Computational design of a novel multi-epitope vaccine candidate against group A rotavirus

Abstract

Rotaviruses A (RVA) are the most common cause of diarrhea-related death in children under the age of five. Because RV vaccines are live attenuated, their use is limited. This work aimed to develop a multi-epitope vaccination against RVA using reverse vaccinology approaches. The viral protein 6 (VP6) was targeted for predicting B-cell and T-cell epitopes, and the best epitopes from its conserved regions were linked by appropriate linkers; additionally, 50 S ribosomal protein L7/L12 was inserted as an adjuvant to the vaccine’s N-terminus. The designed vaccine revealed satisfactory antigenicity, allergenicity, toxicity, and physicochemical characteristics. The molecular docking and molecular dynamics (MD) simulation showed strong binding interactions between the vaccine and toll-like receptor 4 (TLR4), signifying improved antigen presentation efficacy. The vaccine immunity simulation showed a significant rise in immunoglobulins and cytokines. Furthermore, the vaccine candidate showed a high likelihood of successful expression in Escherichia coli (E. coli). Our findings suggest that the multi-epitope vaccine candidate exhibits significant potential; however, experimental evaluations are necessary to determine its ability to stimulate the immune system.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12985-026-03099-0.

Introduction

Rotaviruses (RVs) are prevalent intestinal pathogens that can cause gastroenteritis in humans, animals, and birds [1]. It is a major cause of childhood illness and mortality worldwide, resulting in over 500,000 deaths per year and more than 200,000 deaths in low-income countries (LICs) [2]. The prevalence of diarrhea diminishes with advancing age [3]. RVs in sewage or contaminated water can infect humans and animals even at low levels (< 100 viral particles) due to their high infectivity [4]. Human infection may occur directly via the fecal-oral route or indirectly through the ingestion of untreated or partially treated water and contaminated raw food [5, 6]. RV disease symptoms range from transient loose stools to severe diarrhea and vomiting, which can cause dehydration, electrolyte abnormalities, shock, and death if not treated. In most cases, after a 1–3 day incubation period, the disease manifests abruptly, with fever and vomiting followed by watery diarrhea. Gastrointestinal symptoms typically resolve after 3 to 7 days, but may persist for 2 to 3 weeks [7]. RV antigens in stool specimens are identified using enzyme-linked immunosorbent assay (ELISA) and immunochromatography (ICG), whilst viral RNA is detected using reverse transcription polymerase chain reaction (RT-PCR), allowing for a more certain diagnosis of RV infection [8, 9].

RVs are members of the Sedoreoviridae family and are classified into nine species: RV A-D and F-J, or RVA-RVD and RVF-RVJ, based on the antigenic properties of their viral protein 6 (VP6) [10]. RVA is the predominant etiological cause of diarrhea-related morbidity and mortality in children under five years of age [11, 12]. The RV genome, encapsulated within a triple-layered icosahedral capsid, comprises double-stranded RNA (dsRNA) consisting of eleven segments that encode six structural proteins (VP1-VP4, VP6, and VP7) and five or six nonstructural proteins (NSP1-NSP5/6) [13]. Certain RVs possess an extra open reading frame for NSP6 within the NSP5-encoding segment [14]. The coding regions are surrounded by non-coding regions (NCRs) at the 5’ and 3’ ends of the genome segments. These regions contain nucleotide positions that are highly conserved among RVs of the same species [15]. The three capsid levels are: outermost (VP4 and VP7), middle (VP6), and innermost (VP2). Both VP7 and VP4 participate in viral entry and possess antigenic regions that elicit neutralizing antibody responses [16–18].

Vaccines are one of the most cost-effective public health strategies for severe infectious diseases. The ultimate goal of vaccine development is to create a viable candidate that can robustly elicit a strong humoral and cell-mediated immune response against a specific disease. Historically, the bulk of vaccines were created using heat-killed or attenuated microorganisms [19]. Recent advances in reverse vaccinology, immunoinformatics, and computational biology have the potential to significantly cut the time and cost required to manufacture safe and effective vaccines [20, 21]. Reverse vaccinology is a groundbreaking computational approach that uses genomic data to develop vaccine candidates without the requirement for pathogen cultivation. This approach, which analyzes protein sequences, identifies several epitopes that elicit both cellular and humoral immune responses while limiting side effects [22, 23].

Currently, two vaccines for RV have received approval from the Food and Drug Administration (FDA): RotaTeq^® (Merck & Co.) and Rotarix^® (GlaxoSmithKline [GSK]). These live attenuated vaccines, derived from various strains of RVs, have demonstrated substantial decreases in RV-associated mortality, severe RV disease, and hospital admissions [24]. However, their use is currently limited by several factors, including the inability to administer vaccines to immunocompromised individuals due to their live attenuated characteristics and the possibility of reversion to a pathogenic strain, age restrictions on usage, the risk of intussusception, the vaccine’s high cost, and the need for a cold chain [24, 25]. Ongoing research aims to address these limitations, currently emphasizing the development of innovative oral and injectable vaccines. Therefore, in this study, we aimed to design a multi-epitope vaccine against RVA by employing a combination of reverse vaccinology and immunoinformatics methodologies. The computational workflow utilized is illustrated in Fig. 1.

Fig. 1.

The procedure for developing an RVA multi-epitope vaccine candidate using a computational approach

Materials and methods

Identification, evaluation, retrieval, multiple sequence alignment, and phylogenetic analysis of target protein sequences

To develop multi-epitope vaccines, it is necessary to identify suitable antigens for epitope prediction. After all antigens have been identified, a systematic screening method with limits and filters was used to identify the best antigens [26]. The approach commenced with the identification of the protective antigen RVA using the Protegen database [27]. Protegen is a database and analytical system that organizes, maintains, and analyzes protective antigens. Fundamental antigen data and empirical evidence are compiled from peer-reviewed publications. Comprehensive gene and protein information is automatically retrieved from existing databases utilizing internally designed algorithms. The DeepLoc 2.0 server [28] was employed to determine the subcellular localization of the identified proteins. Because signal peptides are eliminated during post-translational modification (PTM), we must first identify and remove signal peptides from target proteins before predicting epitopes [29]. We used the SignalP 6.0 server [30] to analyze the protein sequences and identify possible signal peptides. The TMHMM 2.0 server [31] was used to determine the number of transmembrane helices in the identified proteins. Proteins with a lower number of transmembrane helices are easier to express and clone [32]; therefore, proteins with transmembrane helix ≤ 1 are appropriate for epitope prediction [33–35]. A fundamental need for epitope-based vaccine development is to ensure nonhomology to the host proteome in order to reduce the risk of autoimmune [36]. To evaluate the homology between the target antigen and the proteome of the host Homo sapiens (taxid: 9606), we used the BLASTp provided by the National Center for Biotechnology Information (NCBI). Target proteins must be antigenic, with no allergenic or toxic effects in the body. Therefore, the antigenicity, allergenicity, and toxicity of the identified proteins were evaluated using the VaxiJen v2.0 server [37–39], AllerTOP v2.1 server [40], and CSM-Toxin server [41], respectively. VaxiJen is the first server capable of predicting protective antigens independently of alignment. It was designed to facilitate antigen classification exclusively based on the physicochemical characteristics of proteins, independent of sequence alignment. To predict antigenicity using this server, the target organism and antigenicity threshold were set to “virus” and 0.4, respectively. The AllerTOP v2.1 server employs a method that utilizes the automatic cross-covariance (ACC) transformation to convert protein sequences into uniform vectors of equal length. CSM-Toxin server provides an in silico classifier for protein toxicity based solely on the primary sequence of proteins. This server effectively detects peptides and proteins with potential toxicity, attaining a Matthews Correlation Coefficient (MCC) of up to 0.66 in both cross-validation and various non-redundant blind tests. All accessible sequences for the target protein were acquired in FASTA format from the NCBI. Subsequently, multiple sequence alignment (MSA) was conducted using the Clustal Omega program [42] to identify conserved regions across all target protein sequences. We did this by selecting the sequence type as “Protein”, uploading the sequence file in TXT format to the server, and selecting the output format as “ClustalW with character counts”. Clustal Omega is a new MSA program that employs seeded guide trees and HMM profile-profile methodologies to produce alignments among three or more sequences. To investigate evolutionary relationships, all selected proteins were phylogenetically analyzed using the Molecular Evolutionary Genetics Analysis-X (MEGA-X) software [43]. To do this, the ClustalW command was conducted in MEGA-X software, and the phylogenetic tree was constructed using the Maximum Likelihood (ML) method with a Bootstrap Replication (BR) value of 1000 [44].

Epitope mapping and screening

A crucial element in combating infections is cell-mediated immune responses, particularly those that are mediated by cytotoxic T-cells (CTLs). Pathogen-derived antigens presented by major histocompatibility complex (MHC) class I molecules are identified in these responses. MHC class I molecules process and present the CTL epitopes [45]. Helper T-cells (HTLs) are crucial in the immune response to pathogen infections by promoting the production of interferon-gamma (IFN-γ). Furthermore, HTLs can trigger and maintain CTL responses and antibody synthesis [46]. The presentation of HTL epitopes by MHC class II molecules, which stimulate and activate CD4⁺ T cells, underscores the significance of HTL epitope prediction in vaccine development [47]. B-cells constitute the primary component of humoral immunity. Linear B cell epitopes are responsible for generating antigen-specific antibodies [48]. To predict all three types of epitopes, we used the conserved regions obtained from MSA as the target sequence. The MHC-I Binding Prediction tool from the Immune Epitope Database (IEDB) [49] was utilized for CTL epitope prediction. CTL epitopes for the HLA allele reference set were predicted by employing the NetMHCpan 4.1 EL approach [50]. The IEDB MHC-II Binding Prediction tool was used for predicting HTL epitopes. HTL epitopes showing binding affinity to the full HLA reference set were predicted with the NetMHCII pan 4.1 EL approach [50]. IEDB assigns a percentile rank to each CTL and HTL epitope, indicating its binding affinity to MHC Class I and MHC Class II molecules, respectively. The lower percentile rankings reveal that epitopes have a higher binding affinity. T-cell epitopes with a percentile rank ≤ 2 and binding affinity to at least two alleles were chosen for further screening. Utilizing the Bepipred Linear Epitope Prediction 2.0 approach [51], we predicted linear B-cell epitopes through the antibody epitope prediction tool provided by IEDB.

The antigenicity of all predicted epitopes was then assessed using the VaxiJen v2.0 server for the target organism “virus” at a threshold score of 0.4. The allergenicity of the epitopes was also checked using the AlgPred server [52]; the threshold value for epitope allergenicity in this prediction was set at 0.5. AlgPred server allows the prediction of allergens via a support vector machine (SVM) module based on amino acid or dipeptide composition. The toxicity of the epitopes was evaluated using the SVM (Swiss-Prot) based method in the ToxinPred server [53]. The ToxinPred server enables users to predict the toxicity of peptides and offers methods to detect mutations that may increase or decrease peptide toxicity. It generates all potential mutants of the specified peptide and predicts the toxicity of each mutant, together with essential physicochemical parameters like hydrophobicity, charge, isoelectric point (pI), etc. HTL cells produce cytokines such as interferon-gamma (IFN-γ) and interleukin-4 (IL-4), which activate cytotoxic T-cells, stimulate macrophages to phagocytize pathogens, and promote B cells to generate antibodies, rendering them highly effective in adaptive immunity, thereby enhancing adaptive immunity [54, 55]. Thus, alongside the criteria being evaluated, the HTLepitopes were assessed for their ability to induce IFN-γ and IL-4 production via the IFNepitope server [56] and IL4pred server [57], respectively. The prediction from the IFNepitope server was conducted on the model of IFN-γ versus other cytokines and the Hybrid approach that combines Motif and SVM. The IL4pred server utilized a combined approach of SVM and Motif for its predictions, applying an SVM threshold value of 0.2.

Population coverage analysis

Due to ethnic and regional differences, the expression and distribution of HLA alleles may differ worldwide. Therefore, in order to determine the efficacy of multi-epitope vaccines across varied global populations, it is essential to assess epitope population coverage [58]. Population coverage estimates for MHC class I, MHC class II, and combination classes across 16 distinct continents and worldwide were meticulously computed using the IEDB population coverage tool [59]. For each class, selected epitopes were uploaded to the server, and the associated alleles were selected.

Design of multi-epitope vaccine construct

To prevent duplication in the multi-epitope vaccine composition, we eliminated epitopes that overlapped with other epitopes. The multi-epitope vaccine was designed by fusing the selected epitopes with an appropriate adjuvant using linkers. The 50 S ribosomal protein L7/L12 (Locus RL7_MYCTU), with accession number P9WHE3 (sequence: MAKLSTDELLDAFKEMTLLELSDFVKKFEETFEVTAAAPVAVAAAGAAPAGAAVEAAEEQSEFDVILEAAGDKKIGVIKVVREIVSGLGLKEAKDLVDGAPKPLLEKVAKEAADEAKAKLEAAGATVTVK), was added to the N-terminal end of the vaccine construct as an adjuvant to enhance the immunogenicity of the vaccine candidate. The 50 S ribosomal protein L7/L12 (toll-like receptor 4 (TLR4) agonist) regulates dendritic cell (DC) maturation and proinflammatory cytokine production (tumor necrosis factor-α (TNF-α), IL1β, and IL-6) via the TLR4/MyD88 signaling pathway [60]. Furthermore, subsequent to naive T cell activation, there is polarization of CD4⁺ and CD8⁺ T cells, induction of T cell-mediated cytotoxicity, and activation of IFN-γ-producing cells [61]. A variety of linkers were employed to interconnect the components of the vaccine construct, including the EAAAK (Glu-Ala-Ala-Ala-Lys) linker for the first CTL epitope and the adjuvant, AAY (Ala-Ala-Tyr) linkers for subsequent CTL epitopes, GPGPG (Gly-Pro-Gly-Pro-Gly) linkers for the HTL epitope, and the KK (Lys-Lys) linker for the linear B-cell epitope and the final HTL epitope. The fundamental function of the EAAAK linker is to allow for flexibility and separation of the adjuvant and antigenic components, hence retaining their structural integrity and immunogenicity [62, 63]. The AAY linker functions as a proteasome cleavage site in mammalian cells, boosting the vaccine’s immunogenicity and promoting the formation of natural epitopes [64, 65]. The GPGPG linker, consisting of tiny amino acids like glycine and serine, ensures sufficient separation between epitope domains while restricting junctional epitope formation, hence facilitating the mobility of protein domains and promoting proper protein folding [66, 67]. The KK linker acts as a cleavage site for the lysosomal protease cathepsin B, which is crucial for MHC-II-mediated antigen processing [68]. A 6xHis tag was appended to the C-terminal end of the vaccine to improve vaccine purification and identification [69].

Determination of antigenicity, allergenicity, toxicity, solubility, and physicochemical properties of the vaccine

The antigenicity of our vaccine candidate was predicted utilizing the VaxiJen v2.0 server with a threshold value of 0.4, as well as the ANTIGENpro server [70]. ANTIGENpro is a protein antigenicity predictor that operates on sequences, does not require alignment, and is pathogen-independent. In this server, the predictions are made by a two-stage architecture based on multiple representations of the primary sequence and five machine learning algorithms. The AllerTOP v2.1 and AllergenFP v1.1 [71] servers were employed to evaluate the allergenic potential of the designed vaccine. The AllergenFP v1.1 server uses a method in which the amino acids in protein sequences are described by five E-descriptors and then transformed into uniform vectors using ACC transformation. The vaccine’s toxicity was evaluated using the CSM-Toxin server. We used the SOLpro [72] and Protein-Sol [73] servers to assess vaccine solubility. The SOLpro predicts a protein’s solubility upon overexpression in Escherichia coli (E. coli) via a two-stage SVM approach that incorporates various representations of the main sequence. In the Protein-Sol server, the population average for the experimental dataset (PopAvrSol) is 0.45; hence, any scaled solubility value more than 0.45 is predicted to be more soluble than the average soluble E. coli protein. Furthermore, various physicochemical characteristics of the multi-epitope vaccine, including molecular weight, theoretical isoelectric point, formula, half-life, instability index, aliphatic index, and grand average of hydropathicity (GRAVY), were determined using the Expasy ProtParam tool [74].

Secondary structure analysis of the vaccine construct

To precisely estimate the distribution of alpha helices, extended strands, and random coils in the vaccine’s secondary structure, we used the Prabi server. This server employs the GOR IV approach [75] to predict the secondary structure of proteins. The GOR IV uses all potential pair frequencies within a window of 17 amino acid residues and has a mean accuracy of 64.4%.

3D model prediction, refinement, and validation of the vaccine construct

The 3D model of the vaccine construct was predicted using Iterative Threading ASSEmbly Refinement (I-TASSER) server [76–78]. Upon submission of an amino acid sequence, the I-TASSER server first attempts to obtain template proteins with analogous folds from the PDB library using LOMETS, a locally implemented meta-threading method. In the second stage, replica-exchange Monte Carlo simulations are used to reassemble the continuous fragments excised from the PDB templates into full-length models, with threaded unaligned sections (mostly loops) created using ab initio modeling. If LOMETS does not identify an adequate template, I-TASSER will construct the entire structure by ab initio modeling. The third step involves re-running the fragment assembly simulations from the SPICKER cluster centroids using the spatial constraints derived from the LOMETS templates and the PDB structures via TM-alignment. The decoys created in the second simulation are then clustered, and the structures with the lowest energy are chosen. We used the GalaxyRefine server [79] to improve the predicted 3D structure to make it more similar to native protein structures. The GalaxyRefine server employs a refinement approach that was effectively validated in CASP10. This method initially reconstructs side chains, followed by sidechain repacking and overall structural relaxation through molecular dynamics (MD) simulation. To assess the quality of the initial and refined models of the vaccine’s three-dimensional structure, the PROCHECK tool [80, 81], which is available on the SAVES v6.1 server, and the ProSA-web server [82, 83] were used. The PROCHECK thoroughly evaluates a protein structure’s stereochemistry. The ProSA-web server provides a Z-score for a specified input structure. The Z-score assesses the model’s overall quality and is presented in a graphic featuring the Z-scores of all experimentally determined protein chains. Diverse colors signify separate groupings of structures in this figure, sourced from multiple modalities, including X-ray and NMR. This server is employed to ascertain whether the Z-score of the specified structure lies within the range typically observed for native proteins of similar size.

Discontinuous B‑cell epitopes prediction

The ElliPro tool from IEDB [84] was used to identify discontinuous B cell epitopes in the vaccine’s refined 3D structure. The ElliPro assigns a score to each predicted epitope, which is defined as a PI (Protrusion Index) value averaged over epitope residues. The prediction was based on a minimum score of 0.5 and a maximum distance of 6 angstroms (Å).

Disulfide engineering of the vaccine candidate

Disulfide bridges are crucial in vaccine development since they provide significant stability to the vaccine construct [85]. Disulfide by Design 2 v2.13 server [86, 87] was utilized to conduct disulfide engineering on the vaccine’s refined three-dimensional structure. This server can readily detect residues containing cysteine mutations or those capable of forming disulfide bridges.

Molecular Docking analysis

Using the Cluspro 2.0 server [88–92], we conducted ligand-receptor docking research to assess the vaccine’s interaction with the human TLR4. The TLR4 (4G8A) was downloaded in PDB format from the RCSB Protein Data Bank (RCSB PDB) [93]. Before docking, ligands were prepared using the UCSF Chimera 1.10.2 software [94], and co-crystallized ligands and water molecules were eliminated. Docking was completed with the default parameters. The docked complex with significant affinity was visualized using the Chimera 1.10.2 software. The PDBsum server [95] was utilized to analyze the interactions between the vaccine and TLR4.

Molecular dynamics simulation

The selected docked complex was subjected to 250 nanoseconds (ns) of MD simulation using the assisted model building with energy refinement (AMBER) v.20 program’s SANDER module [96]. The inhibitor settings and intricate molecular interactions were defined using AMBER’s leap and antechamber modules. The complex was immersed in a TIP3P water box with a defined margin distance of 12 Å between the complex and the box’s boundary. The system was minimized, heated, and equilibrated, followed by a production run in the NPT ensemble at 300 K and 1.0 atm with a time step of 2 fs. The SHAKE method [97] was employed to constrain all hydrogen bonds (H-bonds), while Langevin dynamics facilitated control of temperature. We performed two independent MD simulations of the selected docked complex with different initial velocities (Ig = 109858 and 118245) to validate the complex’s stability. Each simulation was also replicated to further validate the results.

We also used the iMod server (iMODS) [98–100] to further analyze the MD and stability of the selected docked complex. The iMODS simulates macromolecule vibrations, motion, and trajectories while projecting molecular dynamics. Normal mode analysis (NMA) is used by the server to determine the protein’s internal coordinates for stability assessment. NMA in internal (dihedral) coordinates readily reproduces the collective functional movements of biological macromolecules. The PDB file of the selected docked complex was sent to iMODS, and the results were obtained using the default values for all parameters.

Binding free energy analysis

The binding free energies of the vaccine-TLR4 complex were computed using the molecular mechanics Poisson–Boltzmann surface area (MM-PBSA) and molecular mechanics generalized Born surface area (MM-GBSA) approaches. This was achieved using the MMPBSA.py package [101] of the AMBER v.20 program. For the binding energy calculations, 100 frames were selected and evaluated at consistent intervals from simulated trajectories.

Immune response simulation

The C-ImmSim server [102] was used to assess the immunogenic potential of the proposed vaccine candidate. In this server, a novel, rapid Position Specific Scoring Matrix (PSSM) based approach was employed, with little compromises for performance prediction. In compliance with the recommendation for a minimum four-week gap between vaccine doses, three doses were administered during the simulation, with a four-week gap between each [103]. In this prediction, the simulation steps were set to 1050 (350 days). The time steps were set to 1, 84, and 168, with each interval equivalent to 8 h in real life. Other parameters were left at their default state. C-ImmSim, while a useful tool for in silico immunological simulations, has limitations. As with many in silico models, it simplifies the immune system’s complicated interconnections. The model may not accurately represent the intricate connections between immune cell types, cytokine signaling, and other biological processes. This simplification can result in contradictions between simulation results and real-world immune responses, particularly when the immune responses are complex or diverse [104, 105]. In silico models do not fully account for the impact of environmental factors on the immune system. Diet, lifestyle, and exposure to allergens, toxins, and pathogens all influence immune responses, and their absence can limit simulation accuracy [106, 107].

Codon optimization and in Silico cloning of the vaccine candidate

The Java Codon Adaptation Tool (JCat) [108] was employed for the translation of protein to nucleotide sequences and the optimization of codons in vaccine constructions, with the E. coli K 12 strain as the host organism. This tool provides the codon adaptation index (CAI) value and guanine-cytosine (GC) content to assess vaccine expression. Restriction enzyme sites for HindIII and BamHI were added at the N- and C-terminals of the optimized nucleotide sequence, respectively. The optimized gene was ultimately inserted into the multiple cloning site (MCS) of the pET28a (+) expression vector using SnapGene 3.2.1 software.

Results

Identification, evaluation, retrieval, multiple sequence alignment, and phylogenetic analysis of target protein sequences

According to the Protegen database, the VP6 protein was selected as the target protein for epitope prediction. The DeepLoc 2.0 server predicted that the VP6 protein is localized in the virus’s plasma membrane. The SignalP 6.0 server indicated that the VP6 protein is devoid of a signal peptide (Supplementary Figure S1). On the other hand, the TMHMM 2.0 server predicted that the VP6 protein lacks the transmembrane helix (Supplementary Figure S2). The BLASTp analysis indicated that the VP6 protein has no similarity to the human proteome and is considered as a non-homologous protein. VaxiJen v2.0, AllerTOP v2.1, and CSM-Toxin servers predicted the nature of the VP6 protein to be antigenic (with a score of 0.4957), non-allergenic, and non-toxic, respectively. A total of 82 appropriate sequences for the VP6 protein from various RVA strains were identified. Clustal Omega then executed MSA to discern the conserved portions of the target proteins (Supplementary File). The conserved areas were identified based on the absence of gaps in the protein sequence and the maximum number of similar amino acids. The phylogenetic tree was constructed using the VP6 protein sequences of different strains of the RVA species (Supplementary Figure S3). The results indicated that all VP6 protein sequences have been clustered into three distinct groups and have close relationships with one another.

Epitope mapping and screening

Following the identification of the conserved regions of the VP6 protein, a comprehensive epitope prediction and screening procedure was conducted. Antigenic epitopes devoid of toxicity and allergenicity were chosen. The capacity to induce the synthesis of IFN-γ and IL-4 of HTL epitopes was evaluated. This study finally identified 11 CTL epitopes (Table 1), 2 HTL epitopes (Table 2), and 2 linear B-cell epitopes (Table 3) for vaccine design.

Table 1.

A list of predicted CTL epitopes. The selected epitopes are highlighted in bold

CTL epitope	Allele	VaxiJen score	AlgPred score	Toxicity
AHDNLMGTMW	HLA-B*44:02, HLA-B*44:03, HLA-B*58:01, HLA-A*24:02	0.0788	0.33	Non-Toxin
EFLLNGQII	HLA-A*24:02, HLA-A*23:01	0.2596	0.3	Non-Toxin
FLLNGQIINT	HLA-A*02:01, HLA-A*02:03, HLA-A*02:06	0.3239	0.29	Non-Toxin
FPYSASFTL	HLA-B*35:01, HLA-A*26:01, HLA-A*68:02, HLA-B*08:01, HLA-B*07:02, HLA-B*53:01, HLA-B*51:01	0.9064	0.35	Non-Toxin
FPYSASFTLN	HLA-B*51:01, HLA-B*07:02, HLA-B*53:01, HLA-B*35:01	0.8573	0.36	Non-Toxin
FTLNRSQPA	HLA-A*02:06, HLA-B*08:01, HLA-A*68:02	0.2115	0.41	Non-Toxin
FTVASIRSM	HLA-A*26:01, HLA-B*51:01, HLA-A*32:01, HLA-B*53:01, HLA-B*15:01, HLA-A*02:06, HLA-B*58:01, HLA-B*57:01, HLA-B*35:01, HLA-A*68:02	1.0626	0.24	Non-Toxin
FTVASIRSML	HLA-A*68:02, HLA-A*26:01	0.5901	0.19	Non-Toxin
GLLGTTLLNL	HLA-A*02:01, HLA-A*02:06, HLA-A*02:03	1.1525	0.36	Non-Toxin
GPVFPPGMNW	HLA-B*53:01, HLA-B*15:01, HLA-A*30:02, HLA-A*23:01, HLA-B*44:03, HLA-B*51:01, HLA-B*44:02, HLA-A*26:01, HLA-B*07:02, HLA-B*35:01, HLA-B*58:01, HLA-B*57:01, HLA-A*32:01	-0.0981	0.38	Non-Toxin
HDNLMGTMW	HLA-B*44:02, HLA-B*57:01, HLA-B*53:01, HLA-B*58:01, HLA-B*44:03	0.1227	0.31	Non-Toxin
IFPYSASFTL	HLA-B*51:01, HLA-B*35:01, HLA-B*53:01, HLA-A*23:01, HLA-B*07:02, HLA-A*24:02	0.7051	0.32	Non-Toxin
IINTYQARF	HLA-A*32:01, HLA-B*53:01, HLA-B*57:01, HLA-A*30:02, HLA-A*26:01, HLA-A*24:02, HLA-A*23:01, HLA-B*58:01, HLA-B*15:01	0.2540	0.34	Non-Toxin
LITNYSPSR	HLA-A*68:01, HLA-A*31:01, HLA-A*33:01	-0.0408	0.34	Non-Toxin
LLGTTLLNL	HLA-A*02:01, HLA-A*02:06, HLA-A*02:03	0.8640	0.36	Non-Toxin
LLNGQIINT	HLA-A*02:03, HLA-A*02:06, HLA-A*02:01	0.2277	0.3	Non-Toxin
LLNGQIINTY	HLA-B*15:01, HLA-A*03:01, HLA-A*32:01, HLA-A*26:01, HLA-A*01:01, HLA-A*30:02	0.0377	0.29	Non-Toxin
LNGQIINTY	HLA-A*30:02, HLA-A*01:01, HLA-A*26:01, HLA-B*58:01, HLA-B*35:01, HLA-B*15:01	-0.0259	0.28	Non-Toxin
NIFPYSASF	HLA-A*26:01, HLA-B*07:02, HLA-B*58:01, HLA-A*30:02, HLA-A*02:06, HLA-B*51:01, HLA-A*68:02, HLA-A*24:02, HLA-B*08:01, HLA-A*23:01, HLA-B*53:01, HLA-B*15:01, HLA-B*35:01, HLA-A*32:01	0.5373	0.32	Non-Toxin
NSADGATTW	HLA-B*58:01, HLA-A*24:02, HLA-A*23:01, HLA-B*44:03, HLA-A*01:01, HLA-B*44:02, HLA-B*35:01, HLA-A*26:01, HLA-A*32:01, HLA-B*57:01, HLA-B*53:01	0.3402	0.34	Non-Toxin
NYSPSREDNL	HLA-A*24:02, HLA-A*23:01	0.8388	0.32	Non-Toxin
PVFPPGMNW	HLA-B*58:01, HLA-B*44:02, HLA-A*24:02, HLA-A*26:01, HLA-A*23:01, HLA-B*53:01, HLA-B*57:01, HLA-A*32:01	0.0499	0.37	Non-Toxin
QIINTYQAR	HLA-A*33:01, HLA-A*26:01, HLA-A*11:01, HLA-A*31:01, HLA-A*68:01	0.2883	0.31	Non-Toxin
QIINTYQARF	HLA-A*26:01, HLA-B*15:01	0.2151	0.34	Non-Toxin
RSQPAHDNL	HLA-B*58:01, HLA-A*24:02, HLA-A*30:02, HLA-B*07:02, HLA-A*30:01, HLA-B*57:01, HLA-A*32:01	0.5058	0.34	Non-Toxin
RSQPAHDNLM	HLA-B*58:01, HLA-A*30:02, HLA-B*57:01	0.3845	0.3	Non-Toxin
SPSREDNLQR	HLA-A*31:01, HLA-A*68:01	-0.1825	0.35	Non-Toxin
SQPAHDNLM	HLA-B*15:01, HLA-B*40:01	0.0488	0.28	Non-Toxin
TLNRSQPAH	HLA-A*03:01, HLA-A*30:02, HLA-B*15:01	0.2432	0.36	Non-Toxin
TVASIRSML	HLA-A*68:02, HLA-A*02:06, HLA-A*02:03, HLA-A*32:01, HLA-B*07:02, HLA-A*26:01	0.3326	0.22	Non-Toxin
YSASFTLNR	HLA-A*68:01, HLA-A*01:01, HLA-A*03:01, HLA-A*31:01, HLA-A*33:01, HLA-A*11:01	0.4322	0.33	Non-Toxin

Table 2.

A list of predicted HTL epitopes. The selected epitopes are highlighted in bold

HTL epitope	Allele	VaxiJen score	AlgPred score	Toxicity	FN-γ inducer	IL-4 inducer
ASFTLNRSQPAHDNL	HLA-DRB3*02:02, HLA-DRB1*04:01, HLA-DRB1*13:02	0.4114	0.36	Non-Toxin	Positive	Non-IL4-inducer
FPYSASFTLNRSQPA	HLA-DRB3*02:02, HLA-DQA1*04:01/DQB1*04:02	0.5709	0.37	Non-Toxin	Positive	IL4-inducer
LITNYSPSREDNLQR	HLA-DRB1*09:01	-0.0201	0.33	Non-Toxin	Positive	Non-IL4-inducer
LNGQIINTYQARFGT	HLA-DRB1*15:01, HLA-DRB1*12:01, HLA-DQA1*04:01/DQB1*04:02	0.3279	0.32	Non-Toxin	Positive	Non-IL4-inducer
PYSASFTLNRSQPAH	HLA-DRB3*02:02, HLA-DRB1*13:02, HLA-DQA1*04:01/DQB1*04:02	0.5704	0.33	Non-Toxin	Positive	IL4-inducer
SASFTLNRSQPAHDN	HLA-DRB3*02:02, HLA-DRB3*01:01, HLA-DRB1*08:02, HLA-DRB1*07:01, HLA-DRB5*01:01, HLA-DQA1*04:01/DQB1*04:02, HLA-DRB1*01:01, HLA-DRB1*04:01, HLA-DRB1*13:02	0.4128	0.29	Non-Toxin	Positive	Non-IL4-inducer
YSASFTLNRSQPAHD	HLA-DRB3*02:02, HLA-DRB5*01:01, HLA-DRB1*07:01, HLA-DRB1*01:01, HLA-DQA1*04:01/DQB1*04:02, HLA-DRB1*04:01, HLA-DRB1*13:02	0.3767	0.31	Non-Toxin	Positive	Non-IL4-inducer

Table 3.

A list of predicted linear B-cell epitopes. The selected epitopes are highlighted in bold

linear B-cell epitope	Length	VaxiJen score	Allergenicity score	Toxicity
SQRNGIAP	8	1.1633	0.41	Non-Toxin
LNRSQPAHDNLM	12	0.1421	0.29	Non-Toxin
YSPSREDN	8	0.5984	0.32	Non-Toxin

Population coverage analysis

Population coverage was assessed for selected CTL and HTL epitopes, individually and in combination (Fig. 2). The evaluation of population coverage revealed that the CTL and HTL epitopes each included 97.77% and 23.14% of the global population, respectively. Nonetheless, when these epitopes were utilized in combination, the population coverage escalated to 98.43%. The population coverage of the chosen CTL epitopes is maximal in Europe (99.39%) and minimal in Central America (7.76%) (Supplementary Table S1). The chosen HTL epitopes have the highest population coverage in South America (49.16%) and the lowest in Southeast Asia (5.48%) (Supplementary Table S2). In all geographic regions, excluding Central America (41.07%), the combined chosen CTL and HTL epitopes exhibited population coverage exceeding 880% (Supplementary Table S3).

Fig. 2.

Population coverage percentages for MHC I and MHC II epitopes, both individually and combined

Design of multi-epitope vaccine construct

Our multi-epitope vaccine was designed by combining 6 CTL epitopes, 2 HTL epitopes, and 1 linear B-cell epitope with 1 adjuvant sequence (50 S ribosomal protein L7/L12 (Locus RL7_MYCTU)), 1 EAAAK linker, 5 AAY linkers, 2 GPGPG linkers, 1 KK linker, and 1 6xHis tag (Fig. 3). The 264 amino acids make up the multi-epitope vaccine candidate.

Fig. 3.

The schematic representation of the multi-epitope vaccine candidate. The multi-epitope vaccine consists of 1 adjuvant (black), 6 CTL epitopes (cyan), 2 HTL epitopes (red), 1 linear B-cell epitope (violet), linkers (1 EAAAK, 5 AYY, 2 GPGPG, 1 KK), and 1 6xHis tag

Determination of antigenicity, allergenicity, toxicity, solubility, and physicochemical properties of the vaccine

The vaccine candidate achieved a VaxiJen score of 0.5126 for the virus model and 0.4923 for the bacteria model, with a threshold of 0.4; its antigenicity was evaluated at 0.8162 by the ANTIGENpro server. Analysis on the AllerTOP v2.1 and AllergenFP v1.1 servers showed that the designed vaccine is probably a non-allergen. Furthermore, the CSM-Toxin server predicted that the engineered vaccine is non-toxic. In the solubility analysis, the Protein-Sol server predicted a solubility score of 0.454, which is greater than 0.45 and indicates higher solubility than typical E. coli proteins (Fig. 4). On the other hand, using the SOLpro server found that the vaccine is soluble, with a solubility value of 0.954574. Table 4 presents the findings from the Expasy ProtParam tool about the physicochemical characteristics of the proposed vaccine.

Fig. 4.

The solubility plot of the multi-epitope vaccine. The solubility score is more than 0.454, indicating greater solubility than typical E. coli proteins

Table 4.

The physicochemical characteristics of the multi-epitope vaccine candidate

Parameters	Assessment
Molecular weight	27.85 kDa
Theoretical pI	6.07
Total number of negatively charged residues (Asp + Glu)	29
Total number of positively charged residues (Arg + Lys)	25
Formula	C1256H1954N332O378S3
Total number of atoms	3923
Half-life	30 h (mammalian reticulocytes, in vitro) > 20 h (yeast, in vivo) > 10 h (E. coli, in vivo)
Instability index	37.69
Aliphatic index	82.69
GRAVY	-0.089

Secondary structure analysis of the vaccine construct

The Prabi server determined that 147 (55.68%), 21 (7.95%), and 96 (36.36%) amino acids in the vaccine structure formed the alpha helices, extended strands, and random coils, respectively (Fig. 5).

Fig. 5.

The multi-epitope vaccine construct’s secondary structure includes alpha helices (h), extended strands (e), and random coils (c)

3D model prediction, refinement, and validation of the vaccine construct

The I-TASSER server predicted five three-dimensional structures for the vaccine candidate, with C-scores ranging from − 2.98 to -4.54, using different threading templates (PDB IDs: 1RQU, 8UGC, 2FTC, 1DD4, 8HC0, 1DD3). The C-score generally ranges from [-5, 2], with higher values indicating greater model confidence and lower values indicating the opposite; Model 2, with a C-score of -2.98, was chosen according to the aforementioned factor (Fig. 6). The selected model was refined using the GalaxyRefine server. The GalaxyRefine server provided five refined models, which were evaluated using a variety of quality evaluation criteria, including GDT-HA, RMSD, MolProbity, Clash score, Poor rotamers, and Rama favored (Supplementary Table S4). Refined model 2 (Fig. 6) was chosen for further investigation due to its high GDT-HA (0.9536) and Rama favored (94.3) scores, as well as low RMSD (0.408), MolProbity (2.552), Clash score (27.0), and Poor Rotamers (2.0). According to the Ramachandran plot analysis, 78.5%, 17.1%, 3.5%, and 0.9% of residues were identified in most favoured, additional allowed, generously allowed, and disallowed regions, respectively, in the initial 3D model (Fig. 7A). In the refined 3D model, 91.2%, 7.9%, 0.9%, and 0.0% of residues were found in the most favoured, additional allowed, generously allowed, and disallowed regions, respectively (Fig. 7B). The initial model’s ProSA Z-score of -2.92 changed to -3.53 after refining (Fig. 7C, D).

Fig. 6.

The 3D models for the multi-epitope vaccine. The initial 3D model is depicted in purple, while the refined 3D model is shown in yellow. We stacked the two models to show the similarities and differences between them

Fig. 7.

The examination of the quality of the multi-epitope vaccine’s initial and refined 3D models. The Ramachandran plot shows that the proportion of residues in the most favored regions increased from 78.5% in the initial 3D model (A) to 91.2% in the refined 3D model (B). The initial 3D model has a ProSA Z-score of -2.92 (C), while the refined 3D model has a ProSA Z-score of -3.53 (D)

Discontinuous B‑cell epitopes prediction

The ElliPro tool identified three discontinuous B-cell epitopes in the vaccine’s refined 3D structure, ranging in size from 45 to 48 residues and scoring from 0.679 to 0.725 (Fig. 8, Supplementary Table S5).

Fig. 8.

The predicted discontinuous B-cell epitopes in the vaccine’s refined three-dimensional structure. (A) Discontinuous B-cell epitope with 48 residues and a score of 0.725. (B) Discontinuous B-cell epitope with 45 residues and a score of 0.702. (C) Discontinuous B-cell epitope with 45 residues and a score of 0.679. Discontinuous B-cell epitopes are shown as purple spheres, while other vaccine residues are shown as gray sticks

Disulfide engineering of the vaccine candidate

The Disulfide by Design 2 v2.13 server evaluated the vaccine’s refined 3D structure and identified 34 residue pairs capable of forming disulfide bonds (Supplementary Table S6). Peaks of χ3 in 1505 native disulfide links among 331 non-homologous proteins have been recorded at -87 and + 97 degrees, and roughly 90% of naturally occurring disulfide bonds exhibit an energy value below 2.2 kcal/mol [86]. According to the mentioned parameters, only three residue pairings were selected for disulfide bond formation: TYR159-GLY213, TYR176-ALA237, and ALA196-ALA257 (Fig. 9).

Fig. 9.

The disulfide engineering of the vaccine’s refined 3D structure. (A) The wild type. (B) The mutant type, with the three disulfide bonds illustrated by yellow sticks

Molecular Docking analysis

The interaction between the vaccine’s refined 3D structure and TLR4 was investigated using docking analysis on the Cluspro 2.0 server. It predicted 30 poses for the vaccine-TLR4 docked complex (Supplementary Table S7). Among these predicted docking results, cluster 0 with more members (75) and a higher negative energy score (-986.4 kcal/mol) was chosen for further investigation, and the selected complex was displayed using Chimera 1.10.2 software (Fig. 10A). The PDBsum server indicated that in the chosen vaccine-TLR4 docked complex, 32 residues from the TLR4 chain B interacted with 22 residues from the vaccine, whereas 7 residues from the TLR4 chain D interacted with 6 residues from the vaccine (Fig. 10B). The amino acid LYS91 from the TLR4 chain D established a H-bond with amino acid PRO230 from the vaccine, measuring 2.66 Å in distance. The amino acids involved in the H-bond interactions between the vaccine and TLR4 chain B, together with their distances in the docked vaccine-TLR4 complex, are enumerated in Table 5.

Fig. 10.

Molecular docking analysis of the multi-epitope vaccine construct with TLR4. (A) The vaccine-TLR4 docked complex. (B) The interaction map between the multi-epitope vaccine and TLR4

Table 5.

A list of residues participating in H-bond interactions within the docked complex of the multi-epitope vaccine and TLR4 (chain B), accompanied by their respective distances

TLR4 residues (Chain B)	Vaccine residues	H-bond distances (Å)
GLU24	LYS249	2.72
GLU24	ARG253	2.75
GLU27	ARG253	2.69
GLU31	GLN245	2.97
GLU31	GLN245	2.96
ARG355	GLN226	2.66
ARG382	SER225	2.81
TYR403	GLN226	3.18
TYR403	GLN226	2.88
TYR451	SER225	2.87
GLU474	ARG166	3.02
LYS477	THR221	2.64
THR499	ARG166	2.91
THR499	ARG166	2.57
GLN523	ARG166	2.59
GLN547	TYR159	2.81
ARG598	THR203	2.95
ARG598	ASN207	2.71
GLN599	ASN148	2.91
GLU603	ASN207	3.02

Molecular dynamics simulation

Evaluating the root mean square deviation (RMSD), root mean square fluctuation (RMSF), and radius of gyration (Rg) plots for the selected vaccine-TLR4 docked complex offered a thorough insight into its structural dynamics and stability during MD simulation. Since we performed simulations of the selected docked complex with different initial velocities (Ig = 109858 and 118245), even in the same system, we observed some variations in the results, but the results obtained from each initial velocity were very consistent with their replicates. The RMSD plots (Ig = 109858) showed an increasing trend from 0 to 5.2 Å during the first 130 ns of the simulation, followed by stable fluctuations between 3.2 and 4.8 Å until the simulation’s end. On the other hand, the RMSD plots (Ig = 118245) showed a rapid increase from 0 to 5.9 Å within the first 30 ns, subsequently stabilizing with slight fluctuations between 4 and 5.9 Å until the end of the simulation (Fig. 11A). The analysis of the RMSF plots of both initial velocities (Ig = 109858 and 118245) revealed that the TLR4 backbone has small fluctuations, indicating significant structural stiffness, whereas the vaccine area has greater local flexibility, particularly at the start and terminal residues, where fluctuations are larger (Fig. 11B). The Rg plots (Ig = 109858) demonstrate that the docked complex showed Rg values between 25.5 and 26.75 Å, whereas the Rg plots (Ig = 118245) revealed Rg values from 21.5 to 22.1 Å, indicating that the docked complex at Ig = 109,858 is more compact during the simulation (Fig. 11C).

Fig. 11.

MD simulation analysis of the vaccine-TLR4 docked complex in a 250 ns period with different initial velocities (Ig = 109858 and 118245). (A) RMSD plots. (B) RMSF plots. (C) Rg plots

Additionally, the iMODS analyzed the structural dynamics of the selected docked complex and reported the results in the form of a deformability graph, B-factor value, eigenvalue, variance, covariance matrix, and elastic network. Main-chain deformability estimates a molecule’s ability to deform at each of its residues. The complex’s deformability graph has peaks indicating highly malleable residues. The chain’s flexible segments (hinges/linkers) have higher deformability values than the rigid portions of the main chain residues (Fig. 12A). The B-factor represents the relative amplitudes of atomic displacement around equilibrium positions. In the B-factor graph, as in the deformation graph, high values indicate flexible regions and low values indicate rigid regions (Fig. 12B). The eigenvalue indicates the structure’s stiffness or resistance to deformation. A lower eigenvalue indicates simpler deformation. The selected docked complex has an eigenvalue of 2.886652e − 05 (Fig. 12C). Each normal mode’s variance is inversely related to its eigenvalue. Individual variance is represented by purple bars on the variance diagram, while cumulative variance is shown with green bars (Fig. 12D). The coupling between residue pairs in the selected docked complex was assessed using a covariance matrix, with correlated, uncorrelated, and anti-correlated motions shown in red, white, and blue, respectively (Fig. 12E). The elastic network analysis was also conducted; it defines which pairs of atoms are linked by spring. Each dot on the graph indicates a spring between the matching atoms. Dots are colored according to their stiffness; darker greys indicate stiffer regions, while lighter dots suggest more flexible ones. Our complex appears to follow a dark gray paradigm that emphasizes stability (Fig. 12F).

Fig. 12.

The results of the MD simulation of the vaccine-TLR4 docked complex obtained by iMODS. (A) Deformability graph. (B) B-factor graph. (C) Eigenvalue graph. (D) Variance graph with purple bars showing individual variances and green bars depicting cumulative variances. (E) Covariance map, with red representing correlated motions, white representing uncorrelated motions, and blue representing anti-correlated motions. (F) Elastic network model, where darker grays represent stiffer regions and lighter dots indicate more flexible regions

Binding free energy analysis

The MM-PBSA and MM-GBSA approaches were employed to ascertain the binding free energies of the vaccine-TLR4 complex. The MM-PBSA analysis revealed that the vaccine-TLR4 complex possesses a total binding free energy of -693.6 kcal/mol, whereas the MM-GBSA analysis indicated a total binding free energy of -688.85 kcal/mol for the same complex. The diverse contributions to the total binding free energy in both methods indicate that the formation of the vaccine-TLR4 complex is predominantly dependent on van der Waals energy. The energy components of the vaccine-TLR4 complex, computed using the MM-PBSA and MM-GBSA approaches, are presented in Table 6.

Table 6.

The energy components of the vaccine-TLR4 complex

Energy Parameters	Vaccine-TLR4 Complex
MM-PBSA
Van der Waals energy	-654.15
Energy electrostatic	-89.60
Total gas phase energy	-743.75
Total salvation energy	50.15
Net energy	-693.6
MM-GBSA
Energy van der Waals	-654.15
Energy electrostatic	-89.60
Total gas phase energy	-743.75
Total energy salvation	54.90
Net energy	-688.85

Immune response simulation

The immunological response profiles of the vaccine candidate were established using the C-ImmSim server. The first dose led to an elevation in antigen levels, whereas the second and third doses caused a decline. Following each vaccine dosage, the immune response was activated, resulting in the production of immunoglobulins (Ig). Both the second and third dosages resulted in elevated levels of IgM, IgG1, and IgG2 antibodies, with the increase being more pronounced after the third dose (Fig. 13A). After the second and third doses, there was an increase in the number of B cell isotypes, memory B cells, and active B cells. Among isotype B cells, this rise was considerable in IgM but very slight in IgG2 (Fig. 13B, C). After the second and third vaccine doses, there was an increase in the number of TH memory cells, TH active cells, and TH resting cells. In TH memory cells and TH active cells, the rise was bigger in the third dose than in the second dose, although in TH resting cells, it was greater in the second dose (Fig. 13D, E). Similarly, with each dose, the number of TC memory cells and TC active cells increased (Fig. 13F, G). The population of natural killer (NK) cells fluctuated continuously throughout these three doses, with the lowest levels observed following the third dose (Fig. 13H). Macrophage active cells and macrophage resting cells rose following the first and second doses; however, after the third dose, macrophage active cells decreased and macrophage resting cells increased (Fig. 13I). Additionally, all three vaccine doses resulted in considerable cytokine production, including IL-2, IL-10, IL-18, IFN-γ, and transforming growth factor-beta (TGF-β). The second dose produced higher levels of IL-2 and TGF-β compared to the first and third doses (Fig. 13J).

Fig. 13.

Computational modeling of the immune response to the multi-epitope vaccine candidate following three administrations. (A) Antibody and antigen levels. (B) B cells population. (C) B cells population per state. (D) TH cell population. (E) TH cell population per state. (F) TC cell population. (G) TC cell population per state. (H) NK cell population. (I) Macrophage cells population per state. (J) Induced cytokine levels

Codon optimization and in Silico cloning of the vaccine candidate

The multi-epitope vaccine was reverse-translated into a nucleotide sequence comprising 792 base pairs (bp) and optimized for the E. coli K 12 strain, utilizing JCat. The optimized sequence revealed a CAI of 1 and a GC content of 52.15%. The vaccine sequence termini were modified to incorporate two restriction sites, HindIII (173) and BamHI (971). A successful clone of 6142 bp was ultimately achieved by inserting the segment into the pET28a (+) expression vector with SnapGene 3.2.1 software (Fig. 14).

Fig. 14.

Plasmid map for the cloned pET28a (+) expression vector. The multi-epitope vaccine sequence is depicted in purple, encircled by HindIII (173) and BamHI (971), while the pET28a (+) expression vector is illustrated with black lines

Discussion

According to global surveillance data, RV is responsible for more than 40% of pediatric hospitalizations due to diarrhea and remains a leading cause of death in children under the age of five [109, 110]. RV diarrhea is more severe in low- and lower-middle-income countries (LMICs) and can occur at any time of year, but it is more common in high-income countries (HICs) in the autumn and winter [111]. Similarly, the majority of severe RV gastroenteritis infections occur in infants and children in LMICs compared to HICs, making them more vulnerable to dehydration and death [112]. The introduction of RotaTeq^® and Rotarix^® vaccines has significantly reduced the number of RV fatalities. Nonetheless, due to its limited accessibility and diminished efficacy in LICs, numerous preventable fatalities continue to occur [24].

Multi-epitope vaccines provide numerous benefits in the realm of vaccine development. They can elicit strong and varied immunological responses. These vaccines employ immunoinformatics to forecast epitopes for CTLs, HTLs, and B cells, thereby enhancing the vaccine’s immunogenicity. Multi-epitope vaccines can elicit specific expansion of CD4⁺ and CD8⁺ T cells, enhance cytokine production, and stimulate memory T cell proliferation, resulting in robust immune responses against diverse pathogens. Moreover, these vaccines can confer enduring protection, diminish the likelihood of vaccine escape mutants, and present advantages in manufacturing and genetic adaptability, rendering them promising candidates for future vaccine development [113–115]. Nonetheless, a significant limitation of multi-epitope vaccines is their inadequate immunogenicity. Consequently, these vaccine formulations must incorporate adjuvants that can augment immunity post-vaccination [116].

Several studies have employed reverse vaccinology to develop potential vaccines targeting RVA. Usman et al.. (2023) designed a vaccine with multiple epitopes derived from the VP4 and VP7 proteins of RVA. Epitopes were tested for antigenicity, allergenicity, homology with human proteins, and anti-inflammatory properties. The vaccine has four B-cell, three CTL, and three HTL epitopes linked together by linkers and an N-terminal Arginine-Glycine-Aspartate (RGD) motif adjuvant [117]. Kuri et al.. (2024) utilized the conserved sequences of VP4, VP6, and VP7 to predict HTL, CTL, and linear B-cell epitopes. Twenty epitopes were found following the assessment of criteria, including antigenicity, non-allergenicity, non-toxicity, and stability. The epitopes were subsequently joined utilizing appropriate linkers and an N-terminal beta defensin adjuvant [118]. Sharma et al.. (2025) utilized reverse vaccinology to identify epitopes from the VP4, VP7, NSP2, and NSP5. The optimal epitopes were fused with the EAAAK, GGS, and AAG linkers, as well as the N-terminal and C-terminal regions of the flagellin protein from Salmonella typhimurium and cholera enterotoxin subunit B as adjuvants, to construct a multi-epitope vaccine [119]. Our study also focused on designing a multi-epitope vaccine against RVA using advanced immunoinformatics tools [119].

Our workflow in this study aligns with the latest standard pipeline for multi-epitope vaccine design. According to the Protegen database, VP6 protein of RVA is a protective antigen. RV VP6 was chosen as a viable antigen for epitope prediction because, in addition to having high antigenicity and no adverse effects on the body, it is located in the plasma membrane, lacks a signal peptide and transmembrane helix, and has no significant similarity with the human proteome. Our study is the first one in the field of rotavirus vaccine design that assessed the target protein for all of the aforementioned criteria prior to epitope prediction. Diverse vaccination strategies encoding the RV VP6 antigen, including DNA vaccines [120, 121], subunit vaccines comprising recombinant VP6, and self-assembled structures have demonstrated the capacity to provoke immunological responses or confer protection in animal models [122–124]. To develop a universal vaccine that comprehensively targets all RVA variants, we conducted epitope predictions based on conserved regions in the VP6 protein sequences of all RVA variants available in NCBI.

This study’s vaccine structure comprised 6 CTL epitopes, 2 HTL epitopes, and 1 linear B-cell epitope that were antigenic, non-allergenic, non-toxic, and capable of inducing IFN-γ and IL-4 production. The proposed vaccine achieved a population coverage of 98.43% for both CTL and HTL epitopes, which was very close to the population coverage (98.47%) of the vaccine designed in Usman et al. [117]. , indicating a broad protective effect.

The engineered vaccine construct, including 264 amino acids, exhibits favorable characteristics, including robust antigenicity, non-allergenicity, non-toxicity, and acceptable solubility. The antigenicity score for our vaccine candidate in the VaxiJen v2.0 server was predicted to be 0.5126 and 0.4923 for the virus and bacteria models, respectively. This score was lower (0.4807) than our vaccine for the vaccine designed in the study of Sharma et al. [119]. , but higher (0.7698) for the vaccine designed in the study of Kuri et al. [118]. The Expasy ProtParam tool projected that the multi-epitope vaccine would have a molecular weight of 27.85 kDa, and given that it is less than 110 kDa, it may be an acceptable and efficient vaccine due to its ease of handling and purification during the experiment [125]. The vaccine’s theoretical pI was calculated to be 6.07, indicating its acidic characteristics, whereas the vaccines designed in the Sharma et al. [119]. and Usman et al. [117]. studies had theoretical pIs of 9.40 and 8.60, respectively, demonstrating their alkaline properties. The half-life of our vaccine in mammalian reticulocytes was determined at 30 h, whereas the vaccine designed by Usman et al. [117]. has a half-life of one hour in the same tissue, indicating that our vaccine is exposed to the immune system for a more extended duration than those designed by Usman et al.. The vaccine’s instability index, crucial for maintaining structural integrity during storage and administration, was below 40, indicating that the vaccine construct is stable [126]. The aliphatic index, reflecting temperature stability, was determined to be 82.69, signifying that the vaccine shows thermal stability [127]. The vaccine candidate’s negative GRAVY score (-0.089) suggested that it is extremely hydrophilic, which means it can interact with water molecules [128].

The vaccine’s three-dimensional structure was modeled and refined. Quality assessment of the models revealed that the percentage of vaccine residues in the most preferred regions of the Ramachandran plot increased from 78.5% in the initial 3D model to 91.2% in the refined 3D model, demonstrating the projected model’s high accuracy and dependability. The initial 3D model and refined 3D model have ProSA Z-scores of -2.92 and − 3.53, respectively, which are within the range of scores for native proteins. Disulfide engineering was used to improve the vaccine structure’s stability, boost its ability to endure environmental stress, preserve structural stability during transit and storage, and stabilize the protein in biological environments [129, 130].

TLR4 identifies the 50 S ribosomal protein L7/L12 as an agonist capable of inducing DC maturation [61]; therefore, TLR4 was selected as the receptor for the vaccine docking process. Molecular docking analysis revealed a high binding affinity of -986.4 kcal/mol. H-bond interactions analysis confirmed the vaccine construct’s strong binding affinity to TLR4. The MD simulation demonstrated that the vaccine-TLR4 interaction exhibited satisfactory stability and thermodynamic characteristics. Based on the MM-PBSA and MM-GBSA calculations, the vaccine-TLR4 complex shows a significantly negative net energy, confirming its stability [131].

The immunological simulation results indicate that the proposed vaccine elicits significant humoral and cellular immunity. The vaccine’s ability to induce a significant and long-lasting immune response is supported by the continuous increase in IFN-γ and IL-2 levels following vaccination. In the study conducted by McNeal et al. [132]. IFN-γ was the sole cytokine detected in activated CD4 + T cells from immunized mice immunized with chimeric maltose-binding protein (MBP)-VP6 and the adjuvant E. coli heat-labile toxin (LT-R192G), which effectively inhibited RV replication. Yan et al. [133]. assessed the immunogenicity and protective effectiveness of the virus-like particles (VLPs) co-displaying VP2, VP6, and VP7 by immunizing BALB/C mice with four incremental doses of the VLPs, varying from 5 to 40 µg of VLP protein each dose. ELISA-based evaluations of porcine RV (PoRV)-specific antibodies and T cell cytokines, such as IL-4, IL-2, and IFN-γ, indicated that immunization with VP2-VP6-VP7 VLPs can successfully provoke both humoral and cellular immune responses in mice, leading to a significant induction of neutralizing antibodies.

We predict that our vaccine candidate will be effectively expressed in E. coli, as the optimized sequence has a GC content of 52.15%, falling within the recommended range of 30–70% for optimal translational efficiency [134], and the CAI is calculated at 1, which falls within the acceptable range of 0.8-1.0 [135].

Conclusion

This study used bioinformatics tools to identify appropriate epitopes of VP6 from RVA and assembled them into a vaccine construct employing various linkers and adjuvant (50 S ribosomal protein L7/L12). Favorable results were achieved from the assessment of the vaccine’s physicochemical properties, antigenicity, allergenicity, and toxicity. The vaccine’s 3D structure showed appropriate folding and established a robust interaction with TLR4. Our vaccine candidate has the potential to elicit both humoral and cellular immune responses; nevertheless, laboratory experiments are required to validate its efficacy against RVA.

Author contributions

OP: Writing – original draft, Methodology, Visualization. AGL: Writing – original draft, Methodology, Visualization. SA: Software, Data curation. MM: Visualization, Validation, Software. GM: Supervision, Project administration. SS: Supervision, Project administration, Writing – review & editing.

Funding

This study was supported by a grant from Semnan University of Medical Sciences (Grant Number: 4151).

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files.

Declarations

Ethics approval and consent to participate

The ethical committee of Semnan University of Medical Sciences approved this study with the number: IR.SEMUMS.REC.1403.242.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.