Broad Protection Against Clade 1 Sarbecoviruses After a Single Immunization with Cocktail Spike-Protein-Nanoparticle Vaccine

The 2002 SARS outbreak, the 2019 emergence of COVID-19, and the continuing evolution of immune-evading SARS-CoV-2 variants together highlight the need for a broadly protective vaccine against ACE2-utilizing sarbecoviruses. While updated variant-matched formulations such as Pfizer-BioNTech’s bivalent vaccine are a step in the right direction, protection needs to extend beyond SARS-CoV-2 and its variants to include SARS-like viruses. Here, we introduce bivalent and trivalent vaccine formulations using our spike protein nanoparticle platform that completely protected hamsters against BA.5 and XBB.1 challenges with no detectable virus in the lungs. The trivalent cocktails elicited highly neutralizing responses against all tested Omicron variants and the bat sarbecoviruses SHC014 and WIV1. Finally, our 614D/SHC014/XBB trivalent spike formulation completely protected human ACE2-transgenic hamsters against challenges with WIV1 and SHC014 with no detectable virus in the lungs. Collectively, these results illustrate that our trivalent protein-nanoparticle cocktail can provide broad protection against SARS-CoV-2-like and SARS-CoV-1-like sarbecoviruses.


Introduction
Although it has been more than three years since the identification of SARS-CoV-2, 38 COVID-19 continues to cause considerable morbidity and mortality worldwide. Many vaccines 39 against SARS-CoV-2 have been approved and been successful in reducing the cases of serious 40 disease. 1 The majority of these vaccines elicit a neutralizing antibody response that targets the 41 SARS-CoV-2 spike (S) protein. [2][3][4][5][6] The S protein binds to the angiotensin-converting enzyme 2 current vaccines or previous infection continue to emerge, such as the Omicron variants BA.5, 48 BQ.1, and XBB. 9-11 While efforts are currently focused on developing a pan-SARS-CoV-2 49 vaccine to protect against current and future variants, the discovery of a large reservoir of ACE-2 50 binding sarbecoviruses circulating in bats has prompted interest in developing pan-sarbecovirus as 51 well as eventually pan-betacoronavirus vaccines. 12 52 There are two main approaches used to design vaccines to induce broad protection against 53 highly variable viruses. One of these approaches is to focus the immune response on conserved 54 portions of the antigen to elicit cross-reactive antibodies. For SARS-CoV-2, the highly conserved 55 S2 subunit of the S protein has been utilized as an antigen in broadly protective vaccines. 13 Since authorization of the bivalent mRNA boosters, many studies have analyzed the 64 efficacy of an additional dose of the bivalent vaccines compared to a booster of the original booster. Relative to BA.5, neutralization titers decreased 3-fold against BA.2.75.2, 4-5-fold 73 against BQ.1.1, 6-fold against XBB, and 8.5-fold against XBB.1. 18,19 The reduction in 74 neutralization efficiency against these new Omicron variants compared to ancestral SARS-CoV-2 75 is more drastic, even for groups receiving three doses of monovalent vaccine plus bivalent 76 booster. Neutralization titers decreased 37-fold against BA.2.75.2, 41-50-fold against BQ.1.1, and 77 85-100-fold against XBB. 1. 9,19 Together these results present a clear need for an updated vaccine 78 against recent and emerging Omicron variants. 79 While vaccines that provide pan-SARS-CoV-2 immunity would be of immediate interest were not tested because these variants had not yet been reported at the time of the study. While BA.5 as well as bat CoVs SHC014 and WIV1. We demonstrate that the selected trivalent 123 formulations consistently elicited robust neutralization titers against several Omicron variants, 124 SHC014, and WIV1. Additionally, immunization with this cocktail vaccine provided complete 125 protection against challenges with BA.5, XBB.1, SHC014, and WIV1, with no detectable viral 126 titers in the lungs. Collectively, these results strongly suggest that VLP-S cocktail vaccines have 127 the potential to provide broad protection against all significant clade 1 sarbecoviruses.

129
Selection of S proteins for immunization 130 First, an appropriate mix of S proteins had to be selected for a preliminary evaluation of 131 immunogenicity. Sarbecoviruses can be divided into multiple clades based on their RBDs ( Figure   132 1a). 26 Clade 1 contains the sarbecoviruses known to cause disease in humans, including SARS- 133 CoV-1 (Clade 1A) and SARS-CoV-2 (Clade 1B), as well as the sarbecoviruses considered to have 134 the greatest potential to cause future pandemics 12,27-29 . Our priority in this work, therefore, was to 135 generate a cocktail vaccine that elicited a broad neutralizing antibody response against viruses 136 from Clades 1A and 1B. We selected S antigens based on an analysis of the phylogenetic tree 137 ( Fig. 1a and 1b) and their amino acid sequence homology ( Fig. 1c and 1d). As potential S 138 antigens from Clade 1A, we chose SARS-CoV-1 and bat CoV SHC014 due to their low amino 139 acid homology with each other relative to other Clade 1A S proteins (Fig. 1c)

Generation and characterization of S nanoparticle-based vaccines
We previously developed VLPs displaying the 614D HexaPro 30 S protein. 25 The VLPs are 147 composed of 90 homodimers of the bacteriophage MS2 coat protein 31 with an AviTag inserted 148 into a surface loop and self-assemble into an icosahedral structure. As previously described, 149 AviTagged MS2 VLPs were biotinylated and then mixed with a large excess of streptavidin (SA) 150 to produce streptavidin-coated VLPs (MS2-SA). Biotinylated S proteins of each sarbecovirus 151 were also produced as previously described. In brief, HexaPro variants of each S protein with a C-152 terminal trimerization domain, AviTag, and his-tag were expressed in Expi293F mammalian 153 cells. S proteins were purified by immobilized metal affinity chromatography (IMAC), 154 biotinylated in vitro, and purified by size exclusion chromatography (SEC). VLP-S particles were 155 produced by mixing the appropriate ratio of MS2-SA to biotinylated S protein (Fig. 2a). This ratio 156 corresponded to the mixture of MS2-SA and S with the lowest amount of MS2-SA that did not 157 produce a peak indicating excess S on an SEC chromatogram. 158 The S antigens and VLP-S vaccines were characterized by several techniques. Prior to 159 displaying the S proteins on MS2-SA, the purity of the biotinylated S was verified by SDS VLP-SHC014-S elicited higher neutralizing antibody titers against WIV1 compared to VLP-188 SARS-CoV-1-S despite the WIV1 S being more homologous to the SARS-CoV-1 S than to the 189 SHC014 S (Fig 1c). BA.5 -did not protect against a BA.5 challenge, with viral lung titers comparable to those in 201 hamsters immunized with the control VLP (Fig. 3b). 202 Antigenic cartography to select vaccine cocktails 203 Next, we applied antigenic cartography to our neutralization data (Fig. 3c). Using Fc goat antibody (MP Biomedical, 1:5000) diluted in PBST with 1% BSA was added to all wells. 6 The wells were washed three times with PBST after a 1-hour incubation. 100 μL of TMB 7 (Thermo Scientific) were added to each well and allowed to develop for 3 minutes. The reaction 8 was stopped with 160 mM sulfuric acid, and the absorbance at 450 nm was read with a 9 Spectramax i3x plate reader (Molecular Devices) and Gen5 2.07 software (BioTek). Amino Acid Identity and Phylogenetic Trees 10 Amino acid sequences for all included sarbecoviruses were retrieved from GenBank (Table S2).

Antigenic Cartography
Antigenic cartography was performed as has been previously described using the Racmacs R 1 package. 43 In brief, FRNT50 values were calculated from each sera sample against each virus.

2
The dissimilarity Dij between each serum i and virus j is defined as where Hij is the FRNT50 value of serum i against virus j and Hi,max is the maximum 4 FRNT50 value from serum i. The error function for each pair is then is the Euclidean distance on the two-dimensional map between serum i and virus j. For FRNT50 6 values below a detection threshold (e.g., <20), the error function was instead calculated as = Forsythe test, respectively. All statistical analysis was carried out using Prism 9 (GraphPad).