Computational design of an mRNA vaccine targeting antifungal-resistant Lomentospora prolificans

Abstract

Lomentospora prolificans is an emerging opportunistic pathogen that predominantly affects immunocompromised individuals, as well as healthy individuals, often leading to disseminated disease with high mortality rates. Effective treatment is challenging due to its high intrinsic resistance to antifungal agents. To address this, we employed subtractive proteomics and reverse vaccinology approaches to identify potential antigenic proteins for the design of an mRNA-based multi-epitope vaccine (MEV). Our study identified four antigenic proteins as promising vaccine targets. A vaccine construct was developed using a combination of twelve cytotoxic T lymphocyte (CTL), nine helper T lymphocyte (HTL), and five linear B lymphocyte (LBL) epitopes. These epitopes were connected using appropriate linkers (AAY, GPGPG, and KK) and adjuvants to enhance antigenicity and immunogenicity. The vaccine construct was rigorously evaluated for its physicochemical properties, demonstrating high antigenicity, non-toxicity, non-allergenicity, stability, and solubility. Molecular docking studies were conducted to validate the interactions between the vaccine construct and the human toll-like receptor (TLR4). Immune simulation studies further confirmed the vaccine’s potential to elicit a robust immune response. Additionally, molecular dynamics (MD) simulations, principal component analysis (PCA), dynamic cross-correlation matrix (DCCM) analysis, and binding free energy calculations were performed to assess the stability and efficacy of the vaccine-receptor complex. Codon optimization and in-silico cloning were carried out to ensure efficient expression of the vaccine in Escherichia coli strain K12. The findings of this study suggest that the proposed vaccine construct holds significant promise as a novel mRNA-based therapeutic candidate against L. prolificans infections. Further experimental validation is recommended to advance this vaccine toward clinical application.

Introduction

Lomentospora prolificans is an emerging opportunistic pathogen that predominantly infects immunocompromised individuals, often resulting in disseminated disease with high mortality rates. However, it can also cause infections in healthy populations¹. Effective treatment is challenging due to the pathogen’s high intrinsic resistance to antifungal agents. Successful management relies on three key pillars: rapid and accessible diagnostic methods, aggressive surgical debridement when applicable, and prompt administration of effective antifungal therapy¹. Sewage, contaminated waterways, and soil contain the environmental fungus L. prolificans. It is also spread by inhalation of spores and direct inoculation². Disseminated infection following a lung transplant and myeloblastic leukemia has also been linked to L. prolificans-induced morbidity. It was reported to be the source of bone infection after trauma and ocular infection during a lawn mower mishap in healthy people³. L. prolificans primarily affects individuals with weakened immune systems, causing a range of severe diseases, including pulmonary and disseminated infection.

This fungus was first identified by Malloch and Salkin in 1984 from an immunocompetent patient with osteomyelitis, marking its initial recognition as a human pathogen⁴. This was the first reported instance of a nosocomial outbreak of L. prolificans infection, occurring during hospital renovation activities. At that time, patients were treated in a temporary facility that lacked standard protective measures for granulocytopenic individuals. They were kept in solitary rooms with restricted access and required masks, wedges, and dressing gowns. This zone lacked air conditioning, and potted plants were isolated from the renovation zone. The four incidents happened in two adjacent rooms occupied by four patients over 28 days⁵. The role of L. prolificans in invasive fungal infections is becoming more widely acknowledged in regions like Australia, the US, and some countries of Europe.

Antifungal medications have not been linked to a reduced chance of mortality⁶. As L. prolificans is resistant to various antifungal drugs, it causes infections that are difficult to treat, usually with adverse consequences. Early identification and vigorous care are crucial because of the high fatality rate, which is particularly severe in cases of disseminated infection or delayed treatment. Amphotericin B-based formulations along with surgery are advised as current therapy guidelines. Frequently, amphotericin B-based formulations (AmB) fail to work on L. prolificans⁷. The major chronic side effect of using amphotericin B-based formulations is nephrotoxicity. AmB likely causes kidney damage through a variety of methods⁸. A combination of voriconazole and terbinafine can effectively control widespread L. prolificans infections, but certain strains of the infection are resistant to this medication, and this resistance has the potential for side effects like liver toxicity and visual disturbances with an overall mortality rate of 87.3%⁴.

Current research is focused on developing effective treatments with minimal adverse effects to enable significant advancements in the near future. This is due to the complex nature of diseases, which necessitates a comprehensive understanding of their pathogenesis and prognosis. New sequencing methods have proliferated recently, allowing scientists to make significant advancements in the area of vaccine production. In recent years, subtractive proteomics and immunoinformatics approaches have gained popularity as effective and cost-efficient strategies for designing vaccines against a wide range of infectious diseases^9,10.

Therefore, to combat infections caused by L. prolificans, the current study employed computer-aided approaches to design a multi-epitope mRNA vaccine based on immunogenic B- and T-cell epitopes, aiming to elicit both humoral (e.g., B-cell activation and antibody production) and cellular (e.g., T-helper cell and cytotoxic T-cell activation) immune responses. The interacting ability of the proposed vaccine was confirmed through molecular docking with TLR4, MD simulations, PCA analysis, DCCM, and binding free energy analysis. Using in-silico cloning, the vaccine was reverse-translated and expressed in E. coli (K12 strain). This study intends to produce an efficient and dynamic vaccine against L. prolificans using a unique complex of the fungus’ antigenic peptides and immunoinformatic approaches for further experiments¹⁰. There is no commercially available mRNA vaccine that targets fungal pathogens (e.g., Candida albicans). The experimental mRNA vaccine candidates against Candida have mostly targeted single antigens and have shown protective immune responses in preclinical models. However, there is little information on long-term immunity, and these vaccines have not yet been clinically approved¹¹. On the other hand, our multi-epitope mRNA vaccine design includes immunodominant epitopes that are expected to stimulate strong humoral (IgM, IgG1, IgG2) and cellular (CD4⁺ and CD8⁺ T-cell) responses. Simulation results show improved antigen clearance, memory formation, and sustained immunoglobulin and T-cell activity after booster doses. Similarly, the in-silico approach assumes linear epitope processing but does not account for mRNA kinetics such as lipid nanoparticle transport or intracellular antigen processing. These are the primary limits of in silico strategies. However, this technique of developing mRNA vaccines based on predicted linear epitopes has been successfully used in prior studies¹². However, experimental validation is required to check the efficacy and effectiveness of the constructed vaccine.

Materials and methods

The following workflow (Fig. 1) depicts the subtractive proteomics and reverse vaccinology technique to develop mRNA vaccines against L. prolificans infections.

Fig. 1.

The schematic diagram illustrates the steps employed to develop the mRNA vaccine construct.

Proteome retrieval

The reference proteome of L. prolificans (Proteome ID: UP000233524) was retrieved from the UniProtKB database. The proteome underwent subtractive proteomics analysis to identify promising candidates for the construction of a vaccine against L. prolificans.

Human non-homologous protein identification

To minimize the risk of autoimmune reactions, it was crucial to select antigens that do not resemble human proteins when developing a vaccine. Using BLASTp analysis, the entire proteome of L. prolificans was compared to the human proteome with cutoff thresholds of pident < 35, query coverage < 35, bitscore < 100, and e-value > 1e-4^13,14.

Identification of gut non-homologous proteins

Numerous beneficial bacteria in the host body protect it against external contaminants¹⁵. The relationship between humans and gut flora is mutualistic and symbiotic rather than just commensal. Proteins in this microbiota may be accidentally blocked or inhibited, which could negatively affect the host. For this purpose, the BLAST analysis of human non-homologous proteins against the NCBI gut metagenome was performed to obtain gut non-homologous proteins with the previously defined parameters, such as cutoff values of bitscore < 100, e-value value > 1e-4, pident < 35%, and query coverage < 35%^14,16,17.

Identification of essential and virulent proteins

Essential proteins or genes play critical roles in pathogen survival and virulence¹⁸. Their conserved condition lowers the risk of resistance, and they might elicit strong and long-lasting immune responses. These factors make essential proteins crucial to vaccine development. The BLASTp was used to perform sequence alignment between the prioritized proteins from the above step and proteins in the Database of Essential Genes (DEG) based on the filtering criteria of bitscore > 100, query coverage > 35, pident > 35, and e-value value < 1e-4^19,20. A database called DEG holds regularly updated information on the important proteins found in eukaryotes. Moreover, virulent proteins are of great importance in prioritizing vaccine candidates due to their pathogenic and host invasion properties. The Pathogen–Host Interactions Database (PHI-base) provides information about virulence factors from diverse pathogens²¹. The VFDB database was subjected to BLASTp against VFDB by using cutoffs of pident > 35, query coverage > 35, bitscore > 100, and expectation value < 1e-4 to explore the virulent proteins of L. prolificans by identifying homologous proteins to gut non-homologous proteins²².

Determining protein subcellular localization

In bioinformatics, the subcellular localization of proteins is indispensable for understanding biological functions, predicting protein function, and searching for disease mechanisms. Proteins perform functions according to their location. The extracellular and outer membrane proteins were prioritized for vaccine candidates, while cytoplasmic and mitochondrial proteins were considered as putative drug targets. The Euk-mPLoc 2.0, CELLO v.2.5, and WoLF PSORT servers were utilized to identify the subcellular location of the target proteins. A multi-class support vector machine (SVM) categorizing system called CELLO v.2.5 was utilized to predict the location of fungal proteins in a cell. Euk-mPLoc 2.0²³is more potent and adaptable technique for determining the subcellular localization of eukaryotic proteins²⁴.Moreover, the WoLF PSORT categorizes proteins into more than ten localization sites, including proteins that move between the cytosol and nucleus²⁵.

Screening of potential vaccine targets

The prioritized vaccine proteins were then subjected to the VaxiJen v2.0 server²⁶ for antigenicity prediction. Prediction of antigenicity is an important step because it can generate an immune response. The AllerTOP v2.0 server was utilized to determine these proteins’ allergenic profile²⁷.Additionally, the TMHMM server²⁸ was utilized to verify the topological values of the proteins as 0 or 1. The ProtParam Expasy server was employed to examine their physicochemical properties, like molecular weight, GRAVY value, theoretical pI, number of amino acids, and stability²⁹.

MHC-I and MHC-II epitope prediction

Epitopes are specific regions of antigenic protein molecules that can attach to receptors on immune cells and trigger an immunological reaction. MHC I molecules, which are crucial in presenting peptide antigens to cytotoxic T lymphocytes, are present in all nucleated cells. The MHC I molecules on cytotoxic T cells present peptides from intracellular and extracellular sources, triggering an immune response to destroy the cell. In acquired or adaptive immunity, HTL cells are essential because they help cytotoxic T lymphocytes (CTLs) to destroy infected cells and help B cells to produce and release antibodies. After helper T cells are activated and presented on antigen-presenting cells (APCs), they act as effectors and develop into a particular subgroup needed for a targeted immune response³⁰. Moreover, the full HLA reference allele set was used to ensure that the final epitopes covered the HLA present worldwide, and protein sequences were used as the input for the IEDB-recommended³¹ method of MHC-I and MHC-II epitope forecasting. Immune Epitope Database is a well-established and widely used manually curated database of experimentally characterized immune epitopes. IEDB’s prediction algorithms are trained and benchmarked on large sets of experimentally validated binders and non-binders, with negative control data incorporated during development and validation³².Based on their percentile ranks (< 0.5%), the resulting epitopes were ranked.

Prediction of linear B-cell epitopes

B-cell epitopes are essential for both the development of vaccines and the induction of an adaptive immune response. A bioinformatics tool called Bepipred Linear Epitope Prediction 2.0 was utilized to predict antigenic epitopes, or antibody-binding sites, within protein sequences. Its main goal was to determine the linear B-cell epitopes, or regions of proteins that antibodies can recognize.

Analysis of determined MHC-I, MHC-II, and B-cell epitopes

Antigenic, non-toxic, soluble, and non-allergic epitopes must be used in the development of a vaccine. The VaxiJen v 2.0 was used to screen all T-cell and B-cell epitopes that had been prioritized to determine their antigenicity score. While the server, AllerTop2.0 was utilized to predict the allergenicity of these epitopes. Moreover, the solubility and toxicity of these epitopes were examined using the INNOVAGEN³³ and ToxinPred servers³⁴respectively.

Multi-epitope vaccine construction

To create a vaccine construct, linkers were utilized to join the selected HTL, CTL, and LBL epitopes³⁵. Specifically, the AAY, GPGPG, and KK peptide linkers were utilized to connect the putative CTL, HTL, and LBL epitopes to construct a peptide vaccine³⁶. Linkers are short amino acid fragments that connect adjacent epitopes and adjuvants, enhance immune responses, and restrict epitope folding³⁰. An immunogenic linker such as EAAAK was used to attach the adjuvant, beta defensin, to the epitopes of B and T cells.

mRNA vaccine design

From N-terminal to C-terminal, the final vaccine construct contains a signal tPA peptide attached to the N-terminus to facilitate the vaccine’s exit from the cell. Furthermore, the vaccine design should have a suitable Kozak sequence that starts with the start codon³⁷. Subsequently, the sequence for the 5′ cap and 5′ untranslated domains was added to the N-terminal end of the vaccine to make it transcriptionally stable, while the 3′ untranslated domain and poly-A tail were placed at the C-terminal end. Additionally, the Mfold³⁸ and RNAfold³⁹ servers were utilized to forecast the anticipated vaccine’s mRNA secondary structure based on its minimum free energy. In general, the mRNA structure is considered more stable with a lower free energy⁴⁰.

Analysis of designed vaccine

The designed construct of the vaccine was examined for its antigenicity, topology, solubility, allergenicity, and physicochemical characteristics. For this purpose, VaxiJenv2.0 and ANTIGENpro⁴¹were utilized to determine the vaccine’s antigenicity. The TMHMM server and AllerTOP 2.0 assessed the topology and allergenicity, respectively. The ProtParam tool of the Expasy server assessed distinct physicochemical properties of the designed vaccine construct, such as its aliphatic index (AI), molecular weight (MW), theoretical isoelectric point (PI), instability index, and grand average of hydropathicity (GRAVY). The SOLpro tool and the INNOVAGEN server were utilized to forecast the vaccine’s solubility.

Prediction of 2D and 3D structure

PDBsum and SOPMA servers were utilized to predict the 2D structure of the designed vaccine^42,43. These self-optimized prediction techniques may evaluate characteristics including beta sheets, random coils, and bend areas. Using the trRosetta server⁴⁴ a 3D model of the vaccine structure was determined. An online resource named PROCHECK, which describes the stereochemical characteristics of the protein, was used to evaluate the most favored residues in the vaccine through the Ramachandran plot. Moreover, using a Ramachandran plot, the phi/psi angles were evaluated in order to have a thorough grasp of the protein backbone. The SAVES v6.1 server was used to project an ERRAT quality factor value. The ERRAT plot additionally serves as a confirmation procedure for protein structures with non-bonded connections, useful for tracking the progression of crystallographic modeling. The SAVES v6.1 server suggested that the higher the ERRAT score, the better the quality of protein. Moreover, MolProbity is another tool that was used to further validate a vaccine’s structure⁴⁵. MolProbity examines the molecule’s overall stereochemistry and can discover any potential steric conflicts or other structural concerns⁴⁶. Furthermore, the developed 3D model was subsequently evaluated using the Swiss Model Structure Assessment Tool, which validates the vaccine design using the QMEAN score⁴⁷. The QMEAN value provides a composite quality estimate that includes both global and local analysis of the model⁴⁸.

Post-translational modification analysis, globular regions, and proteasomal cleavage analysis

Post-translational modifications (PTMs) of vaccine candidates, peptides, and other types are crucial steps for improving vaccine immunogenicity⁴⁹. A posttranslational modification (PTM) influence the entire protein structure and function in a variety of biological processes, including transcription, translation, apoptosis, cell signaling, and replication. Protein phosphorylation and glycosylation are considered some of the most common and important PTMs since they activate or deactivate different enzymes and receptors. NetNGlyc 1.0⁵⁰ and NetPhos 3.1 servers⁵¹ were used to anticipate locations suggesting posttranslational modifications in proteins, namely N-glycosylation and phosphorylation, respectively. The GlobPlot 2.3 server was used to determine the globular and disordered regions of the vaccine⁵². Disordered regions in proteins frequently contain short linear peptide motifs that are critical for protein activity⁵³. Moreover, the proteasomal cleavage analysis was conducted to analyze the vaccine’s capacity to activate cytotoxic T-cells. This analysis was performed by the Proteasome Cleavage Prediction Server of the Immunomedicine group (http://imed.med.ucm.es/Tools/pcps/).

Aggrescan3D and CABS-flex analysis

When designing vaccines, achieving structural stability is essential. Aggrescan3D was used to examine aggregation-prone areas, followed by the coarse-grained simulation for the flexibility of the vaccine modes using CABS-flex. A total of 50 cycles and 7494 RNG seeds were selected. The global side chain and global C-alpha constraints were set at 1.0⁵⁴.

Conformational B-cell epitopes prediction

The ElliPro⁵⁵ tool from the IEDB server was utilized to determine the conformational B-cell epitopes of the vaccine. A conformational or discontinuous B-cell epitope is the arrangement of amino acids or subunits that make up an antigen and can bind directly with an immune system receptor⁵⁶. ElliPro assigns a protrusion index (PI) score to each epitope that is anticipated. The PI scores of residues serve as the criterion for recognizing conformational B-cell epitopes.

Population coverage analysis

HLA allele dispersion and expression differ by ethnicity and geography around the world, influencing the successful production of vaccines. The population coverage was estimated using the IEDB population coverage tool, which took into account specified MHC I and MHC II epitopes, as well as matching HLA-binding alleles. Based on the distribution of human MHC-binding alleles, this tool forecasts the population coverage of each epitope for various locations around the world⁵⁷.

Prediction of binding pockets

The CASTp server⁵⁸ was used to determine the active sites or binding pockets of proteins. The CASTp offers accurate, comprehensive, and quantitative data on the topographical features of a protein. Active pockets on protein surfaces and the inside of three-dimensional structures were determined, which is crucial for forecasting the protein regions that engage in ligand-protein interactions.

Molecular Docking between vaccine and receptors

A molecular docking study was carried out using the ClusPro 2.0 server⁵⁹to determine the binding affinity between the developed vaccine (as a ligand) and TLR4 (PDB ID: 4G8A), as a receptor. Toll-like receptor 4 is a key lipopolysaccharide (LPS) receptor that identifies pathogen infections and activates inflammatory responses. It plays a vital role in the innate immune response. Three sequential phases are combined by ClusPro, an excellent tool for protein-protein docking, to yield binding affinity: rigid-body docking, lowest energy structure clustering, and energy-minimization-based structural refinement. The PyMol software was utilized to visualize the interactions between the vaccine and the receptor.

Normal mode evaluation

The dynamic motion of the docked complex was further investigated utilizing the iMOD server⁶⁰. For the normal mode simulation study, the docked complex PDB file with the lowest binding energy score was used. This tool utilizes deformability, covariance, B factor, eigenvalue, variance, and elastic network to forecast the range and direction of fundamental protein-protein complex motions.

MD simulation analysis

Assisted Model Building with Energy Refinement, suite v20 (AMBER20) (https://ambermd.org/) was used to conduct the MD simulation analysis as performed previously^61,62. The receptor atoms that were missing were added using AMBER20’s LEaP algorithm. Each complex was submerged in TIP4P water in a truncated octahedron shell with a border distance of 10.0 Å⁶³. Before running the MD simulations, energy minimization on each system was performed to avoid any steric conflict that could emerge during system setup. Initially, all counterions (Na + or Cl-) and water molecules were refined, while the ligand and protein were frozen with a limitation potential of 500 kcal/(molŲ). The protein’s potential was determined using the AMBER20 force field⁶⁴whereas the ligand’s potential was determined using the general AMBER force field (GAFF)⁶⁵. Protein residue side chains have been loosened, and backbone heavy atoms were restricted by a 5 kcal/(molÅ2) constraint force. Lastly, the complete system was refined or optimized with no constraints. In each phase, structural optimization was performed using 2500 steps of steepest descent and 5000 steps of the conjugate gradient technique. The NVT ensemble used energy minimization to gradually heat each system from 0 to 300 K over 200 ns, using a force constant of 10 kcal/(molÅ2) on the protein-ligand complex. The system proceeded through seven rounds of equilibrations at 300 K in 1 ns interval, with declining restraint weights of 5, 3, 1, 0.5, 0.3, 0.1, and 0 kcal/(molÅ2) and no restrictions on the solvation environment. Finally, each system in the NPT ensemble underwent 200 ns MD productions at 300 K and 1.0 atm pressure⁶⁶. Moreover, MD simulation was also performed for 100 ns, in which each complex was submerged in TIP3P water for better comparison with extensive MD simulation at 200 ns and the TIP4P model.

MD trajectories

The simulated trajectories were utilized to compute the solvent-accessible surface area (SASA), radius of gyration (Rg), root mean square deviation (RMSD), and root mean square fluctuation (RMSF)⁶⁷. The stability of each system was assessed using RMSD from the initial equilibrated spots of TLR backbone atoms and vaccine chains. The oscillations in the side chain atoms of toll-like receptors and vaccine chains were also investigated using RMSF. The system’s compactness during MD simulation was examined using the Rg. Furthermore, SASA was determined to get insight into the common SASA between the TLR4 and vaccine elements⁶⁸.

Binding free energy

To compute the system’s binding free energies, MD simulation trajectories were run with the AMBER20 MMPBSA program. MMPBSA determines free energies by comparing the free energy difference between the complex, protein, and ligand alone. The overall binding free energy is computed as the difference between the free energy of the complex (Gcomplex) and the addition of the free energies of the individual proteins (Gprotein) and ligand (Gligands). It can be determined with the following equation Eq. 1:

1

In Eq. (2), the solvation-free energy (ΔGsol) is calculated by adding the polar and nonpolar portions.

2

Compared to total binding free energy estimations, per-residue free energy breakdown can disclose each residue’s influence on the ligand. This allows for a more in-depth analysis of each ligand’s binding capability and selectivity. Equation (3) can be used for calculating it.

3

In the previous equation, ΔEvdW and ΔEele are van der Waals and electrostatic interactions estimated by AMBER20⁶⁹.

PCA and free energy landscape analysis

Molecular dynamics (MD) simulations allow microscopic analysis of the structure and behavior of molecular systems. Principal component analysis (PCA) has become one of the most extensively utilized approaches to evaluating the mobility of proteins within this framework⁷⁰. PCA is a statistical approach used to reduce the number of dimensions needed to explain protein dynamics systematically. This is accomplished using a decomposition technique that filters observed motions from largest to smallest spatial scales. Principal component analysis is a linear development that extracts the most significant data elements from a covariance or correlation matrix. Two principal components (PC1 and PC2) were selected for examination. These two PCs were then utilized to construct and examine the free energy landscape (FEL). In FEL plots, the deep valleys reflect the lowest energy states, and the boundaries between the deep valleys show intermediate conformations.

DCCM analysis

The dynamic cross-correlation matrix (DCCM) analysis was used for insights into the correlated movements of residues in the V1-TLR4-docked complex during a 200 ns molecular dynamics (MD) simulation. The trajectory information gathered during the MD simulation was used to compute the cross-correlation coefficients between the variations of each residue pair⁷¹. This study identified locations within the complex that displayed substantial correlated or anti-correlated movements, offering observations into the vaccine-receptor-docked complexes’ dynamic behavior and potential functional connections.

Immune simulation

The C-ImmSim server was employed to perform computational immune simulations in order to evaluate the vaccine’s efficacy and immunological characteristics. According to the real-world applications, the smallest suggested gap between the first and second doses of most vaccines is 4 weeks. Three injections were given four weeks apart for our immune simulation. For calculating simulation durations, the C-ImmSim server employed a time-step scale. Each time step on this scale corresponds to 8 h in real life. The total number of antigenic steps for simulation was customized to 1050, and the injection points were set at time steps 1, 84, and 168 ⁷². Thus, the vaccine was administered on days 1, 28 and 56 ⁷³, and also these steps correspond with the clinical trials of mRNA vaccines⁷⁴. The remaining parameters were at their default values.

In-silico cloning

Codon adaptation is a strategy in which codons in the construct’s cDNA sequence are modified to maximize the model vaccine’s expression in a suitable expression system⁷⁵. In order to get Codon Adaptation Index (CAI) values, the Optimizer tool employed an algorithm to back translate the vaccine amino acid sequences to DNA. E. coli expression levels were anticipated using the average GC content and CAI values of the modified patterns. A CAI value of 1.0 is seen to be ideal, and the GC content varies between 30% and 70%. To produce a recombinant plasmid, codon-adapted sequences were added to the plasmid vector pET28a (+). SnapGene 8.0.2 software (www.snapgene.com) was used for this objective.

Identification and analysis of novel drug targets

The mitochondrial and cytoplasmic proteins predicted by subcellular localization were subjected to a BLASTp analysis against the FDA-approved database. DrugBank provided a designed database for FDA-approved drug targets. Proteins shortlisted in the subcellular localization step were examined against DrugBank data using BLASTp analysis, with criteria of qcovs < 35, pident < 35, bitscore < 100, and e-value value > 1e-4^17,20. Targets that had no substantial sequence similarity with DrugBank data were prioritized for further analysis in order to predict novel drug targets. The TMHMM–2.0 server calculated transmembrane topology for proteins. Transmembrane proteins, which span the lipid membrane, are potential drug targets. The physicochemical parameters of the finalized proteins, such as molecular weight, theoretical pI, instability index, and grand average hydropathicity, were investigated using the ProtParam tool⁷⁶.

String database analysis

The STRING 11.5 server was utilized to analyze the interaction network of proteins from different drug target proteins⁷⁷. The system configuration was found to have the lowest interaction score with medium confidence of “0.400”. All other parameters remained unaltered from the system’s default configuration.

3D model prediction and drug pocket screening

The 3D structure of the finalized proteins was predicted using the SWISS-MODEL server⁷⁸. To validate the most suitable model, the ERRAT value and Ramachandran plot of each model were assessed using the SAVES v6.0 server. The selected probable druggable proteins were then analyzed to identify putative binding pockets using DoGSiteScorer⁷⁹. DoGSiteScorer evaluated the druggability of the 3D structures. This tool calculates the pocket residue and druggability score, which ranges from 0 to 1. A score closer to 1 indicates a highly druggable protein cavity⁸⁰.

Results and discussion

Proteome retrieval and non-homologous protein identification

The entire proteome of L. prolificans contained 8533 proteins; after removing duplicates, only 8341 proteins were left behind. These proteins were comprehensively studied utilizing the subtractive proteomics workflow to identify and characterize putative immune activation targets. To verify that these targets do not have invasive immunogenic effects by cross-reacting with human proteins. BLASTp analysis against the human proteome revealed 2075 proteins inside the pathogen’s proteome that were classified as host non-homologous. Moreover, non-homology analysis with the human gut microbiome was carried out to analyze their possible interactions and compatibility with the human gut microbiome. BLASTp analysis revealed that a total of 2048 proteins were non-homologous to human gut microbiota proteins, showing that these proteins are structurally or functionally distinct from those found in the human gut microbiome (Figure S1.)

Identification of essential and virulent proteins

It is believed that essential proteins control the necessary processes and are vital for the survival of pathogens. Regarding the development of vaccines, such proteins raise the possibility of affecting pathogen virulence. Vaccine development based on these targets might result in a stronger immune response and broader defense against several strains. It also reduces the possibility of genetic mutations that may affect vaccine effectiveness⁸¹. The database of essential genes was used to identify 22 essential proteins, while the others were discarded. These 22 important proteins were used in the next steps to identify effective vaccine targets.

Virulent proteins have been shown to play a vital role in the pathogenesis, making them a promising vaccine candidate. Virulent proteins might elicit an immune response; hence, virulent protein analysis was undertaken. The virulent protein analysis revealed 07, and these proteins were employed in subsequent analysis. Both essential and virulent proteins were combined, and duplicates were discarded. A total of 26 proteins were selected in this step for further analysis (Fig. S1).

Subcellular localization of proteins

One of the most crucial aspects of therapeutic targets is protein localization⁸². Firstly, the WoLF PSORT, CELLO v.2.5, and Euk-mPLoc 2.0 bioinformatic tools were employed to determine the subcellular localization of these 26 proteins (Table 1). The 3 proteins were linked to the inner membrane, and 4 proteins served as either drug or vaccine targets based on their subcellular localization and functionality; the protein “L. prolificans peptidase M4 C-terminal domain-containing protein” would be surface-exposed and used for extracellular proteolysis, hence immune-accessible and drug-inhibition accessible⁸³. Histidine kinases possess sensor domains that are membrane-bound and kinase domains within the cytoplasm, thus may be available for both vaccine and drug applications in virtue of their involvement in signal transduction⁸⁴. While L-ornithine N (5)-monooxygenase (PvdA) is not a surface-exposed protein but plays a central role in siderophore biosynthesis and iron acquisition that contributes to virulence, addition of the epitopes of this protein may enhance the protection by disrupting iron metabolic pathways and enhancing host immunity through immune responses⁸⁵. Moreover, TauD/TfdA-like domain-containing proteins perform fundamental metabolic functions; they are essential for the pathogen’s ability to survive in the host because of their critical roles in oxidative stress adaptation and sulfur metabolism⁸⁶. Additionally, prior studies have shown that intracellular enzymes, particularly through MHC-I presentation pathways can support T-cell-mediated immune responses⁸⁷and 19 were found in the cytoplasm and mitochondria and are regarded as drug targets (Fig. 2). They are reachable to the host immune system, and proteins located in the extracellular compartment, outer membrane, and periplasm have been explored for determining vaccine candidates. These proteins are vital to host-pathogen interactions and are typically exposed to the outside environment³⁰.

Table 1.

Subcellular localization of 26 combined proteins.

Sr. No	GI numbers	Protein name	Localization
1	A0A2N3N042	SIS domain-containing protein	Cytoplasmic
2	A0A2N3NHG3	Thiamine phosphate synthase/TenI domain-containing protein	Cytoplasmic
3	A0A2N3NES6	Phosphoribosyltransferase domain-containing protein	Cytoplasmic
4	A0A2N3MXY2	Carbonic anhydrase	Cytoplasmic
5	A0A2N3NAZ6	Isocitrate lyase	Cytoplasmic
6	A0A2N3N8S9	AB hydrolase-1 domain-containing protein	Cytoplasmic
7	A0A2N3N1L7	TauD/TfdA-like domain-containing protein	Cytoplasmic, Extracellular
8	A0A2N3N082	Isocitrate lyase	Cytoplasmic
9	A0A2N3NL25	TauD/TfdA-like domain-containing protein	Cytoplasmic
10	A0A2N3MYD7	Nickel/cobalt efflux system	Membrane
11	A0A2N3NJ79	beta-glucosidase	Cytoplasmic
12	A0A2N3N8J1	NADH: flavin oxidoreductase/NADH oxidase N-terminal domain-containing protein	Cytoplasmic
13	A0A2N3N2I7	Carbohydrate kinase PfkB domain-containing protein	Mitochondria
14	A0A2N3MZ68	PrpF protein	Mitochondria
15	A0A2N3NAY7	AB hydrolase-1 domain-containing protein	Mitochondria
16	A0A2N3NIY2	Cation/H + exchanger domain-containing protein	Membrane
17	A0A2N3N9S5	Fatty acid synthase subunit alpha	Cytoplasmic
18	A0A2N3N399	anthranilate synthase	Cytoplasmic
19	A0A2N3MZG5	pyridoxal 5’-phosphate synthase (glutamine hydrolyzing)	Cytoplasmic
20	A0A2N3NEZ7	Phospho-2-dehydro-3-deoxyheptonate aldolase	Cytoplasmic
21	A0A2N3N5D6	Fructose-bisphosphate aldolase	Cytoplasmic
22	A0A2N3NKV7	Tryptophan synthase	Cytoplasmic
23	A0A2N3NFZ1	Peptidase M4 C-terminal domain-containing protein	Cytoplasmic, Extracellular
24	A0A2N3NAB3	NodB homology domain-containing protein	Extracellular
25	A0A2N3N0P6	Histidine kinase	Cytoplasmic, Extracellular
26	A0A2N3NFW9	L-ornithine N(5)-monooxygenase	Cytoplasmic, Extracellular

Fig. 2.

The schematic presentation of subcellular localization of proteins using the CELLO v.2.5, Euk-mPLoc 2.0, and WoLF PSORT servers.

Analysis of prioritized vaccine candidates

Employing the VaxiJen v2.0 and AllerTop 2.0 servers, the antigenicity and allergenicity profile of the vaccine candidate proteins was determined. A protein’s ability to enhance an effective immune response and build protective immunity against infections depends on its antigenicity. The 5 proteins out of 7 vaccine targets (A0A2N3N1L7, A0A2N3MYD7, A0A2N3NAB3, A0A2N3N0P6, and A0A2N3NFW9) were determined to be non-allergenic and antigenic, according to the result of the analysis. In contrast, it has been observed that 2 proteins (A0A2N3NFZ1 and A0A2N3NIY2) are neither allergic nor antigenic. The TMHMM server was used to detect the presence of hydrophobic α-helices in the transmembrane. The presence of transmembrane hydrophobic α-helices in membrane-bound proteins aids in their membrane embedding⁸⁸. The GRAVY value determines protein polarity, the aliphatic index provides thermostability, and the instability index indicates whether the protein is stable or not. Protein stability is influenced by molecular weight and theoretical pI. After the analysis, four proteins (A0A2N3N1L7, A0A2N3NAB3, A0A2N3N0P6, and A0A2N3NFW9) were finalized as vaccine candidates, and they were antigenic, non-allergenic, highly stable, and hydrophilic with a topological value of 0 (Table 2).

Table 2.

Physicochemical properties of prioritized vaccine candidates.

Protein ID	Proteins Names	No. of A.A	Topology Value	Mol. Wt	Theoretical Pi	Aliphatic Index	Gravy	Allergen	Antigen	Instability index
A0A2N3N1L7	TauD/TfdA-like domain-containing protein	382	0	42589.44	5.81	75.79	-0.543	No	Yes	Stable
A0A2N3MYD7	Nickel/cobalt efflux system	459	o33-55i99-121o141-163i210-232o242-259i312-334o361-383i	49665.17	6.02	116.01	0.421	No	Yes	Unstable
A0A2N3NIY2	Cation/H + exchanger domain-containing protein	876	o47-69i76-95o105-127i140-162o175-197i209-231o246-268i288-310o335-357i364-386o396-415i427-449o	94214.21	6.16	113.63	0.417	No	No	Unstable
A0A2N3NFZ1	Peptidase M4 C-terminal domain-containing protein	397	0	44514.12	6.6	74.21	-0.52	No	No	Stable
A0A2N3NAB3	NodB homology domain-containing protein	424	0	44920.04	5.38	67.45	-0.331	No	Yes	Stable
A0A2N3N0P6	Histidine kinase	730	0	81107.73	5.55	92.59	-0.216	No	Yes	Stable
A0A2N3NFW9	L-ornithine N(5)-monooxygenase	582	0	63605.62	6.05	84.97	-0.252	No	Yes	Stable

Prediction of CTL and HTL epitopes

Identifying CTL and HTL epitopes is essential to construct a vaccine that is effective. Based on the receptors on their membranes, T-cells are categorized as either CD4+ (HTLs) or CD8+ (CTLs). These cells attach themselves to MHC-I and MHC-II molecules’ epitopes. HTLs interact with MHC-II molecules, while CTLs attach to MHC-I epitopes. MHC molecules and antigenic epitopes can interact to generate a potent immune response. The primary function of MHC1 cells is to eliminate tumor cells that have the proper antigens and to destroy cells infected by foreign substances. MHC-II cells are crucial for initiating and optimizing the immune response. By instructing other cells to carry out these functions, they “mediate” the immune response and control the kind of immunological response that arises⁸⁹. Thus, predicting high-affinity epitopes is critical⁹⁰. The IEDB-recommended NetMHCpan EL 4.1 approach was implemented to predict highly immunogenic CTL epitopes with the percentile rank score of < 0.5%. The top 10 non-overlapping epitopes with a low percentile rank score were selected for every protein. The targeted epitopes were checked for their antigenicity, toxicity, allergenicity, and water solubility. To create the vaccines, 12 CTL epitopes from 4 proteins were chosen that were non-allergenic, highly antigenic, highly non-toxic, and good water-soluble. The 12 epitopes DYKEITTAR, RTHPVTGEK, HESGAATSL, STAEATTAR, KENGVVATF, VSKGLVEQW, KVTRDLTEW, AELRLISAY, TIKTRTPSL, YQYSPTERF, RTAHLTILK, and KVLSINHPR were selected as promising candidates for vaccine designing (Table S1).

The predicted results for MHC-II epitope analysis identified 9 unique epitopes with a percentile rank of less than 2.0, indicating high interaction affinity. These epitopes were further assessed for their antigenicity, non-allergenicity, nontoxicity, and water solubility. The 9 epitopes, FTRNIVGLKKEESDA, QGDYKEITTARYSDE, ELGYHVTNYNLDTKD, QSGRVTTQTPQTESA, LKNDFLANMSHEIRT, MDDYIAKPVNKQLLA, VEQWNRTEGEFIGRD, RSRFTFLNYLHENNR, and IAPVTGDDEPAVPEE, were selected as potential vaccine candidates (Table S2).

Prediction of linear B-cell epitopes

Prediction of B-cell epitopes are important for comprehending the immune response because they highlight the areas of antigens that antibodies can recognize. By emphasizing epitopes that could enhance strong antibody responses, these predictions improve vaccine design⁹¹. From the epitope output dataset, 11 B-cell epitopes were finalized, ranging in length from 10 to 40 amino acids. Five epitopes were selected for vaccine development based on their high antigenicity, non-allergenicity, low toxicity, and good water solubility. These epitopes were the following: GVQLSKLSLAGRDQL, RNIVGLKKEESDA, ENMETVGFEERRNQI, RLAAGGQVGSEAR, and DNNGATQTRTL (Table S3).

Population coverage analysis

HLA allele dispersion varies across ethnic groups and geographical regions around the world. One of the most important factors in creating a vaccine construct is predicting population coverage. The current research evaluated the collective population coverage of the shortlisted HTL and CTL epitopes and their matching HLA alleles. It was discovered that about 99.74% of the global population consists of selected epitopes. The highest population coverage was stated in Europe, where combined coverage was 99.96%, and the lowest population coverage was found in Central America, where combined coverage was 53.8%, followed by North Africa, North America, and the West Indies, with reported coverage of 99.01%, 99.89%, and 99.69%, respectively. Furthermore, the population coverage in South Asia, South Africa, South America, and Oceania was determined to be 98.7%, 95.27%, 95.15%, and 97.88%, respectively. Moreover, the population coverage in West Africa, Southwest Asia, Southeast Asia, Northeast Asia, East Asia, East Africa, and Central Africa was 98.43%, 95.79%, 97.66%, 97.88%, 99.67%, 97.08%, and 94.79%, respectively. Briefly, this investigation showed that the identified epitopes might be promising candidates to be considered for the construction of an vaccine construct (Fig. S2).

Development of vaccine construct and analysis

The vaccine construct was created by the use of finalized CTL, HTL, and B cell epitopes; these highly antigenic, non-allergic, and non-toxic epitopes were then linked together using the appropriate linkers, AAY (Alanine-Alanine-Tyrosine), which serve as restriction sites for proteasomes in cells of mammals. This enables effective separation of epitopes within cells: GPGPG (Glycine-Proline-Glycine-Proline-Glycine), They are renowned for their capacity to induce helper T lymphocyte (HTL) responses and KK(Lysine-Lysine), which have the potential to enhance immunogenicity⁹²respectively. Linkers are useful because they aid in vaccine epitope presentation while also preventing junctional epitope development. These linkers were used because of their extensive use in similar multi-epitope vaccine designs⁴⁰ and have been shown to promote appropriate epitope separation, processing, and immune recognition. Since this is an in-silico study and our focus was on computational modelling, there is a need for an experimental validation step of these linkers for the specific vaccine construct. The linker EAAAK was chosen to join adjuvant with construct because it can increases overall structural consistency. The vaccine was designed using twelve MHC class-I, nine MHC class-II, and five B cell epitopes. Only one multi-epitope vaccine construct (V1) was developed by using beta-defensin adjuvant (Fig. S3), which regulates the immune response by coupling with vaccine protein. It was used as an adjuvant because of its dual role in the immune system: direct antimicrobial activity and activation of immune response via TLR4 signaling. It is well reported that it can produce broad-spectrum antimicrobial activity, including against fungal pathogens such as Candida albicans⁹³. Moreover, it is experimentally reported that β-defensins function as endogenous ligands for TLR4, stimulating innate immune responses⁹⁴.

The VaxiJen v2.0 score for the V1 vaccine construct was 0.98, exceeding the antigenic threshold. The vaccine design showed the highest antigenicity, with an estimated ANTIGENpro score of ~ 0.92. The predicted molecular weight of the suggested vaccine construct was 49 kDa, which is the optimal molecular weight for commercial production. The vaccine construct was highly soluble, with a SOLpro value of ~ 0.70. The calculated GRAVY value was − 0.74, indicating the hydrophilic nature of the model. The theoretical PI score was 9.43, the aliphatic index was 62.38, and the instability index indicated a stable vaccine design (Table 3).

Table 3.

Physicochemical analysis of vaccine construct.

Sr. No.	Parameters	V1
		Measurement	Indication
1	ANTIGENpro	0.923202	Antigenic
2	VaxiJen v2.0	0.9873(Antigen)	Antigenic
3	Allergenicity	Non-allergen	Non-allergen
4	Topology	0	Appropriate
5	SOLpro	0.70827	Soluble
6	Instability index	30.37	Stable
7	Molecular weight	49038.06	Appropriate
8	GRAVY value	-0.74	Hydrophilic
9	Amino acid length	453	Appropriate
10	Aliphatic index	62.38	Thermostable
11	Theoretical Pi	9.43	Basic

mRNA vaccine design

To make the mRNA vaccine construct, from N-terminal to C-terminal, the final vaccine construct contains the following: UTR-Kozak sequence-tPA (5’ m7GCap-5’) (signal peptide)-MITD sequence-peptide vaccination − (3’UTR)2, the poly(A) tail (Fig. 3). The Kozak sequence, which starts protein translation in many eukaryotic transcripts, the 3` and 5` beta-globin UTRs, the 5` cap, which stops mRNA degradation, and the tPA signal sequence, which points the target protein toward the cellular secretion pathway, are all included in the construct. To enhance mRNA robustness and translation rate, a 120-base poly(A) tail was inserted at the 3`end (Fig. 3). The RNAfold and Mfold servers displayed the secondary structure of the mRNA vaccine sequence. The minimum free energy scores of -798.50 kcal/mol (optimal secondary structure) and − 508.15 kcal/mol (centroid secondary structure) were predicted by the RNA fold server. Additionally, the Mfold server was also used to calculate the minimum free energy of the vaccine mRNA’s most favorable secondary structure, which was − 765.40 kcal/mol. This conclusion aligns with previous studies indicating that the predicted mRNA secondary structure (Fig. S4) of the current vaccine construct is likely to remain stable following its delivery, transcription, and expression within the host⁹⁵.

Fig. 3.

The schematic diagram shows the final mRNA vaccine construct. The amino acid-designed vaccine comprises 5′ Cap, 5′ UTR, a Kozak sequence, tPA, and EAAAK. MITD sequence, 3′ UTR, poly-A tail, with HTL, CTL, and B-cell epitopes linked by the linkers and adjuvant.

2D and 3D structure prediction and validation

A secondary structure of protein contains alpha helices and beta sheets that play a vital role in maintaining a precise organization of amino acids and proper folding. The PDBsum server determined the secondary structure of the vaccine construct. It indicates which sections of the protein are different (blue) and which are highly similar (red)⁹⁶. The SOPMA server was used to learn about the helices, strands, and coils. The secondary structure of V1 is comprised of 75.06% coils, 20.97% helices, and 3.97% strands. Protein functioning is determined by its three-dimensional structure (Figure S5). The organization and shape of an amino acid residue are critical for its interaction with diverse molecules such as enzymes, receptors, and antibodies. The trRosetta server was utilized in the development of the tertiary structure vaccine. The plot analysis of V1 published by Ramachandran revealed that vaccine residues are concentrated in highly preferred zones, with 91.2% in most favored regions, 7.5% in additional allowed regions, and 0.8% in generously allowed regions. Saves Server was also used to predict the ERRAT score, and results revealed that the quality factor of V1 was 92.3. The quality of non-bonded atomic interactions is determined by the ERRAT’s “quality factor,” where a higher percentage denotes higher quality⁹⁷. For high-quality models, ERRAT yields an overall quality factor of > 50 (Fig. 4). Moreover, for further validation, MolProbity scores contributed to the structure’s overall quality. In general, a MolProbity score below 2 is considered good and acceptable⁹⁸ and our model showed a MolProbity score of 1.22, which indicates minimal steric conflicts and a well-optimized structure. According to the QMEAN score of 0.69, the predicted model scored as an outstanding model.

Fig. 4.

The three-dimensional structure of the subunit vaccine and model validation. (A) 3D vaccine structure predicted by the trRosetta server (B) The Ramachandran plot of the vaccine (C) Graph with ERRAT score, which measures the quality of the 3D model.

PTMs, globular regions and proteasome cleavage analysis

PTMs significantly impact protein stability and function. PTMs can be classified into reversible and irreversible alterations, with some modifications, such as N-linked glycosylation, occurring before protein folding. Glycosylation refers to the enzymatic binding of an oligosaccharide to a protein residue. This mainly happens in the endoplasmic reticulum (ER) and Golgi apparatus for protein binding for secretion or cell surface expression; however, typical cytoplasmic and nuclear glycoproteins are known⁹⁹. In terms of PTM sites, vaccine V1’s sequence contains no N-glycosylation sites. Phosphorylation may trigger substantial structural modifications, resulting in noticeable differences in T-cell recognition with a minor impact on the peptide sequence¹⁰⁰. For vaccine V1, the NetPhos-3.0 service predicted 45 phosphorylation sites (13 serine, 21 threonine, and 11 tyrosine) inside the construct. Phosphorylated proteins or epitopes have a high affinity for cytotoxic T cells and play a significant role in promoting targeted and strong immune responses. In vaccine V1, eight disordered areas were evaluated, beginning with the last MHC-I epitope and covering all MHC-II epitopes. B-cell epitopes lacked disordered regions. Except for the last epitope, two possible globular domains have been identified that cover the adjuvant, B-cell epitopes, and MHC-I epitopes (Fig. S6). Furthermore, our designed vaccine needs to generate peptides capable of binding to MHC I and activating cytotoxic T-cells. The Proteasome Cleavage Prediction Server conducted the proteasome cleavage analysis, and a total of 177 immunoproteasomal cleavage sites were identified. These results indicate that our vaccine has the capacity to activate cytotoxic T-cells. Proteasomes play an important role in class I antigen presentation in mammals, where there are two types of proteasomes: the constitutive proteasome and the immunoproteasome¹⁰¹. Most cells express the constitutive or standard proteasome, but when exposed to proinflammatory stimuli, they switch to express the immunoproteasome¹⁰². Some immune cells, particularly dendritic cells, express the immunoproteasome constitutively¹⁰³.

Aggrescan3D and CABS-flex analysis

The vaccine candidate had a total score of -484.6305, a minimum score of -5.1426, a maximum score of 3.5926, and an average score of -1.0698, indicating that it was not prone to aggregation. Fig. S7A displays the Aggrescan3D overlaid structures. The vaccine candidate was also discovered in ten simulation-generated models. A minimal RMSF of 0.2 Å and a maximum RMSF of 4.92 Å were observed for the vaccine candidate (Fig. S7B), indicating that while the core of the vaccine candidate remained stable during the simulation, certain surface loops or termini displayed expected flexibility. Protein stability and solubility must be balanced in order to maintain protein function and prevent aggregation¹⁰⁴. High normalized and global solubility are supported by the negative average and total aggregation scores, respectively. In general, the more negative the value, the higher the global solubility.

Conformational B-cell epitopes prediction

Conformational B-cell epitopes typically consist of one to five linear sections of amino acid residues. These irregular, unevenly spaced epitopes form interaction sites with antibodies due to their proximity in space rather than order. B-cell epitopes generate antibodies and cytokines, inducing an immunological response. The Ellipro server determined the vaccine residues that form the discontinuous B-cell epitopes. For vaccine complex (V1), 225 residues with scores ranging from 0.502 to 0.919 were evaluated to be present in seven conformational B-cell epitopes (Table 4). The epitopes varied in size from 8 to 54 residues (Fig. S8).

Table 4.

Confirmational B-cell epitopes of vaccine V1.

Sr. No.	Residues	Number of residues	Score
1	A: G1, A: I2, A: I3, A: N4, A: T5, A: L6, A: Q7, A: K8, A: Y9, A: Y10, A: C11, A: R12, A: V13, A: R14, A: G15, A: G16, A: R17, A: C18, A: A19, A: V20, A: L21, A: S22, A: C23, A: L24, A: P25, A: K26, A: E27, A: E28, A: Q29, A: I30, A: G31, A: K32, A: C33, A: S34, A: T35, A: R36, A: G37, A: K39, A: C40, A: C41, A: R43, A: K44	42	0.919
2	A: E46, A: A47, A: A48, A: A49, A: K50, A: D51, A: Y52, A: K53	8	0.764
3	A: T152, A: P153, A: S154, A: L155, A: A156, A: A157, A: Y158, A: Y159, A: Q160, A: Y161, A: S162, A: P163, A: T164, A: E165, A: R166, A: F167, A: A168, A: A169, A: Y170, A: R171, A: T172, A: A173, A: A211, A: G212, A: P213, A: G214, A: P215, A: G216, A: Q217, A: G218, A: D219, A: Y220, A: K221, A: Y245, A: N246, A: L247, A: D248, A: T249, A: K250, A: D251, A: G252, A: P253, A: G254, A: P255, A: G256, A: Q257	46	0.758
4	A: E54, A: I55, A: T56, A: T57, A: A58, A: R59, A: A60, A: A61, A: Y62, A: R63, A: T64, A: H65, A: P66, A: V67, A: T68, A: G69, A: E70, A: K71, A: A72, A: A73, A: Y74, A: H75, A: E76, A: S77, A: G78, A: A79, A: A80, A: T81, A: S82, A: L83, A: A84, A: A85, A: Y86, A: S87, A: T88, A: A89, A: E90, A: A91, A: T92, A: T93, A: A94, A: R95, A: A96, A: A97, A: Y98, A: K99, A: E100, A: N101, A: G102, A: V103, A: V104, A: A105, A: T106, A: F107	54	0.727
5	A: P273, A: G274, A: P275, A: G276, A: L277, A: N279, A: D280, A: F281, A: L282, A: A283, A: N284, A: M285, A: S286, A: H287, A: E288, A: I289, A: R290, A: T291, A: G292, A: P293, A: G294, A: P295, A: G296, A: M297, A: D298, A: D299, A: Y300, A: A367, A: V368, A: P369, A: E370, A: E371, A: K372, A: K373, A: G374, A: V375, A: S378, A: K379, A: L380, A: S381, A: L382, A: A383, A: G384, A: R385, A: D386, A: K389, A: K390, A: G412, A: E414, A: E415, A: N418, A: K421, A: K422	53	0.631
6	A: L424, A: A425, A: A426, A: G427, A: G428, A: Q429, A: V430, A: G431, A: S432, A: E433, A: A434, A: R435, A: K436	13	0.63
7	A: M408, A: E409, A: T410, A: T447, A: L448, A: E449, A: A451, A: A452, A: K453	9	0.502

Molecular docking of MEVC with TLR4

Coordination of a targeted immune response against infectious pathogens depends on the precise relationship between human TLRs and vaccine components. TLRs regulate both innate and adaptive immunity, which in turn regulates the activation of important cytokines and APCs¹⁰⁵. There are ten known TLRs in humans (TLR1–TLR10). The most well-characterized of them is TLR4, which has been said to be essential for identifying fungal PAMPs¹⁰⁶. The production of essential immunological chemicals, which draw in and activate a variety of immune cells, depends on this reaction. This coordinated activation provides long-term immunity to pathogens. Protein internal cavities, binding sites, surface structural pockets, area, and volume of each pocket were identified and determined using the Castp server¹⁰⁷. The best pocket of the vaccine complex (V1) has a volume of 19481.835 ų and an area of 29646.045Ų (Figure S9).

Molecular docking experiments were employed to examine the interaction of the proposed vaccine with the TLR4. Docking is an in-silico approach that predicts the binding affinity and direction of a receptor and its ligand¹⁰⁸. The interactions between vaccine V1 and Toll-like Receptors (TLR4) were discovered, providing useful insights into potential implications for vaccine design and immune response control. TLR4 and V1 got the lowest score of -1085.9 kcal/mol, as calculated. PyMol v2.4 software was utilized to investigate the interactions between ligand and toll-like receptor (TLR4). A total of 21 interactions were evaluated that indicate the strongest hydrogen bonding between the ligand and receptor. Overall bond lengths of docked complex V1 rely on 1.7 to 2.5, and residues that were involved in docking are as follows: LYS-39, THR-268, ARG-391, SER-570, ASN-35, ARG-330, THR-499, ARG-290, GLU-474, ASN-279, ASP-280, ASP-502, LYS-477, GLN-430, ASN-44, GLU-130, ARG-67, TRP-119, GLU-89, GLN-118, TYR-122, PHE-63, ARG-87, GLY-111, ARG-126, THR-125, GLU-42 (Fig. 5).

Fig. 5.

Visual representation of molecular docking of vaccine construct V1 (pink) with TLR4 receptor (golden). Interacting amino acids of TLR4 with V1 residues are represented by blue and pink colors, respectively.

Normal mode analysis of vaccine

To confirm the stability of the vaccine, NMA evaluation of V1 was performed utilizing the IMODs server. The results include graphs displaying the deformability, covariance matrix, and elastic network (Fig. S10). These graphs show how stable vaccine-receptor complexes are. The deformability, covariance matrix, B factor, eigenvalue, variance, and elastic network graphs show the key stability of the vaccineV1with TLR4. Deformability indicates a molecule’s ability to change its residues. It is shown as peaks. Deformability increases with peak height (Fig. S10A). The coupling between pairs of residues, or whether they experience correlated (red), uncorrelated (white), or anti-correlated (blue) motions, is represented by the covariance matrix (Fig. S10B). The B-factor values acquired from the root mean square (RMS) and NMA were consistent (Fig. S10C). The related eigenvalue for each normal mode indicates the rigidity of motion; lower eigenvalues indicate simpler structural deformations, which are associated with reduced energy consumption (Fig. S10D). The variance of each normal mode is inversely proportional to its eigenvalue; the individual and cumulative variances are shown by blue and green color bars, respectively (Fig. S10E). The elastic network model recognizes which atom pairs have springs connecting them. A spring between the corresponding atoms is portrayed by each dot on the graph. Dots are colored according to how stiff they are; stiffer springs are represented by darker grays, and vice versa (Fig. S10F).

MD simulation analysis

RMSD

The root mean square deviation (RMSD) was determined for the ligands, pocket, and protein to determine dynamic stability and sampling pattern. The molecules’ original shapes were used to calculate the RMSD over 200 ns of molecular dynamics simulation¹⁰⁹.The protein RMSD (grey; black highlighted peaks) showed fluctuations initially during the equilibration phase. After this the peaks of protein stabilized with minor alterations within the range of RMSD between 3.0 and 4.5 ± 0.5 Ǻ, suggesting that the TLR4 structure remained stable throughout the simulation period with no significant conformational changes (Fig. 6). Moreover, the pocket RMSD (olive green; red highlighted peaks) showed an almost similar trend to protein’s RMSD during simulation; indicating the structural stability of binding pocket. The RMSD of pocket remained stable after the initial deviations within the range of 2.0 and 4.0 ± 1 Ǻ during the simulation, the RMSD increased in a significant manner in the first 100 ns and then decreased for structural rearrangements for binding, indicating that the pocket remained intact during the vaccine binding. Furthermore, the RMSD of the ligand (pink and blue highlighted peaks) remains steady throughout the simulation, with minimal alterations ranging from 0.5 to 1.5 ± 0.5 Ǻ. The ligand had the lowest RMSD in the V1-TLR4-bonded system, indicating that it is tightly complexed with the receptor (TLR4) (Fig. 6). Overall, the RMSD trajectory demonstrated that the vaccine-TLR4 complex maintained the structural integrity and binding stability.

Fig. 6.

RMSD conformation values of the ligand, protein, and receptor binding pocket during the corresponding 200 ns MD simulation.

RMSF

Root mean square fluctuations (RMSF) serve to explain the protein area fluctuations that occur during the simulation. Each residue’s flexibility is thus assessed in order to have a greater understanding of how protein binding affects the flexibility of the complex. The existence of intrinsically flexible areas led to significant variations in RMSF values within the complex. This is closely related to the loop regions that are joined at the end terminals and contain adjuvants along with inserted epitopes and linker sequences. RMSF provides an explanation for the local variations along the protein chain. Peaks in the RMSF graphs indicate the areas of the protein that fluctuate the most throughout the simulation. The residual flexibility analysis was conducted in order to have a better understanding of the developed complex’s stability. Figure 7 shows that the V-TLR4 had lower flexibility values, confirming the stability and effectiveness of our study.

Fig. 7.

RMSF values for protein structure residue for the protein-ligand complex obtained throughout a 200 ns MD simulation.

Radius of gyration and SASA analysis

The overall compactness and conformational stability of the vaccine-receptor complexes over a 200 ns trajectory of molecular dynamics simulation were revealed by the Rg plots. The Rg plot exhibited minor deviations within the range of around 28.8 to 30.4 ± 0.2 Ǻ. No significant alterations in peaks suggested that the complex remained structurally compact and stable. The Rg plot demonstrated that the vaccine candidate maintained structural stability in the TLR4 complex (Fig. 8).

Fig. 8.

The Radius of Gyration (Rg) analysis for the docked vaccine and TLR4 complex, obtained by running the MD simulation for 200 ns.

Moreover, the SASA analysis of the V1-TLR4 was performed to determine the extent of surface exposure to solvent molecules throughout the simulation period, as illustrated in Fig. 9. The SASA values remained stable for the vaccine and receptor complex, fluctuating within a narrow range (1900 A²-2080 A²), indicating that the complex maintained a consistent surface structure without undergoing significant conformational unfolding or rearrangements. This stability indicated that binding of V1 to TLR4 might not cause significant structural disruptions or changes in the exposure of amino acid residues to the solvent environment. Overall, the stable SASA profile demonstrated the structural stability of the complex and indicated their potential as immunologically effective candidates.

Fig. 9.

The solvent-accessible surface area analysis for the docked vaccine and TLR4 complex, obtained by running the MD simulation for 200 ns.

Binding free energy

Binding free energies were estimated to confirm the intermolecular stability¹¹⁰. Several binding interaction energies of the vaccine and TLR4 complex, including van der Waals, electrostatic, polar solvation, and non-polar solvation. The binding energies were − 223.89 kcal/mol, -2003.78 kcal/mol, 2053.54 kcal/mol, and − 29.98 kcal/mol. The total gas phase energy (ΔGgas) was − 2227.67 kcal/mol. Four energy components (ΔEele, ΔEvdW, ΔGnonpol, and ΔGele (PB)) of the V1-TLR4 complex were examined to better understand their impact on binding. Both the van der Waals interaction ΔEvdW (-223.89 kcal/mol) and the nonpolar solvation contribution ΔGnonpol (-29.98 kcal/mol) play a crucial function in the binding free energy. Furthermore, the sum of electrostatic interactions, such as (ΔEele + ΔGele, sol(PB)) and (ΔEele + ΔGele, sol(GB)) were 78.41 kcal/mol and 216.11 kcal/mol, respectively. Additionally, it should be considered that ΔEvdW is substantially stronger than ΔGnonpol, showing that van der Waals interactions may contribute primarily to ligand binding to the TLR4 receptor. MEVC and TLR4 have significant binding interactions, with estimated total binding free energies ΔGpred (PB) and ΔGpred (GB) of -174.13 kcal/mol and − 37.75 kcal/mol, respectively (Fig. 10). This leads to an effective immunological response.

Fig. 10.

Binding free energy calculation for the vaccine-TLR4 complex during the MD simulation.

The significant positive contributions (ΔG ≥ 0.5 kcal/mol) of several residues highlight their importance at the interface (Fig. 11A). The Fig. 11B also shows that (electrostatic and van der Waals) interactions stabilized the binding, with van der Waals being important for hydrophobic residues and electrostatics for charged ones.

Fig. 11.

Total binding free energy per residue and energy contribution graphs. (A) Per-residue energies and labeled significant contributors (B) Energy decomposition per residue.

A comparative analysis highlighted the dynamic behavior of the V1-TLR4 complex over both 100 ns (with the TIP3P model) and 200 ns (with the TIP4P model) simulations. The RMSD graphs demonstrated that at 100 ns (Fig. S11) the complex stabilized earlier, but at 200 ns it reached a more consistent and stabilized RMSD (Fig. 6), suggesting fewer conformational changes throughout the simulation. RMSF analysis explains the protein area fluctuations that occur during the simulation better at 200 ns (Fig. 7) as compared to 100 ns (Fig. S12). Moreover, the Rg remained stable in both simulations, but at 200 ns the complex showed improved compactness, especially in the last 50 ns as compared to 100 ns (Fig. S13), suggesting consistent structural integrity. Similarly, the SASA analysis revealed stabilized peaks in both simulations, with the 200 ns simulation showing better results in comparison with 100 ns (Fig. S14). Overall, the MD simulation performed at 200 ns with the integration of the TIP4P model led to improved structural stability and more consistent interaction patterns, particularly in the last 50 ns of the trajectory.

PCA analysis and free energy landscape analysis (FEL)

Principal component analysis (PCA) is a method for reducing dimensionality of complex motion data without eliminating significant data. PCA simplifies overall motion by focusing on the primary directions of atomic displacement, notably the first two principal components (PC1 and PC2), which represent a significant portion of structural alterations. In the PCA plot (Fig. 12A), the progression of simulation time is indicated by a color gradient ranging from blue (early phase) to red and orange colors (later phase). The transition from blue to red indicated a smooth conformational change toward a stable and highly favorable bound state. The presence of densely packed red/orange clusters suggested stable conformations, indicating that the complex has attained a stable structural state, likely suitable for immune recognition.

Fig. 12.

Principal component analysis and free energy landscape projection. (A) The PCA plot color gradient from blue to orange/red represents the simulation’s time progression, with blue indicating the early phase and orange/red indicating later stages. (B) The FEL plot is represented in the form of ΔG◦ values in the FEL.

Furthermore, the FEL analysis was employed to analyze the conformational stability of vaccine-receptor complexes, with the first two eigenvectors (PC1 & PC2) generated from the MD simulation trajectory¹¹¹.

In the FEL map, regions with lower free energy suggest stable states, indicated by warm colors, and areas with higher free energy indicate transitional states, indicated by light colors. The FEL map for the complex V1-TLR4 complex also revealed a clear transition from high-energy to low-energy regions. This suggested that the complex undergoes early structural rearrangements before establishing a stable interaction with TLR4 (Fig. 12B).

DCCM

Significant information about the dynamics of structure and interactions within the complex is provided by the dynamic cross-correlation matrix (DCCM) analysis of the V1-TLR4-docked complex across a 200 ns MD simulation. The DCCM displays a color range from − 0.6 (blue) to 1.00 (brown), which represents different patterns of positive and negative correlations within the residues. Certain domains or parts of the complex move in a highly coordinated manner, as indicated by the large brown blocks along the diagonal and off-diagonal areas. Significantly, areas surrounding residues ~ 0–250 and 600–1050 exhibit strong positive correlations, suggesting that these areas move in tandem during the simulation and may represent structurally or functionally related domains. Conversely, smaller blue and yellow portions show areas where residues flow in opposing directions, suggesting possible functional antagonistic sites or mechanical linkage. The complex’s dynamic equilibrium and functional interaction depend on these oppositely linked motions, which are rare but essential. The presence of both correlated and anti-correlated motions indicates that the V1-TLR4 complex needs a high degree of structural orchestration in order to function. All things considered, the DCCM provided a thorough examination of residue interactions and dynamic activity, offering crucial information for upcoming functional and mechanistic studies of the V1-TLR4 docked complex (Fig. 13).

Fig. 13.

A dynamic cross-correlation matrix of Ca atoms around mean positions for molecular dynamics simulations spanning 200 ns was calculated. Color-coding the extents of correlated and anti-correlated motions from brown to blue indicates positive and negative correlations, respectively.

Immune simulation

The immune simulation profile of the vaccine (V1) was evaluated using the C-ImmSim server. Immune simulations involve the injection of vaccines in doses. Typically, three injections are given at different intervals to boost immunogenicity, which peaks after the third dose. In immunological simulations, the vaccine was administered in three doses at time step intervals of 01, 84, and 168. Following each of the three vaccination doses, C-ImmSim found that the progressive rise in immunoglobulin levels, including IgG, IgG1, IgG2, and IgM, greatly boosted the primary immune response. After the third dose, the immunoglobulin concentration was substantially higher. The antigen concentration, on the other hand, dropped during and following the second and third doses of the vaccine, as seen by Graph “A” in Fig. S15. The graph “B” in Fig. S15 depicts the B-cell immune response, indicating an active B-cell response that remains consistently high after the third injection. Similar to graph “B”, graph “C” in Fig. S15 depicts how the entire B-cell profile changes after vaccination. Graph “D” displays the cytotoxic T-cell duplication profile, indicating a good immunological response. Graph “E” demonstrates a significant rise in helper T-cell generation, whereas the concentrations of active and resting helper regulatory T cells increased following the initial vaccination and progressively declined over time, as seen by Graph “F” in Fig. S15. These findings indicate an optimal immunoregulatory mechanisms, increased antigen clearance, and successful immune memory development after every injection.

Immune response activation

A vaccine with several epitopes can elicit humoral and cellular immunity. A key element of effective vaccination against infections is adaptive immune responses. The two cell types that make up the adaptive immune system are B-cells, which mature into plasma cells to produce antibodies, and antigen-specific T-cells, which are stimulated to multiply by APCs. CD8 + has cytotoxic properties. T-cells are principally in charge of eliminating tumor cells that express the appropriate antigens and cells infected by external sources like fungus. When peptides connected to MHC class I molecules engage with their TCR, they are activated. For the immune response to be established and optimized, CD4 + Th-cells are essential. Since these cells are neither cytotoxic nor phagocytic, they are unable to eradicate infections or kill infected cells directly. However, by directing other cells to do these tasks and controlling the kind of immunological response that arises, they “mediate” the immune response. TCR recognition of antigens coupled to class II MHC molecules activates T-helper cells. Th-cell stimulation results in the release of cytokines that control the activity of several cell types, including APCs, which in turn stimulate the production of memory cells and antibodies (Fig. 14)¹¹¹.

Fig. 14.

Proposed mechanism of immunity induced by the vaccine.

Codon adaptation and in-silico cloning

Codon adaptation is an approach that increases the efficiency of codons, leading to higher rates of expression. The OPTIMIZER tool was used to adjust the codons in the MEV sequence. A simple method for adapting the target gene’s codon usage to the expression host is the OPTIMIZER tool. To assess the construct’s potential for protein expression, OPTIMIZER estimates the codon adaptation index (CAI) and percentage CG content. The SnapGene 8.0.2 software 8.0.2 (www.snapgene.com) was used to construct the cloned structure of the vaccine into the pET-28a (+) expression vector. The optimized DNA sequence was added into the E. coli vector pET-28a(+) using the PsiI and BstZ17I restriction sites (Fig. 15). A GC concentration of 30–70% and a codon adaptation index (CAI) value greater than 0.7 are considered adequate. According to OPTIMIZER, the vaccine complex had a CAI of 1.0 and a GC content of 56.9%, which indicates a strong potential for effective translation in the host organism.

Fig. 15.

The optimized vaccine construct (V1) was virtually cloned into the pET-28a (+) expression vector.

Identification of novel drug targets

After subcellular localization, a total of 26 proteins were finalized, with 3 proteins serving as vaccine targets and 19 as drug targets. However, 4 proteins had the highest probability of being employed as either drug or vaccine targets based on the predicted localization of three servers. Out of the 23 drug target proteins, 12 have similarities with FDA-approved targets, with the remaining 11 proteins being suggested for further investigation. Expasy’sProtParam program was used to evaluate the physicochemical characteristics of these 11 proteins, including their instability index, molecular weight (MW), theoretical pI, and grand average of hydropathicity (GRAVY). The transmembrane topology of proteins was also determined. Six proteins with molecular weights < 100 kDa were identified as potential drug targets. The theoretical Pi scores ranged from 5.5 to 6.4, the instability index values ranged from 28.36 to 38.91, and the estimated GRAVY values ranged from − 0.125 to -0.415, indicating the hydrophilic nature of the final drug targets, and their topological values were 0.

3D structure prediction and drug pocket screening

The tertiary structures of the six proteins were predicted using the SWISS model server. To determine the most suitable 3D model, the ERRAT value and Ramachandran plot of each model were analyzed using the SAVES v6.0 server. All Ramachandran plot residues were in the most favorable regions. The ERRAT analysis of all proteins revealed an overall quality factor of more than 90%. To bind small compounds, a protein must be able to hold a pocket, which is known as pocket of druggability and is considered a critical step during drug formulation. All druggable targets were rated powerful, having a druggability score greater than 0.5, and these targets were determined by the DoGSiteScorer. Protein 1 contains two binding pockets (P_0, P_2) with surface areas of 1994.59 and 1036.45, respectively. Protein 2 has four binding pockets (P_0, P_1, P_2, and P_7), with surface areas of 1249.64, 935.49, 676.62, and 315.99, respectively. Protein 3 contains two binding pockets (P_0, P_2) with surface areas of 1586.92 and 1045.54, respectively. Protein 4 contains three binding pockets (P_0, P_1, P_2) with surface areas of 4841.39, 1181.35, and 1110.27, respectively. Protein 5 has two binding pockets (P_0, P_2) with surface areas of 948.07 and 660.14, respectively. Protein 6 contains two binding pockets (P_0, P_1) with surface areas of 1300.9 and 913.61, respectively (Table 5; Fig. 16).

Table 5.

Proteins in favorable regions of the Ramachandran plot and ERRAT quality factor.

Sr. No.	Accession numbers	Protein description	ERRAT quality factor	A.A in favored regions
1	A0A2N3NKV7	Tryptophan synthase	97.6	93.6
2	A0A2N3N5D6	Fructose-bisphosphate aldolase	90.9	86.6
3	A0A2N3NEZ7	Phospho-2-dehydro-3-deoxyheptonate aldolase	98	91.3
4	A0A2N3NFW9	L-ornithine N(5)-monooxygenase	96.6	88.7
5	A0A2N3NAZ6	Isocitrate lyase	96.6	94.7
6	A0A2N3N399	anthranilate synthase	98.2	96.1

Fig. 16.

Drug pocket screening of novel drug targets from the genome of L. prolificans (A) A0A2N3NKV7 (B) A0A2N3N5D6 (C) A0A2N3NEZ7 (D) A0A2N3NFW9 (E) A0A2N3NAZ6 (F) A0A2N3N399.

STRING interaction analysis

The STRING v10.5 databases were used to characterize selected targets and estimate protein-protein interactions (Fig. S16). PPI studies comparing the nominated proteins to others from the same strain can help identify the most appropriate therapeutic targets¹¹². PPI studies with a medium confidence score (0.400) were selected to minimize false-positive results. Hub proteins with node degree (K) ≥ 5 comprised a significant number of interactions. Six proteins were selected as appropriate targets for this characterization.

Conclusion

L. prolificans is responsible for a wide range of infections, including life-threatening pulmonary, disseminated, and CNS infections. Currently, there is no existing FDA-licensed vaccine to prevent L. prolificans infections, indicating the need for a vaccine. In this study, a computational approach was utilized on the reference genome of L. prolificans, resulting in four highly conserved and prospective vaccine candidates that met numerous vaccine candidacy characteristics. Additionally, CTL, HTL, and LBL epitopes were identified from these proteins to produce the multi-epitope mRNA vaccine. The resulting mRNA vaccine was found to be soluble, hydrophilic, and acidic. The structural analysis demonstrated that the vaccine was stable. Moreover, our study showed that the vaccine has a significant affinity with the TLR-4 receptor, validated by MD simulation, MMPBSA, Rg, SASA, PCA, and DCCM analysis. The vaccine may trigger both humoral and adaptive immune responses. This mRNA vaccine might be stable enough after entering, being transcribed, and being expressed in the host. Our study provides a rationally designed, linear-epitope-based mRNA vaccine construct as a starting step, with an expectation that experimental testing will be required to validate its immunogenicity and protective effectiveness.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

MBIR: Formal analysis, Investigation, Writing – original draft. FA: Investigation, Writing – original draft. MUK: Formal analysis, Software, Writing – original draft. HE: Investigation, Data curation. UN: Data curation, Writing–review & editing. AA: Visualization, Resources, Funding acquisition, Writing–review & editing. RU: Investigation, Data curation, Visualization. KC: Visualization, Writing–review & editing. SCO: Conceptualization, Investigation, Writing–review & editing. MS: Conceptualization, Investigation, Supervision, Resources, Writing–review & editing.

Funding

The authors wish to thank Princess Noura bint Abdulrahman University Researchers Supporting Project Number (PNURSP2025R33), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia, for the financial support.

Data availability

The molecular dynamics (MD) simulation trajectory data, the predicted epitopes and vaccine 3D model have been deposited in the public repository Figshare under the DOI: http://doi.org/10.6084/m9.figshare.29613395.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-14907-y.