Retrieval of the proteome sequence

An in-depth survey of the literature was carried out to find the protein necessary to constitute a subunit vaccine. An enterotoxin protein, Stx1 (GenBank: AQZ85745.1) of E. coli O157: H7 (str. EDL933) was selected, and its amino acid sequence was retrieved in FASTA format from the NCBI protein database (http://www.ncbi.org/). Immunogenic potential was assessed for extracellular Shiga toxin (Stx1), with a value of 0.5137 (VaxiJen v2.0) and antigenicity of 0.834022 (ANTIGENpro). The selected Shiga toxin was 315 amino acids long.

Detection and assessment of CTL, HTL, and LBL epitopes

Developing a sound vaccine requires accurate epitope predictions, which also lessen the need for exhaustive experimental testing. T-cell-dependent immunity is primarily triggered by T-cell epitopes. In bacterial infections, T-cells recognize specific peptide antigens, then multiply and transform, forming the basis of the primary immune response. We predicted six epitopes, each consisting of nine residues, evaluating them based on binding affinity scores with respective MHC I supertypes. Subsequently, the predicted epitopes were screened based on antigenicity, allergenicity, and toxicity. The two CTL epitopes (RLSSVLPDY, RTNNVFYRF) were selected for the final vaccine construct, as shown in Table 1. In humoral and cell-mediated immune responses, HTLs (T-lymphocytes) play a critical role, suggesting that HTL receptor-specific epitopes are vital for the development of immunotherapeutic and prophylactic vaccines. Initially, a total of 246 unique HTL epitopes of 15 residues each were identified, with only 6 HTL epitopes screened due to their cytokine-inducing properties, specifically IL-4, IL-10, and IFN-γ. Subsequently, three among these six HTL epitopes were found suitable for vaccine development based on antigenicity, allergenicity, and toxicity predictions (Table 2). B cell epitope prediction is crucial to induce strong antibody responses. Antibody production initiates when B cell receptors recognize epitopes, resulting in plasma and memory cell formation. Plasma cells produce antibodies during initial infections, while memory cells assist in future infections. Using the ABCPred server, linear B cell epitopes (LBL) in the target protein were identified for vaccine design. Epitopes with a score above the threshold of 0.51 were selected, resulting in a total of 32 predicted epitopes, each with a length of ten residues. Among these, two epitopes met the specified criteria and were selected for vaccine design, as shown in Table 3. The selected CTL and LBL epitopes were preferred as they exhibited 100% conservation and sequence identity across sequenced clinical isolates of E. coli O157:H7, high predicted antigenicity scores, and a protective profile (non-allergenic and non-toxic). Critically, the local sequence alignment using NCBI BLAST exhibited ≤ 30% identity with human proteins and human gut commensals (e.g., Lactobacillus species), reducing the risks of autoimmune cross-reactivity and gut microbiota dysbiosis.

Table 1 CTL epitopes chosen as the final components for the vaccine constructTable 2 HTL epitopes chosen as the final components for the vaccine constructTable 3 LBL epitopes chosen as the final components for the vaccine constructMulti-epitope mRNA vaccine construct design

To boost the immune response and improve the efficacy of the multi-epitope vaccine, an adjuvant is needed, as many subunit vaccines often lack robust immunogenicity. RpfE (O53177) was chosen as an adjuvant to improve the vaccine candidate’s immunogenicity (Islam et al. 2022c). The proposed structure of the mRNA vaccine construct is arranged from the N- to the C-terminus as follows: 5′ UTR, Kozak sequence, signal peptide (tPA), EAAAK linker, adjuvant (RpfE) linked by an “EAAAK” linker, followed by B cell epitopes connected with “KK” linkers, and HTL and CTL epitopes joined by “AAY” and “GPGPG” linkers, respectively (Fig. 2). Each linker aids in structural stability or immune response, helping in expression, immune processing, and presentation. The resulting vaccine is 184 amino acids long. The spatial arrangement of the vaccine construct was designed to increase expression, stability, and antigen processing. The N-terminal tPA signal peptide was integrated to direct the synthesized peptide to the host cell secretory pathway, aiding extracellular secretion. The adjuvant RpfE was placed at the N-terminus and separated from the antigen core by a rigid EAAAK linker to protect against steric hindrance and to protect individual folding domains. In this antigen core, HTL and CTL epitopes were separated using AAY and GPGPG linkers, respectively. These linkers are highly recognized by mammalian proteasomes and lysosomal cathepsins, ensuring efficient proteolytic cleavage and presentation of individual epitopes. Flexible KK linkers were positioned between the linear B-cell epitopes to maintain structural flexibility and facilitate optimum antibody recognition.

Fig. 2

Schematic representation of the study methodology and a graphical representation of the formulated multi-epitope vaccination architecture

Physicochemical and immunological profile

Stability and the ability to trigger an immune response are directly affected by the vaccine’s physicochemical attributes. The vaccine construct’s physicochemical parameters are displayed in Table 4. The molecular weight of the construct was found to be 19969.45 Da. The vaccine construct consists of 2780 atoms with a molecular formula of C882H1369N267O258S4. The theoretical isoelectric point (pI) of 10.31 indicated that the vaccine construct is notably of a basic nature. The half-life estimates were 1.9 h in mammalian reticulocyte cells (in vitro), over 20 h in yeast (in vivo), and over 10 h in E. coli (in vivo). An instability index of 31.24 indicated protein stability. The aliphatic index is a measure of the thermal stability of a protein. The calculated aliphatic index of 68.53 and a GRAVY score of -0.287 suggested that the vaccine is both thermostable and hydrophilic. The ANTIGENpro and VaxiJen v2.0 servers also predicted that the constructed vaccine exhibited strong antigenic properties (antigenicity, 0.9278) in a bacterial model. Moreover, it also showed no allergenic or toxic potential. These findings indicated that the multi-epitope mRNA vaccine developed here could emerge as an effective candidate for combating E. coli.

Table 4 Physicochemical and immunological characteristics of the vaccine constructSecondary and tertiary structure prediction and validation

The PSIPRED and SOPMA servers were used to anticipate the secondary structure of the vaccine. Sequence analysis revealed that 21.74% of the structure was composed of alpha-helices, corresponding to 40 residues. Random coils made up 72.28% of the sequence, consisting of 133 residues, and 5.98% of the vaccine’s construct showed a β-strand structure, including 11 amino acids (Table 5 & Fig. 3). The selected model (model 1) showed a GDT-HA score of 0.8940, RMSD of 0.537, MolProbity score of 2.496, clash score of 41.0, and poor rotamers score of 0.7. Before refinement, the Ramachandran plot showed 86.3% favorable residues, with 11.8% in allowed regions and 0.5% in disallowed regions. Upon refinement of Model 1, 91.5% of residues were positioned in the favorable region, 7.8% in the allowed regions, and 0.0% in the disallowed regions of the Ramachandran plot. The Z-score, which represents the overall structural quality of the model, was calculated using the ProSA-web server. The improved model had a Z-score of − 4.06, whereas the crude model had a Z-score of − 2.94 (Fig. 4). We utilized the Scratch protein predictor server to identify the most suitable tertiary structure template. The predicted tertiary structure was refined by the Galaxy-Refine server, and the model with the lowest C-score was selected based on server recommendations (Fig. 5).

Table 5 Vaccine construct’s secondary structural characteristicsFig. 3

Secondary structure elemental prediction of a proposed multi-epitope vaccine construct. (A) SOPMA secondary structure prediction; (B) PSIPRED secondary structure; (C) structural distribution percentage

Fig. 4

Analysis of Ramachandran plots by the PROCHECK server. (A) Before refinement; (B) After refinement; (C) Validation of the model with the Pro-SA server

Fig. 5

The 3D structure of the designed vaccine construct

Molecular docking studies

The docking of the vaccine (ligand) with TLR2 and TLR4 (receptor) aimed to predict binding affinity and interactions. Using ClusPro v2.0, ten docking configurations were generated. The complex with the lowest energy score and optimal binding interactions was selected. Model 7 and Model 0 met the criteria and were chosen as the best vaccine–TLR2 and vaccine–TLR4 complex, with energy scores of − 1096.2 and − 1110.5, respectively. These low energy scores suggest a high binding affinity and stable complex formation. The vaccine–TLR2 complex showed 20 hydrogen bonds, all classical, with Asp30, Arg95, Tyr97, Arg112, Gln402, Ala143, Ser427, Tyr144, Ser279, Asn252, Thr501, Glu76, Ser477, His453, Asn330, Asp354, Arg147, His352, and Arg140 from the vaccine involved in CHB interactions. Figure 6 shows the relevant TLR2 active site residues. Additional interactions included 4 electrostatic salt bridges, zero disulfide bonds, and 132 single non-bonded contacts. In the selected complex model 0 of the vaccine–TLR4 complex, 20 classical hydrogen bonds were found in the interaction surface. Vaccine residues Arg264, Glu178, Asn265, Arg159, Leu156, His159, Thr150, Arg234, Asp84, Asp60, Arg128, Pro107, Asn35, Glu76, Asp379, Tyr403, Arg112, Arg355, Phe103, Gln523, Ser62, Gln63, contributed to conventional hydrogen interactions (Fig. 7). Other interactions included 8 salt bridges, zero disulfide bonds, and 211 single non-bonded contacts.

Disulfide bond engineering of the constructed vaccine

Disulfide engineering identified 30 amino acid pairs as potential candidates for disulfide mutations using the Disulfide by Design v2.12 web server. After analyzing χ3 angles and energy scores, only two pairs met the criteria: χ3 angle between − 87° and + 97°, and energy score ≤ 2.2 kcal/mol. Mutations were generated for Phe113-Ala134 and Thr152-Ala155, with χ3 angles of + 97.69° and − 95.76° and energy scores of 1.70 and 0.96 kcal/mol, respectively (Fig. 8).

Normal mode analysis of vaccine-receptor docked complex

The deformability of residues in the vaccine-receptor docked complexes was assessed through normal mode analysis (NMA) using iMODS (Figs. 9 and 10). Deformability refers to how easily a protein can change its 3D structure, and peaks in the graph correspond to regions that can undergo significant structural changes.

Fig. 6

TLR2 (receptor) and the vaccine construct docked complex, and their interaction. (A) 3D structure of TLR2 (chain A, light purple and chain B, light green) and vaccine construct (pink colour, chain C) and docked complex; (B) interacting residue networks

Fig. 7

Vaccine construct (ligand) and TLR4 (receptor) docked complex and their interaction. (A) 3D structural model of the TLR4-vaccine complex; (B) close-up view of hydrogen bonds and salt bridge interactions at the contact interface

The B-factor graph comparing the PDB and NMA data showed that NMA predicted higher B-factors, indicating more flexibility than the experimentally obtained PDB values. The eigenvalue represented the energy needed to distort the protein structure, with lower eigenvalues confirming the flexibility and stability of the structure.

The variance graph, which is inversely related to the eigenvalue graph, uses red for individual variance and green for cumulative variance. The covariance matrix illustrated residue relationships, with red indicating correlated residues, white for uncorrelated, and blue for anti-correlated. The docked complex displayed stronger residue correlations, contributing to its stability.

Fig. 8

Disulfide bond engineering. (A) Stability of the vaccine protein before disulfide bond engineering in the original form (the form before substitution of amino acids by cysteine); (B) The mutant form (the form after substitution of amino acids by cysteine) with five pairs of disulfide bond formation. The disulfide pairs were shown by yellow sticks and pointed by black arrows

Fig. 9

Molecular dynamics and simulation analysis of the vaccine construct and the TLR2 receptor complex. (A) Vaccine construct and TLR2 docking complex (ClusPro); (B) deformability; (C) B-factor; (D) eigenvalue; (E) covariance; (F) elastic network (iMODS)

Fig. 10

Molecular dynamics and simulation analysis of the vaccine construct and the TLR4 receptor complex. (A) Vaccine construct and TLR4 docking complex (ClusPro); (B) deformability; (C) B-factor; (D) eigenvalue; (E) covariance; (F) elastic network (iMODS)

Immune response simulation

The immunological simulation study showed that the vaccine is predicted to effectively trigger an immune response similar to the natural immune system. Each of the three sequential injections strongly activated primary immune responses, with a gradual increase in primary response and activation of the secondary response. However, compared to the primary immune response, the secondary and tertiary responses were more robust (Fig. 11). These responses exhibited increased antibody levels (IgG1 + IgG2, IgM, and IgG + IgM), indicating that memory cells had developed, which led to faster antigen clearance upon subsequent exposures (Fig. 11A). Elevated levels of active B-cells, plasma B-cells, helper T-cells, and cytotoxic T-cells were observed, suggesting a strong immune memory and enhanced antigen clearance (Fig. 11B-D). In comparison to the Th1 immune reaction (Fig. 11I), the Th0 showed a lower percentage and fewer cells (cells/mm³). The vaccine also stimulated production of cytokines such as IFN-γ, IL-23, IL-10, and IFN-β, which are crucial for immune defense. Macrophage movement was enhanced during the presentation, with dendritic cell movement being predicted (Fig. 11F and H). Overall, the study predicted that the polyvalent vaccine may generate robust immune responses, including high levels of immunoglobulins, APCs, cytokines, and active B- and T-cells, indicating its promising potential.

Codon optimization and in silico cloning

Codon optimization tools were used to improve the translation of the mRNA vaccine construct within host cells. The GenSmart Codon Optimization tool (G.S.) was employed to optimize the sequence for efficient expression in human cells. The coding sequence (CDS) had 552 nucleotides, and the rare codon analysis tool from G.S. was used to evaluate the quality of the construct. The CAI value of 0.87 (Fig. 12A) was above the acceptable 0.8 threshold, suggesting efficient expression. The GC content was analyzed, with the ideal range being 30–70%, and the optimized construct had a GC content of 56%. These analyses indicated that the codon-optimized mRNA vaccine construct may be efficiently expressed in human cells. The free energy of the mRNA vaccine construct was assessed using the RNAfold server. The results indicated that the mRNA vaccine construct was stable with the MFE of -204.10 kcal/mol (Fig. 12C). In addition, the secondary centroid structure had a free energy of -211.91 kcal/mol (Fig. 12B). These findings suggested that the mRNA vaccine construct can be efficiently manufactured and is structurally stable, potentially enhancing its efficacy as a vaccine (Fig. 12D). Furthermore, the vaccine construct was specifically engineered to insert PstI and NruI restriction sites as the start and end cut points, accordingly, to insert the modified sequence into the pET-28a (+) vector. The final clone had a total of 5276 base pairs, with the target sequence covering 552 base pairs. The vector contributed the remaining base pairs, and the target sequence was marked in red in Fig. 13.

Fig. 11

Immune simulation showing primary, secondary, and tertiary immune responses to the vaccine construct. (A) Immunoglobulin production post-injection; (B) B-cell population after three injections; (C) B-cell states; (D) helper T-cell population; (E) helper T-cell states; (F) cytotoxic T-cell states; (G) macrophage states; (H) dendritic cell states; (I) cytokine/interleukin production and Simpson Index

Fig. 12

Vaccine mRNA codon and structure optimization. (A) CAI Score; (B) Optimal structure; (C) Centroid structure; (D) MFE mountain plot and thermodynamic ensemble; (E) Positional entropy (RNAfold)

Fig. 13

Computational modeling of the vaccine construct cloning into the pET-28a (+) vector