Immunogenicity and pre-clinical efficacy of an OMV-based SARS-CoV-2 vaccine

The vaccination campaign against SARS-CoV-2 relies on the world-wide availability of effective vaccines, with a potential need of 20 billion vaccine doses to fully vaccinate the world population. To reach this goal, the manufacturing and logistic processes should be affordable to all countries, irrespectively of economical and climatic conditions. Outer membrane vesicles (OMVs) are bacterial-derived vesicles that can be engineered to incorporate heterologous antigens. Given the inherent adjuvanticity, such modified OMVs can be used as vaccine to induce potent immune responses against the associated protein. Here we show that OMVs engineered to incorporate peptides derived from the receptor binding motif (RBM) of the spike protein from SARS-CoV-2 elicit an effective immune response in immunized mice, resulting in the production of neutralizing antibodies. The immunity induced by the vaccine is sufficient to protect K18-hACE2 transgenic mice from intranasal challenge with SARS-CoV-2, preventing both virus replication in the lungs and the pathology associated with virus infection. Furthermore, we show that OMVs can be effectively decorated with RBM peptides derived from a different genetic variant of SARS-CoV-2, inducing a similarly potent neutralization activity in vaccinated mice. Altogether, given the convenience associated with ease of engineering, production and distribution, our results demonstrate that OMV-based SARS-CoV-2 vaccines can be a crucial addition to the vaccines currently available.


Introduction
The outbreak of SARS-CoV-2 has generated a pandemic event which caused almost 200 million documented infections worldwide and 4 million deaths to the current date 1 . Despite early attempts to contain the infection within the area where the first human cases were reported, the etiological agent of COVID-19 has spread worldwide and became the eighth human coronavirus. Unlike SARS-CoV and MERS, the current prevalence and the high rate of asymptomatic or pauci-symptomatic infections will favour SARS-CoV-2 circulation in adults and children in the future and will establish the infection as endemic, similarly to other human respiratory coronaviruses which cause flu-like pathologies (OC43, HKU1, 229E and NL63). Developing a long-term effective vaccination strategy is crucial to build the best possible immunity against SARS-CoV-2 in the human population, following the continuous evolution of the virus driven by the adaptation into the host and the selective pressure posed by the immune system. Reminiscent of the yearly influenza prophylaxis, the need of periodic updating and re-administration of the SARS-CoV-2 vaccine, depending on the extent of viral genetic drift, becomes a likely prospect for the future.
Currently, more than 280 candidate vaccines are under development at different stages, and, among these, 102 are in clinical phase 2 . While vaccines based on protein subunits are reported to represent the largest fraction of those investigated, vaccines based on mRNA and viral vectors were the first to obtain emergency use authorization by the major agencies and were massively distributed and administered world-wide. As a result, the world witnessed an unprecedented technical and economical challenge, given the need of manufacturing, distributing, storing, and administering billions of doses of vaccines in every country, irrespectively of climatic, social and economic conditions. To overcome such challenges and provide a sustainable long-term prophylaxis, the vaccine production should rely on a process easily scalable and low-cost to make the doses affordable to the less favoured regions of the world. Importantly, to facilitate distribution and storage in all countries, the vaccine should not depend on the cold chain, which is impractical and logistically and economically prohibitive particularly in several parts of the world.
Among the several technologies available for vaccine development, outer membrane vesicles (OMVs) have emerged in recent years as an attractive tool capable of coupling excellent built-in adjuvanticity provided by the bacterial pathogen-associated-molecular patterns (PAMPs) associated with the vesicles, and an easily scalable production and purification process 3 4 5 . While OMVs purified from different pathogens are known to induce potent protective immunity against the same pathogens from which they are derived from (accordingly, anti-Neisseria meningitidis OMV-based vaccine are currently available for human use 6 7 8 ), we have recently developed a strategy which allows the engineering of E.
coli OMVs selectively loaded with heterologous antigens 9 10 . This strategy has been successfully demonstrated to induce potent immunity against foreign bacterial components and against human tumour-specific antigens. 11,12 Given that OMVs are readily phagocytosed, antigens carried by the vesicles are efficiently presented by professional antigen presenting cells, leading to efficient induction of a sustained Th1 response as well as an optimal humoral response. Since clinical evidence demonstrates that an accelerated induction of a Th1 cell response is associated with less severe cases of COVID-19 13,14 and that convalescent individuals develop strong memory CD4+ and CD8+ T cells 15 , the ability of OMVs to trigger Th1 represents a desired feature. Crucially, in addition to the simple and cost-effective setup required to produce and purify OMVs, the antigen-decorated vesicles are extremely stable for long-term storage at room temperature, making it a convenient vaccine to distribute all over the world 10 .
All SARS-CoV-2 vaccine candidates under development are designed to induce antibodies specific for the spike (S) protein. Neutralizing titres found in vaccinated individuals, as well as in convalescent patients, correlate strongly with antibody binding to the spike receptor binding domain (RBD), which interacts with the angiotensin-converting enzyme 2 (ACE2) on the host cell membrane 16 . Accordingly, the most potent monoclonal antibodies isolated from convalescent patients recognize epitopes located in the RBD interface with ACE2 17 . The ability to specifically concentrate the immune response against these epitopes would therefore exclusively elicit neutralizing antibodies, while minimizing the risk of generating non-neutralizing or poorly neutralizing immunoglobulins binding to irrelevant spike regions.
The OMVs offer the unique opportunity to display short and defined epitopes to B-and Tcells in a highly immunogenic context provided by the bacterial vesicle. Taking advantage of such potential and the availability of the crystal structure of the RBD in complex with ACE2, we engineered the OMVs with peptides derived from the RBM alpha helices which form the interface with ACE2. Here we show that RBM-derived peptides can be correctly associated with OMVs, inducing neutralizing antibodies titres sufficient to fully protect K18-hACE2 transgenic mice from challenge with SARS-CoV-2. Given the efficacy of the vaccine in the animal model, the ease of its engineering, the cost-effective production process, and the stability at room temperature, we propose the OMV-based vaccine as a promising candidate to continue the vaccination campaign against SARS-CoV-2.

Design and construction of the OMV-based vaccine
As shown by the 3D structure of the SARS-CoV-2 RBD in complex with ACE2, the concave receptor binding motif (RBM) of the spike protein is organized in two discrete ordered chains incorporating the 5 and 6 strands which cross each other and include most of the residues contacting ACE2 ( Figure 1A). Patient-derived monoclonal antibodies (mAbs) binding epitopes in these chains of the RBD potently neutralize virus cell entry in vitro and are currently in clinical use or in advanced clinical development 18 . In the attempt of producing a vaccine eliciting neutralizing immunity against SARS-CoV-2, we generated OMVs decorated with the polypeptides derived from the RBM, fused with FhuD2, a Staphylococcus aureus lipoprotein shown to efficiently deliver heterologous protein domains to the E. coli outer membrane and to the vesicular compartment 10 .
To this end, the nucleotide sequence coding for 5 (RBM438-462) and 6 (RBM467-509) strands or a combination of the two (RBM438-509) were fused to the 3' end of the sequence encoding FhuD2 in plasmid pET-FhuD2 12 , thus generating the fusion proteins FhuD2-RBMs ( Figure   1B). The resulting plasmids were used to transform E. coli60, a hyper-vesiculating E. coli BL21(DE3) derivative recently created in our laboratories 10 .
From the culture supernatant of one recombinant clone each (E. coli60(pET-FhuD2-RBMs)) OMVs were concentrated and purified. In order to verify the correct association of the fusion proteins with the bacterial vesicles, FhuD2-RBMs-OMVs were subjected to SDS-PAGE analyses, which revealed the presence of protein species compatible with the predicted molecular weights of the FhuD2-RBM fusion proteins ( Figure 1C). Based on the intensity of the bands visualized by Coomassie staining, the incorporation of the fusion protein into OMVs appears to be similar to the incorporation level of FhuD2 alone, indicating that the addition of the SARS-CoV-2 RBM polypeptides is well tolerated and compatible with efficient transport into the bacterial vesicles.

Immunogenicity of FhuD2-RBMs-OMVs
Having proven efficient association with OMVs, we next tested the capacity of FhuD2-RBMs-OMVs to induce the production of antibodies capable of recognizing the RBD in the context of the SARS-CoV-2 spike protein. To this aim, five CD1 mice were immunized with each construct three times at two-week intervals with 10 μg of each engineered OMV preparation combined with aluminum hydroxide (Alum) (2 mg/ml, Figure 2A

Protective activity of FhuD2-RBM438-509-OMVs in hACE2 transgenic mice
Having established that the RBMs OMVs are effective at inducing neutralizing immunity in mice, and that vesicles which incorporate the combination of two RBM strands elicit higher immunity compared with those displaying either the 5 or 6 strands alone, we selected the former to explore the degree of protection induced in vivo against SARS-CoV-2 challenge.

Elicitation of neutralizing antibodies with OMVs decorated with the RBM438-509 epitope from the P.1 isolate.
In order to test the adaptability of the OMV-based vaccine to RBM from different SARS-CoV-2 variants, we engineered the vesicles using the combination of 5 and 6 strands derived from the P.1 variant (B.1.1.28.1) which carries two changes (E484K, N501Y) in the RBM compared to the Wuhan-1 isolate (Fig. 4A). The peptide, fused to FhuD2, could be expressed in E. coli and was associated efficiently to OMVs, as observed by SDS-PAGE  Figure   4C). Of note, while ELISA plates were coated with RBD derived from the Wuhan-1 isolate, the amount of IgGs in the animal sera capable of binding the RBD was similar to that induced by OMV decorated with the RBM derived from the Wuhan-1 isolate ( Figure 4C), despite the presence of two polymorphic amino acids. Moreover, sera collected from mice immunized with OMVs decorated with the RBM438-509 from the P.1 variant neutralized SARS-CoV-2 Wuhan-1 spike pseudotyped vectors with similar potency respect to sera from mice immunized with OMVs engineered with the RBM438-509 epitope from the Wuhan-1 isolate ( Figure 4D). In conclusion, our results prove that the OMVs can be successfully engineered to induce neutralizing immunity with RBM form different SARS-CoV-2 variants. With this complex scenario, we believe that an OMV-based vaccine could offer interesting advantages and solutions. First of all, OMVs appear amenable to display crucial peptides of the Spike glycoprotein. Of note, while properly folded eukaryotic glycoproteins can normally only be expressed in a eukaryotic expression system, requiring a labor-intensive protein purification process, the crucial portion of the spike RBM, which contacts directly ACE2, can be efficiently expressed and incorporated into OMVs while maintaining a natural

Discussion
conformation. An effective immunity is in fact elicited in vaccinated mice, with efficient production of neutralizing antibodies, at levels sufficient to protect the animals from infection.
Accordingly, the reduced viral replication in the lungs of vaccinated mice was associated with a favorable clinical score and a severely reduced recruitment of inflammatory monocytes.
Second, OMVs are extremely easy to produce. The separation of the biomass from the culture supernatant and an ultrafiltration step to concentrate OMVs and eliminate contaminants released into the supernatant is essentially all is needed for vaccine production 22 . The process is amenable to large scale production and can be easily transferred to different sites to expand vaccine production. Moreover, the production yields make the vaccine costs particularly affordable. Under laboratory conditions, we reproducibly obtain more than 5.000 OMV-based vaccine doses/liter of culture. Practically speaking, it means that using a 1.000-liter fermentation unit associated to a tangential flow ultrafiltration devise is sufficient to provide 5 million doses of vaccine/week at costs that are expected to be well below 1 USD/dose. Third, although not yet experimentally demonstrated, theoretically speaking our OMV vaccine should be particularly indicated as a booster vaccine both in the case the boost is needed to reinvigorate the immune response against the original vaccine strain and in the case the immune response has to be potentiated against a new viral variant. Accordingly, OMVs can be easily and quickly modified to incorporate updated antigens.
In the first case, as the data here presented demonstrate, our vaccine has the unique property to elicit antibodies which specifically bind the RBM region interacting with the ACE2 receptor. Therefore, when the immune response goes below a protective threshold, it could be convenient to boost those memory B and T cells which are particularly relevant for protection. This would avoid the "dilution" of the response toward irrelevant epitopes.
Moreover, the potent adjuvanticity of the OMVs should further enhance the optimal expansion of functional B and T cells.
As far as the use of a booster dose to cope with the spreading of new variants is concerned, theoretically speaking our OMV-based vaccine should mitigate the risk of the so call "original antigenic sin (OAS)". OAS refers to the fact that once the immune system is exposed to a certain pathogen, a second infection by a slightly different version of the same pathogen preferentially triggers the immunological memory generated by the first infection 23  In conclusion given the convenience associated with the engineering, production and storage, the efficacy as a vaccine against SARS-CoV-2 makes E. coli60 OMVs a valuable tool for the future management of COVID19.  Biol. 498, 91-103 (2009)). Briefly, pET21-FhuD2 plasmid was linearized by PCR, using FhuD2-v-R and pET-V-F primers (Table 1). In parallel, the synthetic DNA encoding the RBD of SARS-CoV-2 was synthesized by GeneArt (Thermo Fisher Scientific, Waltham, MA, USA) and used as template for the amplification of the three RBM epitopes (Table 2). More in detail, RBM438-462 and RBM467-509 were amplified by PCR with the forward 2-F and 1-F and the reverse 2-R and 1-R primers, respectively ( Table 1). The PCR products and the linearized plasmid were mixed together and used to transform E. coli HK100 strain.

Engineering BL21(DE3) E. coli strains with SARS-CoV-2 neutralizing epitopes
RBM438-509, the combination of the two epitopes, was assembled with two steps of PCR in succession. In the first step two different fragments carrying an overlapping sequence were amplified with 2-F/R-1 and F-1/1-R primers ( Table 1). In the second step the two fragments were eluted from Agarose gel, mixed together and used as template for a final PCR reaction with the primers 2-F and 1-R. This final product and the linearized plasmid were mixed together and used to transform the E. coli HK100 strain.
The RBM-BV438-509 gene, encoding for the RBM438-509 epitope from the P.1 isolate, a variant of the RBD carrying the E484K and N501Y mutations, was synthesized by GeneArt (Thermo     paraformaldehyde for 20 min and stained with 0.05% (wt/vol) crystal violet in 20% methanol.

RNA extraction and qPCR
Tissues homogenates were prepared by homogenizing perfused lung using gentleMACS  Flow cytometry analysis were performed on a BD FACSymphony A5 SORP and analyzed with FlowJo software (Treestar).

Confocal immunofluorescence histology and histochemistry
Lungs of infected mice were collected and fixed in 4% paraformaldehyde (PFA). Samples were then dehydrated in 30% sucrose prior to embedding in OCT freezing media (Bio-Optica). 20 m sections were cut on a CM1520 cryostat (Leica) and adhered to Superfrost

Clinical score
Mice were observed daily for clinical symptoms. Disease severity was scored as reported in Table 4.

Preparation of viral pseudotypes and neutralization assays.
SIV-based lentiviral particles pseudotyped with SARS-Cov-2 spike were produced in 10 cm Fitted sigmoidal curves and IC50 were obtained using Prism (GraphPad) with the least square variable slope method and using the dose-normalized response protocol.

Human sera
The sera of 50 healthcare workers with pre-existing immunity for SARS-CoV-2 (as per