Phylogeny-guided multi-epitope vaccine design targeting conserved and divergent regions of Brucella OMP2, OMP25, and OMP31: an immunoinformatics study

Abstract

Brucella is a zoonotic intracellular bacterial pathogen. Currently, no human vaccines for brucellosis are approved, and only animal vaccines are available. Despite the widespread use of live attenuated vaccines, several drawbacks have prompted investigation into outer membrane proteins (OMPs) as potential vaccine candidates. This study employs immunoinformatics to design a multi-epitope vaccine targeting conserved regions of the OMP2a, OMP2b, OMP1, OMP25, and OMP31 antigens against four common Brucella species. Helper T-lymphocyte (HTL), Cytotoxic T-lymphocyte (CTL), and B-cell epitopes of OMP31, OMP25, OMP2a, and OMP10 were predicted and analyzed. The selected epitopes form a multi-epitope design featuring a flexible linker, “GGSSGG,” which connects the epitopes. This construct was fused to the Cholera Toxin B subunit (CtxB) at the N-terminus as an adjuvant, while a His-tag at the C-terminus facilitates purification. The tertiary structure of the multi-epitope was then modeled and refined. The 3D structure and its sequence were evaluated for physicochemical properties, structural features, in silico cloning, B epitope prediction, immune response simulation, and molecular docking with toll-like receptors (TLRs). The stability of the construct-TLR interactions was assessed through molecular dynamics (MD) simulations. The final vaccine construct demonstrated antigenicity and non-allergenicity, along with satisfactory physicochemical properties, as it binds to TLR2/4 receptors and induces an in-silico immune response. MD simulations confirmed the stability of the docked complexes. While this multi-epitope vaccine candidate shows promise in eliciting various immune responses against brucellosis in silico, it is crucial to validate its efficacy through laboratory and animal model testing.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12865-026-00806-y.

Introduction

Brucellosis is an infection caused by Brucella, a zoonotic Gram-negative bacterium, which is a facultative intracellular bacterial pathogen. Brucella melitensis, Brucella suis, Brucella abortus, and Brucella canis are four health-threatening & common human pathogen serotypes of Brucella whose primary hosts are caprine, swine, bovine, and canidae, respectively [1]. The disease is listed among the “seven neglected endemic zoonoses” by the World Health Organization (WHO) [2]. In recent years, the number of brucellosis patients has been on the rise in some populous countries, such as China [3]. Moreover, it has been shown that despite livestock vaccination against Brucella, dairy products remained contaminated in some cases [4].

Brucella initially adheres to and penetrates mucosal surfaces such as the respiratory airways, abraded or ulcerated skin, and the digestive tract. An in vitro analysis revealed that in eukaryotes, sialic acid residues serve as a base for Brucella adhesion to the cell surface of phagocytic cells [5]. Macrophages may serve as sites for Brucella replication. Bacteria that survived free radicals and bactericidal chemicals secreted by macrophages after opsonization through an IgG-mediated mechanism can cause a long-term or chronic infection [6]. Furthermore, Brucella virulence is mainly dependent on intracellular replication, persistence in the reticuloendothelial system, and the ability to invade the host [7]. Both in vitro and in vivo evaluations revealed that dendritic cells (DCs) are another site of proliferation for this pathogen. It leads to low expression of major histocompatibility complex (MHC) class II, CD80, and CD86, and to inhibition of functional maturation, characterized by the absence of proinflammatory cytokines such as IL-12 and TNF-α, and by inefficient antigen presentation [8].

Similar to other intracellular bacterial pathogens, Brucella species can modulate host cell apoptosis, a typical response of mammalian cells to a wide variety of bacterial infections, thereby allowing bacterial replication and survival [9]. Infections are manifested by nonspecific symptoms (Flu-like manifestations) in addition to sacroiliitis, neurobrucellosis, orchitis, spondylitis, and endocarditis [10, 11].

Laboratory diagnosis of human Brucellosis includes nucleic acid amplification tests (NAATs), serological tests, and cultures, with NAATs being the most accurate. They are more specific than the serological tests and more sensitive than conventional cultures currently available for diagnosing human Brucella infection [12].

Doxycycline-gentamicin and doxycycline-streptomycin are the two most commonly prescribed antibiotics for human brucellosis. However, in addition to their side effects, they’re facing many obstacles, such as therapeutic failure, relapses, reinfection, and the eruption of resistant Brucella strains [13]. Currently, no human brucellosis vaccines are available, and control relies primarily on animal vaccination. Live attenuated vaccines, such as the B. melitensis strain Rev1 and the B. abortus strain S19, are widely used as anti-brucellosis agents. Still, their disadvantages include the induction of abortion in pregnant animals, the cost of persistent serological responses, virulence to humans, and antibiotic resistance to brucellosis treatment [14].

Outer membrane proteins (OMPs) perform critical roles in Brucella’s structure, morphology, nutrient acquisition, colonization, invasion, and protection against external toxic threats such as antibiotics [15]. The 31-KDa OMP (OMP31) is an important virulence factor of Brucella species, and can induce both humoral immunity and T-helper (Th1) cell response dramatically [16]. In addition, OMP31 inhibits the nuclear factor (NF)-κB p65 signaling pathway, leading to autophagy in BV-2 Microglia [17]. Except for B. abortus, OMP31 is expressed in the other three common Brucella species [9].

The 25-KDa OMP (OMP25) reduces pro-inflammatory cytokine secretion, disrupts DC activation [20], attenuates IFN-β production, and enhances adaptability [18]. It can also downregulate TNF-α production in human macrophages during infection [19]. Interestingly, a significant homology (34% identity) was found between Brucella’s OMP31 & OMP25 [9]. OMP10 is also a kind of outer membrane lipoprotein that is necessary for Brucella species virulence [18]. At last, OMP2a & OMP2b, porin membrane proteins, are channels with selective permeability of small hydrophilic solutes that can act as bacteriophage receptors and resist environmental conditions [20, 21].

Previous multi-epitope vaccine studies against Brucella have generally targeted only a single species or relied on a limited selection of antigens, which restricts cross-protective potential [22, 23]. In contrast, the present work integrates conserved epitopes from five well-characterized OMPs (OMP2a, OMP2b, OMP10, OMP25, and OMP31) across the four most clinically relevant Brucella species [24]. This multi-species, multi-antigen design strategy provides broader epitope conservation, enhances immune-stimulatory potential, and represents a more comprehensive approach than earlier Brucella vaccine constructs. All potential antigenic determinants were assessed by a series of immunological and toxicological evaluations after epitope mapping of each protein. Next, the epitopes were conjugated to each other by peptide linkers with Cholera Toxin B subunit adjuvant in the N-terminal. Lastly, the interaction of the vaccine construct with Toll-like receptors (TLR)2/4 was evaluated using molecular docking, as they play an essential role in activating the host’s immune response [25].

Methods

Retrieval of antigen sequences

We initially evaluated a broad panel of Brucella outer membrane and virulence-associated proteins. All candidates underwent a stepwise filtering workflow including antigenicity, allergenicity, toxicity prediction, transmembrane topology for surface exposure, and conservation analysis across Brucella species. OMP2a, OMP2b, OMP10, OMP25, and OMP31 were selected because they demonstrated high conservation across the four major Brucella species, strong antigenic potential, and favorable immunological and physicochemical profiles. Proteins that did not meet these criteria were excluded from downstream epitope analyses. The proteins’ sequence of OMP2a (entry: Q45079), OMP2b (entry: Q45078), OMP10 (entry: A9MDJ4), OMP25 (entry: Q45110), and OMP31 (entry: Q93E87) were retrieved from the Uniprot database (https://www.uniprot.org). To design a vaccine targeting conserved regions of candidate proteins, we aligned the Uniprot sequences of each surface protein using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/), and the region containing conserved residues across the four species was identified for further epitope prediction. While several approaches exist for conserved sequence analysis in multi‑epitope vaccine design, including the IEDB Epitope Conservancy Tool and custom algorithm‑based pipelines, the present study employed multiple sequence alignment using Clustal Omega to identify conserved regions at the full‑length protein level. Unlike the IEDB Epitope Conservancy Tool, which requires predefined epitope inputs and evaluates conservation post hoc, Clustal Omega enables a priori identification of conserved motifs across homologous proteins without restricting analysis to preselected epitope regions [26, 27]. This strategy allows direct visualization of conserved residues, preserves positional context within the protein sequence, and promotes transparency and reproducibility of conserved region selection. In contrast to custom‑developed algorithms or proprietary software, Clustal Omega is a widely validated, open‑source, and freely available tool with robust performance for large‑scale comparative protein analyses [28]. Therefore, its use provides a pragmatic balance between methodological rigor, reproducibility, and biological interpretability for identifying conserved antigenic regions suitable for downstream epitope prediction in multi‑epitope vaccine design [29].

Phylogenetic tree of candidate protein

Five Brucella outer membrane protein (OMP) sequences—OMP2a (UniProt ID: Q45079), OMP2b (Q45078), OMP10 (A9MDJ4), OMP25 (Q45110), and OMP31 (Q93E87)—were retrieved from the UniProt database and used for phylogenetic analysis. An unrooted phylogenetic tree was generated using the NGPhylogeny.fr “One Click” pipeline [30]. The workflow included multiple sequence alignments with MAFFT, followed by automatic model selection, which identified LG + G4 as the best-fit evolutionary model. Phylogenetic reconstruction was performed using 1,000 ultrafast bootstrap replicates to assess branch support. Branch lengths were calculated to represent evolutionary distance in substitutions per site, and the final tree topology was visualized with an evolutionary scale bar (0.2 substitutions/site).

Prediction of helper T-Lymphocyte (HTL) and cytotoxic T-Lymphocyte (CTL) epitopes

The conserved sequences were submitted to the IEDB web server for both HTL and CTL epitope prediction. OMPs’ HTL epitope predictions were carried out using the IEDB MHCII binding predictor tool (http://tools.iedb.org/mhcii) with the consensus 2.22 prediction method. Each epitope was assigned a percentile rank indicating its affinity for the MHC, with lower ranks indicating greater affinity. Fifteen amino acid (aa) epitopes with a percentile rank below five were selected for further analysis.

IEDB’alignmentss MHCI tool (http://tools.iedb.org/mhci) was used for CTL epitope mapping of each OMP. The nine amino acid lengths with percentile ranks below five were considered favorable. The prediction was based on consensus 2.18. The FASTA format was used for protein sequences during the prediction process. All available human leukocyte antigen (HLA) alleles were selected for epitope mapping of both MHC I and II.

B-cell linear epitope prediction

B lymphocyte linear epitope prediction was performed to predict potential sequences that may trigger the humoral immune response. All OMPs were analyzed individually using the Bepipred Linear Epitope Prediction 2.0 tool of the IEDB (http://tools.iedb.org/bcell/) to predict their potential epitope regions, with the default parameters. BepiPred 2.0 uses a Random Forest algorithm trained on epitopes and non-epitope amino acid residues found from crystal structures [31]. It reports its analyses through charts, where regions with scores exceeding the threshold are defined as potential epitopes.

Investigation of Antigenicity, toxicity, allergenicity, and immunogenicity

Immunological assessments were performed for both T- and B-cell epitopes. T lymphocyte epitopes with more HLA allele hits were given priority for immunological analyses. First, the antigenicity of epitopes was evaluated by VaxiJen v2.0 (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html) using default parameters for all HTL, CTL, and B-cell linear epitopes. The target organism was bacteria with a threshold of 0.4. Next, all epitopes were submitted to the ToxinPred server for toxicity prediction (http://crdd.osdd.net/raghava/toxinpred/). The SVM (Swiss-Prot) method was chosen for prediction, and other methods remained the default. Subsequently, allergenicity was checked for B-cell and HTL epitopes using the AllerTOP v. 2.0 tool (https://www.ddg-pharmfac.net/AllerTOP) and AlgPred 2.0 server (https://webs.iiitd.edu.in/raghava/algpred2/). The parameters were the default as well. The server classifies a given sequence as an allergen or a non-allergen based on the total set of allergens trained on it [26].

Immunogenicity is an essential property of CTL epitopes, as it describes their potential to trigger an immune response. The IEDB server Class I Immunogenicity tool was used to evaluate the immunogenicity (http://tools.iedb.org/immunogenicity). Sequences with a score greater than zero are described as immunogenic, while scores below zero belong to non-immunogenic sequences. Additionally, the higher the positive score, the more potent the immunogenic characteristics. The preferred setting for the analysis parameters was the default.

HTL, CTL, and B-cell linear epitopes that were recognized as antigens, non-toxic, and non-allergenic were filtered for subsequent analysis. Moreover, CTL epitopes must be immunogenic to be preferred. Finally, a total of 10 epitopes were included in the vaccine construct: five CTL epitopes, five HTL epitopes, and two B-cell linear epitopes to stimulate both cellular and humoral immune responses.

T lymphocyte cytokine induction analyses

Interferon gamma (IFNγ) is released in large amounts by activated cytotoxic T cells and T helper cells [32]. In addition, pretreatment with IFN-γ activates cellular defense mechanisms that enhance intracellular killing of Brucella and prevent bacterial replication [33]. The induction of IFN-γ was assessed using the IFNepitope web server (http://crdd.osdd.net/raghava/ifnepitope/) using the default mode. The Motif and SVM hybrid method was used to predict IFNγ induction, and sequences were classified as either IFN-γ or non-IFN-γ inducers.

IL-4 induction was also evaluated, as it inhibits pro-inflammatory cytokine secretion and acts as a B-cell mitogen. It can also act as an enhancer for IgG & IgE secretion [34]. The IL4pred server (https://webs.iiitd.edu.in/raghava/il4pred/design.php) was used for this approach. Interleukin-10 is another cytokine that limits the host’s immune response to pathogens, thereby preventing tissue damage [35]. Therefore, epitopes must be IL-10 inducers, and this was predicted with the IL-10Pred tool (https://webs.iiitd.edu.in/raghava/il10pred/predict3.php), which divides epitopes as “Inducer” or “Non-inducer”. For both IL-4 and IL-10 predictions, the parameters were set to their default values.

Prediction of Immunoglobulin secretion

Antibody secretion is one of the most critical functions of B cells during the immune response, and the selected B-cell linear epitopes were investigated by the IgPred server (https://webs.iiitd.edu.in/raghava/igpred) for this. This server can detect the peptide potency in triggering IgA, IgG, or IgE secretion. The threshold for all of the immunoglobulins was 0.9, and sequences with values above indicated an immunoglobulin inducer.

Investigation of population coverage

An epitope can elicit an immune response in the body when its target MHC allele is expressed among a population. The IEDB provides a tool, Population Coverage, that measures epitope-related allele frequency of MHC class I and II in different populations (http://tools.iedb.org/population). Targeted MHC alleles of epitopes for both class I and II MHCs were submitted separately to determine the worldwide and continental frequencies, with percentages.

Constructing the multi-epitope vaccine

The desirable epitopes that passed assessments were fused with a TLR-4 agonist and flexible linkers. The flexible linker sequence used throughout the construct was GGSSGG (Gly–Gly–Ser–Ser–Gly–Gly). Glycine–serine–rich linkers are widely used to provide conformational flexibility and solubility between adjacent epitope modules, thereby reducing steric interference and facilitating the independent folding of domains [36, 37]. A rigid α-helical spacer EAAAK (used between the adjuvant and the epitope cassette) was chosen to physically separate the cholera toxin B subunit (CtxB) from the multi-epitope region and to reduce undesired domain–domain interactions; the EAAAK motif has been validated experimentally as a helix-forming rigid linker in fusion proteins [38]. The adjuvant selected was the CtxB (mature monomer ≈ 103 aa; UniProt entry: P01556). CtxB is non-toxic, forms a pentamer that binds GM1 ganglioside on epithelial and antigen-presenting cells, and enhances antigen uptake, dendritic-cell targeting, mucosal and systemic antibody responses, and T-cell responses, properties previously exploited in bacterial and subunit vaccine designs. For these reasons, CtxB was appended to the N-terminus via the EAAAK linker [39]. A tag of six histidine residues was also used at the C-terminal of the vaccine sequence, as it helps the purification process [40].

Protein sequence and human similarity

To avoid autoimmune reactions, the vaccine sequence was compared with human proteins using the Protein Information Resource (PIR) (available at https://research.bioinformatics.udel.edu/peptidematch/batchpeptidematch) to assess the resemblance between the vaccine construct and human proteins, with Homo Sapiens as the reference organism.

Secondary structure prediction & 3D modeling

The vaccine sequence was analyzed by the PSIPRED 4.0 server (http://bioinf.cs.ucl.ac.uk/psipred) for secondary structure and aa content (e.g., hydrophobic, polar).

To predict the tertiary structure of the vaccine construct, I-TASSER (https://zhanggroup.org/I-TASSER) was used, which returned five structures in PDB format, requiring further evaluation to select the best. Structural validations were performed using SAVES v6.0 (https://saves.mbi.ucla.edu/), which includes ERRAT, Verify3D, and PROCHECK tools. ERRAT identifies regions of error in protein crystal structures by examining the statistics of pairwise atomic interactions [41], and areas with higher scores are more stable. Verify3D assesses how well an atomic 3D model matches its amino acid sequence by assigning a structural class based on its location and environment, and comparing the results with established structures; a desirable structure should pass this assessment [42].

The Ramachandran plot, generated by PROCHECK, also shows the statistical distribution of the combinations of backbone dihedral angles ϕ and ψ. The Ramachandran plot visualizes the energetically allowed and disallowed regions of dihedral angles. The threshold used in this study for disallowed areas was 2.0; structures above this threshold were no longer considered desirable. Along with a lower disallowed region, the higher allowed region percentage was also considered for structural validation.

The ProSA-web server was another indicator of a valid structure. It computes a Z-score, a measure of protein model quality. The 3D models that fell out of the normal range depicted by the plot or other structural errors underwent protein refinement using the GalaxyRefine web server (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=REFINE). GalaxyRefine server output contains five remodeled structures, and all models were reassessed using SAVES v6.0 tools and the ProSA-web server. The model with the best results was selected for subsequent analyses.

Epitope-HLA Docking

GalaxyPepDock (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=PEPDOCK) was used to simulate each epitope with its target HLA allele binding modality. Initially, the PDB format of target HLAs available in the RCSB [43] was retrieved, and ligands or water molecules were removed using Chimera 1.16 [44]. Lastly, the epitopes were docked using GalaxyPepDock [37], and the Gibbs free energy change (ΔG) and the dissociation constant (Kd) were reported.

Discontinuous B-cell epitope mapping

Since B-cell epitopes are generally discontinuous (conformational) and comprise multiple sequential segments [37], the 3D structure of the protein was evaluated by ElliPro (http://tools.iedb.org/ellipro/) for conformational epitope prediction. The results include both linear and structural epitopes, shown in JSmol in three dimensions, along with the sequences from the provided PDB structure. However, these results were further assessed for the immunological characteristics.

Physicochemical and immunological features analysis

The ProtParam tool on the Scratch server (https://web.expasy.org/protparam/) was used to characterize the vaccine construct. The molecular weight, isoelectric point (pI), instability index, aliphatic index, and grand average of hydropathy (GRAVY) index were measured in this server. Furthermore, the vaccine solubility upon overexpression in Escherichia coli and the antigenicity were assessed with SOLPro and ANTIGENpro [45, 46]. Additionally, the vaccine construct antigenicity was rechecked with VaxiJen v2.0, and allergenicity was also predicted using the AllerTOP v. 2.0 tool.

The TLR-Protein interaction & related investigations

The interaction of TLR2 and TLR4 with the vaccine construct was evaluated, as they play a critical role in immune responses. The Cluspro 2.0 (https://cluspro.bu.edu/peptide/index.php) web server was chosen for this approach. The 3D structure of the vaccine construct was docked separately to the crystal structures of human TLR2 (PDB code: 6NIG) and TLR4 (PDB code: 3FXI) retrieved from the RCSB database. Models with interactive residues for TLR2 and TLR4 were considered favorable. The binding energy and interactive residues were analyzed by PRODIGY server (https://wenmr.science.uu.nl/prodigy) and PDBsum Generate (http://www.ebi.ac.uk/thornton-srv/databases/pdbsum/Generate.html), respectively.

Immune simulation

An in silico method was used to simulate immune responses using the C-IMMSIM server (https://kraken.iac.rm.cnr.it/C-IMMSIM/), which reported the immune response profiles of the anti-brucellosis vaccine constructs and the immunogenicity of the multi-epitope vaccine. It is an agent-based model that uses a position-specific scoring matrix (PSSM) and machine learning algorithms to predict the mammalian immune system’s humoral and cellular responses. According to most vaccines currently in use, the minimum interval between first and second doses is 4 weeks. The simulation time steps are 10, 94, 178, 262, 346, 430, 514, 598, 682,766, 850, 934 h (8 h corresponds to one cell division cycle in real life and the time step 10 is injection at time = 0 according to Kaba et al. (2018) [47] were selected along with homozygous host haplotypes (HLA-A*0101, HLA- B* 0702, HLA-B*B1501, HLA-DRB1*0101), while other general simulation parameters were: random seed = 12345 with vaccine injection not containing LPS, simulated volume: 10, simulation steps 2400, Adjuvant = 100.

Molecular dynamics of vaccine construct

The most potent vaccine-receptor complexes with the lowest free energy, highest number of hydrogen and hydrophobic bonds, and best ligand-receptor affinity were selected for further investigation. The stability of the construct-TLR2 and construct-TLR4 complexes, and of TLR2, TLR4, and the construct unbound, was evaluated by MD simulation. The GROMACS 2018 package and the OPLS-AA force field were used for this purpose [48]. The structures were placed in a triaxial box. They were 1 nm from all edges [48]. TIP3P water molecules filled the simulation box. In the simulation, if the system is positively charged, it is neutralized with chlorine ions; if negatively charged, with sodium ions [49]. The energy minimization process for the simulated features has been divided into two parts: first, equilibration with NVT (constant particle number, volume, and temperature) at 300 K and 100 ps, and second, equilibration with a Parrinello-Rahman barostat at 300 K and 1.0 bar and 100 ps. A Parrinello-Rahman barostat was used to equilibrate the system to NPT (constant particle number, pressure, and temperature) at 300 K and 1.0 bar for 100 ps [50]. The Ewald mesh particle algorithm was used to handle long-range electrostatic charges. Its no-effect distance was set at 10 angstroms [51]. To calculate van der Waals interactions, a distance of 1 nm was used. To limit the length of covalent bonds, a linear constraint algorithm was employed. A 100-nanosecond (ns) simulation was used for the selected complexes by the molecular docking method, and the designed protein was applied after applying the necessary balances [52]. The root mean square fluctuations (RMSF), root mean square distance (RMSD), and radius of gyration (RG) were calculated to evaluate the stability of the vaccine-receptor complexes.

Calculation of free binding energy

The binding free energy between the vaccine construct and the TLR receptors was calculated using the Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PBSA) method [53]. MM-PBSA analysis was performed using the Hawdock webserver, considering the following equation:

HawkDock implements the MM-PBSA method using the standard theoretical framework shared with GROMACS-based calculations.

Codon optimization and in Silico cloning of the designed construct

Codon adaptation was performed to improve vaccine expression in E. coli. The vaccine protein sequence was submitted to the Java Codon Adaptation Tool (JCat) (http://www.jcat.de/) [54] for reverse translation and codon optimization according to the codon usage of E. coli (strain K12). Finally, the optimized vaccine sequence, with restriction sites BamHI and HindIII added to the N- and C-termini, respectively, was cloned into the pET-28a(+) plasmid vector using the SnapGene tool (https://www.snapgene.com/try-snapgene/) to validate vaccine expression.

Results

Retrieval of antigen sequence

The sequences of B.canis OMP2a, OMP2b, OMP10, OMP25 and OMP31 have been retrieved with Q45079, Q45078, A9MDJ4, Q45110 and Q93E87 entry, respectively. The conserved OMP sequence across the four targeted species was identified for epitope mapping using Clustal Omega (Supplementary data 1).

Phylogenetic analysis of protein candidates

The phylogenetic tree revealed apparent clustering among the analyzed Brucella OMP sequences (Fig. 1). OMP2a and OMP2b are grouped as a tightly clustered paralogous pair, reflecting their high degree of sequence similarity. In contrast, OMP10, OMP25, and OMP31 displayed progressively greater divergence, with OMP31 occupying the most distant position on the tree. The distribution of branch lengths indicated substantial evolutionary separation between the OMP2 clade and the remaining OMPs. This topology directly informed epitope selection for downstream immunoinformatics analysis: conserved regions within the closely related OMP2a/OMP2b cluster were prioritized, while divergent motifs from OMP10 and OMP31 were incorporated to improve antigenic breadth and maximize potential cross-protective immune responses.

Fig. 1.

Phylogenetic tree of the five selected Brucella outer membrane proteins using reference sequences from UniProt: OMP2a (Q45079), OMP2b (Q45078), OMP10 (A9MDJ4), OMP25 (Q45110), and OMP31 (Q93E87). The unrooted tree was generated using the NGPhylogeny.fr “One Click” pipeline, which includes MAFFT alignment, automatic model selection (LG + G4), and 1,000 ultrafast bootstrap replicates. Branch lengths represent evolutionary distance in substitutions per site (scale bar: 0.2). OMP2a and OMP2b form a tight paralogous cluster, indicating high sequence similarity, while OMP10, OMP25, and OMP31 diverge progressively, with OMP31 being the most distant. This topology informed epitope selection by prioritizing conserved regions within the OMP2 clade and including divergent motifs from OMP10 and OMP31 for broad cross-protection

Prediction of HTL and CTL epitopes

The HTL epitopes were predicted using the IEDB tool with the consensus 2.22 prediction method and 15-aa-length epitopes. The CTL MHC-I Binding Prediction tool was used to predict nine-aa-length epitopes using a consensus prediction method. As previously mentioned, the sequences were evaluated for all HLA alleles in both CTL and HTL epitope prediction. Moreover, epitopes with a percentile rank < 5.0 and diverse target HLA alleles were favorable.

The CTL epitopes with precedence HLA-allele coverage were assessed for antigenicity, toxicity, and immunogenicity. Candidates from OMP25 & OMP31 (Table 1) were selected for the vaccine construct because they showed higher antigenicity and immunogenicity, with no toxicity. Subsequently, all targeted OMPs underwent epitope mapping. Epitopes with lower percentile rank and higher allele coverage were evaluated for antigenicity, toxicity, allergenicity, and cytokine induction (including IFNγ, IL-4, and IL-10). A desirable HTL epitope for a vaccine construct should be predicted to be an antigen, non-toxic, and an IFNγ inducer. IL-4 & IL-10 induction is also considered a good indicator of the efficacy of the chosen epitopes. Table 2 shows the final candidates for the HTL epitope.

Table 1.

Predicted cytotoxic T lymphocyte (CTL/CD8⁺) epitopes of Brucella proteins with their corresponding immunological properties. Epitopes were predicted using the IEDB MHC-I binding tool with a percentile rank below five, conservancy across selected species, population coverage (for the indicated HLA alleles), and predicted antigenicity, immunogenicity, and toxicity are shown

Protein Name	Peptide Sequence	MHC-I	MHC-I allele	Percentile rank	HLA population (World)	Antigenicity	Toxicity	Immunogenicity score
OMP25	AFAGWNFQK	78–86	A*03:01	2.45	9.77	0.520/Antigen	Non-Toxin	0.287/Immunogenic
			A*11:01	0.97	8.99
			A*29:02	3.25	2.18
			A*31:01	1.39	3.02
			A*68:01	4.85	3.29
			E*01:03	2.9	47.30
	LYLGYGWNK	62–70	A*11:01	3.15	8.99	0.527/Antigen	Non-Toxin	0.241/Immunogenic
			A*23:01	1.4	3.06
			A*24:02	3.7	12.59
			A*31:01	3.15	3.02
			C*14:02	3.55	1.73
	FRVGWTAGA	165–173	B*27:05	1.0	3.17	1.355/Antigen	Non-Toxin	0.352/Immunogenic
			B*39:01	1.15	1.81
			C*06:02	2.65	9.13
			C*07:01	2.61	11.57
OMP31	VQAGYNWQL	86–94	A*02:06	0.35	24.39	1.843/Antigen	Non-Toxin	0.156/Immunogenic
			A*02:01	1.8	1.09
			B*40:01	1.98	2.13
			B*38:01	4.4	2.02
			E*01:01	2.96	50.32
	WTLGAGAEY	185–193	A*01:01	0.94	10.09	1.165/Antigen	Non-Toxin	0.192/Immunogenic
			A*25:01	3.31	1.88
			A*26:01	0.5	3.28
			A*29:02	0.28	2.18
			A*30:02	0.36	1.36
			B*15:01	2.9	5.65
			B*15:02	3.23	1.62
			B*18:01	1.35	4.88
			B*35:01	1.4	5.63
			B*46:01	1.75	2.02
			B*57:01	2.2	2.58
			B*58:01	4.3	2.26
			C*03:03	1.56	4.68

Table 2.

Predicted helper T lymphocyte (Th/CD4⁺) epitopes of Brucella proteins and their respective immunological characteristics. Epitopes (15-mer peptides) were predicted using the IEDB MHC-II binding tool with a percentile rank below five. Percentile rank, major HLA-DR/DQ/DP allele coverage, conservancy across selected species, and population coverage are presented, in addition to immunological and toxicological properties

Protein Name	Position	Epitope Sequence	MHC-II alleles	Adjusted rank	Antigenicity	Toxicity	Allergenicity	IFN-γ induce	IL-10 induce	IL-4 induce
OMP2b	171–185	GDVINDDVISAGSYR	DRB1*03:01	2.20	0.47/Antigen	Non-toxin	Non-allergen	+	-	+
			DRB1*03:05	0.71
			DRB1*03:06	0.40
			DRB1*03:07	0.40
			DRB1*03:08	0.40
			DRB1*03:09	1.80
			DRB1*11:07	0.37
	189–203	ISYTFTGGNGFSAVI	DRB1*09:01	2.60	0.93/Antigen	Non-Toxin	Non-allergen	+	+	+
			DQA1*05:01/DQB1*03:01	4.3
OMP10	27–41	SGNAPIIAHTPAGIE	DRB1*09:01	3.80	0.59/Antigen	Non-Toxin	Non-allergen	+	-	+
			DRB1*13:02	2.0
	112–126	SSAGSRFSLTRRNAG	DRB1*11:01	1.5	1.08/Antigen	Non-Toxin	Non-allergen	+	+	-
OMP31	8–22	SIAAMFATSAMAADV	DQA1*01:02	0.77	0.42/Antigen	Non-Toxin	Non-allergen	+	-	+
			DQA1*03:01	4.1
			DQA1*04:01	2.2
			DQB1*03:01	4.1
			DRB1*01:01	3.5
			DRB1*09:01	1.4

B-cell linear epitope prediction

All OMPs were investigated for B-cell linear epitope mapping. The IEDB server with the Bepipred Linear Epitope Prediction 2.0 method and 0.5 thresholds was used. Each aa with a score above the threshold was recognized as an epitope and analyzed for antigenicity, toxicity, allergenicity, and immunoglobulin induction. A desired epitope should be recognized as an antigen, non-toxic, and non-allergenic with an IgG induction potential, since it’s the primary antibody secreted in response to pathogens. As described in Table 3, only OMP2a and OMP25 of B. canis contained potential B-cell linear epitope sequences.

Table 3.

Linear B lymphocyte epitopes from OMP25 and OMP2a. The table lists the sequence of each epitope and its corresponding allergenicity, antigenicity, toxicity, and Immunoglobulin type. Both predicted epitopes were non-allergenic, antigenic, non-toxic, and IgG inducers

Protein Name	Position	Epitope Sequence	Epitope type	Allergenicity	Antigenicity*	Toxicity	Antibody Class Prediction/Score
OMP25	178–203	EYRYTQYSNKNYDLAGTTVRNKLDTQ	Linear	Non-allergen	1.054/Antigen	Non-Toxin	IgG inducer/0.930
OMP2a	54–80	TSNNSRHDGQYGDFSDDRDVADGGVST	Linear	Non-allergen	0.930/Antigen	Non-Toxin	IgG inducer/1.057

Investigation of population coverage

The population coverage of a vaccine’s target HLAs determines its efficacy in the population. Before population coverage analysis, the potential epitopes were screened from selected HTL and B-cell linear epitopes. Similar evaluations were also conducted for each screened epitope and sorted in Supplementary data 2. Their allele frequency was also included in the investigation of total population coverage. Moreover, the population coverage of CTL & HTL epitope target alleles is depicted in Supplementary Fig. 1 (Figure S1). The results were illustrated in six different categories, based on the researchers’ preference. On the IEDB web server, the continents are not defined separately; consequently, the average coverage across different sections was treated as the population coverage of the whole continent. The CTL epitopes’ overall population coverage across HLA alleles was 100% worldwide, while coverage across continents varied, as detailed in Figure S1. Iran’s CTL epitopes’ population coverage (96.77%) was also reported, as this study was conducted in Iran.

The global population coverage was 91.47% for the HTL epitope target alleles; other regions are shown separately in Figure S1. The lowest CTL and HTL epitope HLA-allele population coverage was observed in Central America and South Africa, at 4.91% and 32.1%, respectively.

Constructing the multi-epitope vaccine

The vaccine construct was made by stacking all selected CTL, HTL, and B-cell epitopes with a flexible GGSSGG linker at the N-terminal in addition to the CtxB as a natural adjuvant. Lastly, the His-tag was placed at the C-terminus, resulting in a 353 aa vaccine sequence. The final construct and epitopes’ order of placement are shown in Fig. 2.

Fig. 2.

Primary structure and domain organization of the multi-epitope vaccine construct against Brucella. A Amino acid sequence of the final vaccine construct. CTL epitopes are highlighted in yellow, HTL epitopes in green, B-cell linear epitopes in red, and GGGSSGG linkers in gray. B Schematic linear representation of the vaccine polypeptide. From N- to C-terminus: EAAAK linker (black), cholera toxin B subunit adjuvant (CtxB, blue), multiple CTL epitopes (yellow), HTL epitopes (green), B-cell epitopes (red), flexible GGGSSGG (gray), OMP2b signal sequence (light gray), and a C-terminal 6×His tag (black). Legend at the bottom indicates color coding

Protein sequence and human similarity

The PIR server was used to predict the similarity of vaccine sequences with human proteome sequences. The results indicated no similarity between the vaccine sequence and the Homo sapiens proteome (Figure S2). Therefore, there is probably no risk of autoimmune responses from the vaccine administration.

Secondary structure prediction & 3D modeling

The secondary structure and aa composition of the vaccine sequence were investigated using PSIPRED V4. As represented in Figure S3, the sequence contained 49.29% small nonpolar, 14.16% hydrophobic, 9.63% aromatics plus cysteine, and 26.91% polar amino acids. Moreover, the secondary structure consisted of 17.56% helix, 28.61% strand, and 53.83% coil chains.

The I-Tasser server was also utilized for homology modeling. The web server provides the five best 3D models, along with their PDBs, which were further analyzed using the ERRAT, ProSA-web, Verify3D, and PROCHECK tools. Since the assessments of the given models were unsatisfactory, the best model with the most desirable scores underwent protein refinement. The non-refined best model had passed Verify3D, with a Z-score of −6.62 and an ERRAT of 73.6. However, the most favored region of the Ramachandran plot was only 54.1%, with the disallowed region accounting for 1.9%. In contrast, the best model among the refined structures using GalaxyRefine had 70.37% ERRAT, a −6.82 Z-score, 1.9% in the disallowed region, and 76.7% in the allowed region, and it also passed Verify3D. Although the ERRAT score showed a marginal decrease after refinement, this change mainly reflects the presence of flexible linker and junction regions, whereas significant improvements were observed in the Ramachandran favored regions and overall structural quality. The final model was significantly improved after refinement, and its plots are presented in Fig. 3.

Fig. 3.

Structural validation and quality assessment of the predicted 3D vaccine model. A Ribbon representation of the final refined tertiary structure (green). B Ramachandran plot showing φ/ψ angle distribution of the model; most residues lie in favored and allowed regions. C ProSA-web Z-score plot; the black dot at − 6.82 indicates the model quality is within the range of native proteins of similar size. D ERRAT plot showing local error values across residue windows; red and yellow bars correspond mainly to flexible or non-globular regions, while gray bars indicate regions with acceptable quality. E Per-residue score profiles of Verify3D analysis (averaged in blue, raw in green), confirming overall good stereochemical and energetic quality of the model

Peptide-HLA Docking

Epitope-HLA docking was performed with GalaxyPepDock to investigate epitope-HLA interactions. The docking was performed between the epitope and one of its widespread HLA alleles retrieved from the RCSB. The server output contained five docked models, each one presented with its respective ΔG & Kd. The docked models with lower ΔG and Kd were chosen because they represented more stable interactions and are depicted in detail in Figure 4. Docking details are represented within a table in Supplementary data 3.

Fig. 4.

Representative molecular docking complexes of selected CTL and HTL epitopes with common human HLA class I and II alleles. CTL epitopes with their corresponding HLAs (A-E) and HTL epitopes with their corresponding HLAs (F-I). Vaccine-derived epitopes are shown as sticks, while HLA molecules are depicted in cartoon representation. Detailed binding energies and dissociation constants for all epitope–HLA interactions are summarized in Supplementary data 3

Physicochemical features

The ProtParam tool analysis of physicochemical properties revealed that the molecular weight and pI of the vaccine sequence are 35.11 KDa and 6.49, respectively. The instability index was 34.82, indicating the vaccine is stable (proteins < 40 are considered stable), and the aliphatic index was 57.20. Moreover, the GRAVY index reported a value of −0.388, indicating a non-polar construct. Additionally, SOLpro analysis of vaccine solubility showed that the protein is soluble during overexpression in E. coli. The antigenicity of the whole vaccine construct was double-checked with Vaxijen v.2.0 and ANTIGENpro, and both servers confirmed that the given sequence is an antigen. Lastly, an allergenicity assessment with AllerTOP showed a “Non-allergen” vaccine. The physicochemical properties are summarized in Table 4.

Table 4.

Physicochemical and immunological features of the vaccine construct. The vaccine indicated a MW of ~ 35 kDa and a 353 Aa sequence. Moreover, the antigenicity was preserved in the entire sequence, and the vaccine construct remained non-allergenic. Also, it is soluble upon overexpression in E. coli

Features	Assessment
Length	353 amino acids
Mw	35117.11
pI	6.49
Instability index	34.82/Stable
Aliphatic index	57.20
GRAVY index	−0.388
Antigenicity	1.2193 (Vaxijen v.2.0)/Antigen
	0.916566 (ANTIGENpro)/Antigen
Allergenicity	Non-allergen
Solubility	Soluble − 0.539175

Discontinuous B-cell epitope prediction & B lymphocyte epitopes surface accessibility

The vaccine construct was assessed for B lymphocyte conformational epitopes using the ElliPro tool. It predicts both linear and conformational B-cell epitopes with a score ranging from 0.5 to 6.0. It uses a 3D structure of the given construct to predict potential B-cell epitopes. However, the predicted epitopes (both linear and conformational) were verified to be part of the targeted proteins, as the immune response against the vaccine construct is not valuable if it’s not directed against the target proteins. All predicted discontinuous and linear epitopes were evaluated for antigenicity, allergenicity, toxicity, and immunoglobulin secretion, but none of the linear epitopes passed the indicators. A B-lymphocyte conformational epitope, with ‘’EYRYTQYSNKNYDLAGTTVR’’ sequence, which is a part of the OMP25 B-cell linear epitope, was validated as a potential conformational epitope. It was an antigenic epitope, non-toxic and non-allergenic, and an IgG secretor (score: 0.929). All of the ElliPro outputs are available in Supplementary data 4.

The B-cell linear epitopes in the vaccine sequence were examined for surface accessibility, as epitopes located in the protein core can be ineffective [55]. The two linear epitopes of B-cell epitopes in the vaccine construct are depicted in Fig. 5, showing that they are accessible to B-cell receptors.

Fig. 5.

Predicted conformational B-cell epitope exposure of the vaccine construct. The overall protein is shown as a blue transparent surface with a cartoon backbone. Predicted conformational B-cell epitopes are highlighted in bright yellow, demonstrating their prominent surface exposure and ensuring accessibility to antibodies

The TLR-vaccine construct interaction & related investigations

The interactions between TLR2 and TLR4 and the vaccine construct were analyzed using the molecular docking tool on the Cluspro 2.0 server. The TLR2 PDB with “6NIG” and the TLR4 PDB with “3FXI” were retrieved, and ligands and water molecules were removed. The docking results contained clusters of binding modalities; however, the binding modes were filtered to interact with the TLR2/4 binding sites. The favorable models were first submitted to the PRODIGY server for investigation of ΔG and Kd. Next, the involved residues in the interactions of the TLR2-vaccine construct interaction (Fig. 6A) and the TLR4-construct (Fig. 6C) were evaluated using PDBsum Generate. The ΔG and Kd at 25℃ were − 12.4 kcal.mol^− 1 and 8.30.10^− 10 M, respectively, with the hydrogen bonds (H-bond) in ‘’Glu264 and Ser201”, “Asp263 with Gln3”, “Asn294 with Thr1”, “Arg321 with Asn4” (2 H-bonds), and “Tyr323 with Phe109 & Ala110” between TLR2 and vaccine construct, respectively, as interaction hotspots (Fig. 6B). Moreover, the TLR4-construct ΔG and Kd at 25℃ were − 17.9 kcal.mol^− 1 and 7.50.10^− 14 M, respectively. (Fig. 6D). The hotspot residues of the vaccine construct interacted with TLR4, forming 34 H-bonds, as shown in Fig. 6D.

Fig. 6.

Molecular docking of the multi-epitope vaccine construct with human TLR2 and TLR4. The vaccine construct (red) forms stable complexes with both TLR2 (top) and TLR4 (bottom). A Vaccine–TLR2 complex. B Detailed TLR2 interface residues on the vaccine construct. C Vaccine–TLR4 complex. D Detailed TLR4 interface residues on the vaccine construct. E Legend for interaction types: salt bridges (red), hydrogen bonds (cyan), hydrophobic contacts (yellow), non-bonded contacts (gray). The docking predicts strong recognition of the vaccine by both TLR2 and TLR4, supporting its potential adjuvant activity and ability to trigger innate immune responses

Immune simulation study of vaccine construct

The C-IMMSIM immune simulation server was employed to evaluate the in silico immune response profile of the designed vaccine. Within the first week after administration, a marked rise in IgM levels indicated a typical primary immune response. In subsequent secondary and tertiary responses, elevated concentrations of IgG1, IgG2, and IgM, along with combined IgG + IgM activity, were observed, accompanied by a decline in antigen concentration (Fig. 7A).

Fig. 7.

In silico immune simulation following three intramuscular injections of the multi-epitope vaccine. A Dynamics of B-cell populations (total, memory, and isotype-specific). Rapid expansion of IgM followed by strong IgG1 and IgG2 responses indicates effective class switching and development of long-term humoral immunity. B Dynamics of T-helper (Th) and cytotoxic T-cell (CTL) populations showing sustained expansion of memory Th and CTL cells. C Serum antibody titers (IgM + IgG, IgM, IgG1, IgG2, IgG1 + IgG2, IgG2) with repeated peaks after each injection and persistent high IgG levels, consistent with real-world secondary and tertiary immune responses. D Cytokine profile showing high and sustained levels of IFN-γ, IL-2, IL-12, and other Th1/Th2 cytokines, confirming a strong, balanced cell-mediated and humoral immune response induced by the vaccine construct

The increase in B-lymphocyte populations suggested isotype switching and the establishment of humoral memory responses (Fig. 7 C). In parallel, the helper T-cell (TH) population demonstrated a sustained increase in memory cell counts (green), confirming the induction of long-term cellular immunity (Fig. 7B). Moreover, enhanced cytokine secretion, including IFN-γ and IL-2, during antigenic exposure (Fig. 7D) further supported the construct’s ability to stimulate the cellular immune response. Collectively, these simulation results indicate that the designed vaccine has the potential to elicit robust humoral and cellular immune responses against Brucella species. The complete output data are provided in Supplementary Data 5.

Molecular dynamics of vaccine construct

MD simulations were performed for 100 ns to verify the stability of the TLRs and the designed construct over time. The RMSD analysis showed that the designed construct reached stability after about 20 ns and maintained it to the end of the simulation, with a mean RMSD of 0.54 nm (Fig. 8a). The TLR2 and TLR4, with average RMSD values of 0.33 and 0.23 nm, respectively, maintained a stable structure throughout the simulation. In addition, the construct-TLR2 and –TLR4 complexes, with average RMSDs of 0.54 and 0.36 nm, respectively, indicated stable binding during the simulation (Fig. 8 A). As shown in the graph, the Rg of the construct, TLR2, and TLR4 alone and in interaction with each other indicate favorable stability, as lower Rg values indicate a more compact structure (Fig. 8B).

Fig. 8.

Molecular dynamics simulation (100 ns) assessing stability of the vaccine construct and its complexes with TLR2 and TLR4. A Root-mean-square deviation (RMSD) of the free vaccine construct (yellow) and complexes with TLR2 (blue) and TLR4 (orange), showing all trajectories remain below 0.8 nm, indicating high structural stability. B Radius of gyration (Rg) showing compact and stable folding of the vaccine alone and in complexes throughout the simulation. C Root-mean-square fluctuation (RMSF) per residue for the free vaccine construct and the two complexes. Lower fluctuations in the interface regions confirm the vaccine’s stable binding to both TLR2 and TLR4

The average fluctuation of each residue during the simulation was plotted using RMSF. As shown in Fig. 8 C, the RMSF values show small fluctuations (less than 3 Å) for most amino acids in the protein-TLR4 and protein-TLR2 complexes and unbound structures. These results show that the designed construct was stable in interaction with the immune system receptors, TLR2 and TLR4, and could potentially trigger an immune response.

Binding free energy calculation

The MM-PBSA analysis demonstrated favorable binding of the designed vaccine to both TLR2 and TLR4, with total binding free energies of − 85.79 kJ/mol and − 75.95 kJ/mol, respectively (Table 5).

In both complexes, the dominant stabilizing forces were van der Waals interactions and electrostatic contributions, reflected by their large negative values. Although the polar solvation energy (GB term) contributed positively and opposed binding, the strong vdW and electrostatic components outweighed this effect, resulting in overall negative binding energies for both receptors.

The vaccine exhibited slightly stronger affinity for TLR2 than for TLR4, as indicated by a more negative ΔG (–85.79 vs. − 75.95 kJ/mol, respectively). The high electrostatic stabilization observed in the TLR2 complex (–343.56 kJ/mol) suggests more favorable charge–charge interactions relative to the TLR4 complex. Overall, the MM-PBSA results support the thermodynamic stability of vaccine–TLR interactions and confirm that both TLR2 and TLR4 can effectively recognize and bind the vaccine construct, consistent with the expected mechanism of innate immune activation.

Reverse translation, codon optimization, and in Silico cloning

Codon adaptation of the designed vaccine construct was performed using the JCat server based on the codon usage of the E. coli strain K12. The total length of the vaccine sequence was 1059 nucleotides. Moreover, the Codon Adaptation Index (CAI) of the improved sequence was 0.97, indicating a high expression rate in the vector. The Guanine-Cytosine (GC)-content was estimated at 51.65%, which is within the optimal range (30–70%) for a desirable expression [56], and shows a high probability of expression in the E. coli K12 strain (Supplementary data 6). The adapted vaccine sequence was inserted into the pET28a (+) expression vector between the BamHI and HindIII restriction sites using the SnapGene tool (Fig. 9).

Fig. 9.

In silico cloning of the epitope vaccine sequence within the pET28a (+) vector through the SnapGene software free trial. The segment colored in red is the nucleotide sequence of the Brucella vaccine, and the rest of the circle indicates the backbone of pET28a (+)

Discussion

Brucellosis, also known as Malta fever, causes approximately 500,000 cases globally each year. However, it is believed that the actual incidence ranges from 5,000,000 to 12,500,000 [57]. This vast discrepancy underscores the urgent need for effective preventive strategies. Although vaccination remains the most reliable approach for controlling brucellosis transmission, no licensed human vaccine is currently available [58, 59]. In the current study, we designed a multi-epitope vaccine against four common Brucella pathogen species, with a focus on the conserved regions of the targeted proteins.

Multiple strategies exist for antigen selection in multi-epitope vaccine development, ranging from genome-wide reverse vaccinology to targeted antigen selection [60]. Genome-wide approaches are relatively unbiased and can identify novel candidates, but they often produce large datasets requiring extensive computational and experimental validation [61]. In contrast, targeted strategies focus on well-characterized antigens, offering stronger biological confidence but potentially limiting innovation. High-dimensional omics technologies can further refine antigen discovery, although their application in bacterial vaccines is limited by cost, complexity, and translational challenges [62]. In this study, a targeted antigen selection strategy was employed, integrating a comprehensive literature review with systematic in silico immunoinformatics filtering. OMP2a, OMP2b, OMP10, OMP25, and OMP31 were selected due to their phylogenetic tree, immunogenicity, association with virulence, and conservation across Brucella species. This approach balances feasibility with biological relevance, aiming to enhance translational potential while avoiding the limitations of purely genome-wide or overly restrictive single-antigen strategies [60].

Several recent studies have proposed multi-epitope vaccines against Brucella. Still, they generally target a single species, use only a limited number of proteins, or do not fully integrate advanced immunoinformatics workflows. For instance, the multi-epitope vaccine presented by Li et al. (2022) [59] focused on B. melitensis-specific epitopes and did not emphasize cross-species conservation. In contrast, our vaccine design uniquely integrates conserved regions from OMP2a, OMP2b, OMP10, OMP25, and OMP31, producing a construct with high similarity (> 70–97%) across four pathogenic Brucella species. Moreover, while a study by Wu et al. [63] integrated multi-species epitope mapping of candidate proteins, it didn’t focus on the conserved sequences of these candidates across species to enhance their protective efficacy. The use of multiple sequence alignment makes the design unique, enabling the development of a low-cost vaccine with broader efficacy by simultaneously targeting multiple species rather than focusing on just one [23, 59, 64]. Some previous studies have assessed epitope conservation using the IEDB Epitope Conservancy Tool or custom computational algorithms; however, these approaches typically rely on predefined epitope inputs or require specialized software development. By contrast, the alignment-based strategy employed in this study enables unbiased identification of conserved regions across Brucella species at the protein level, before epitope mapping. This ensures that epitope prediction is restricted to evolutionarily stable regions, thereby increasing the likelihood of broad cross-species protection. The alignment outcomes revealed over 70% similarity in OMP31, more than 80% in OMP2a, and more than 97% similarity in the OMP2b, OMP25, and OMP10 sequences among the four targeted species. This provides a broader protective spectrum compared with previously published constructs.

The phylogenetic analysis supported our antigen-selection strategy. OMP2a and OMP2b formed a tightly clustered pair, reflecting their high similarity, whereas OMP10, OMP25, and especially OMP31 showed greater divergence. This pattern guided epitope selection by allowing us to prioritize conserved regions within OMP2a/OMP2b while incorporating divergent motifs from OMP10 and OMP31 to broaden antigenic coverage across Brucella species.

Consistent with previous computational studies [48–51], our approach integrated T-helper, cytotoxic T-cell, and B-cell epitope mapping to activate cellular and humoral responses. However, unlike earlier studies, which lacked extensive cytokine-induction prediction, our analysis evaluated the induction potential of IL-2, IL-4, IL-10, and IFN-γ for each epitope, providing a more comprehensive immunogenicity profile. Extended HLA coverage analysis further strengthens the translational potential of our construct. Compared with earlier multi-epitope vaccines [23, 59, 63, 65], we have conducted population coverage analysis, which provides valuable information on vaccine efficacy. Compared with a study by Tarrahimofrad et al. [59], our vaccine showed broader global coverage of MHCI, while revealing similar population coverage of MHCII.

The vaccine was constructed using “GGSSGG” as a flexible linker between epitopes and “EAAAK” as a flexible linker at the N-terminus between the vaccine sequence and the adjuvant. Subsequently, the vaccine sequence was modeled, refined, and docked to TLR2 & TLR4 to evaluate their potential to start an immune response. The docking results indicated a high affinity between the vaccine construct and the TLR2 & TLR4 binding sites, with numerous hydrogen bonds. Furthermore, the vaccine’s similarity to the human proteome was investigated by PIR, which reported no similarity between the vaccine sequence and the human proteome, thereby avoiding autoimmune responses. Additionally, immune simulation analysis was conducted to predict cytokine and immunoglobulin induction. The vaccine construct successfully activated lymphocytes and led to primary and secondary immune responses against Brucella [66].

In addition to immunological features, the physicochemical properties of the construct were evaluated and reported. The Brucella vaccine construct exhibited high antigenicity, solubility, and favorable non-allergenicity, as supported by the findings. Based on Expasy ProtParam analysis, the predicted half-life of the vaccine was 7.2 h in mammalian reticulocytes, over 20 h in yeast, and 10 h in E. coli. Moreover, with an instability index of 34.41, the protein was classified as stable.

MD simulations revealed stable RMSD, Rg, and RMSF values for both the vaccine construct and its complexes with TLR2 and TLR4 throughout the 100-ns simulation. These results indicate strong structural integrity and stable receptor engagement. The MD simulation-assessed values were quite similar to those reported in previous studies [65, 66]. MM-PBSA results further confirmed the stability of vaccine–TLR interactions. The construct showed favorable binding to both receptors, with total binding energies of − 85.79 kJ/mol for TLR2 and − 75.95 kJ/mol for TLR4. In both complexes, van der Waals and electrostatic forces were the main contributors to stabilization, overcoming the unfavorable polar solvation term. Overall, the designed construct maintained stable conformations in isolation and when interacting with TLR2 and TLR4, suggesting sufficient structural integrity to elicit an immune response. Finally, codon optimization yielded a CAI of 0.97 and a GC content of 51.65%, suggesting efficient expression of the vaccine sequence in the E. coli K12 strain.

Conclusions

Current Brucellosis vaccines, while effective in reducing the incidence of invasive Brucella infections, have certain limitations. This study aimed to design a novel epitope-based vaccine to address these shortcomings. Immunodominant, epitope-rich regions from highly protective Brucella antigens—Omp2a, Omp2b, Omp10, Omp25, and Omp31—were identified using immunoinformatics tools and linked together with suitable linkers. Additionally, the potential adjuvant CtxB was incorporated to enhance the vaccine’s immunogenicity. The final construct was predicted to be antigenic, non-allergenic, and exhibited favorable physicochemical properties. Molecular docking analyses confirmed its binding affinity for TLR2 and TLR4 receptors, while in silico immune simulations predicted a robust immune response following vaccination. MD simulations further validated the stability of the docked complexes, and codon optimization suggested efficient expression of the vaccine. Although these in-silico results are promising, the efficacy and safety of the designed vaccine require confirmation in laboratory and animal studies.

Authors’ contributions

Author contributions. A.D. Writing the manuscript draft, data collection, and analysis. H. M. Thesis supervisor, conceptualizing, designing, and study implementation, data collection and analysis, writing and editing the manuscript. A.S.H.‘s contribution in immune simulation, M.R. and M.M.‘s writing and editing of the manuscript, and K.A. and B.S.‘s Collaboration in MD simulation and analysis.

Funding

This project was approved and supported by the Semnan University of Medical Sciences and Student Research Committee (grant No. 3848).

Data availability

Data is contained within the article or supplementary material.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Conflict of interest

All authors declare that there are no competing or financial interests, and the authors confirm that this article’s content has no conflict of interest.

Competing interests

The authors declare no competing interests.