Tuberculosis mRNA vaccine candidate: predicted immune responses and structures vary with antigen order
Background: Can the same antigens yield different results in a different order?
In an mRNA vaccine design combining three Mycobacterium tuberculosis antigens, predicted antigenicity and immune responses varied with antigen order. The researchers selected Pattern 3, PPE68–IrtA–PE9, as a candidate for further evaluation, although it did not perform best on every measure. The study, published in Scientific Reports, was limited to computational design and analysis. It did not test immune responses after vaccination or protection against tuberculosis.
BCG offers some protection against severe tuberculosis in children, but its effectiveness against adult pulmonary tuberculosis varies widely. Identifying antigens that can elicit protection remains a challenge in developing new vaccines. The researchers compared three fusion constructs combining the Mycobacterium tuberculosis H37Rv antigens IrtA, PE9, and PPE68 with the same amino acid composition. Pattern 1 was IrtA–PE9–PPE68, Pattern 2 was IrtA–PPE68–PE9, and Pattern 3 was PPE68–IrtA–PE9. The study did not examine every possible order.
Key findings: Comparative advantages differed by measure
Tools from the Immune Epitope Database (IEDB) and BepiPred-3.0 were used to predict candidate regions recognized by T cells and B cells, and AlphaFold 3 was used to model the constructs’ three-dimensional structures. VaxiJen predicted antigenicity scores of 0.6103, 0.6031, and 0.6122 for Patterns 1, 2, and 3, respectively, with Pattern 3 scoring highest. This difference does not establish a difference in the magnitude of an actual immune response or a statistically significant difference in efficacy. The reported value of 0.865 for Pattern 3 was the ElliPro score of its top predicted B-cell epitope, located at amino acid residues 274–372. It should not be interpreted as an epitope proportion or an antibody response rate.
Molecular docking with toll-like receptors (TLRs) also yielded different comparative advantages. For TLR4/MD-2, Pattern 2 had the lowest predicted binding free energy, −14.4 kcal/mol; for TLR2, Pattern 1 had the lowest, −13.2 kcal/mol. Pattern 3’s values were −12.2 and −10.4 kcal/mol, respectively. In the TLR4/MD-2 complex, Pattern 3 was predicted to form 12 hydrogen bonds, forming hydrogen bonds with both TLR4 and MD-2. In a separate TLR2 complex, 9 hydrogen bonds were predicted. These results alone do not demonstrate receptor activation or intracellular signaling.
The immune simulation assumed three doses given four weeks apart. The three antigen orders showed similar patterns in memory cytotoxic T-cell responses and interferon gamma (IFN-γ) levels. Pattern 2 maintained higher memory helper T-cell levels, while Pattern 3 was predicted to have a higher peak antibody titer and a more sustained response. All of these simulated results must be distinguished from responses measured after administration to animals or people.
Implications and outlook: Validation remains after candidate selection
The minimum free energy predicted by RNAfold for the codon-optimized mRNA encoding Pattern 3 was −2,217.20 kcal/mol. This is an energy value for a calculated secondary structure. It cannot serve as evidence of reduced intracellular degradation, increased protein expression, manufacturing stability, or easier distribution.
Structural and physicochemical predictions also revealed development constraints. Pattern 3 had greater uncertainty in the relative positioning of domains than the other antigen orders, and all three constructs had a low predicted solubility score of 0.361. Computational classifications of allergenicity and toxicity also cannot substitute for biological safety testing. These findings provide a basis for including antigen order as a comparison variable in follow-up experiments. Actual expression levels, antigen processing, immune activation, and protection in the body remain to be assessed. The study methods and comparative results are available in the full paper.
Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a major global health challenge due to the limited efficacy of the Bacillus Calmette-Guérin (BCG) vaccine. Using an immunoinformatics-driven strategy, we designed and evaluated three distinct multi-epitope vaccine constructs (MEVCs) derived from PPE68, IrtA, and PE9, which were subsequently developed into an mRNA vaccine construct. T-cell and B-cell epitopes were predicted using IEDB tools and BepiPred-3.0, and the designed constructs were systematically evaluated for antigenicity, allergenicity, toxicity, and physicochemical characteristics. Structural modeling with AlphaFold3, followed by epitope mapping and molecular docking with TLR2 and TLR4/MD-2, identified Pattern 3 (PPE68-IrtA-PE9) as the most promising construct. It exhibited the highest antigenicity score (0.6122), a high abundance of B-cell epitopes (0.865), and demonstrated predicted binding to the TLR4/MD-2 complex (ΔG = - 12.2 kcal/mol), forming 12 hydrogen bonds and engaging both receptor components, as well as to TLR2 (ΔG = - 10.4 kcal/mol) with nine hydrogen bonds. In silico immune simulations of Pattern 3 predicted strong T-cell responses, elevated IFN-γ levels, and high IgG1, IgG2, and IgM titers, while the codon-optimized mRNA exhibited a stable secondary structure (ΔG = - 2,217.20 kcal/mol). These results suggest that antigen domain arrangement may influence predicted immunogenicity and structural stability, and exhibit a favorable in silico safety profile, supporting PPE68-IrtA-PE9 as a promising mRNA vaccine design for further experimental evaluation.
Vaccine development teams can treat antigen order as an experimental variable even when fusing the same antigens. For example, comparing expression levels, antigen processing, and immune responses for the three orders under the same conditions could help assess whether computational differences translate into actual biological differences. In this study, the candidate with the highest antigenicity score differed from those with the most favorable receptor-binding energies. Candidate selection therefore requires considering multiple measures together. Pattern 3’s predicted antibody response should be considered alongside its structural uncertainty and low predicted solubility. These comparisons can help prioritize follow-up experiments, but they do not yet provide a basis for judging whether the candidate could replace BCG or prevent tuberculosis in people.