key: cord-0994849-w2fyxzbn
authors: Li, Lei; Honda-Okubo, Yoshikazu; Huang, Ying; Jang, Hyesun; Carlock, Michael A.; Baldwin, Jeremy; Piplani, Sakshi; Bebin-Blackwell, Anne G.; Forgacs, David; Sakamoto, Kaori; Stella, Alberto; Turville, Stuart; Chataway, Tim; Colella, Alex; Triccas, Jamie; Ross, Ted M; Petrovsky, Nikolai
title: Immunisation of ferrets and mice with recombinant SARS-CoV-2 spike protein formulated with Advax-SM adjuvant protects against COVID-19 infection
date: 2021-08-03
journal: Vaccine
DOI: 10.1016/j.vaccine.2021.07.087
sha: f918b8e1f9f00f303d14a194d9973e0db99a2a37
doc_id: 994849
cord_uid: w2fyxzbn

The development of a safe and effective vaccine is a key requirement to overcoming the COVID-19 pandemic. Recombinant proteins represent the most reliable and safe vaccine approach but generally require a suitable adjuvant for robust and durable immunity. We used the SARS-CoV-2 genomic sequence and in silico structural modelling to design a recombinant spike protein vaccine (Covax-19™). A synthetic gene encoding the spike extracellular domain (ECD) was inserted into a baculovirus backbone to express the protein in insect cell cultures. The spike ECD was formulated with Advax-SM adjuvant and first tested for immunogenicity in C57BL/6 and BALB/c mice. Covax-19 vaccine induced high spike protein binding antibody levels that neutralised the original lineage B.1.319 virus from which the vaccine spike protein was derived, as well as the variant B.1.1.7 lineage virus. Covax-19 vaccine also induced a high frequency of spike-specific CD4+ and CD8+ memory T-cells with a dominant Th1 phenotype associated with the ability to kill spike-labelled target cells in vivo. Ferrets immunised with Covax-19 vaccine intramuscularly twice 2 weeks apart made spike receptor binding domain (RBD) IgG and were protected against an intranasal challenge with SARS-CoV-2 virus given two weeks after the last immunisation. Notably, ferrets that received the two higher doses of Covax-19 vaccine had no detectable virus in their lungs or in nasal washes at day 3 post-challenge, suggesting that in addition to lung protection, Covax-19 vaccine may have the potential to reduce virus transmission. This data supports advancement of Covax-19 vaccine into human clinical trials.

recombinant proteins [7] . Although several vaccines have received emergency-use authorisation, ongoing questions include likely duration of vaccine protection, long-term safety, potential for antibody-enhanced disease, activity against variant strains and immune correlates of vaccine protection [8] [9] [10] .

Recombinant or inactivated protein vaccines are a safe and reliable approach but generally suffer from weak immunogenicity unless formulated with an appropriate adjuvant [11] . Adjuvants induce higher and more durable immune responses and can also be used to impart a relevant T helper bias to the immune effector response [12] and can overcome immune impairment seen with advancing age or chronic disease [13] . Advax-SM is a combination adjuvant developed by our team that consists of delta inulin polysaccharide particles (Advax™) formulated with a Toll-like receptor 9 (TLR9)-active oligonucleotide, CpG55.2. A similar adjuvant approach provided enhanced protection of recombinant spike protein vaccines against severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS), coronaviruses [14, 15] . The Th1-bias imparted by the adjuvant also prevented the eosinophilic lung immunopathology otherwise seen after immunisation with SARS spike protein alone or with alum adjuvant [14] . Advax adjuvant has been shown to be safe and effective in human vaccines against seasonal and pandemic influenza [16, 17] and hepatitis B [18] .

Advanced computer modelling techniques may be useful to accelerate pandemic vaccine design. This study describes how we used a range of approaches including computer modelling to characterise the SARS-CoV-2 spike protein from the genomic sequence, and then used a modelled 3-D structure to identify angiotensin converting enzyme 2 (ACE2) as the relevant human receptor. We then utilised our computer model to design a vaccine from the extracellular domain (ECD) of the SARS-CoV-2 spike protein, to test the hypothesis that this antigen when formulated with Advax-SM adjuvant would induce neutralising antibodies able to block the binding of the SARS-CoV-2 virus to ACE2 thereby preventing infection. While our study was in progress others confirmed that ACE2 was indeed the receptor for the spike protein and viral entry into host cells was further enhanced by priming of the spike protein by transmembrane protease serine 2 (TMPRSS2) [19] . Our subsequent results confirmed that our computationally-designed spike antigen when formulated with Advax-SM adjuvant induced antibodies against spike protein that were able to neutralise wildtype lineage B.1.319 SARS-CoV-2 virus as well as cross neutralising the variant B.1.1.7 lineage virus. Covax-19 vaccine induced memory CD4 and CD8 T cell responses with a Th1 phenotype and this translated into the ability to kill spikelabelled target cells in vivo. Notably, Covax-19 vaccine provided lung protection in immunised ferrets but also inhibited day 3 nasal virus shedding, suggesting an additional potential to impact on virus transmission.

Vaccine Design:

In mid-January 2020, we identified the putative spike protein from the SARS-CoV-2 genome sequence in NCBI (accession number: NC 045512) [20] . Given the homology of the spike proteins (76.4% sequence identity), SARS-CoV-1 was used as a template to model the SARS-CoV-2 spike protein. We performed a PSI-BLAST search against the Protein Data Bank (PDB) Database for 3D modelling template selection. Using the SARS-CoV-1 structure (PDB-ID 6ACC) [21] we performed structural homology modelling using Modeller9.23 (https://salilab.org/modeller/) to obtain a 3D structure of SARS-CoV-2 spike protein ( Figure 1A ). The quality of the spike protein model was evaluated using GA341 and DOPE score, and the model was assessed using the SWISS-MODEL structure assessment server (https://swissmodel.expasy.org/assess). To help identify the putative cellular receptor for SARS-CoV-2, the crystal structure of human ACE2 (PDB-ID 3SCI) [22] was retrieved, and using HDOCK server, the spike protein was then docked against human ACE2 protein (http://hdock.phys.hust.edu.cn/) [23] . The docking poses were ranked using an energy-based scoring function and the docked structure analysed using UCSF Chimera. The high binding score predicted human ACE2 as the entry receptor for SARS-CoV-2 spike protein, confirming spike protein suitability for vaccine design [24] . The docked model was optimised using AMBER99SB-ILDN force field in Gromacs2020 (https://www.gromacs.org/). Molecular dynamic simulation (MDS) was carried out for 100 ns using a GPU-accelerated version of the program (Supplementary video 1). The structural stability of the complex was monitored by the root-mean-square deviation (RMSD) value of the #CLS3985) at 5x10 3 cells per well in 40 μL. After 1 h of virus-serum coincubation, 40 μL were added to the cell-plate for a final well volume of 80 μL. Plates were incubated for 72 h until readout (37°C, 5% CO 2 , >90% relative humidity), which occurred by staining cellular nuclei with NucBlue dye (Invitrogen, #R37605) and imaging the entire well's area with a high-content fluorescence microscopy system (IN Cell Analyzer 2500HS, Cytiva Life Sciences). The number of cells per well was determined using InCarta image analysis software (Cytiva). The percentage of viral neutralisation for each well was calculated with the formula N = (D-(1-Q))x100/D, where "Q" is the well's nuclear count divided by the average nuclear count of the untreated control wells (i.e. without virus or serum), and "D" equals 1 minus the average Q-value for the positive infection control wells (i.e. cells + virus, without serum). Therefore, the average nuclear counts for the infected and uninfected cell controls are defined as 0% and 100% neutralisation levels, respectively. The threshold for determining the neutralization endpoint titre of diluted serum samples mixed with virus was set to N≥50%.

A non-replicative SARS-CoV-2 Spike pseudotyped lentivirus-based platform was developed to evaluate neutralisation activity in infected/convalescent sera in a Biosafety Level 2 (BSL2) facility. The human ACE2 open reading frame (Addgene# 1786) was cloned into a 3rd generation lentiviral expression vector pRRLSIN.cPPT.PGK-GFP.WPRE (Addgene# 122053), and clonal HEK 293T cells stably expressing ACE2 were generated by lentiviral transductions as described previously [25] , followed by single cell sorting into 50% HEK 293T conditioned media (media conditioned from 50% confluent HEK 293T cultures). Lentiviral particles pseudotyped with SARS-CoV2 Spike envelope were produced by cotransfecting HEK 293T cells with a GFP encoding 3rd generation lentiviral plasmid HRSIN-CSGW (a gift from Camille Frecha [26] ), psPAX2 and plasmid expressing codon optimised but C-terminal truncated SARS-CoV-2 S protein (pCG1-SARS-2-S Delta18 [27] , herein Spike Delta18) courtesy of Professor Stefan Polhman using polyethylenimine as described previously [25] . Neutralisation activity of donor sera was measured using a single round transduction of ACE2-HEK 293T with Spike pseudotyped lentiviral particles. Briefly, virus particles were pre-incubated with serially diluted donor sera for 1 h at 37°C.

Virus-serum was then added onto ACE2-HEK 293T cells seeded at 2,500 cells per well in a 384-well tissue culture plate a day before. Following spinoculation at 1200xg for 1 h at 18°C, the cells were moved to 37°C for a further 72 h. Entry of pseudotyped particles was assessed by imaging GFP-positive cells and total cell numbers imaged through live nuclei counter staining using NucBlue (Invitrogen).

Total cell counts and % GFP-positive cells were acquired using the InCell imaging platform followed by enumeration with InCarta high content image analysis software (Cytiva). Neutralisation was measured by reduction in % GFP expression relative to control group infected with the virus particles without any serum treatment.

BALB/c and BL6 mice were sacrificed, and individual spleens were collected one to two weeks after the last immunisation. Single-cell suspension in sterile PBS+3% FCS was prepared using a 70 µm easy strainer (Greiner Bio-One) with a 5 ml syringe plunger. Isolated spleen cells were pelleted and incubated in red blood cell (RBC) lysis buffer. For Cytometric Bead Array (CBA) assay, splenocytes were cultured at 5 x 10 5 cells/well in 96-well plates with 3 µg/ml of rSp antigen [corresponding to SARS-CoV-2 (Wuhan) reference sequence Q13 to P1209] at 37°C and 5% CO 2 . Two days later, the supernatants were harvested and cytokine concentrations determined by mouse Th1/Th2/Th17 CBA (BD) and analysed by FCAP array Software (BD). In addition to CBA assay, enzyme-linked immune absorbent spot (ELISPOT) assay was performed using mouse Interlukin-2 (IL-2), Interlukin-4 (IL-4) or and anti-mouse CD8-APC (both from BD) and analysed on a FACSCanto II (BD). T-cell proliferation was expressed as the ratio of divided daughter cells to total T-cells, expressed as a percentage, by analogy to calculation of a stimulation index in thymidine proliferation assays.

Functional CD8 + T cell response was determined by performing in vivo CTL assays, as described earlier [28] . Briefly, naïve syngeneic target spleen cells were left unpulsed or pulsed for 2 h in humidified CO 2 incubator at 37°C with 5µM H-2K b -restricted Sp 539-546 (VNFNFNGL) synthetic peptide [29] (DGpeptide, Hangzhou, China). Unpulsed (control) and peptide (antigen)-pulsed spleen cells were labelled with 0.5 µM CFSE (CFSE low ) and 5 µM CFSE (CFSE high ), respectively. Then, naïve syngeneic and immunised mice were adoptively transferred with 4×10 6 cells of a 1:1 mix of control-to-antigen-pulsed target spleen cells. Eighteen (18) hours later, adoptive transfer recipient mice were euthanised, their splenocytes isolated and resuspended in PBS for acquisition on a BD FACSCanto-II instrument. To evaluate the percentage of antigen-specific target cell killing, the ratio of CFSE high /CFSE low in survivors was compared to the ratio in transferred naive control mice.

Fitch ferrets (Mustela putorius furo, spayed female, 6 to 12 months of age), were purchased from 

At day 28, all ferrets were infected intranasally with SARS-CoV-2 virus (1 x 10 5 PFU) and were monitored daily during the infection for adverse events, including weight loss and elevated temperature for 10 days. At day 3 post infection, nasal swabs were collected from all animals, and three animals from each group, except for the Advax-SM only control group, were humanely euthanised, and lung tissue was collected. Three lobes from the right lung of each animal was formalin fixed for histopathology. Two lobes from the left side of lung from each animal were snap frozen and homogenised using 1 ml DMEM, and the supernatant was collected and kept frozen at -80˚ for viral titres. The dilution curve was plotted, and the area under the curve was calculated and multiplied by 1,000 to give standard units.

Nasal washes were titrated in quadruplicates in Vero E6 cells. Briefly, confluent VeroE6 cells were inoculated with 2-fold serial dilutions of sample in DMEM containing 2% FBS, supplemented with 1% penicillin-streptomycin (10,000 IU/ml). At 3 days post infection (dpi), virus positivity was assessed by reading out cytopathic effects. Infectious virus titres (TCID 50 /ml) were calculated from four replicates of each nasal wash using the Reed-Muench method.

To assess the viral replication and pathological effect of infection, ferrets (n=3) were euthanised 3 days post infection. The right lung lobes were taken for viral plaques, the incision was tied with surgical suture, and the lung was inflated with 10 ml formalin. Lungs were removed and placed into formalin for 1 week prior to paraffin embedding. Ferret lungs were embedded into paraffin and were cut using a Lecia microtome. Transverse 5 µm sections were placed onto Apex superior adhesive glass slides (Leica biosystem Inc, IL, USA), which were coated for a positive charge, and were processed for H&E staining. Briefly, sections were deparaffinised in xylene and hydrated using different concentrations of ethanol (100%, 95%, 80% and 75%) for 2 min each. Deparaffinised and hydrated lung sections were stained with hematoxylin (Millipore sigma, MA, USA) for 8 min at RT, differentiated in 1% acid alcohol for 10 sec, and then counterstained with eosin (Millipore sigma, MA, USA) for 30 sec. Slides were then dehydrated with 95% and 100% ethanol, cleared by xylene, and mounted using Permount® mounting media (Thermo Fisher scientific, MA, USA). Lung lesions were scored by a board-certified veterinary pathologist blinded to the study groups as follows: Alveolar (ALV) score: 1 = focal, 2 = multifocal, 3 = multifocal to coalescing, 4 = majority of section infiltrated by leukocytes; Perivascular cuffing (PVC) score: 1 = 1 layer of leukocytes surrounding blood vessel, 2 = 2-5 layers, 3 = 6 -10 layers, 4 = greater than 10 cells thick; Interstitial Pneumonia (IP) score: 1 = alveolar septa thickened by 1 leukocyte, 2 = 2 leukocytes thick, 3 = 3 leukocytes, 4 = 4 leukocytes.

For lung immunohistochemistry, the deparaffinised and hydrated lung tissue sections were subjected to antigen retrieval by sub-boiling in 10 nM sodium citrate buffer at pH6 for 10 min and then incubated in 3% fresh-made hydrogen peroxide for 10 min to inactivate endogenous peroxidase at RT. The lung sections were blocked with 5% horse serum in PBS for 1 h at RT, incubated with SARS-CoV-2 

GraphPad Prism 8.3.1 for Windows was used for drawing graphs and statistical analysis (GraphPad Software, San Diego, CA, USA). The differences of antibody levels were evaluated by the Mann-Whitney test, and other differences between groups were evaluated by two-tailed Student's t-test.

ANOVAs with Dunnett's test was used for weight loss with a statistical significance defined as a pvalue of less than 0.05. Limit of detection for viral plaque titres was 50 pfu/ml for statistical analysis.

Limit of detection for neutralisation is 1:10, but 1:5 was used for statistical analysis. Geometric mean titres were calculated for neutralisation assays. For all comparisons, p<0.05 was considered to represent a significant difference. In figures * = p < 0.05; ** = p < 0.01; and *** = p < 0.001. All error bars on the graphs represent standard mean error.

The mouse studies were performed at Flinders University, Australia. 

Based on the high binding score (-57.6 kcal/mol) seen from docking our 3D-model of spike protein to several putative receptors, we predicted ACE2 as the human entry receptor for SARS-CoV-2 ( Figure   1B ) [24] . This was soon confirmed by other groups using in vitro assays [19] . Based on this spike protein model, we sought to design a stable soluble secreted spike protein trimer for use as our vaccine immunogen. We designed a synthetic gene comprising the spike protein extracellular domain (ECD) together with N-terminal honeybee melittin signal sequence (HBMss) to ensure protein secretion and attached a hexa-histidine tag at the C-terminal end to assist with protein purification ( Figure 1A ).

Molecular dynamic simulation performed on the spike protein ECD vaccine construct confirmed its ability to form a stable trimer despite the lack of the transmembrane and cytoplasmic domains and the absence of any large trimerisation domain tag as used by others ( Figure 1C ). The spike ECD gene construct was constituted into a baculovirus backbone, and the subsequent virus then used to transfect two insect cell lines (SF9 and Tni). While both cell lines successfully secreted the protein construct, higher protein expression was obtained in the Tni cells and these were used for subsequent production of a recombinant spike protein which was purified using a nickel affinity column and sterile filtration. The final protein product had a purity of ~ 90% by SDS-PAGE (Figure 1D & E) and was sterile with a low endotoxin and residual DNA content (data not shown).

The serum anti-spike protein response of BL6 mice immunised with rSp alone was dominated by IgG1, a T helper 2 (Th2) isotype, whilst in Advax-SM adjuvanted groups the response was characterised by a switch to more IgG2b/c and IgG3 against spike ( Figure 2B) . Overall, Advax-SM adjuvant was associated with a much higher anti-spike IgG2/IgG1 ratio consistent with a Th1-biased response Figure 1) .

Spike RBD-binding antibodies have been reported to correlate with SARS-CoV-2 virus neutralisation [30] . There was a high correlation for each IgG subclass between the level of spike and RBD binding antibodies by ELISA, suggesting a significant proportion of spike antibodies induced by our rSp antigen were directed against the RBD region. RBD-binding IgG was almost undetectable in mice immunised with rSp alone, although these mice did exhibit some RBD-binding IgM. Notably, the Advax-SM adjuvant increased the spike IgG response but particularly favoured production of RBD-binding antibodies when expressed as a ratio of the total spike IgG response ( Figure 2C) . Both BL6 and BALB/c mice immunised with Advax-SM adjuvanted rSp produced antibodies able to neutralise live SARS-CoV-2 virus. In BL6 mice, the highest neutralising antibodies were seen after immunisation with rSp 5µg+Advax-SM (GMT 3,712), then rSp 1 µg with Advax-SM (GMT 1088) and then rSp alone (GMT 736) (Figure 3B) . The same trends were seen in BALB/c mice with highest response for rSp 5µg+Advax-SM (GMT 4,352), then rSp 1 µg with Advax-SM (GMT 960) and finally rSp alone (GMT 512). To evaluate potential cross-protection against a variant strain, sera were also tested for ability to neutralise live SARS-CoV-2 "Alpha" variant of concern (lineage B.1.1.7, or "UK-strain").

Only sera from BL6 or BALB/c mice immunised with Advax-SM adjuvanted rSp were able to neutralise the B.1.1.7 variant virus, with no neutralisation activity seen in mice immunised with rSp alone ( Figure   3C ).

We next asked whether there was any correlation between total spike or RBD antibody levels and pseudotype or live virus neutralisation titres. In BL6 mice, there was a positive correlation between spike and RBD binding IgG and pseudotype and live virus neutralisation titres, with the highest correlation between spike IgG and pseudotype neutralisation (r 2 = 0.49, p<0.0035), followed by RBD IgG and pseudotype neutralisation (r 2 = 0.38, p<0.015) (Supplementary Figure 3A) . Interestingly, there were only weak non-significant correlations between spike IgG and live virus neutralisation titres (r 2 = 0.20, p<0.09) or RBD IgG (r 2 = 0.22, p<0.07).

In BALB/c mice, there was a positive correlation for spike IgG with pseudotype neutralisation titres (r 2 = 0.46, p<0.005) (Supplementary Figure 3B) . There was also a positive correlation for spike IgM with both pseudotype (r 2 = 0.42, p<0.009) and live virus (r 2 = 0.4, p<0.01) neutralisation. However, there was no correlation between RBD IgM and either pseudotype or live virus neutralisation, suggesting that IgM in BALB/c might neutralise SARS-CoV-2 through an RBD-independent mechanism.

Interestingly, there was only a weak correlation between pseudotype and live virus neutralisation (r 2 = 0.17, p<0.02) (Supplementary Figure 4) . This could reflect that pseudotype neutralisation assays are unique at two levels, they utilise greater numbers of viral particles to enable cellular transduction and GFP expression, and only measure the consequence of a single round of spike-driven cellular fusion.

By contrast, the live virus neutralisation assay measures inhibition of viral entry and productive infection over a 3-day period with repeated rounds of viral replication. Hence, each assay measures different but important parameters of viral infection, providing clues as to the ability of immune sera to neutralise first-round viral entry vs. a replicative infection

Cytokine production was measured in culture supernatants of rSp-stimulated splenocytes obtained from immunised mice. In BL6 mice, rSp-stimulated IL-2, IFN-γ and TNF-α was significantly higher in the Advax-SM group, consistent with their Th1 bias ( Figure 4A-C) . Similarly, in BALB/c mice, there was higher rSp-stimulated IFN-γ and TNF-α in the Advax-SM group (Figure 4A-C) . In BALB/c mice, rSpstimulated IL-4, IL-6 and IL-10 production was highest in the rSp-alone immunised group, which also exhibited low IFN-γ and TNF production, consistent with a Th2 bias (Figure 4D-F) . IL-17 was modestly increased in Advax-SM adjuvanted rSp groups in both BL6 and BALB/c mice ( Figure 4G) . Overall, rSpalone groups exhibited a Th2 cytokine bias, while Advax-SM groups exhibited a Th1 bias with an increased IFN-γ/IL-4 ratio ( Figure 4H ). ELISPOT assays on splenocytes from immunised mice confirmed significantly higher frequencies of IL-2 and IFN-γ secreting T cells in response to rSp stimulation in the Advax-SM groups (Figure 5A-B) . Anti-spike IL-4-producing T cells were significantly higher in BALB/c mice, consistent with their Th2 bias ( Figure 5C ). Anti-spike IL-17-producing T-cells were also higher in the Advax-SM group in BL6 mice ( Figure 5D ).

Spike-specific CD4+ and CD8+ T cell memory cell population were further assessed using a CFSE-dye dilution proliferation in response to rSp stimulation. Notably, anti-spike CD8 T cell responses were markedly increased in both BALB/c and BL6 mice that had received Advax-SM adjuvanted rSp ( Figure   6 ), consistent with the high levels of anti-spike CTL activity also seen in mice receiving this formulation (Supplementary Figure 1) . There was also a clear trend to higher anti-spike CD4 T cell responses in mice that had received Advax-SM adjuvanted rSp, although this difference only reached significance in the BALB/c group (Figure 6 ).

Having confirmed that the formulation of rSp with Advax-SM adjuvant gave optimal immunogenicity whether measured by neutralising antibody, T cell cytokines or CTL responses in mice, we next moved to test the efficacy of this optimised formulation in a ferret infection challenge model. 

Vaccines normally take 10-15 years from discovery to final market approval [31] . To accelerate our COVID-19 vaccine development we made use of a well-validated protein manufacturing platform complemented by in silico modelling analyses. In this way, as soon as the SARS-CoV-2 genome sequence became available in Jan 2020 [20] , we were able to identify the putative spike protein, model its structure and use docking programs to predict human ACE2 as the main receptor for the virus, as then confirmed by others [24] . This facilitated our rapid design of a recombinant spike protein antigen able to be produced as a soluble secreted protein in insect cells, which we named Covax-19™ vaccine.

We first evaluated the immunogenicity of Covax-19 vaccine in BL6 and BALB/c mice. Spike protein alone was effective in inducing antigen-specific IgG and IgM antibodies and potent neutralisation activity whether measured by pseudotype or wildtype virus neutralisation assays indicating the recombinant vaccine antigen produced insect cell-based baculovirus expression vector system (BEVS) is a good immunogen for a COVID vaccine. The benefits of the BEVS is that it enables eukaryotic-like glycosylation whilst mitigating the risk of adventitious virus contamination associated with mammalian cell systems, provides high yields, and can be rapidly scaled [32] to respond to pandemic needs for rapid vaccine production. However, immunisation with spike protein alone induced a heavily Th2 biased antibody response and only a weak T-cell response, indicating the need to add an appropriate adjuvant.

Advax-SM adjuvant has been under development by our team for some time and has shown utility in the past when used in vaccines against other coronaviruses, namely SARS and MERS [14, 15] . In the current study, the addition of Advax-SM to recombinant spike protein increased antibody production, created a more balance Th1/Th2 response, induced cytotoxic T cells able to kill spike protein labelled targets and enhanced antibody neutralization activity. Notably sera from Covax-19 immunised mice were able to cross-neutralise the B.1.1.7 virus variant of concern. This confirmed that Advax-SM as a suitable adjuvant for our Covid-19 vaccine candidate. Given the need for speed in development of this Covid-19 vaccine we did not conduct an extensive comparison of other adjuvants with our spike protein, as most of these are either not freely available for commercial vaccine use, are not available as GMP products, or are experimental and have not yet been shown to be safe and effective in humans and hence did not meet key criteria for use in our intended human vaccine.

SARS-CoV-1 and SARS-CoV-2 viruses target interferon pathways [33, 34] . Hence, coronavirus vaccines should ideally prime a strong memory Th1 and interferon response with CD8+ T cells playing a critical role in detection and silencing of virus-infected cells [35] . In mice, the Advax-SM adjuvated rSp vaccine induced a strong Th1 response characterised by a switch from IgG1 to IgG2 and IgG3 IgG isotypes together with an increased frequency of IL-2, IFN-γ and TNF-α secreting anti-spike T cells, and a high level of CTL killing of spike-labelled target cells, in vivo. By contrast, immunisation with rSp alone (or formulated with alum adjuvant) induced a predominantly Th2 antibody and T cell response against spike protein, with lower levels of neutralising antibody against the wildtype virus and no neutralising activity against the B.1.1.7 virus variant and no in vivo CTL activity against spike-labelled targets.

Immunisation with Advax-SM adjuvanted rSp induced a high frequency of spike-specific memory CD4+ and CD8+ T cells, which were not seen in mice immunised with rSp alone. This suggests that the Advax-SM adjuvant was able to induce effective dendritic cell cross-presentation of spike protein to CD8 T cells, with CD8 T cell priming to exogenous antigens typically requiring activation of CD8+ dendritic cells [36] . Notably, this CD8 T cell cross-presentation was associated with significant in vivo CTL activity against spike-labelled targets suggesting that our vaccine should be able to robustly control infection, not just through induction of neutralising antibody but also through induction of CTLs able to efficiently identify and kill any residual virus-infected cells in the body. It has been difficult to identify non-reactogenic adjuvants that induce strong CD8+ T cell responses, making this a potential key advantage of Advax-SM when used in viral vaccines where strong CD8+ T cell responses are likely to be important to protection. Overall, this demonstrated that our insect cell expressed spike ECD construct when formulated with Advax-SM adjuvant is an effective immunogen against SARS-CoV-2.

Whereas other adjuvant platforms might provide some nonspecific antiviral protection via activation of the innate immune system, this has not been a feature of the Advax-SM adjuvant. Notably, there was no suggestion of disease reduction in the ferrets that were injected with Advax-SM adjuvant plus an irrelevant influenza vaccine. Similarly, in a past SARS CoV vaccine study we did not see any nonspecific protection in mice that received Advax-SM alone [14] , nor did we see nonspecific protection of Advax-CpG adjuvant in a ferret H5N1 influenza study [37] . Hence, this data all supports the enhanced protection of Advax-SM adjuvanted vaccine being mediated by its ability to enhance the adaptive immune response to the co-administered spike antigen.

The role of RBD-binding and neutralising antibodies in SARS-CoV-2 protection remains unclear.

Initially, there were concerns of the possibility of antibody-mediated disease enhancement (ADE), as seen in SARS, dengue, Respiratory Syncytial Virus and other viral diseases [38] . Reassuringly to date, there have been no reports of ADE in COVID-19 patients, although those with the most severe COVID-19 illness often have high RBD and neutralising antibodies [39] , suggesting neutralising antibody may not be enough, by itself, and other mediators like CTLs may also be required to fully control SARS-CoV-2 infection. Furthermore, a spike protein vaccine using a large Human Immunodeficiency Virusderived protein trimerization tag and formulated with MF59 squalene adjuvant was shown to induce serum neutralising antibody, but provided no protection against nasal virus replication in either the ferret or hamster challenge models [40] . This suggests, at a minimum, that serum neutralising antibody is not able to prevent nasal virus replication. Furthermore, convalescent plasma has not proved effective when administered to severely ill patients but instead can induce immune escape variants [41] . Hence antibodies by themselves may not be sufficient to prevent or reverse COVID-19 disease. In Phase 2 trials, LY-CoV555, a cocktail of two IgG1 antibodies appeared to accelerate the decline in viral load over time but ultimately did not demonstrate clinical benefit [42] with other experimental monoclonal treatments still undergoing human testing [43] .

The gold standard for antibody assessment remains live wildtype virus neutralisation assays, as these directly measure the ability of antibody to block cellular infection. However, different cell types may be infected via different mechanisms, so use of different cell lines in these assays could still give varying results. VeroE6 is frequently used in virus neutralisation assays with viral entry in these cells primarily endosomal and driven through cathepsin cleavage of the spike protein. In contrast, entry of SARS-CoV-2 into nasopharyngeal cells is driven through TMPRSS2-mediated cleavage of spike [44] .

Whilst primary ciliated or goblet cells from nasopharyngeal tissue might be the most physiologically relevant cell type to use in neutralisation assays, high-throughput serology screening using airinterface cultures is not feasible. Whilst pseudotyping assays can be performed outside of a BSL3 facility, they measure only a single round of spike-mediated cellular fusion and, hence, do not mimic a natural infection where there are multiple rounds of entry and replication. RBD-binding antibody assays work on the presumption that antibodies that block spike protein from binding to ACE2 should stop virus infectivity. However, in animal studies vaccines that have been shown to induce anti-RBD IgG titres have not prevented virus replication in the nasal mucosa, suggesting either that such antibodies fail to prevent virus binding or entry or that they fail to get access to the nasal epithelium.

How do results compare for these assays? Previous studies on convalescent patients have reported a positive correlation between spike-specific IgG and both pseudotype virus and live virus neutralisation. In our study, there was a poor correlation between the pseudovirus and live virus assays, suggesting they measure different determinants of neutralisation. The live virus assay measures the ability to block cell infection by a small pool of viral particles across 3 days of culture.

The pseudotype assay uses a large pool of virus particles as a surrogate for a single spike-driven fusion event. In our study, total spike antibody ELISA predicted pseudotype neutralisation better than the RBD-binding ELISA. Interestingly, in BALB/c but not BL6 mice, there was a positive correlation between total anti-rSp IgM (but not anti-RBD IgM) with neutralisation titres. Elite donors with high neutralisation titres in human convalescent cohorts surprisingly achieved this via anti-viral IgM (S.

Turville, personal communication). There is still no established correlate of COVID-19 protection that has been confirmed in either animal models or humans. The fact that different assays seem to yield different results suggests that the identification of a correlate based upon simple antibody protection may not be straightforward.

Unless a vaccine is able to induce potent sterilising immunity, some SARS-CoV-2 virus will inevitably enter cells in the nasal mucosa, where antibodies will not be able to reach it and begin to replicate. In the face of uncertainty over antibody protection and rapidly waning circulating SARS-CoV-2 antibody levels, a strong CD8 T cell response with interferon production and CTL activity is likely to be important for virus control. A large body of clinical data demonstrates that reduced T-cell responses and production of Th1 cytokines, such as interferon and IL-2, are seen in patients with severe COVID-19 disease [45] [46] [47] [48] . Moreover, the mode of action and protection of several SARS-CoV-2 vaccines has been linked to induction of type I interferon secretion by amplifying T cell memory formation and promoting B cell differentiation and survival [49] . Notably, our Covax-19 vaccine imparted a strong Th1 bias and robust T cell responses by virtue of the Advax-SM adjuvant. By contrast, alum and squalene emulsion adjuvants induce a strong Th-2 bias, which may not be as beneficial for COVID-19

virus control [50] . COVID-19 vaccine with alum adjuvant demonstrated a Th2-biased response with a low IFN-γ/IL-4 ratio [51] , and we similarly saw a strong Th2 bias of alum for spike protein in the current study. Notably, alum-and squalene-adjuvanted COVID-19 vaccines were both ineffective against nasal virus replication [40] . This contrasts strongly, with the ferret protection data shown here, where Advax-SM adjuvanted rSp, completely prevented both lung and nasal virus replication, an exciting finding as prevention of nasal virus replication could be the key to prevention of virus transmission.

We are currently do not know the mechanism for the prevention of nasal virus replication in the ferrets by Advax-SM adjuvanted rSp, with the possibilities that is it due to CTL induced by the vaccine migrating to the nasal mucosa where they might then rapidly eradicate virus infected cells, or an ability of the adjuvant to induce neutralising antibodies with different functional properties that are better able to access the nasal environment and prevent infectivity of the virus, or both. Future studies will attempt to explore these mechanisms further.

A limitation of the current study was that ferrets do not exhibit weight loss or other signs of SARS-CoV-2 clinical infection [52] , with no animal models fully reproducing the features of severe SARS-CoV-2 clinical infection in humans. Ongoing studies are testing our Covax-19 vaccine in other species including hamsters and non-human primates to see whether the effects of the vaccine on inhibition of nasal virus replication extends to other species. The current study also only assessed short term protection immediately after immunisation and it will also be important in the future to assess longterm protection.

The COVID-19 pandemic represents a significant evolving global health crisis. The key to overcoming the pandemic lies in the development of an effective vaccine against SARS-CoV-2 that ideally prevents transmission as well as serious disease. Recombinant protein-based approaches to COVID-19 offer benefits over other technologies including a 40-year record of safety and effectiveness including in very young infants, together with reliable large scale manufacture and high stability under typical refrigerated conditions [53] . By contrast, other available technologies, including nucleic acid and adenoviral vector platforms have a high level of reactogenicity and pose cold chain and other distribution challenges [54, 55] . This study showed that an Advax-SM adjuvanted rSp vaccine (Covax-19 vaccine) when administered as two sequential intramuscular doses several weeks apart induces strong anti-spike antibody and T cell responses in mice and was able to protect ferrets against SARS-CoV-2 infection. The lack of day 3 nasal virus shedding in the immunised ferrets suggests Covax-19 vaccine may additionally be able to reduce virus transmission. Planned clinical trials will assess how this promising animal data translates into human protection. : Antigen-specific cytokine-producing cells were evaluated following re-stimulation with rSp of splenocytes using anti-mouse cytokine antibody-coated plates by ELISPOT. Statistical analysis was done by Mann-Whitney test. *; p < 0.05, **; p < 0.01, ***; p < 0.001, ns; not significant. Figure 6 : Antigen-specific proliferation was measured by FACS using CFSE-labelled splenocytes after culture for 5 days with rSp antigen. Results are presented as mean + SD. Statistical analysis was done by Mann-Whitney test. *; p < 0.05, **; p < 0.01, ***; p < 0.001, ns; not significant. 

Characteristics of SARS-CoV-2 and COVID-19

Weekly Operational Update on COVID-19

Pathology and pathogenesis of severe acute respiratory syndrome. The American journal of pathology

Covid-19: Is a second wave hitting Europe?

Fast-spreading UK virus variant raises alarms

Evolution of the COVID-19 vaccine development landscape

The COVID-19 vaccine development landscape

Next-generation vaccine platforms for COVID-19

Out of the frying pan and into the fire? Due diligence warranted for ADE in COVID-19. Microbes and Infection

COVID-19 vaccine safety

Vaccine adjuvant systems: enhancing the efficacy of sub-unit protein antigens

Targeting the innate immune response with improved vaccine adjuvants

Adjuvant strategies to improve vaccination of the elderly population

Severe acute respiratory syndrome-associated coronavirus vaccines formulated with delta inulin adjuvants provide enhanced protection while ameliorating lung eosinophilic immunopathology

Middle East respiratory syndrome coronavirus spike protein vaccine in dromedary camels and alpacas

Human Phase 1 trial of low-dose inactivated seasonal influenza vaccine formulated with Advax™ delta inulin adjuvant

Randomized clinical trial of immunogenicity and safety of a recombinant H1N1/2009 pandemic influenza vaccine containing Advax™ polysaccharide adjuvant

Immunogenicity and safety of Advax™, a novel polysaccharide adjuvant based on delta inulin, when formulated with hepatitis B surface antigen: a randomized controlled Phase 1 study

SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

A new coronavirus associated with human respiratory disease in China

Cryo-EM structure of the SARS coronavirus spike glycoprotein in complex with its host cell receptor ACE2

Mechanisms of host receptor adaptation by severe acute respiratory syndrome coronavirus

The HDOCK server for integrated protein-protein docking

In silico comparison of spike protein-ACE2 binding affinities across species; significance for the possible origin of the SARS-CoV-2 virus

Mobilization of HIV spread by diaphanous 2 dependent filopodia in infected dendritic cells

Efficient lentiviral transduction of Herpesvirus saimiri immortalized T cells as a model for gene therapy in primary immunodeficiencies

SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

Advax delta inulin adjuvant overcomes immune immaturity in neonatal mice thereby allowing single-dose influenza vaccine protection

Identification of murine CD8 T cell epitopes in codon-optimized SARS-associated coronavirus spike protein

Convalescent plasma anti-SARS-CoV-2 spike protein ectodomain and receptor-binding domain IgG correlate with virus neutralization

Vaccine development for an imminent pandemic: why we should worry, what we must do

Baculovirus as versatile vectors for protein expression in insect and mammalian cells

Activation and evasion of type I interferon responses by SARS-CoV-2

Type I interferon susceptibility distinguishes SARS-CoV-2 from SARS-CoV

CD8+ T cell efficacy in vaccination and disease

Adjuvants enhancing cross-presentation by dendritic cells: the key to more effective vaccines?

Delta inulin polysaccharide adjuvant enhances the ability of split-virion H5N1 vaccine to protect against lethal challenge in ferrets

Antibody-dependent enhancement and SARS-CoV-2 vaccines and therapies

Disease severity dictates SARS-CoV-2-specific neutralizing antibody responses in COVID-19

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

Effect of convalescent plasma therapy on viral shedding and survival in patients with coronavirus disease 2019

Covid-19: Eli Lilly pauses antibody trial for safety reasons

COVID-19 antibodies on trial

SARS-CoV-2 entry factors are highly expressed in nasal epithelial cells together with innate immune genes

Clinical and immunological features of severe and moderate coronavirus disease 2019

COVID-19 severity correlates with airway epithelium-immune cell interactions identified by single-cell analysis

Elevated exhaustion levels and reduced functional diversity of T cells in peripheral blood may predict severe progression in COVID-19 patients

High IL-6/IFN-γ ratio could be associated with severe disease in COVID-19 patients

COVID-19 vaccines: modes of immune activation and future challenges

Potential adjuvants for the development of a SARS-CoV-2 vaccine based on experimental results from similar coronaviruses

Construction and immunogenic studies of a mFc fusion receptor binding domain (RBD) of spike protein as a subunit vaccine against SARS-CoV-2 infection

Infection and rapid transmission of SARS-CoV-2 in ferrets

Towards effective COVID-19 vaccines: Updates, perspectives and challenges

Viral vectors as vaccine platforms: deployment in sight

DNA and RNA-based vaccines: principles, progress and prospects

The following reagent was deposited by the Centers for Diseases Control and Prevention and obtained through BEI Resources, NIAID, NIH: SARS-Related Coronavirus 2, Isolate USA-WA1/2020, NR-52281. The authors would like to thank the University of Georgia Animal Resources staff, technicians, and veterinarians for animal care. We also acknowledge the expert assistance of Johnson Fung and King Ho Leong with the endotoxin and CTL assays.

YHO, LL, JB, and NP are affiliated with Vaxine Pty Ltd which holds the rights to COVAX-19™ vaccine and Advax™ and CpG55.2™ adjuvants.