial model for 3D classi cation using C1 symmetry and the most ordered class was further re ned using the RELION 3D auto-re ne procedure with C3 symmetry. CTF re nement and particle polishing were performed using RELION. The nal half-maps were masked with a soft, extended mask and FSC calculated using the gold standard FSC cutoff of 0.143.

Thermal stability testing
Puri ed antigen was sterile-ltered and diluted in PBS to a nal concentration of 0.18 mg/ml. 250 µl aliquots were added to 1.5 ml sterile microcentrifuge tubes which were then stored at either 4 °C, 25 °C  Blood was collected via the tail vein prior to each vaccination and by cardiac puncture at the study end point. After overnight incubation of blood samples at 4 °C, serum was collected via centrifugation for 10 min at 10000xg, heat inactivated at 56 o C for 30 min and stored at -80 °C prior to analysis. BAL collection was performed by inserting a catheter into the trachea and ushing the lungs with 400 µl of PBS. After centrifugation at 300xg for 7 min at 4 °C, clari ed BAL supernatants were frozen at -80 °C prior to analysis.
To assess T cell responses in mice at the study end point, splenocyte suspensions depleted of red blood cells (RBC) were analysed using ICS or the FTA as described below. ELISA A capture ELISA was used to screen ExpiCHO-S supernatants and puri ed proteins for expression of Sclamp vaccine candidates. Nunc MaxiSorp ELISA plates were coated with 2 µg/mL of the anti-Clamp mAb HIV1281 in PBS overnight at 4˚C. Plates were then blocked with 150 µl/well of 5% KPL Milk Diluent/Blocking Solution Concentrate (SeraCare) in PBS with 0.05% Tween 20 for 1 h at room temperature (RT). Blocking buffer was removed and plates were incubated with serial dilutions of harvested ExpiCHO-S supernatant for 1 h at 37˚C. Plates were washed three times with water before incubation with 5 µg/mL of an in-house produced recombinant CR3022 mAb 10 with a mouse IgG1 backbone for 1 h at 37˚C. Following another wash step, plates were incubated with a 1:2000 dilution of a horse radish peroxidase (HRP)conjugated goat anti-mouse secondary antibody (#A16072) for 1 h at 37˚C. After a nal wash, plates were developed for 5 min using TMB Single Solution chromogen/substrate (#002023) before the reaction was stopped by addition of 2N H 2 SO 4 . Absorbance at 450 nm was then read on a Spectramax 190 Microplate reader (Molecular Devices).
For ELISA analysis of mouse serum from placebo-or Sclamp-immunised mice, Nunc Maxisorp ELISA plates were coated with 2 µg/ml of antigen and blocked as above. The blocked plates were incubated with serial dilutions of each mouse serum at 37 °C for 1 h. The plates were then washed, developed and read as described above. EC 50 values were calculated by three parameter curve tting using nonlinear regression in GraphPad Prism (version 8.3.1). The Limit of Detection (LoD) was de ned as the reciprocal of the highest concentration of sera tested and any values falling below the LoD were reported as ½ LoD.
For depletion ELISA analysis of mouse serum, 10 µg/ml of the depleting antigen was added to titrated mouse serum in a 96-well round bottom plate. Antigens used for depletion included Sclamp, an alternate clamp stabilised viral glycoprotein (i.e. in uenza virus haemagglutinin (HA)clamp) and SARS-CoV-2 RBD. Sera and depletion antigen were incubated at 37 °C for 1 h prior to addition to the ELISA plate and analysis as described above. The ELISA plate setup included separate rows coated with the depletion antigen as a control for incomplete depletion of domain-speci c responses.

MN assay
Neutralising activity against infectious SARS-CoV-2 was assessed by a traditional MN assay as described for SARS-CoV 35 . Brie y, SARS-CoV-2 isolate CoV/Australia/VIC01/2020 36 was passaged in Vero cells and stored at -80°C. Serum samples were heat-inactivated at 56°C for 30 min and serially-diluted 1:20 to 1:10240 before addition of 100 TCID 50 of SARS-CoV-2 in MEM/0.5% BSA and incubation at RT for 1 h. Residual virus infectivity in the plasma/virus mixtures was assessed in quadruplicate wells of Vero cells incubated in serum-free media containing 1 µg/ml of TPCK trypsin at 37°C/5% CO 2 ; viral cytopathic effect (CPE) was read on day 5. The neutralising antibody titre was calculated using the Reed/Muench method as previously described 35 . LoD was de ned as the reciprocal of the highest concentration of sera tested and any values falling below the LoD were reported as ½ LoD.

Depletion MN assay
The depletion MN assay was conducted as above with the following modi cations; serum dilutions were prepared in 62.5 µl MEM in 96-well round bottom plates and mixed 1:1 with 62.5 µl of the depletion antigen ( nal concentration 10 µg/ml). Sera and depletion antigen were incubated at 37 °C for 1 h before addition of virus and further incubation for 1 h at RT. Each sample was then added to quadruplicate wells of Vero cells and CPE was assessed as for the standard MN assay described above.
PRNT 50 analysis Two SARS-CoV-2 isolates generated from infected patients in Queensland, Australia were used for the PRNT assay, QLD02/2020-30/1/2020 (GISAID accession EPI_ISL_407896) and QLDID935/2020-25/03/2020 (GISAID accession EPI_ISL_436097) and were referred here as 614D and 614G, respectively. Both isolates were passaged three times in Vero E6 cells and titrated by focus-forming assay on Vero E6 cells. Serial dilutions of mouse sera and BAL were incubated with ~ 6000 focus-forming units per ml of SARS-CoV-2 for 1 h at 37 °C. Serum-virus mixtures were added to Vero E6 cell monolayers (pre-seeded at 40,000 cells/well in 96-well plates and incubated overnight) and incubated at 37 °C for 30 min. Subsequently, cells were overlaid with 1% (w/v) medium viscosity carboxymethyl cellulose in M199 (Gibco) supplemented with 2% heat-inactivated fetal bovine serum supplemented with 1% Penicillin-Streptomycin (Sigma-Aldrich) and incubated at 37 o C in 5% CO 2 for 14 h.
After incubation, overlay was removed and cells xed with 80% cold-acetone in PBS for 30 min at -20 °C. Plates were then dried, blocked with blocking buffer (1xKPL in 0.1% PBS-Tween 20) for 1 h and then incubated with 1 µg/ml of CR3022 anti-S mAb and 0.2 µg/ml IR-Dye800conjugated goat anti-human IgG (LI-COR) in blocking buffer. Plates were washed 3 times after antibody incubations by submerging in PBS with 0.1% Tween-20. Plates were then dried prior to visualising using Odyssey (LI-COR). Immunoplaques were manually counted in blinded fashion.

Preparation of formalin-inactivated virus
To prepare formalin-inactivated virus, SARS-CoV-2 was cultured in Vero E6 cells for 3 days before harvesting supernatant and clari cation via centrifugation. Formalin was added to the virus stock at 1:4000 and incubated at 37 °C for 3 days. The formalin-treated virus stock was centrifuged using Millipore Amicon lters to concentrate the virus prior to removal of formalin and buffer-exchange using PBS. Virus was passaged on Vero E6 cells to con rm loss of infectivity. Virus protein concentration was quanti ed by BCA analysis.

Safety and toxicology
A repeat dose toxicity study was conducted in Sprague Dawley rats in accordance with current guidelines 39,40 . Three IM doses of either saline or 20 µg SARS-CoV-2 Sclamp in a 200 µl bolus with or without MF59C.1 adjuvant were administered to adult rats at fortnightly intervals over a four-week period. This dose level corresponds to 40% of the intended maximum clinical dose of 50 µg of SARS-CoV-2 Sclamp antigen. At 24-48 h following administration of the nal dose, terminal examinations were conducted on ten rats per sex from each treatment group. Five additional rats per sex, from each group were allowed a treatment-free recovery period of a further two weeks prior to necropsy. Animals were observed daily over the course of the study for changes in skin and fur, eyes and mucous membranes, respiratory and circulatory function, gait and posture, behaviour, tremors or convulsions and any other abnormal ndings. These observations also included daily examination of the injection site for reactions (oedema, erythema and eschar). Upon sacri ce, terminal blood samples were collected for haematology, coagulation, biochemistry and SARS-CoV-2 IgG antibody titre assessments. Necropsy procedures included a detailed examination of the external appearance of the animal and internal organs, weights of critical organs, and collection and preservation of the tissues and organs. All tissues and gross lesions were subjected to microscopic examination.
Hamster challenge study Syrian hamsters (n = 10 per group) were vaccinated with either a single dose or two doses as shown in Fig. 4A and 4B. Vaccination with PBS alone was used as a placebo and vaccination with 50 µg of formalin-inactivated SARS-CoV-2 virus formulated with Alhydrogel was included as a comparator vaccination. All vaccines were delivered by IM immunisation. For formalin-inactivated SARS-CoV-2 + Alhydrogel vaccination 100 µl total volume was delivered per dose, while all other vaccines comprised a total volume of 50 µl per dose. For an infection and recovery comparator group, hamsters were housed separately in a BSL-3 containment isolator and infected with 10 2 TCID 50 of SARS-CoV-2 (BetaCoV/Munich/BavPat1/2020) diluted in PBS to a dose volume of 100 µl delivered via IN administration.
At 4 weeks after the single dose or prime/boost vaccination, or 7 weeks after infection in the infection and recovery group, hamsters were challenged with 10 2 TCID 50 of SARS-CoV-2 (BetaCoV/Munich/BavPat1/2020) diluted in PBS to a dose volume of 100 µl delivered via IN administration. Following challenge, animal weight loss was recorded daily and throat swabs taken for virus quanti cation by TCID 50 and qPCR. At the time of peak (day 4) or resolution (day 8) of infection, animals were sacri ced and tissue samples collected from the lung lobes and nasal turbinate to measure virus titre by TCID 50 and qPCR. Samples were also xed and sectioned to enable the assessment of lesions by gross pathology as well as histopathological assessment for evidence of congestion, emphysema, presence of foreign body, alveolar hemorrhage, bronchioloalveolar hyperplasia and in ammation and edema. Alhydrogel was included as a comparator vaccination. At 4 weeks after dosing ferrets were infected IN with 10 6 TCID 50 of SARS-CoV-2 (BetaCoV/Munich/BavPat1/2020) diluted in 300 µl of PBS. Following challenge, animal weight loss was recorded daily and nose and throat swabs taken for virus quanti cation by TCID 50 and qPCR. At 4 days post-challenge animals were sacri ced and the lung lobes and nasal turbinate samples collected to measure virus titre by TCID 50 and qPCR. Samples were also xed and sectioned to enable the assessment of lesions by gross pathology as well as histopathological analysis as described above.

Statistical analysis
Statistical analysis of the data was performed using the IBM SPSS Statistics Software (version 25) or GraphPad Prism (version 8.3.1). P values are reported for statistically signi cant comparisons with p < 0.05. Non-signi cant data with p > 0.05 are denoted 'ns'. Figure 1 Antigen design and analysis. A, In vitro screening of S1/S2 linker modi cations for yield and CR3022 a nity (KD). B, Linear representation of recombinant spike antigen, C, In vitro screening of signal sequence changes yield and CR3022 a nity (KD) D, In vitro screening of C-terminal length for yield and CR3022 a nity. E, Cryo-TEM reconstruction of antigen structure, tted with prefusion Spike structure PDB:6VXX41 shown in rainbow cartoon for one S monomer. F, Production yield from CHO cell culture transient expression, stable pools and clones in asks or bioreactors. Red -antigen concentration in cell culture supernatant estimated by BiaCore. Black -Protein recovery following immunoa nity puri cation measured by absorbance at 280 nm. G, Stability of stable pool derived Sclamp as assessed by CR3022 a nity (KD) following incubation at 4°C, 25°C, or 40°C for up to 8 weeks. H, Stability of stable pool derived Sclamp as assessed by percentage trimer analysis using SE-HPLC following incubation of the protein at 4°C, 25°C, or 40°C for up to 8 weeks.

Figure 2
The antibody response following SARS-CoV-2 Sclamp vaccination in BALB/c mice. A, Prime/boost vaccination and bleed schedule for the study. B, SARS-CoV-2 Sclamp-speci c IgG EC50 titre (reciprocal EC50) in vaccinated mice 20-, 35-and 42-days following intramuscular injection of the rst dose. C, MN titre against infectious SARS-CoV-2 (614D). D, Analysis of clamp-speci c, RBD-speci c and non-RBD S-speci c IgG EC50 titre assessed by depletion ELISA. Each line represents the response from a single mouse. E, Analysis of RBD-speci c and non-RBD S-speci c MN titre against infectious SARS-CoV-2 (614D) evaluated using the depletion MN assay, as described in Methods. Each line represents the response from a single mouse. F, MN titre against SARS-CoV-2 614D and 614G isolates evaluated by PRNT50 assay. G, Relative percentage of total IgG response directed to the clamp domain, RBD and non-RBD spike epitopes, calculated based on EC50 titres in D. H, Relative percentage of MN titre directed to the RBD and non-RBD spike epitopes, calculated based on titres in E. Analysis for panels C-H was limited to day 42 sera from Ag + MF59C.1 vaccinated mice only. P values were calculated using: 1) one-way ANOVA with Tukey's multiple comparison post-hoc test for normally distributed and homoscedastic data, 2) Welch's ANOVA with Games-Howell post-hoc analysis for all heteroscedastic data and 3) Pairwise Kruskal-Wallis H test for non-normally distributed and homoscedastic data sets.