A cohort of 500 COVID-19 non-vaccinated patient volunteers were obtained from Asry Medical Center and Awali Hospital in Manama, Bahrain. The selection was based on a positive SARS-CoV-2 PCR test, age range between 18 and 75 years and a clear clinical status defined by the patients’ attending physicians according to World Health Organization (WHO) COVID-19 classification as asymptomatic, mild, moderate and severe forms (https://www.who.int/westernpacific/emergencies/covid-19/information/asymptomatic-covid-19. Information on this cohort is compiled in Table 1 . The patient’s sera were prepared and stored at -80°C. One hundred pre-pandemic historical sera from age-matched healthy individuals were also retrieved and used as negative controls for this study. Informed consent was obtained from all patients and control individuals, and approval was obtained from the Arabian Gulf University Institutional Research and Ethics Committee on December 24, 2020, reference E031-PI-12/20. All methods were performed in accordance with the principles of Helsinki declaration for Medical Research Involving Human Subjects.

Purpose-bred BALB/c mice used in this study were housed in stainless steel ventilated cages under a 12-hour light/12-hour dark cycle at room temperature ranging from 18 to 22°C and a humidity between 50 and 60%. Animals were fed rodent chow and water ad libitum. Experiments were conducted according to national and international guidelines, such as those issued by the US National Institute of Health Office of Laboratory Animal Welfare (OLAW) (https://olaw.nih.gov/home.html) the study is reported in accordance with ARRIVE guidelines (https://arriveguidelines.org). The university Research and Ethics committee approved the protocols (reference E031-PI-12/20).

The coordinates of the SARS-CoV-2 S protein trimeric form (PDB ID: 6VXX) were downloaded from the RCSB database (https://www.rcsb.org). Water and heteroatoms were removed from the coordinates, and the coordinates of the complex were split into separate PDB monomer files using PyMOL software (Delano, 2002) from the Schrodinger software package R. Because the PDB ID: 6VXX model is missing the furin and TMPRSS2 cleavage site loops, the S protein monomer was modelled to include these loops using the Phyre2 web server (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index) (Kelley et al., 2015), and the results were saved on PyMOL software to visualize exposed residues.

The S protein amino acid sequence was submitted to the BepiPred-2.0 Sequential B-Cell Epitope Predictor server (http://www.cbs.dtu.dk/services/BepiPred/cite.php) using the default parameters to predict putative immunogenic epitopes in the S protein monomer surface-exposed regions. BepiPred-2.0 is based on a random forest algorithm trained on epitopes and non-epitopes annotated from antibody-antigen protein crystal structures (Jespersen et al., 2017).

The S protein S1/S2 junction sequence (residues E648-L809) containing the furin and TMPRSS2 proteolysis sites was submitted to the improved Proteasome Cleavage Prediction Server (http://imed.med.ucm.es/Tools/pcps/). This server predicted the cleavage sites generated by the constitutive proteasome/immunoproteasome within the protein sequence using different Ngram models (Garcia-Boronat et al., 2008) and displayed the predicted epitopes.

To predict MHC class 1-binding peptides, we used the NetMHCpan 4.1 (https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/) and NetMHC - 4.0 (https://services.healthtech.dtu.dk/services/NetMHC-4.0/) servers. NetMHCpan and NetMHC servers operate on artificial neural networks (ANNs) to predict the binding of peptides to MHC class I molecules of known sequences. To predict MHC class 2 peptides, we used the NetMHCII 2.3 (https://services.healthtech.dtu.dk/services/NetMHCII-2.3/) (Jensen et al., 2018), which allows for the prediction of peptides binding to HLA-DR, HLA-DQ, and HLA-DP. We analyzed The S protein S1/S2 junction sequence using the default settings of all servers. Peptides were then sorted by the software according to their binding affinity level to MHC molecules. Binding affinity prediction methods for NetMHCII were constructed using an extended data set of quantitative MHC–peptide binding affinity data obtained from the Immune Epitope Database covering HLA-DR, HLA-DQ, HLA-DP and H-2 molecules (Jensen et al., 2018).

We used BioEdit, a multiple sequence alignment (Hall, 1999) software program, to analyze the conservation of the identified epitopes. We used the Epitope Conservancy Analysis tool from the freely accessible Immune-Epitope Database (IEDB; http://tools.iedb.org/conservancy/) to assess the variability of the epitopes based on the sequence alignment of 25 SARS-CoV-2 variants that have circulated or are still circulating since the start of the pandemic, along with SARS-CoV and MERS-CoV sequences. We acquired the amino acid sequences of the extracellular domain of SARS-CoV, MERS-CoV, and SARS-CoV-2 spike protein from the National Center for Biological Information (NCBI) protein ID: YP_009825051.1, YP_009047204.1 and YP_009724390.1, respectively. We reproduced SARS-CoV-2 variants sequences by introducing specific mutations to the Wuhan sequence to generate different variants using published mutations in the covariant (https://covariants.org/) and Stanford University SARS-CoV-2 Variants databases (https://covdb.stanford.edu/variants/omicron_ba_1_3/)

We used the interferon epitope server https://webs.iiitd.edu.in/raghava/ifnepitope/predict.php to identify IFN-γ-inducing epitopes in the designed polypeptide P3/Fur/x3 and protein subunits P3-L and SJ/FT.

For ab initio modelling, we used Iterative Threading Assembly Refinement (ITASSER) (https://zhanglab.dcmb.med.umich.edu). All designed peptides sequences were submitted to the ITASSER server using the default settings. The server identified structural templates from the template library by using the multiple threading approach LOMETS. Functional insights regarding the target were then derived by rethreading the 3D models through the protein function database BioLiP (Yang and Zhang, 2015). The obtained models were visually inspected with PyMOL.

The polypeptides P3/FUR/x3 and P4/TMP/x3 were chemically synthesized by GeneCust (GeneCust, Boynes, France) at a 5 mg scale and 95% minimum purity, as verified by mass spectrometry analysis and high-performance liquid chromatography (HPLC). For the polypeptides used in the indirect ELISA, a biotin tag was alternatively added to allow antigen binding to streptavidin-coated plates.

For the development of the indirect ELISA with the candidate vaccinating antigens, Ninety-six-well ELISA plates (Nunc MaxiSorp, Thermo Fisher Scientific) were coated with 200 ng per well of streptavidin in carbonate-bicarbonate buffer, pH 9.6 (C3041, Sigma-Aldrich), overnight at 4°C. The coated plates were then washed three times with 200 µl of wash buffer (0.1 M phosphate, 0.15 M sodium chloride, pH 7.2, 0.05% Tween 20) and blocked with wash buffer containing 5% nonfat milk “Régilait” powder at room temperature for 1 hour under shaking. Then, the plates were washed as previously described, and 100 µl of biotinylated peptides (200 ng) was added to each well. After peptide coating, the plates were washed again, 100 µl of serum sample diluted to 1:100 in dilution buffer was added to each well, and the plates were incubated at room temperature for 1 h. The plates were then extensively washed 3 times, 10000-fold-diluted horseradish peroxidase (HRP)-conjugated goat anti-human IgG (Abcam) was added, and the plates were incubated for 1 hour at room temperature. TMB substrate (TMB, Sigma-Aldrich) was added to each well, and the plates were then incubated for 20 minutes in the dark. The reaction was stopped by adding 100 μL of 2 N H2SO4 solution. The optical density (OD) was measured at 450 nm using a FLUOstar Omega (BMG LABTECH) absorbance microplate reader. We defined a sample as IgG positive if the OD value was equal to 3 standard deviations (SDs) above the mean of the negative control values (n=100).

To test the cohort of patients with COVID-19 for IgG to the SARS-CoV-2 S protein RBD region, we used a commercial quantitative ELISA kit (RayBio COVID-19 S1 RBD protein Human IgG ELISA kit; RayBiotech; Parkway, USA). The kit provides positive controls to construct a standard curve for IgG titer determination. The ELISA was performed according to manufacturer’s procedure. Briefly plates were received coated with SARS-CoV-2 S1 RBD protein, and incubated with 100µL of serum for 1h at room temperature (RT). Then, the wells are washed four times with 300µL of wash buffer, and 100µL of biotinylated anti-human IgG antibody was added and incubated for 30 minutes at RT. The wells were washed again and the TMB substrate solution was added. The reaction was stopped, and the OD was measured at 450nm.

The multiepitope subunit P3-L was designed to include a homopolymerized (3 times) 32-amino-acid core sequence containing the FCS. The SJ/FT subunit was designed to contain both the furin and TMPRSS2 cleavage motifs and the FP sequence. The two protein subunits were produced as recombinant proteins. The cDNA nucleotide sequences encoding these multiepitope subunits were chemically synthetized by GeneCust (GeneCust, Boynes, France). Each cDNA was inserted into the pET-22b (+) plasmid (Novagen Sigma-Aldrich, USA) NdeI- XhoI multiple cloning site, and the recombinant expression vector was transformed into Escherichia coli BL21 (DE3) cells. Positive colonies were selected on 50 µg/ml ampicillin agarose plates. For each recombinant cDNA, a single positive colony was grown overnight at 37°C on LB medium, the expression of the recombinant protein was induced by adding 0.5 mm IPTG, and bacteria were cultured at 220 rpm for 16 hours at 37°C for P3-L and at 15°C for SJ/FT. The expressed protein was purified from the sonicated bacterial pellet using a Ni-NTA column. The purified protein was dialysed in 20 mM Tris, 4 M urea, pH 8.0, at 4°C. Endotoxin was removed using endotoxin removal beads from Miltenyi Biotec (Bergischgladbach/Germany).

We used the VaxiJen server (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html) and ANTIGENpro module of the SCRATCH protein predictor (http://scratch.proteomics.ics.uci.edu/) to evaluate the immunogenicity of the multiepitope subunits. We used the AllerTOP v. 2.0 (http://www.ddg-pharmfac.net/AllerTOP/) and AlgPred servers (http://crdd.osdd.net/raghava/algpred/) to check for potential allergenicity.

We used the ProtParam tool of the EXPASY database server (http://web.expasy.org/protparam/) to determine the physicochemical parameters (molecular weight, half-life, atomic composition, stability index and mean hydrophilicity) of the vaccine candidates’ antigens.

Cell culture and stimulation: Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood from 10 healthy donors and 10 patients with COVID-19 by density centrifugation through Ficoll-Hypaque (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). PBMCs were cultured in 96-well plates at a concentration of 1 × 10⁶ cells/ml in a final volume of 200 µl of RPMI 1640 medium (Sigma, St. Louis, MO) supplemented with 2 mM L-glutamine (Sigma, St. Louis, MO), 100 U/ml penicillin (Sigma, St. Louis, MO), 100 µg/ml streptomycin (Sigma, St. Louis, MO), 10% (v/v) heat-inactivated human AB serum (Sigma, St. Louis, MO), 1% HEPES (0.01 M), (Invitrogen, Cergy-Pontoise, France), 1% sodium pyruvate (1 mM) (Invitrogen, Cergy-Pontoise, France), 1% MEM nonessential amino acids (Invitrogen, Cergy-Pontoise, France), 2-mercaptoethanol (2−10 M), (Invitrogen, Cergy-Pontoise, France), and 0.2% gentamicin (20 µg/ml) (Invitrogen, Cergy-Pontoise, France). The cells were stimulated with individual proteins or with peptide at a final concentration of 10 μg/ml for 10 days at 37°C in 5% CO₂. Recombinant human IL-2 (BD Bioscience, San Diego, CA) (10 IU/well) was added at days 2, 4 and 7 during culture. For positive controls, cells were stimulated with phytohemagglutinin (PHA) (Sigma-Aldrich, St Louis, MO) (10 μg/ml) for 3 days. Negative control cells were not stimulated. Culture supernatants were then harvested and frozen at -80°C until use for ELISA, whereas cells were used for the ELISpot test.

The ELISpot assay was performed with PBMCs stimulated for 10 days with proteins or peptides (10 μg/ml) or for 5 days with PHA (10 μg/ml) using a human IFN-γ immunospot assay. Briefly, 96-well PVDF filter plates (Millipore) were coated with 100 µl of anti-human IFN-γ monoclonal antibody (mAb) (2 µg/ml) (1-D1K, Mabtech) overnight at 4°C, and free binding sites were blocked with X-VIVO for 1 hour at room temperature. The cells were washed twice with phosphate-buffered saline (PBS 1X), resuspended in X-VIVO™ 15 medium (Lonza, Sweden), re-stimulated with protein or peptide (10 µg/ml) for 30 min at 37°C, and then transferred to coated ELISpot plates. The cells were incubated for 16 hours in a 5% CO2 humidified atmosphere at 37°C. Biotinylated anti-IFN-γ detection antibody (7-B6-1, Mabtech) (1 µg/ml) was added, and the plates were incubated for 2 hours at room temperature. After washing, ExtrAvidin-Alkaline Phosphatase (Sigma) diluted 1/2000 was added, and the plates were incubated for 30 min at room temperature. The spots were revealed by adding the substrate (BCIP/NBT)-containing buffer. Spots were counted using a CTL ImmunoSpot reader (CTL Analysers, Shaker Heights, OH). The results are expressed as spot-forming units (SFU) per 106 PBMCs.

Female BALB/c mice aged 6 to 8 weeks were used for the immunogenicity study. Each group of mice consisted of five animals. For peptides (P3/FUR/x3, P4/TMP/x3), animals were immunized via the subcutaneous or intramuscular route and received three injections at day 0, day 14, and day 28 with 200 μg of P3/FUR/x3 or P4/TMP/x3 mixed with 25 μg of CpG (ODN 2395 VacciGrade, InvivoGen, USA) adjuvant. PBS was used for the immunization of the control group. Blood samples were collected before immunization, then at days 28 and 42 and up to day 100 for some groups. For the recombinant protein subunits (P3-L, SJ/FT), we evaluated the effect of escalating doses (10, 25, and 50 μg), as well as the effect of adjuvants (2 μg of CpG, 5 μg of MPLA, and 50 μg of alhydrogel). Mice received two successive shots at day 0 and day 14, and blood samples were collected 14 days after the second administration. Serum samples were analyzed for specific IgG against antigens with an in-house indirect ELISA. Briefly, mouse sera were diluted at 1:100 and added to the coated plate with the designed peptide/protein antigens. After incubation, HRP-conjugated goat anti-mouse IgG diluted 1:2000 (Abcam) was used as the secondary antibody.

Live virus experiments were performed in approved biosafety level 3 (BSL-3) facilities at Georgia State University and strictly followed the approved standard operation procedures. Neutralizing Abs were measured by plaque reduction neutralization assay (PRNT) using ancestral B.1 Wuhan virus (BEI# NR-52281). For PRNT assay Vero E6-TMPRSS2-T2A-ACE2 cells (BEI NR-54970) were seeded in 6-well plates at 200,000 cells/well in M199 medium with Earle’s salts (10X) supplemented with 5% inactivated fetal bovine serum, buffered with 3% sodium bicarbonate and 1% Penicillin-Streptomycin for 3 days (preformed monolayers). Serum samples were heated at 56oC for 1 hour to inactivate complement. Heat-inactivated serum samples were diluted at 1:10 and were further serially diluted 4-fold from 1:10 to 1:5120. 60μl of serially diluted serum was mixed in 96-well plates with an equal volume of 100 PFU of SARS-CoV-2. Serum/virus mixtures were incubated for 30 mins at 37°C. After incubation, 100 ul of serum/virus mixture was transferred to monolayered cells and incubated for 1 h at 37°C with gentle rocking every 15 minutes. After incubation, 2 ml of 1% low melting agarose media was added to each well and plates were incubated at 37°C for 2 days. After two days, plates were overlaid with 2% neutral red in 1% low-melting agarose for visualizing plaque formation. The number of plaques were counted and recorded in each well. The neutralization titer for each dilution was calculated as follows: NT = [1-(plaque count with sera/plaque count without sera)]x100.

For the analysis of the ELISA results from the cohort of COVID-19 patients, we used R (v. 4.0.5), readxl (v. 1.3.1), dplyr (v. 1.0.6) and stringi (v. 1.5.3) software packages for data handling and analysis. The “disease status” variable was assigned a numeral (integer) and factor (categorical variable) using the following rules: 0=asymptomatic, 1 = mild, 2 = moderate, and 3 = severe. We transformed the OD values obtained from the ELISA experiments from numeric (float, quantitative variable) to factor (categorical variable) data using the following rule: OD < 0.45 = immune (–) and OD ≥ 0.45 = immune (+) for the epitopes displayed by peptide P3/Fur/x3. For calculation of the chi-square test, we used the chisq.test function from the R Stats Package. We used the Monte Carlo simulation method instead of the asymptotic chi-squared distribution of the test statistic to compute the p value because some of the expected values were below 5. For the analysis of IFN-γ data, statistical analyses were performed with GraphPad Prism 9.2.0 (GraphPad Software, San Diego, CA). The results are expressed as medians (interquartile range, IQR, as variances). The Wilcoxon signed-rank test was used to compare the median IFN-γ production between stimulated and unstimulated cells. The Mann-Whitney test was used for intergroup analysis. p values less than 0.05 were considered to indicate significance. Correlations were estimated using the Spearman rank (rs) correlation coefficient.

The first step of our approach to designing a SARS-CoV-2 precision vaccine consisted of delimiting solvent-exposed sequences in the S protein S1/S2 subunit junction that potentially contain immunogenic epitopes. We extracted the structure of the SARS-CoV-2 spike protein monomer from the structural 3D model PDB ID: 6VXX using PyMOL software and visualized it using the Bepipred.2 server. We observed that the S1/S2 junction is mostly solvent exposed ( Figure 1 ). Moreover, the server revealed potential B-cell epitopes in this structural region ( Supplementary Table 1 ). This analysis also showed that furin and TMPRSS2 proteolytic motif sequences lie in two distant solvent-exposed loops ( Figure 1 ) and are part of the predicted potential epitopes, as indicated by the estimated epitope probabilities ( Supplementary Table 1 ). These observations suggest that this important region for virus infectivity can elicit an antibody response. We also submitted the SARS-CoV-2 S protein S1/S2 junction to the improved immunoproteasome cleavage-prediction server to search for putative T-cell epitopes. We identified fourteen immunogenic epitopes that the immunoproteasome of infected cells might generate upon processing of the SARS-CoV-2 S protein ( Figure 2 ). To check whether these predicted epitope sequences are conserved across SAR-CoV-2 variants, we aligned them using BioEdit software (Hall, 1999). Thus, we ran the aligned sequences on the Epitope Conservancy Analysis tool of the freely accessible Immune-Epitope Database (IEDB). The percentage of protein sequence matches showed that these peptides are conserved among the twenty-four SARS-CoV-2 variants that circulated and/or are still circulating ( Supplementary Table 2 ). These epitopes are not conserved in SARS-CoV or MERS-CoV ( Figure 2 ). Peptides 3 and 4 contain residues of the FCS sequence (PRRAR), peptide 11 contains the TMPRSS2 cleavage motif, and the sequence of peptide 12 overlaps with the fusion peptide (FP) sequence ( Supplementary Table 2 ).

For confirmation of the T-cell nature of these epitopes, we used the IFN epitope prediction server (Dhanda et al., 2013) to determine which epitopes potentially induce interferon γ (IFN−γ) . The results of this analysis identified peptides 3 and 11 as IFN-γ inducing epitopes ( Supplementary Table 3 ).

To further investigate the immunogenic epitopes at SARS-CoV-2 for presentation to the immune system by MHC molecules, we used the Net MHC pan, ProPred I, NetMHC II and ProPred II servers to run algorithms that perform epitope prediction with evaluation of antigenic peptide binding to defined MHC I and II molecules. This study showed that numerous predicted peptides from the SARS-CoV-2 S protein S1/S2 junction bind MHC class 1 and class 2 molecules of known sequence ( Supplementary Tables 4–6 ). Several of the predicted peptides contain or overlap amino acids from the FCS or the TMPRSS2 proteolytic site or lie near these sites. Some of these peptides were predicted to have strong binding affinity for specific MHC alleles, while others were predicted to have weak binding affinity ( Supplementary Tables 4–6 ). This finding implies that the predicted epitopes can be presented to the immune system through numerous common MHC 1 and class 2 molecules.

The prediction data show that the SARS-CoV-2 S protein S1/S2 junction contains epitopes that can potentially elicit both T- and B-cell immune responses, including epitopes that overlap with the furin and TMPRSS2 proteolytic sites.

To check if the predicted epitopes overlapping with the sequence involved in SARS-CoV-2 infectivity generate an immune response in patients with COVID-19, we used two synthetic peptides, P3/FUR/x3 and P4/TMP/x3. We designed these peptides by triplicating a sixteen-residue core sequence (Q662-I679) that spans the FCS in P3/FUR/x3 and a nine-residue core sequence (P794-R802) that spans the TMPRSS2 motif in P4/TMP/x3. In each polypeptide, we introduced the five-residue-long flexible sequence GGGGS as a molecular linker between each repeated core sequence ( Table 2 ). This linker brings stability and maintains distance between epitopes for stable and correct folding of the synthetized polypeptide (Patel et al., 2022). We used these polypeptides to investigate the specific antibody response in patients with COVID-19.

To study this antibody response, we developed an indirect enzyme-linked immunosorbent assay (ELISA) using the P3/FUR/x3 and P4/TMP/x3 polypeptides as antigens and examined the cohort of 500 patients with COVID as described in Table 1 , as well as 100 negative control serum specimens collected before the pandemic. The data showed that none of the control sera reacted with P3/FUR/x3 ( Figure 3 -A1) and P4/TMP/x3 ( Figure 3 -A3), while 80% and 86% of the COVID-19 patients’ sera specifically recognized P3/FUR/x3 ( Figure 3 -A2) and P4/TMP/x3 ( Figure 3 -A4), respectively. We detected this specific antibody response in numerous patients’ sera up to 50 weeks after infection. These data show that the S protein S1/S2 junction involved in SARS-CoV-2 infectivity elicits a natural robust and persistent antibody response.

To investigate the nature of this antibody response, we analyzed the ELISA data in the context of disease clinical forms. We observed a highly significant difference (p value = 0.0001) between patients’ antibody response to P3/FUR/x3 based on their disease status. All patients with the asymptomatic or mild form of the disease developed antibodies to epitopes in or around the furin cleavage site sequence ( Figure 3B ). Therefore, the antibody response to the S protein epitopes covering the furin proteolytic cleavage site in patients with COVID-19 significantly correlates with protective immunity to SARS-CoV-2.

Furthermore, the antibody response we observed did not cause antibody-dependent enhancement (ADE) of infection. Indeed, the patients who developed this type of immunity experienced rapid and full recovery, and no COVID-19 was recorded among them ten months after this study.

The study of the cohort of patients with COVI-19 for antibody response to the SARS-CoV-2 S protein RBD using a commercial ELISA kit revealed that 36.4% of the total samples had an IgG ELISA titer less than the cut-off (15 Unit/ml), indicating that these patients were negative for IgG against the RBD region at the time of blood sampling. Indeed, in this group of patients the blood sampling was performed at least 6 months after infection. Meanwhile, out of the 63.6% positive sera, 51.6% developed a low IgG titer, i.e less than three times the cut-off (45 Unit/ml); 18.3% a medium titer between 45 and 100 Unit/ml and 30.1% produced a high COVID-19 IgG titer above 100 Unit/ml. The highest IgG titer was 253.16 units/ml. No correlation was observed between the antibody response to the RDB and the response to the three candidate vaccinating antigens. There was also no correlation between the anti RDB IgG titer and the disease status.

Based on the data from the patient-centered study that showed strong and persistent natural immunity against the epitopes of the S protein covering the furin and TMPRSS2 cleavage sites, we developed two protein subunits, P3-L and SJ/FT, that encompassed the epitopes identified from the S protein S1/S2 junction. P3-L is 107 amino acids long, 11.1 kDa protein engineered by sequentially adding 3 core sequences of 32 amino acids that comprise the furin proteolytic site separated by a GGGGS molecular linker ( Figure 4A ). To design the SJ/FT protein subunit, we selected a sequence of 230 amino acids (24.8 kDa) from the S1/S2 junction that comprises both the furin and TMPRSS2 proteolytic motifs and includes the epitopes predicted earlier by computational analysis.

We selected these two subunits based on their predicted good antigenicity, no allergenicity and stability. Indeed, we used the VaxiJen server to predict the immunogenicity of S1/S2 junction-derived antigens P3/FUR/x3, P4/TMP/x3, P3-L and SJ/FT. The predicted immunogenicity data suggest that these polypeptides and protein subunits are immunogenic ( Supplementary Table 6 ). In addition, the prediction of allergenicity using the AlgPred and AllerTOP computational tools showed that P3/FUR/x3, P4/TMP/x3 and subunits P3-L and SJ/FT are nonallergenic ( Table 3 ), which makes them suitable for vaccination. Furthermore, we determined the physicochemical properties of P3-L and SJ/FT using the ProtParam tool of the EXPASY database server. The relatively low predicted instability index, good solubility values, and estimated half-life ( Supplementary Table 7 ) suggest that the physicochemical properties of the engineered antigens are suitable for production as recombinant vaccine subunits. We produced P3-L and SJ/FT as recombinant proteins in E. coli ( Figure 4B ) and purified them as endotoxin-free proteins for use in immunization studies.

To further investigate the vaccinating potential of the engineered recombinant subunits, we analyzed antibody and T cell responses in patients with COVID-19 antibodies to these antigens. ELISA experiments using P3-L and SJ/FT protein subunits as antigens and the serum of the patients of the COVID-19 cohort showed specific antibody responses in 84.7% and 97%, respectively ( Figure 5A ). To investigate the T-cell response, we studied IFN-γ production upon stimulation of the patients’ immune cells using P3-L and SJ/FT. We used PBMCs from COVID-19 patients with mild disease (MD) or severe disease (SD) and healthy controls (HCs) as defined by the WHO to perform ELISpot experiments. The results showed that the number of cells producing IFN-γ in response to stimulation with the P3-L and SJ/FT proteins, compared to unstimulated cells (NS), was highest for patients with MD (p<0.001) ( Figure 5B ). These observations show that the P3-L and SJ/FT protein subunits contain epitopes that generate antibody and T-cell responses in patients with COVID-19.

To further assess the vaccinating potential of the P3-L and SJ/FT protein subunits and polypeptides P3/FUR/x3 and P4/TMP/x3, we carried out a computational study to check whether these antigens exhibit a conserved or mimicked native 3D structure of the SARS-CoV-2 S protein S1/S2 junction. We constructed 3D models of their structures and compared them with the native 3D structure of the S protein monomer. The predicted structural models show an alpha coil/helix fold that is consistent with the structure of immunogenic epitopes and solvent-accessible protein sequences. A repetitive 3-fold display of the alpha coil/helix fold is observed in the P3/FUR/x3 peptide and protein subunit P3-L ( Figure 6A ), which shows that the unit peptide sequence containing the PRRAR epitope folds independent of the GGGGS linker. Superimposition of the SJ/FT and S protein 3D models ( Figure 6B ) shows that the overall conformation of the sequence containing the immunogenic epitopes is very well conserved with an observed RMSD of 0.912 angstrom and 171 amino acids out of 230 superimposing perfectly on the native S protein corresponding sequence.

The data we gathered suggest strongly that the antigen-presenting cells process P3-L, SJ/FT and P3/FUR/x 3 engineered vaccine candidates’ antigens, the same way they do with the SARS-CoV-2 native S protein. The data also support their use as candidate vaccinating antigens.

To assess the vaccinating potential of the engineered antigens in vivo, we immunized BALB/c mice with different concentrations of P3-L, SJ/FT, P3/FUR/x3 and P4/TMP/x3 ranging from 25 µg to 200 μg. We also used various adjuvants alone or in combination to determine the best adjuvant to use in vaccination. We immunized different groups of 5 mice each by administering 2 doses of the antigens within an interval of two weeks ( Supplementary Table 8 ). We collected sera from week 4 to week 14 following immunization. The results of ELISA performed with the sera collected following immunization of mice with 200 µg of the polypeptide P3/FUR/x3 showed the highest antibody response in the group of mice administered the CpG TLR9 ligand adjuvant ( Figure 7A ). In mice immunized using the P3-L and SJ/FT protein subunits, we observed the highest IgG titres with a combination of CpG and either one of the adjuvants alhydrogel or MPLA ( Figures 7B, C ). P4/TMP/x3 did not induce repeatedly a significant IgG response [data not shown]

We evaluated the neutralization capacity of two pool of sera samples collected from COVID-19 patients with respectively low and medium/high antibody titers against vaccinating antigens but in which antibodies against RBD antigens were not detected. The pool of sera with low antibody titer was from patients with the moderate form of the disease while the pool of sera with the medium/high antibody titer was from asymptomatic patients.

Neutralizing Ab measurements based on PRNT assay revealed that patient sera with medium to high titers against vaccinating antigens have the capacity to neutralize SARS-CoV-2 infectious virus (B.1) and demonstrated an average plaque reduction of 76.5% at 1:10 serum dilution ( Figure 8A ). Consistent with the low antibody levels against vaccinating antigens in some patient sera samples, we detected low neutralization activity against SARS-CoV-2 infectious with an average plaque reduction of 13.7% at 1:10 serum dilution ( Figure 8B ). The sera in this pool were from patient with the moderate form of the disease.

Next, we evaluated sera collected from vaccinated mice 28 days post immunization for neutralization capacity. Neutralizing Ab measurements based on PRNT assay revealed that the three vaccinating antigens elicited antibodies with efficacy to neutralize SARS-CoV-2 infectious virus. High NAb responses were detected at 1:10 dilution with an average plaque reduction of 90.7%. Sera from mice vaccinated with Ag1 (P3) showed the highest neutralization capacity with 95% plaque reduction at 1:10 dilution. Sera from mice vaccinated with Ag2 (P3-L) and Ag3 (SJ/FT) showed 90% and 87% plaque reduction, respectively. Notably, moderate plaque reduction was still detected at 1:160 serum dilution for all three samples ( Figure 8C ).

Through the development of this new approach for designing precision viral vaccines, we used computational tools to generate a data set that identified SARS-CoV-2 immunogenic epitopes in the virus structural region associated with infectivity. We experimentally linked these epitopes to protective immunity in patients with COVID-19. Based on these findings, we engineered vaccinating subunit antigens that we thoroughly tested in vivo by immunizing mice. Through this vaccination study, we determined the best antigen concentration and adjuvants that elicited a robust neutralizing antibody response.