First computational design using lambda-superstrings and in vivo validation of SARS-CoV-2 vaccine

As described in the “Methods” section, the problem of finding optimal weighted λ–superstrings with a maximum value of λ for a given length that can serve as CVs against SARS-CoV-2 was solved using an Integer Programming algorithm (Methods, “Optimization with CPLEX”). This algorithm was fed with three elements, the host string set \(H\), the target string set \(T\) and the weighting function \(w\), obtained as follows:

First, the set \(H\) of host strings was taken as the 22 distinct sequences corresponding to the Surface protein of SARS-COV-2 that appear in the Genbank²⁸ and GISAID²⁹ databases (Methods, “Extraction of the sequences”).

Next, we took as the set \(T\) of target strings (i.e., as potential epitopes), the 9-mers contained in some of the 22 host strings in positions corresponding to residues before the transmembrane domain.

Then, we assessed the weights of the epitopes using a function \(w\) in which the estimation of their immunogenicities and the estimation of the binding affinity to HLA-I was taken into account (Methods, “Weighting of the epitopes”).

Then, we used the algorithm to calculate a weighted λ–superstring with maximum λ for each length between 9 and 280. A scatterplot for the value of λ as a function of the length of the CV is shown in Fig. 1. It can be well fitted by a least square line with the regression line \(\uplambda =-0.579005+0.446982\cdot l\) , where \(l\) is the length of the candidate. The intercept and slope of the line were accurately determined, with a low standard error and a low P-value, as shown in Table 1. The R-squared value of the fit was 0.999668, and the closeness of this value to 1 indicates a good fit. Thus, each one-unit increase in length is associated approximately with an increase of 0.4 in λ all along the range from 9 to 280. Therefore, the map is robust and there is no significant loss in the λ increase per unit length in the considered interval of lengths.

Furthermore, we calculated the VaxiJen overall prediction for each CV (Methods, “Ranking the candidates with Vaxijen”). These optimal weighted λ-superstring, as well as the corresponding λ values and VaxiJen predictions, are shown in the table in Additional file 3. The threshold of 0.4 indicated in VaxiJen for the viral model was surpassed by the candidates with lengths of 22,24,67,68,69,70, and 175, as well as those with a length of at least 184 amino acids (candidates shown in green in the above-mentioned table).

Each λ-superstring can be naturally divided as it constitutes a union of a small number of peptides located in different regions of the protein. These peptides are enumerated, for each λ-superstring, in the fourth column of the table. When a peptide has some intersection with a domain of the protein, the domain is annotated next to the peptide. For the λ-superstrings with lengths from 176 to 183, the third peptide intersects two domains, namely NTD and RBD, and for those with lengths from 237 to 247, the fourth peptide also intersects the same two domains.

The two CVs with the maximum value in the VaxiJen overall predictions are shown in Table 2. The first CV (22 amino acids in length) is contained in the NTD domain, and the second CV (277 amino acids in length) can be divided, as previously described, as it originates a multipeptide of five peptides. In particular, the third and fifth peptides intersect the NTD and RBD domains, respectively, making them appropriate targets for vaccine development against SARS-CoV-2²⁷.

We selected the first CV in Table 2 for further biological assays because it showed the maximum value in the overall prediction in VaxiJen. This peptide STQDLFLPFFSNVTWFHAIHVS is 22 amino acids in length, and is contained in the NTD domain, therefore being a valid candidate antigen for vaccine development^{27, 41, 42}, with an overall prediction of 0.5545 in VaxiJen.

After analyzing our results computationally, we synthesized the 22-amino acid SARS-CoV-2-NTD peptide (designated here as CoVPSA) and performed in vivo experiments to test its immunogenicity and putative efficiency.

A first proof of concept was related with the immunogenicity of CoVPSA, and it was determined using a previously described procedure²² that evaluates the best immunogenic epitopes for preparing vaccines.

Safety was also examined by a cell viability assay after Trypan blue staining and apoptosis induction. Safety for DC vaccine vectors is considered as a percentage of cell viability higher than 95% and apoptosis induction lower than 7%–8%. Table 3 shows that DC vaccines loaded with CoVPSA peptides, or the unrelated bacterial peptide used as negative control, presented 98%–99% cell viability and lower than 4%–5% apoptosis. Therefore, we concluded that the DC vaccines loaded with peptides presented good safety profiles.

Next, we performed immunogenicity assays to measuring the DTH response of the vaccine vector. DCs were loaded with the peptide, then, mice were primed for 7 days intraperitoneally with COVID-19 peptides and then inoculated with the vaccine formulation (DC-CoVPSA) into the left hind footpads, with the non-inoculated right hind footpads acting as basal controls. Forty-eight hours later, we measured the DTH response as the swelling of the left hind footpads compared with the right hind footpads.

We also included empty DCs in these experiments and DCs loaded with a bacterial peptide unrelated to SARS-CoV-2 but with high CV efficiency against the bacterial pathogen²². Analysis of DTH responses (blue bars in Fig. 2) indicated that DCs loaded with COVID-19 designed peptide (DC-CoVPSA bars) elicited significantly stronger immune responses than DCs loaded with the control bacterial peptide (DC-CONT bars) or empty DCs (DC labelled bars). This may be explained by the fact that mice primed and DC-vaccinated with the same COVID-19 peptide produced high DTH responses, while mice primed and DC-vaccinated with different peptides were not able to elicit significant DTH responses (DC-control in Fig. 3). Next, we collected the popliteal lymph nodes and cultured them in vitro with 1 µg/mL of each peptide, CoVPSA, control peptide, or saline for 72 hours to examine the main immune cell populations by flow cytometry.

We observed that the highest percentages of immune cells corresponded to CD19⁺ cells (25,63%) that usually correspond to B cells, followed by MHC-II⁺ cells (27, 45%) that usually label DCs and macrophages, next CD4⁺ T cells (10,39%) and CD8⁺ T cells (14,61%).

The control peptide (CONT) produced no significant immune responses, as we observed only a small percentage of CD19⁺ cells (5,3%) and moderate numbers of MHC-II⁺ cells (13,5%) (DC-CONT bars in Fig. 2). Empty DCs (DC bars in Fig. 2) induced no significant numbers of immune cells.

These results indicated the clear induction of immune cells by DC vaccines loaded with CoVPSA peptide, with immune cells involved in antibody formation, such as B cells, DCs, and CD4⁺ T cells, being stimulated. While not predominant, cytotoxic immune responses caused by CD8⁺ T cells were also induced by DC vaccines loaded with CoVPSA peptide. These results were not surprising since CD4⁺ and CD8⁺ T cell epitopes are recovered from patients with mild and severe COVID-19 that are specific for the Spike protein²⁴.

A second proof of concept was related with the production of cytokines, either anti-viral cytokines, such as TNF-α, IFN-γ, IL-2, KC, and IL-12, or acute Th2 cytokines, such as IL-4, IL-6, MIP-2, or IL-10. The COVID-19 cytokine storm observed in patients with severe disease correlates with high levels of TNF-α, IL-6, IL-4, and IL-10, as well as with a clear deficiency in the production of IFN-related cytokines (i.e., IFN-α, IFN-γ, or IL-12)⁴³.

Our results in Fig. 3 show that DCs loaded with CoVPSA peptide produced mainly Th1-Th7 cytokines, IL-12, IL-17A, and IL-2. However, this DC-CoVPSA vaccine platform did not induce cytokines participating in the COVID-19 cytokine storm, such as IL-6, IL-10, or TNF-α (bars labelled with DC-CoVPSA in Fig. 3).

Interestingly, DC-CoVPSA vaccines induced high levels of IFN-γ (blue bars in Fig. 3) but barely detectable levels of MIP-2, an inflammatory cytokine that recruits inflammatory macrophages (grey bars in Fig. 3). The lack of significant levels of IL-4 (orange bars in Fig. 3) but high levels of IL-2 (red bars) strongly suggested the induction of Th1-Th17-type immune responses, but with no exacerbation of other cytokines, such as TFN-α or MIP-2.

In summary, the high levels of IFN-γ and especially IL-12, involved in vaccine efficiency and anti-viral responses, respectively, prompted us to suggest that CoVPSA peptide might function as an immunogenic epitope. CoVPSA peptide might be a good candidate to prepare vaccine platforms that induce not only antibody production but strong anti-viral T cell responses. We also confirmed these results in samples of human sera that served as a third proof of concept of our vaccine design. We recruited four asymptomatic patients with non-active COVID-19, four vaccinated volunteers (three with the Pfizer vaccine and one with the Moderna vaccine prepared against the RBD region of Spike protein), and four healthy donors that were non-vaccinated and tested negative in a COVID-19 antigen test. We collected blood from these 12 volunteers and compared the titers of different IgG COVID-19 antibodies as follows: (i) IgG antibodies able to neutralize the virus (neutral-RBD column in Fig. 4) were assessed by a neutralization antibody assay, (ii) anti-RBD antibodies that correspond to IgG antibodies recognizing the whole Spike protein including the RBD region binding to the ACE2 receptor (anti-RBD column), and (iii) anti-CoVPSA antibodies that reflect the IgG antibodies against our designed peptide in the NTD region (anti-CoVPSA column). As expected, COVID-19 asymptomatic patients presented medium and varied titers of IgG viral neutralization antibodies and low but significant levels of whole Spike protein anti-RBD antibodies, as previously reported^44,45,46. These COVID-19 patients also presented medium anti-CoVPSA IgG antibody titers (compare columns 1, 2 and 4 of Fig. 4). Interestingly, volunteers vaccinated with mRNA vaccines prepared against the RBD region of the Spike protein presented not only the highest titers of IgG viral neutralization antibodies that correlated with significant antibody titers against the RBD region^44,45,46, but also significant responses against the CoVPSA peptide. Analyses of the cytokine concentrations in the sera of these volunteers indicated that COVID-19 patients presented a storm cytokine pattern with high levels of IL-4, IL-6 and IL-8 and low or insignificant levels of IFN-γ, as previously published^{44,45,46,47,48,49,50}. Interestingly, vaccinated volunteers presented high levels of IFN-γ as well as significant levels of IL-2 and IL-4. This indicates that mRNA vaccines induce good antibody responses, as well as significant anti-viral cellular responses, as measured with neutralizing anti-RBD antibodies, antibodies against other Spike protein regions, such as the NTD region, and high levels of anti-viral cytokines, such as IFN-γ, IL-17A and IL-2.