2.1. Protein Retrieval and Human Homology Screening

The complete proteome of M. tuberculosis strain KZN 1435 was retrieved from the NCBI GenBank database (https://www.ncbi.nlm.nih.gov/genome/; accession ID: GCA_000023625.1). Initially, all predicted protein sequences were subjected to signal peptide screening using five independent web servers: TargetP 2.0, DeepSig, SignalP 6.0, Phobius, and Signal-BLAST [12–16]. Only proteins consistently predicted as secretory by all five tools were retained for further analysis. Only proteins consistently predicted as secretory by all five tools were retained for further analysis [12,15,16]. After applying this initial secretory-protein filtering step, 84 non-redundant secretory candidate proteins were selected for subsequent immunoinformatic analysis. These 84 candidate proteins were screened against the Homo sapiens proteome using BLASTp (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Proteins with E-values <1e−4 were excluded to avoid host–self reactivity.

2.2. Transmembrane Topology Prediction

Filtered proteins were analyzed using DeepTMHMM to determine the number and position of predicted transmembrane helices. Proteins with zero or one predicted transmembrane helix and favorable extracellular or surface-associated topology were prioritized for downstream epitope prediction. This filtering step was used to enrich proteins with potential immune accessibility and to exclude proteins with extensive transmembrane regions that may be less suitable for multiepitope vaccine design [17].

2.3. Epitope Prediction

Linear B-cell epitopes were predicted using ABCPred (https://webs.iiitd.edu.in/raghava/abcpred/) with a 12-mer window to match the epitope length used in the final screening table and vaccine construct [18]. CTL epitopes were predicted using NetMHCpan-4.1 (https://services.healthtech.dtu.dk/service.php?NetMHCpan-4.1) against selected HLA class I supertypes, with peptide length set to 12 amino acids [19]. HTL epitopes were predicted using the IEDB MHC-II binding tool with the IEDB recommended method, and the peptide length was set to 12 amino acids to maintain consistency with the final epitope format used in Table 1 and in the multiepitope construct. Only 12-aa epitope candidates were retained for downstream validation and vaccine construction [20].

Table 1. Summary of high-confidence epitope-protein entries with HLA binding or B-cell prediction scores, antigenicity, allergenicity, and toxicity. [Click here to view]

2.4. Epitope Validation: Antigenicity, Allergenicity, and Toxicity

Each predicted epitope was evaluated using: VaxiJen v2.0 for antigenicity [21], AllergenFP v1.0 (http://ddg-pharmfac.net/AllergenFP/) for allergenicity [22], and ToxinPred (https://webs.iiitd.edu.in/raghava/toxinpred/) for toxicity [23]. Epitopes were retained only if they were predicted to be antigenic, non-allergenic, and non-toxic. HLA-binding results were used to prioritize T-cell epitopes with favorable predicted binding profiles. Population coverage and epitope conservancy were not used as final selection claims unless supported by explicit numerical analysis and reported data [24].

2.5. Vaccine Construct Design

Validated CTL, HTL, and B-cell epitope candidates were evaluated for redundancy before vaccine construction. The screening process yielded eight high-confidence epitope-protein entries corresponding to six non-redundant peptide sequences. Of these six non-redundant sequences, five epitopes were incorporated into the final 126-residue construct used for downstream physicochemical, structural, docking, and codon optimization analyses. The HTL epitope HADRELASQTEA was retained in Table 1 as a high-confidence candidate but was not included in the final construct architecture. Identical peptide sequences predicted from different protein entries were counted as separate epitope-protein entries in the screening table but were not repeated in the final chimeric construct. The final multiepitope vaccine construct was assembled using non-redundant epitope sequences selected for downstream structural, docking, and codon optimization analyses. The human β-defensin 3 (hBD-3) adjuvant sequence was placed at the N-terminus and connected to the first CTL epitope using the rigid EAAAK linker to maintain functional separation between the adjuvant and epitope region. The AAY linker was used to facilitate proteasomal processing of CTL-related regions, whereas GPGPG linkers were used to separate HTL and B-cell epitope regions and improve independent epitope presentation. The final construct architecture was arranged as follows: hBD-3 adjuvant–EAAAK–CTL epitope–AAY–HTL epitope–GPGPG–B-cell epitope–GPGPG–HTL epitope–AAY–B-cell epitope.

2.6. Physicochemical Characterization and Solubility Profiling

The physicochemical properties of the final multiepitope vaccine construct were calculated using the ExPASy ProtParam server [25]. Protein solubility was predicted using SoluProt [26], Protein-Sol [27], and SoDoPE [28] to assess the theoretical feasibility of soluble recombinant expression in Escherichia coli.

2.7. Tertiary Structure Prediction

The tertiary structure of the multiepitope vaccine construct was predicted using the I-TASSER server (https://zhanggroup.org/I-TASSER/), which generates structural models through template identification from the Protein Data Bank followed by iterative fragment assembly simulations [29]. The predicted models were ranked based on C-scores, with higher values indicating greater structural reliability. Model quality was further evaluated using the QMEAN scoring function implemented in the SWISS-MODEL Structure Assessment tool, which provides global and local estimates of structural accuracy based on statistical potentials derived from experimentally resolved protein structures [30]. The QMEAN Z-score was used to assess agreement between the predicted structure and native proteins of similar size, while residue-level quality profiles supported the overall structural consistency of the model.

2.8. Molecular Docking with Immune Receptors

Molecular docking was performed to computationally evaluate the interaction between the multiepitope vaccine construct and innate immune receptors using the CB-Dock server (http://clab.labshare.cn/cb-dock/) [31], which enables automated blind docking through cavity detection and AutoDock Vina-based scoring [31]. The refined vaccine structure in PDB format was docked against the TLR2-containing human TLR1–TLR2 heterodimer complex (TLR2; PDB ID: 2Z7X) and human TLR4/MD-2/LPS complex (TLR4; PDB ID: 4G8A). Docked complexes were ranked according to their docking score or binding energy values, and the top-scoring conformations were selected for further interaction analysis. In this study, the 126-residue vaccine construct was treated as a macromolecular peptide ligand for computational receptor-compatibility screening against TLR2 and TLR4. Intermolecular interaction interfaces between the vaccine construct and TLR receptors were subsequently evaluated using the Protein–Ligand Interaction Profiler server to identify hydrogen bonds, hydrophobic contacts, and other non-covalent interactions [32]. The docking results were interpreted only as computational indications of receptor compatibility and binding tendency and were not considered experimental evidence of TLR activation [32].

2.9. Codon Optimization and in silico Cloning

Reverse translation and codon optimization of the final vaccine construct for theoretical expression in E. coli K12 were performed using the JCat server [33]. The optimized nucleotide sequence was then assessed by in silico cloning simulation using Benchling as an online DNA design and sequence-annotation platform. The optimized insert was virtually mapped into pBR322 for cloning-vector simulation and into pET-28a(+) for expression-cassette simulation. NdeI and SacI restriction sites were introduced at the 5′ and 3′ ends of the optimized gene sequence, respectively, to support directional cloning simulation. The open reading frame, His-tag placement in pET-28a(+), start codon, stop codon, and restriction-site orientation were confirmed through plasmid-map visualization and in silico translation. This analysis was used only to assess the theoretical feasibility of cloning and expression-cassette design; actual recombinant expression, solubility, and protein yield require experimental validation.

3.1. Protein Retrieval and Human Homology Screening

An integrated immunoinformatic pipeline was applied to design a multiepitope vaccine construct targeting M. tuberculosis strain KZN 1435. The complete proteome was first screened for secretory potential using TargetP 2.0, DeepSig, SignalP 6.0, Phobius, and Signal-BLAST. This initial filtering step retained 84 non-redundant secretory candidate proteins for further analysis [12,15,16]. Only proteins consistently predicted as secretory by all five tools were retained. To minimize the risk of host cross-reactivity, homology screening against the H. sapiens proteome was performed using BLASTp, resulting in the exclusion of six homologous proteins. The remaining 78 non-homologous candidate proteins were selected for downstream topology screening and epitope prediction.

3.2. Transmembrane Topology Prediction

Transmembrane topology screening using DeepTMHMM [17] identified 44 proteins containing zero or one predicted transmembrane helix, indicating potential extracellular or surface-associated accessibility. Detailed topology analysis of protein candidate four revealed a single high-confidence transmembrane helix located approximately between residues 20 and 50 (Fig. 1), with the remaining sequence predominantly predicted to be extracellular based on DeepTMHMM analysis [17]. This topology profile was used to prioritize protein regions with potential surface-associated accessibility for downstream epitope mapping. Overall, transmembrane topology prediction using DeepTMHMM provided a reliable framework for prioritizing immunologically relevant surface-associated proteins for downstream epitope mapping.

Figure 1. DeepTMHMM transmembrane topology prediction of protein 4 from M. tuberculosis KZN 1435. [Click here to view]

3.3. Epitope Prediction and Validation

Epitope prediction was performed using complementary immunoinformatic approaches. Linear 12-mer B-cell epitopes were predicted using ABCPred, whereas 12-mer CTL and HTL epitopes were identified using NetMHCpan-4.1 and the IEDB MHC-II binding tool, respectively. All predicted epitopes were further screened for antigenicity, allergenicity, and toxicity using VaxiJen, AllergenFP, and ToxinPred. Following multilayered filtering, 13 proteins were found to contain antigenic, non-allergenic, and non-toxic epitopes. From these candidates, eight high-confidence epitope-protein entries with favorable predicted HLA-binding or B-cell prediction scores were prioritized.

Among these eight entries, two peptide sequences were identified in more than one protein entry. The CTL epitope VMPGDGEVVGVG was detected in protein entries 4 and 29, whereas the HTL epitope IGQTGMPSSFSM was detected in protein entries 8 and 34. Therefore, the screening process yielded eight high-confidence epitope-protein entries corresponding to six non-redundant peptide sequences. Of the six non-redundant peptide sequences identified from the eight epitope-protein entries, five epitopes were incorporated into the final 126-residue construct used for downstream physicochemical, structural, docking, and codon optimization analyses. The HTL epitope HADRELASQTEA was retained in Table 1 as a high-confidence candidate but was not included in the final construct architecture. This clarification avoids ambiguity between epitope entries identified during screening and peptide sequences incorporated into the final multiepitope vaccine construct.

3.4. Vaccine Construct Design

A multiepitope vaccine construct was rationally designed by integrating selected non-redundant CTL, HTL, and B-cell epitopes into a single chimeric polypeptide using an adjuvant sequence and linkers selected to support antigen processing and epitope separation. The N-terminal region incorporated a hBD-3 adjuvant sequence as an immunostimulatory component. The adjuvant was connected to the CTL epitope VMPGDGEVVGVG using an EAAAK linker to maintain structural separation between domains. The remaining epitope regions were arranged sequentially using AAY and GPGPG linkers to support appropriate antigen processing and presentation through MHC class I and II pathways. The final construct included the CTL epitope VMPGDGEVVGVG, HTL epitopes IGQTGMPSSFSM and HHGTDPAAAVGL, and B-cell epitopes PGDLGSTQSKQA and GSIDKIEPAGDK. The full primary amino acid sequence of the final multiepitope vaccine construct is provided below to ensure reproducibility: GIINTLQKYYCRVRGGRCAVLSCLPKEEQIGKCSTRGRKCCRRKKEAAAKVMPGDGEVVGVGAAYIGQTGMPSSFSMGPGPGPGDLGSTQSKQAGPGPGHHGTDPAAAVGLAAYGSIDKIEPAGDK. This modular arrangement of the hBD-3 adjuvant [34], multiclass epitopes, and linkers was designed to theoretically support epitope separation and coordinated cellular [35] and humoral immune recognition against M. tuberculosis.

3.5. Physicochemical Characterization and Solubility Profiling

The designed multiepitope vaccine construct, comprising 126 amino acid residues, was evaluated for key physicochemical and structural properties (Table 2). The predicted instability index was 39.44, suggesting that the construct may be stable under computational assessment. The GRAVY value (–0.404) suggested a hydrophilic profile consistent with favorable solubility. Solubility prediction using SoDoPE yielded a score of 0.809, suggesting theoretical feasibility for soluble expression in E. coli. The predicted half-life in E. coli exceeded 10 hours, indicating favorable theoretical stability in a bacterial expression system. Structural assessment additionally indicated that most residues were surface-exposed, supporting accessibility for immune recognition. These predicted properties suggest that the designed construct has favorable physicochemical features for subsequent experimental evaluation.

Table 2. Key physicochemical and structural features of the multiepitope vaccine construct. [Click here to view]

3.6. Secondary and Tertiary Structure Prediction

Figure 2a shows the refined three-dimensional structure of the multiepitope vaccine construct displayed in cartoon representation with a rainbow gradient from the N-terminus (blue) to the C-terminus (red). The model exhibited a compact predicted folding architecture consisting of α-helices, β-strands, and connecting loop regions, suggesting an acceptable structural arrangement for downstream computational analyses. Figure 2b presents the Ramachandran plot used to evaluate backbone stereochemical quality. The majority of residues were located within favored and additionally allowed regions, with only a small proportion observed in disallowed regions, primarily corresponding to flexible loop or terminal segments. These results suggest an acceptable backbone geometry for the predicted structural model.

Figure 2. Representative refined 3D model of the multiepitope vaccine construct (a) and Ramachandran plot validation (b). [Click here to view]

Figure 3 illustrates residue-level structural confidence together with predicted secondary structure annotation. Lower residue-wise error values across most positions suggest relatively favorable local structural confidence, while the distribution of α-helices, β-strands, and coil regions confirms the presence of ordered structural elements required for maintaining antigen presentation stability. Figure 4 presents the predicted normalized B-factor profile and secondary-structure annotation of the refined vaccine construct. Most regions showed relatively low normalized B-factor values, suggesting limited local flexibility and a generally stable modeled conformation, whereas several peak regions indicated more flexible segments within the construct. The accompanying secondary-structure annotation further illustrates the distribution of predicted helical, strand, and coil regions across the vaccine sequence. Together, these structural features provide computational support for the overall conformational stability and local flexibility pattern of the designed multiepitope construct.

Figure 3. Per-residue confidence score and predicted secondary structure of the vaccine construct. [Click here to view]

Figure 4. Predicted normalized B-factor profile and secondary-structure annotation of the refined multiepitope vaccine construct. [Click here to view]

3.7. Molecular Docking with Immune Receptors

Figure 5 illustrates the selected docking conformations and residue-level interaction profiles of the vaccine construct with the TLR2-containing TLR1–TLR2 heterodimer and the TLR4/MD-2 complex. In the TLR1–TLR2 docking model, hydrophobic interactions were observed with residues ILE114A, LEU117A, LEU138A, and PHE144A at interaction distances ranging from 3.60 to 3.82 Å. These residue-level interactions were generated from the docking and interaction profiling analyses performed in the present study and should therefore be interpreted as original computational results. The observed contacts were located within the predicted receptor–construct interface and may contribute to theoretical receptor accommodation through non-covalent interactions.

Figure 5. Molecular docking visualization of the designed multiepitope vaccine construct with the TLR2-containing human TLR1–TLR2 heterodimer complex (PDB ID: 2Z7X) (a) and the human TLR4/MD-2/LPS complex (PDB ID: 4G8A) (b), including residue-level interaction mapping. [Click here to view]

Similarly, docking with the TLR4/MD-2 complex revealed hydrophobic interactions involving residues PHE358A, ASN361A, ASN365A, and ASN383A, with interaction distances between 3.61 and 3.75 Å. The presence of these predicted contacts suggests theoretical compatibility between the vaccine construct and the selected receptor models. However, experimental receptor-binding and immune activation assays are required to confirm actual receptor engagement and downstream immunological effects.

TLR2 was selected as a primary receptor target due to its established role in recognizing M. tuberculosis-associated molecular patterns and initiating downstream signaling pathways that promote proinflammatory cytokine production and antigen-presenting cell activation [36]. The observed docking profile suggested compatible receptor accommodation of the designed construct. However, these findings remain computational predictions and do not confirm receptor activation, cytokine induction, or downstream adaptive immune stimulation [36]. The complementary docking pattern observed with TLR4 may indicate additional theoretical receptor compatibility, but experimental receptor-binding and immune activation assays are required for confirmation.

3.8. Codon Optimization and in silico Cloning

Codon optimization was performed to adapt the multiepitope vaccine gene for theoretical expression in E. coli. The optimized sequence yielded a codon adaptation index of 1.0 and a GC content of 54.76%, suggesting compatibility with the codon usage pattern of the selected bacterial host [33]. These parameters support the theoretical feasibility of recombinant expression, although actual expression level and protein yield require experimental validation. Virtual cloning of the optimized gene into the pBR322 vector indicated correct insert placement within the simulated recombinant construct. The presence of selectable markers bla and tetR indicates that the simulated plasmid design is compatible with antibiotic-based selection. Simulated restriction digestion using BamHI, EcoP15I, and FspAI produced fragment sizes consistent with correct gene insertion and orientation, supporting the predicted integrity of the simulated recombinant construct (Fig. 6a).

Figure 6. In silico cloning and restriction analysis of the multiepitope vaccine construct in pBR322 (a) and pET-28a(+) (b) expression vectors. [Click here to view]

To further evaluate expression feasibility under inducible high-yield conditions, the optimized gene was also cloned into the pET-28a(+) expression vector downstream of the T7 promoter. This construct contains elements compatible with IPTG-inducible transcription and kanamycin-based selection. Virtual restriction digestion with BamHI and EcoRV generated fragment patterns consistent with successful ligation of the vaccine cassette (Fig. 6b), supporting its suitability for controlled recombinant protein production in E. coli BL21 (DE3). The use of pBR322 and pET-28a(+) provided complementary in silico cloning strategies for evaluating plasmid maintenance-oriented and expression-oriented construct designs. The predicted restriction patterns and regulatory compatibility of both constructs support their theoretical feasibility for downstream experimental testing, including transformation and recombinant expression assays [37].

3.9. Limitations and Future Perspectives

Although this study provides a rational immunoinformatic framework for designing a strain-specific multiepitope vaccine candidate against M. tuberculosis KZN 1435, several limitations should be acknowledged. First, the antigenicity, allergenicity, toxicity, HLA binding, structural modeling, receptor interaction, and codon optimization results were generated using computational prediction tools. Second, the actual expression level, solubility, folding behavior, and stability of the recombinant vaccine protein in E. coli have not yet been validated through in vitro cloning and protein expression assays. Third, the predicted interaction of the vaccine construct with TLR2 and TLR4 provides preliminary computational indications of potential receptor compatibility, but receptor activation, cytokine induction, antigen presentation, and downstream adaptive immune responses remain to be verified using cell-based assays and animal models. Fourth, although the selected epitopes were screened for non-allergenicity and non-toxicity, the safety and immunogenicity of the complete construct must be confirmed experimentally before translational development.

Future studies should therefore focus on in vitro cloning, recombinant expression, purification, and biochemical characterization of the vaccine construct. Subsequent immunological validation should include assessment of cytokine responses, T-cell activation, antibody production, and protective efficacy using appropriate in vitro and in vivo models. These experimental studies will be essential to determine whether the computationally designed construct can elicit protective immune responses against drug-sensitive and MDR M. tuberculosis strains.