Development of a vaccine construct against Pneumocystis jirovecii pneumonia using computational tools Authors: Ragini Mishra, Nahid Akhtar, Jorge Samuel Leon Magdeleno, Abdul Rajjak Shaikh, Manik Prabhu Narsing Rao, Neeta Raj Sharma, Luigi Cavallo, Mohit Chawla Published: 2025 ABSTRACT Abstract Pneumocystis jirovecii poses a significant threat to immunocompromised individuals, necessitating the development of an effective vaccine. This study employs an immunoinformatics approach to design a promising vaccine candidate against P. jirovecii. Utilizing various computational tools, the study identified potential antigenic epitopes capable of eliciting immune responses within the P. jirovecii major surface glycoprotein C. The chosen epitopes were evaluated using computational tools for their allergenicity, interferon-γ and interleukin activation ability, and toxicity, ensuring the selection of immunogenic and safe candidates. These analyses led to the selection of 10 epitopes, which were then linked with adjuvants to model a potential vaccine candidate. Molecular docking and molecular dynamics simulations were performed in a solvent environment to investigate the binding interactions between the vaccine candidate and toll-like receptors, along with calculations of thermodynamic properties. Finally, in silico immune simulations were performed to analyze the immunogenic potential of the vaccine candidate. Future prospects include in vitro and in vivo validation of the vaccine candidate and the exploration of novel adjuvants to enhance its immunogenicity. This study contributes to the ongoing efforts to develop a preventive solution against P. jirovecii infections, addressing a critical gap in the protection of immunocompromised individuals. Introduction In the realm of opportunistic pathogens, Pneumocystis jirovecii stands as an enigmatic and elusive microorganism, challenging both clinicians and researchers alike. This unicellular fungus has garnered increasing attention over the years due to its profound impact on the health of immunocompromised individuals, particularly those with HIV/AIDS, organ transplant recipients, cancer patients, and patients undergoing immunosuppressive therapies or prolonged corticosteroid therapy [1]. Among children, it ranks as a prevalent contributor to pneumonia, with ~25% of pediatric cases exhibiting P. jirovecii detection via polymerase chain reaction. In the context of HIV patients, P. jirovecii is detectable in a range spanning from 20% to 69% of affected individuals [2]. Individuals with healthy immune systems can harbor P. jirovecii in their lungs without experiencing symptoms, potentially serving as carriers and unintentionally spreading the pathogen to those who are immunocompromised [3]. Notably, the introduction of highly active antiretroviral therapy has significantly altered the landscape of Pneumocystis pneumonia (PCP), reducing its incidence in HIV-infected individuals [4]. However, with the growing use of immunosuppressive agents in organ transplantation, cancer treatment, and autoimmune diseases, a new demographic at risk has emerged [5]. The most effective treatment against P. jirovecii infection is trimethoprim-sulfamethoxazole but its use can be limited due to allergic reactions, adverse effects and drug resistance [2, 3, 6]. Alternative therapies, such as pentamidine and primaquine, should be used with caution as they are also associated with adverse effects like renal toxicity, hypoglycaemia, prolonged QT syndrome and increased risk of hemolysis [3, 7, 8]. Similarly, other treatments including clindamycin and dapsone are also associated with adverse effects [6]. Hence, it is important to continuously explore new ways to prevent and treat the P. jirovecii infection. The genes responsible for encoding the major surface glycoproteins (Msg) make up ~3%–6% of the genome. An exceptional diversity of genes encoding the Msg superfamily has been reported, with each Pneumocystis species possessing between 64 and 179 unique genes. Furthermore, the conservation of most Msg families or subfamilies across species highlights the crucial role these encoded proteins likely play in the organism’s survival [9]. Msg is believed to play a pivotal role in interactions between the host and pathogen, potentially facilitating the evasion of host immune responses through antigenic variation. Specifically, P. jirovecii contains a total of 179 Msg genes, which are categorized into five families, namely MsgA, MsgB, MsgC, MsgD, and MsgE [9]. These Msg proteins are recognized by serum antibodies and helper T cells. A prior investigation has indicated that the B cell epitopes responsible for eliciting antibody responses may be found within MsgC [10]. In individuals with HIV who are experiencing active P. jirovecii pneumonia, there is a significant increase in antibody titers against MsgC compared to the HIV patients with pneumonia caused by other factors [11, 12]. In a different research, it was observed that MsgC, a segment of the Msg protein, consistently triggered elevated antibody levels and stood out as the primary antigen recognized by serum antibodies. Additionally, MsgC led to rapid and substantial increases in antibody levels compared to another fragment known as MsgA. Notably, children who experienced more frequent or severe respiratory infections displayed higher antibody levels targeting MsgC, suggesting it serves as a sensitive indicator of frequent or intense clinical exposure [13]. Considering these roles of MsgC in the immune response to P. jirovecii infection, these proteins were targeted for the vaccine design in this study using a computational immunoinformatics approach. Computational vaccine design has emerged as a powerful approach to accelerate the development of candidate vaccines, particularly for pathogens that are difficult to culture or lack effective treatments [14]. Traditional computational vaccine pipelines often rely on standard epitope prediction and docking approaches, but these methods require further optimization to improve accuracy, immunogenicity, and structural stability. This study integrates a multi-tier computational strategy, combining advanced epitope screening, molecular dynamics (MD) simulations, and in silico immune profiling, to design a vaccine candidate against P. jirovecii. Our approach refines epitope selection through multiple immunogenicity and safety filters, ensuring the most promising candidates are identified for vaccine development. This methodological framework enhances the precision of computational vaccine design, provides a reproducible workflow for similar pathogens, and contributes to the broader field of in silico vaccine development and offers a structured framework for future vaccine studies. Methodology Retrieval and analysis of protein sequence The P. jirovecii MsgC sequences were retrieved from GenBank with accession numbers XP_018 228 271 and XP_018230092.1 in FASTA format. Prediction of epitopes and determination of their antigenicity, toxicity, allergenicity, interferon-γ, and interleukin activation potential To anticipate the T-helper cell epitopes capable of binding to MHC class II molecules, the NETMHC 2.3 website was utilized. In contrast, in detecting the T-cytotoxic cell epitopes with potential binding to MHC class I molecules, inducing a cellular immune response, the NETMHC 4.0 website was employed [4]. The protein sequences, presented in FASTA format, were input into both NETMHC 2.3 and NETMHC 4.0 servers, selecting a peptide length of 9. Predictions for CD4 + T cell epitopes involved human leukocyte antigen (HLA) alleles such as DRB1_0101, DRB1_0301, DRB1_0401, DRB1_0701, DRB1_0801, DRB1_0901, DRB1_1001, DRB1_1101, DRB1_1201, DRB1_1301, DRB1_1501, and DRB1_1602. Additionally, predictions for CD8 + T cells considered HLA alleles like HLA-A0101, HLA A0201, HLA-A0301, HLA-A2402, HLA-A260, HLA-B0702, HLA-B0801, HLA-B2705, HLA-B3901, HLA-B4001, and HLA-B5801.The default thresholds for strong and weak binders in NETMHC 2.3 and NETMHC 4.0 were employed [15, 16]. Predicted epitopes were ranked based on their binding affinities, and weak binders were excluded to retain only strong-binding epitopes. Multiple linear B-cell epitopes were predicted using the ABCPred server, with the amino acid sequence input in one-letter code under default parameters [17]. Antigenicity of the forecasted B-cell and T-cell epitopes was assessed using the Vaxijen v2.0 web server, with the target organism set to fungi while keeping all other parameters at their default settings [18]. The AllergenFP, ToxinPred, and IFNepitope web servers were utilized with the default parameters to evaluate the epitopes for antigenic, toxic, allergenic, and interferon-γ activation potential [19–21]. Finally IL4Pred was used to determine the interleukin-4 (IL-4) elicitation potential of the epitopes under default parameters [22]. In the VaxiJen 2.0, AllergenFP, ToxinPred, and IL4Pred web servers, amino acid sequences were entered in single-letter code, whereas in IFNepitope, sequences were provided in FASTA format. Only epitopes fulfilling the following criteria were selected for further analyses: antigenic, non-allergenic, non-toxic, and capable of inducing IL-4 and IFN-γ responses. Epitopes-HLA docking The predicted epitope peptides were modeled using AlphaFold 3, and docking was performed with the HADDOCK 2.4 webserver. To validate the docking protocol and assess its ability to reproduce experimentally determined peptide–HLA interactions, a redocking procedure was first conducted using native co-crystallized peptides from each selected HLA structure. The docking grid was constrained to the space originally occupied by the peptides in their crystal structures to ensure accurate placement and interaction assessment. The resulting docked poses were evaluated based on HADDOCK scores and root-mean-square deviation (RMSD) values relative to the crystallographic conformations. Although HADDOCK scores are not direct measures of binding affinity, they were used as relative indicators of complex stability within each simulation. To further analyze the molecular interactions in the docked complexes, the ProLIF tool [23] was employed to characterize residue-level contacts between the peptides and HLA proteins. The interacting residues were color-coded according to their interaction types for visualization (see Fig. 1). Additionally, to evaluate the dynamic stability of the peptide–HLA complexes, MD simulations were performed following a protocol similar to that described for the TLR5–MEV complex (see MD simulation section 2.7), focusing on hydrogen bonding, receptor flexibility, and peptide mobility. Docking poses and binding energies of the top-ranked epitope–HLA complexes predicted using HADDOCK. Each panel illustrates the best-scoring conformation between selected epitopes and their respective HLA alleles (PDB IDs: 4O2F, 4U6Y, and 6Y2A). The interactions of the epitope peptides with their respective HLA proteins were further analyzed using the ProLIF tool [23], and the residues were color-coded according to their interaction types. Vaccine engineering and physicochemical attributes determination For the final vaccine construct, the selected B-cell and T-cell epitopes were linked to adjuvants, including Salmonella dublin flagellin protein, RS09 (APPHALS), and the Pan HLA DR-binding epitope (PADRE) by employing “GGS” linkers. The physicochemical properties, allergenic potential, and antigenicity of the final P. jirovecii vaccine candidate were assessed by entering the amino acid sequence in single-letter sequence to ExPASyProtParam, AllergenFP, and Vaxijen 2.0, respectively, with default parameters. Computational immunosimulation analysis To explore the immune response profile induced by the P. jirovecii vaccine construct, in silico immune simulation was conducted using the C-IMMSIM online tool [24]. The default settings were applied, with the exception of the time step. It is recommended to administer consecutive vaccine doses with a minimum interval of 4 weeks, although in certain scenarios, intervals of 8 weeks to 6 months may be considered [25, 26]. Therefore, the immune response profile for the P. jirovecii vaccine candidate was assessed by administering three vaccine doses at 4-week intervals. Time steps of 1, 84 (equivalent to 4 weeks), and 168 (equivalent to 8 weeks) were employed for the simulation. Modeling and docking of TLR5 with MEV construct We conducted comprehensive 3D structure prediction for the P. jirovecii multi-epitope vaccine (MEV) and human TLR5 (UniProt ID: D1CS82) using AlphaFold 3.0 [27]. The quality of the predicted tertiary structures of MEV and TLR5 was meticulously scrutinized using the Qualitative Model Energy Analysis webserver [28]. The structural validation of the vaccine construct involved a Ramachandran plot analysis and ProSA analysis [29]. In the docking of MEV and TLR5, we harnessed the HADDOCK server [30]. An information-driven docking strategy was applied, guided by key interacting residues. Peptides LQRVRELAVQ and EILDISRNQL were used to define active site residues corresponding to the binding region of flagellin (within the MEV construct) and human TLR5. Following this approach, the top-ranked cluster with the lowest HADDOCK score emerged as the definitive structure for further detailed analyses. Molecular dynamics simulations study In this study, MD simulations were conducted using the GROMACS 2023 simulation software [31], with parameters adopted from the methodology described by Chawla et al. [32]. The system, comprising the MEV complexed with the TLR5 receptor, was placed in a cubic simulation box and immersed in a solvating environment consisting of atomistic TIP3P and coarse-grain WatFour (WT4) water models [33]. The SIRAH force field [34] was applied to the WT4 coarse-grain water model and ions, while the CHARMM-36 force field [35, 36] was used for the proteins and TIP3P water molecules. To ensure overall neutrality and mimic physiological conditions, Na+ and Cl− coarse-grain ions were added, resulting in a bulk ionic strength of 0.15 M. A two-step minimization approach was employed, consisting of 5000 steps of steepest descent minimization followed by 5000 steps of conjugate gradient minimization. System equilibration was carried out through 1 ns of constant Number of particles, Volume, and Temperature (NVT) and 1 ns of constant Number of particles, Pressure, and Temperature (NPT) equilibration, both conducted without restraints. The production simulations were then performed for 140 ns using the NPT ensemble. Temperature was maintained at 310 K through a velocity rescaling approach with a coupling time of 0.1 ps, while pressure was regulated at 1 atm using the Parrinello–Rahman barostat with a coupling time of 2 ps [36]. The equations of motion were integrated with the leapfrog algorithm using a time step of 2.0 fs. Electrostatic interactions were calculated using the particle mesh Ewald summation method [37], with Coulomb and van der Waals cut-offs set to 1.0 nm. Periodic boundary conditions were applied along the x, y, and z axes to replicate bulk behavior. Bond lengths involving hydrogen atoms were constrained using the LINCS algorithm [38]. The simulation coordinates were saved in trajectory files at 20 ps intervals. Trajectory processing and analyses were performed using GROMACS tools, while visualization and molecular graphics were generated with PyMOL software [39]. Protein–protein interaction maps were created using ProLIF [23], and all the plots were prepared using Matplotlib [40]. To assess the binding energy of MEV in complex with its respective receptors, the MM–PBSA (molecular mechanics Poisson–Boltzmann surface area) method [41] was utilized. The binding free energy (ΔG) was calculated over the final 20 ns of one replica per complex using the gmxMMPBSA software. The MM–PBSA binding energy calculation was performed using the following equation: where: ΔEMM: molecular mechanics energy, which is the sum of bonded (bond, angle, dihedral) and non-bonded (electrostatic, van der Waals) interaction energies between the proteins forming a complex. ΔGsolv: solvation-free energy, which consists of polar and non-polar contributions. MMPBSA provides an efficient way to calculate binding free energies without performing time-consuming free energy perturbation simulations. To ensure reproducibility, two independent MD simulation replicates of the MEV–TLR5 complex were performed, each initialized with different random velocity seeds. Structural convergence and stability were assessed using RMSD analysis, with both replicates showing consistent stabilization patterns and RMSD plateauing after ~40 ns. Based on this, the final 20 ns of a representative replicate was selected for MM–PBSA binding energy calculations. Interaction stability over time was further validated through hydrogen bonding analysis and ProLIF-based interaction profiling. A simulation length of 140 ns was chosen to allow sufficient equilibration and conformational sampling. The 1.0 nm cutoffs for Coulomb and van der Waals interactions were selected following standard GROMACS protocols to ensure accurate modeling of non-bonded interactions. Results Protein sequence retrieval and analysis The selected two sequences of Msg C, having accession numbers: XP_018228271.1 and XP_018230092.1, are 500 amino acids and 289 amino acids long, respectively. Prediction of T-cell and B-cell epitopes and analysis of their antigenic, allergenic, toxic, and interferon-γ activation potential A total of 42 T helper cell epitopes were predicted as strong binders for the selected HLA alleles in the sequence XP_018228271.1 and 120 for the sequence XP_018230092.1. However, only one epitope from XP_018228271.1 and none from XP_018230092.1 were chosen from the 42 T helper cell epitopes as it fulfilled the epitope filtering parameters. Epitopes that are antigenic and fulfilled other filtering parameters, such as antigenic, non-allergenic, non-toxic properties, and interferon-γ and IL-4 eliciting potential, were finally selected for further analysis. The selected epitopes are documented in the Supplementary Table S1. Additionally, 28 T cytotoxic epitopes were predicted as strong binders for XP_018228271.1 and 26 for XP_018230092.1. Among these, only four epitopes from XP_018228271.1 and none from XP_018230092.1 were selected and are listed in Supplementary Table S2. Epitopes predicted as weak binders were excluded from further analysis. For B cell epitopes, 46 were predicted for XP_018228271.1 and 31 for XP_018230092.1 (Supplementary Table S3). However, only four epitopes from XP_018228271.1 and two from XP_018230092.1 were selected for further analysis based on fulfilment of filtering parameters, as outlined in Supplementary Table S3. In summary, ten epitopes (Table 1) that demonstrated characteristics of non-allergenicity, antigenicity, and non-toxicity, with the potential to elicit immune responses such as IL4 and IFN-γ were meticulously chosen for the design of the P. jirovecii vaccine candidate after multiple in silico analyses. Peptide modeling and docking with HLA alleles In order to confirm that the predicted epitope peptides were indeed capable of binding to their respective HLA alleles, molecular docking was performed for each epitope–HLA complex. The docking analysis revealed notable variations in HADDOCK scores across the studied complexes (Fig. 1). To validate the docking protocol and assess the ability of the HADDOCK algorithm to reproduce experimentally determined peptide–HLA interactions, a redocking procedure was first carried out using the native co-crystallized peptides from each selected HLA structure. The original bound peptides were extracted from their crystal structures and re-docked into their respective receptors targeting their respective binding sites. The resulting complexes closely reproduced the experimental poses, confirming the reliability of the docking approach. Specifically, the peptide docked to HLA-B3901 (PDB ID: 4O2F) achieved a HADDOCK score of −169.85 and an RMSD of 0.7 Å relative to the crystallographic conformation. Similarly, HLA-A0201 (4U6Y) and HLA-B*2705 (6Y2A) yielded HADDOCK scores of −321.33 and −277.16, respectively, both maintaining RMSD values of 2.0 Å, demonstrating the robustness of the redocking protocol. Further analysis of individual peptide–HLA interactions showed that RLLGFLCAL docked to HLA-A0201 (4U6Y) with a HADDOCK score of −145.18, while QRLGGHFPL exhibited particularly favorable docking results with HLA-B2705 (6Y2A) and HLA-B3901 (4O2F), with scores of −232.06 and −151.51, respectively. In contrast, HFNGLSVTL showed a relatively higher HADDOCK score (−127.22) with HLA-B3901 (4O2F), indicating a less favorable docking pose. Although HADDOCK scores cannot be directly interpreted as binding affinities, they provide a relative measure of complex stability within a given docking simulation. Collectively, these findings confirm that the HADDOCK redocking protocol accurately reproduces peptide–HLA complex conformations and identifies QRLGGHFPL, particularly in complex with HLA-B*2705, as the most favorable interaction among the peptides analyzed. The interactions of the epitope peptides with their respective HLA proteins were further analyzed using the ProLIF tool [23], and the residues were color-coded according to their interaction types (see Fig. 1). Molecular dynamics simulations of epitope–HLA complexes To further assess the stability of these epitope–HLA complexes, MD simulations were conducted, focusing on hydrogen bonding, receptor flexibility, and peptide mobility (Fig. 2A). The hydrogen bond analysis demonstrated that HLA-A0201 in complex with RLLGFLCAL epitope maintained an average of 3.87 ± 1.83 hydrogen bonds. The HLA-B2705 receptor bound to QRLGGHFPL 6Y2A exhibited the highest hydrogen bond count, averaging 6.21 ± 1.56, which is consistent with its strong docking score. The HLA-B3901 receptor in complex with HFNGLSVTL formed 4.50 ± 2.05 hydrogen bonds, suggesting moderate stability. The specific peptide interactions followed a similar trend, with QRLGGHFPL forming 4.55 ± 2.02 hydrogen bonds when bound to 6Y2A, while the same peptide bound to 4O2F displayed a slightly lower count of 3.51 ± 1.88 hydrogen bonds. The RLLGFLCAL peptide bound to 4U6Y had the lowest hydrogen bonding frequency, averaging only 1.19 ± 0.96, correlating with its weaker docking score. Stability analysis of epitope–HLA complexes during MD simulations. RMSD values demonstrate relative structural stability of each complex, while hydrogen bonding patterns indicate consistent binding of the predicted epitopes with respective alleles. Receptor flexibility, assessed through RMSD measurements, revealed notable differences in structural stability. The HLA-A0201 receptor complex exhibited an average RMSD of 5.61 ± 1.52 Å, indicating significant fluctuations. In contrast, HLA-B27 complex showed much lower structural deviations, with an RMSD of 2.56 ± 0.83 Å, further supporting the stability observed through hydrogen bonding. The HLA-B3901 receptor complex exhibited an intermediate level of flexibility, with an RMSD of 5.28 ± 1.27 Å. The mobility of the peptides themselves also varied among the complexes. The QRLGGHFPL peptide in complex with 6Y2A displayed moderate fluctuations, with an RMSD of 2.10 ± 0.67 Å, while the same peptide in 4O2F exhibited slightly higher flexibility, with an RMSD of 2.40 ± 0.67 Å. The HFNGLSVTL peptide in the 4O2F complex showed a similar level of fluctuation, with an RMSD of 2.45 ± 0.41 Å. Interestingly, the RLLGFLCAL peptide bound to 4U6Y had the lowest structural deviations, with an RMSD of 1.87 ± 0.35 Å, despite exhibiting the weakest hydrogen bonding interactions. Additionally, the structural evolution of the complexes over the course of the MD simulations was analyzed by superimposing frames at 0, 25, and 50 ns (Fig. 2B). This figure highlights the structural deviations over time, illustrating RMSD differences between each frame. The QRLGGHFPL peptide in the 6Y2A complex maintained its structural integrity with minimal deviations across time points, reinforcing its stability. In contrast, the RLLGFLCAL peptide in 4U6Y exhibited greater structural shifts between frames, further supporting its lower binding affinity and reduced stability. These results suggest that the stability of the complexes is influenced by a combination of binding energy, hydrogen bond formation, and structural flexibility. The QRLGGHFPL peptide, particularly in the 6Y2A complex, showed the strongest interactions, supported by high hydrogen bonding, low receptor flexibility, and moderate peptide mobility. In contrast, RLLGFLCAL bound to 4U6Y demonstrated lower stability, with weak hydrogen bonding and high receptor fluctuations. These findings highlight the potential of QRLGGHFPL as a strong binder, particularly within the HLA-B2705 (6Y2A) system, which exhibited the most stable overall interaction profile. Vaccine engineering and physicochemical attributes determination The integration of the 10 selected epitopes and adjuvants utilized GGS linkers to craft a stable, antigenic, and non-allergenic potential vaccine construct for P. jirovecii, totalling 483 amino acids. Additionally, the detailed physicochemical properties of the final vaccine candidate, including theoretical pI, isoelectric point, molecular weight, number of atoms, and aliphatic index, etc., are presented in Table 2. The final vaccine sequence is provided below, in Fig. 3. Schematic representation of the designed 483-amino-acid vaccine construct, showing the arrangement of B-cell and T-cell epitopes linked via GGS linkers to PADRE, RS09, and flagellin adjuvants. Key physicochemical properties including isoelectric point, antigenicity score, and instability index are summarized beside the sequence. Computational immunosimulation analysis The immune response profiles generated by C-IMMSIM represent theoretical simulations of how the designed MEV might interact with the host immune system. Figure 1 displays the anticipated immune response profile of the P. jirovecii candidate vaccine. Notably, computational immune simulation predicted that the second and third doses of the vaccine candidate could increase the concentrations of different immunoglobulins compared to the initial dose (Fig. 4A). This sequential administration also led to heightened total B-cell and B-memory cell populations, indicative of a secondary immune response (Fig. 4B). Although the plasma B lymphocyte population was initially low after the first vaccine dose, subsequent vaccinations led to a significant increase in their numbers (Fig. 4C). The simulations showed that the total Th-cell population, alongside Th memory cells, could increase after the second dose in comparison to the first (Fig. 4D). However, the total Th-cell population remained similar after the second and third injections. Moreover, the injection of the P. jirovecii vaccine candidate demonstrated the potential to stimulate the production of cytokines, including IFN-gamma, transforming growth factor-beta (TGF-β), interleukin-10 (IL-10), and interleukin-12 (IL-12) (Fig. 4E). Predicted immune response to the P. jirovecii vaccine injection (A) antigen count and immunoglobulin titer; (B) population of B-cells; (C) population of plasma B-cell; (D) Th cell population;and (E) population of cytokines. Molecular modeling of vaccine candidate and docking with TLR5 The structures of both TLR5 and the potential MEV construct were modeled using AlphaFold3. The resulting models demonstrated high quality, as reflected by their local quality estimation scores. Ramachandran plot showed that 97.7% of all residues in modeled vaccine fall within the allowed regions, with 11 residues in non-favorable regions. Additionally, the ProSA analysis yielded a Z-score of −7.24, indicating a reliable structural model. Both the Ramachandran plot and ProSA quality curves have been incorporated into the supplementary information (Supplementary Fig. S1). The TLR5 model achieved an average score of 0.62 ± 0.06 (Fig. 5A). (A) The line plot unambiguously presents quality scores for both models. TLR5 (yellow line) demonstrates good quality across the predicted structure, while the MEV (gray line) exhibits a lower score, mainly due to the lack of structural information and the presence of highly flexible regions. (B) The node plot effectively illustrates the interactions between TLR5 and MEV, pinpointing the specific residues involved and categorizing the interactions by type (e.g. hydrogen bonds and cationic interactions). The image shows the TLR5 and MEV complex obtained from molecular docking. (C) The image depicts the final conformation of the TLR5 (purple ribbon) and MEV (pink ribbon) complex obtained from the molecular docking simulation. TLR5’s key residues, shown as light orange spheres, interact with the corresponding key residues of MEV, displayed as blue spheres. To investigate the interaction mechanisms, molecular docking simulations were conducted via the HADDOCK web server. The docking results for the MEV-TLR5 complex revealed key residues at the interaction interfaces. The pose with the lowest binding energy is depicted in Fig. 5C. The stability of this interaction interface is primarily maintained by hydrogen bonds, with a minor contribution from cationic interactions (Fig. 5B). Molecular dynamics analysis of vaccine model and TLR5 MD simulations are an effective method for exploring the stability of biological systems. In this study, GROMACS was utilized to assess the stability of an MEV bound to TLR5 within a solvent environment. Two independent MD simulations, each lasting 140 ns and initiated under distinct conditions, were conducted, yielding highly consistent results. The simulations were performed in the presence of water to accurately reflect the native biological context and evaluate the stability of these complexes. Although Fig. 6A displays data from both replicates, the discussion will primarily focus on the findings from the first replicate for clarity. A) Dynamic behavior and interactions of the TLR5-MEV complex. This figure comprehensively illustrates the dynamic behavior of the TLR5-MEV complex over the simulation period. It includes RMSD which depict the stability of the complex. The buried surface area (BSA) plot shows the extent of the interaction interface between TLR5 and MEV, while the hydrogen bond plot indicates the consistency of hydrogen bonds present at the interface over time. (B) Additionally, the figure features superimposed structural snapshots of the TLR5-MEV complex at 20, 80, 120, and 140 ns, highlighting conformational changes. To assess the stability and conformational changes, we computed the RMSD of the backbone relative to its initial configuration. The MEV-TLR5 complex demonstrated an average RMSD of 14.29 ± 1.33 Å, stabilizing after ~40 ns and remaining stable for the rest of the simulation (Fig. 6A). The TLR5 receptor exhibited an average RMSD of 11.70 ± 2.05 Å, while the MEV, with flexible epitope and linker loops, displayed higher flexibility with an average RMSD of 19.01 ± 2.02 Å (Fig. 6A). Hydrogen bond analysis of the MEV-TLR5 complex indicated stable interactions, with an average of 16.12 ± 3.15 hydrogen bonds over the 140 ns simulation (Fig. 6A). The BSA at the interaction interface between the vaccine construct and TLR5 averaged 854.4 ± 101.4 Å2 and remained stable across both replicates (Fig. 6A). To confirm the structural stability of the MEV-TLR5 complex, we aligned the structures at various time points throughout the simulation. Snapshots at 20, 80, 120, and 140 ns were superimposed to assess conformational changes. The comparisons indicated that while the MEV exhibited some flexibility, particularly in the epitope and linker regions, the overall structure of the complex remained consistent over time. This consistency suggests that the TLR5-MEV interface retains its structural integrity, supporting the stability observed in RMSD and the hydrogen bond analyses (Fig. 6B). TLR5-MEV interface retains its structural integrity, corroborating the stability observed in RMSD and hydrogen bond analyses (Fig. 6B). The key interaction residues at the interface between the MEV and TLR5 molecules have been shown to remain stable after 50 ns of the MD simulation trajectory (Fig. 7A). Further, the calculated ΔG for the MEV-TLR5 complex was −100.33 ± 11.95 kcal/mol, indicating a favorable binding interaction between the MEV and TLR5 (Fig. 7B). This stability is further supported by key interactions involving several critical residues. Lys106 of TLR5 plays a pivotal role by forming hydrogen bonds. Additionally, Lys106 interacts with MEV residues Ser445 and Asp444, bridging the vaccine construct and the receptor. Tyr108 of TLR5 forms a hydrogen bond with the Ser445 of MEV and also engages in a π-stacking interaction with Tyr447 of MEV, further stabilizing the complex. These interactions underscore the importance of specific residues in maintaining the stability and functionality of the MEV-TLR5 complex, with various types of bonds contributing to the overall integrity of the system. The bar plot (A) detailing the key interactions between TLR5 and MEV, emphasizing the residues involved in maintaining the complex stability. This panel (B) presents the result of MM-GBSA calculations and the binding free energy (ΔG) generated for the last 20 ns of simulation. Discussion The use of small, selective epitopes with distinct immunological properties targeting key immune system components has substantially advanced the development of innovative vaccines and immunoinformatics tools [42, 43]. These approaches have proven effective in accurately predicting immune responses through rigorous experimental validation [44]. Identifying optimal features within potential pathogen targets is a critical step in vaccine development, with immunoinformatics playing a key role in identifying both effective and safe vaccine candidates, driving the ongoing quest for protective immunity [45, 46]. Research teams, such as those involved in study [47, 48], have focused on vaccine development for pathogens such as human cytomegalovirus and Fusobacterium necrophorum using computational tools. These efforts emphasize the importance of integrating both B and T cell responses in vaccine design to prevent initial infection, as well as its subsequent dissemination and invasion. Various research groups have proposed alternative vaccination strategies targeting viruses with high morbidity, especially in vulnerable populations such as newborns and infants. A notable example is human parechovirus, a member of the Picornaviridae family, which can cause severe complications such as encephalitis, with the risk of neurological sequelae, including paralysis [49]. Vaccination against P. jirovecii could be advantageous, but as of now, no vaccines have progressed to the clinical trial stage. However, multiple attempts have been made to advance the vaccine development against P. jirovecii. A study revealed that when CD4 + T cell-depleted mice were immunized with the N-terminal portion of a protein known as Pneumocystis cross-reactive antigen 1 (Pca1), the resulting antibodies recognized P. jirovecii and provided protection against Pneumocystis pneumonia [50]. A recombinant P. jirovecii KEX1 peptide vaccine increased the level of plasma anti-KEX1 IgG antibodies and memory B-cells in immunosuppressed macaques and protected the animals from Pneumocystis pneumonia post Pneumocystis exposure [51]. Another research reported that administering the recombinant KEX1 vaccine before the immunosuppression primate model of HIV-related immunosuppression resulted in a strong and enduring antibody response, which persisted throughout the immunosuppression period. Additionally, giving a booster dose of KEX1 during immunosuppression triggered the memory responses recall against the KEX1 antigen [52]. Gingo et al. have reported that the MsgC helps in the development of immune response, specifically increase the level of IgG, in HIV-infected patients after Pneumocystis pneumonia [53]. Previous research has indicated that the serum from HIV-infected individuals who have experienced Pneumocystis pneumonia in the past shows significantly higher recognition of the MsgC fragment compared to those who have not had Pneumocystis pneumonia. This recognition is even more substantial than that observed for other Msg fragments such as MsgA and MsgB [10, 54, 55]. This is the first study exploring the vaccine potential of P. jirovecii MsgC proteins through immunoinformatics. Earlier the immunoinformatics approach has been applied for the development of vaccine candidates against Candida dubliniensis, Candida tropicalis, canine circovirus, rotavirus, dengue virus, and monkeypox virus [56–60]. Additionally, immunoinformatics has been employed to design a novel MEV targeting the rotavirus outer capsid proteins VP4 and VP7, along with the key non-structural proteins NSP2 and NSP5 [61]. Furthermore, this approach has been used to design an mRNA vaccine candidate against Serratia marcescens, a pathogenic Gram-negative bacterium that causes infections in immunocompromised patients [62]. Ten epitopes (Table 1) that demonstrated characteristics of non-allergenicity, antigenicity, and non-toxicity, with the potential to elicit immune responses such as IL4 and IFN-γ were meticulously chosen for the design of the P. jirovecii vaccine candidate after multiple in silico analyses. Among the shortlisted epitopes, QRLGGHFPL exhibited the most favorable docking energies, strong hydrogen bonding interactions, and structural stability during MD simulations. It also fulfilled all immunogenicity and safety filters, marking it as the most promising epitope. Subsequently, the 10 epitopes (Table 1) were interlinked by GGS linkers alongside PADRE, RS09, and the flagellin adjuvants, culminating in the design of the ultimate P. jirovecii vaccine candidate. RS09 serves as a TLR4 agonist, and the flagellin protein functions as a TLR5 agonist [63–65]. TLR agonists play a pivotal role in activating both innate and adaptive immunity. The inclusion of PADRE in the multi-epitope peptide vaccine candidate can enhance both its immunogenicity and efficacy [66]. PADRE demonstrates a high-affinity binding capability to diverse MHC class II molecules, facilitating the development of antigen-specific CD4 + T-cell responses [67]. The finalized vaccine candidate for P. jirovecii is predicted to be stable, antigenic, and non-allergenic. Crucially, molecular docking and MD simulations provided confirmation of the stable interactions between the computationally designed P. jirovecii vaccine construct and human TLR5 receptor in the presence of solvent. Overall, the computational immunosimulation indicated that the second and third injections of the P. jirovecii vaccine candidate that were administered after 4 weeks and 8 weeks could increase the antibody titer, B-cell population, plasma B-lymphocyte population, and Th-cell population in comparison to the first dose. The computational analyses conducted in this study underscore the immunogenic potential of the P. jirovecii vaccine candidate. While our study successfully integrates multiple computational techniques for epitope selection, molecular docking, and immune simulations, there are inherent limitations to relying on these predictive tools. Current in silico models, while highly efficient, still require experimental validation to confirm immunogenic potential. Table 1.List of final epitopes selected for vaccine engineeringEpitope typeEpitope sequenceVaxijen scoreT helperYKIYDPLNG0.5431T cytotoxicRLLGFLCAL0.6085T cytotoxicQRLGGHFPL2.9945T cytotoxicHFNGLSVTL2.3590B cellNNVKKMCEEFDYKIYD0.5830B cellKKHLKSYCNGVKGLDP1.4746B cellEFDYKIYDPLNGHLTS0.9632B cellLFGRCYQDYLNRTAIK0.6484B cellMEKICTKLEPLKVKLE0.8525B cellGNNTEARKEFNEKLLK1.2131The epitopes listed in the table are interferon-γ and IL4 eliciting, non-toxic, non-allergenic, and antigenic. List of final epitopes selected for vaccine engineering The epitopes listed in the table are interferon-γ and IL4 eliciting, non-toxic, non-allergenic, and antigenic. Table 2.Physicochemical attributes of the P. jirovecii candidate vaccineNumber of amino acids483Molecular weight50 545.79AntigenicityAntigen (Vaxijen score: 1.0017)Allergenic potentialNon-allergenTheoretical pI9.50Instability index29.86 (stable)Total number of atoms7099Aliphatic index83.58Extinction coefficient42 650Grand average of hydropathicity (GRAVY)−0.257Estimated half-life30 h (mammalian reticulocytes, in vitro). > 20 h (yeast, in vivo).> 10 h (Escherichia coli, in vivo). Physicochemical attributes of the P. jirovecii candidate vaccine Our integrated immunoinformatics, molecular docking, and MD simulations collectively highlight a potential vaccine construct predicted to elicit an effective humoral and cellular immune response. The predicted engagement of the vaccine construct with human TLR5, supported by MD simulations, suggests a plausible mechanism for innate immune activation that could translate into enhanced adaptive responses in vivo. Although the vaccine construct exhibited favorable binding energies with TLR5 and predicted immune activation profiles, these outcomes represent computational predictions of immunogenic potential rather than empirical validation of efficacy or safety. Furthermore, the immune simulation results, showing elevated antibody titers, memory B cell formation, and cytokine secretion, indicate that the construct may effectively stimulate both humoral and cellular immunity, key requirements for durable protection. Collectively, these findings bridge the gap between computational prediction and biological relevance, positioning this study as a translational foundation for experimental validation and eventual clinical application. Thus, the current work provides preliminary biological insight into the immune mechanisms of P. jirovecii and represents a translational step toward developing an effective vaccine candidate against this pathogen through future in vitro and in vivo experiments. Conclusion The research presents a systematic approach in the rational design of a vaccine candidate against P. jirovecii, a pathogen of significant concern for immunocompromised individuals through computational tools. The immunoinformatics-driven identification of potential antigenic epitopes offers a promising foundation for the development of an effective preventive solution against P. jirovecii infection. The computational analyses have led to the identification of a novel, antigenic and non-allergenic P. jirovecii vaccine candidate. Additionally, the immune simulation also predicted the potential of the designed vaccine in eliciting an effective immune response. The immune simulation results provide preliminary computational insights that warrant experimental testing to confirm immunogenic efficacy. Notably, molecular docking and MD simulations demonstrated strong interactions between the MEV structure and TLR5. Future work will focus on experimental validation of these computational findings through in vitro and in vivo studies aimed at confirming the vaccine candidate’s immunogenicity, safety, and stability. Such investigations will be essential to establish its translational potential and to bridge the gap between computational predictions and clinical application. Supplementary Material