Background

The Junín virus (JUNV) is a member of the Arenaviridae family, renowned for causing Argentine haemorrhagic fever (AHF), which is a severe endemic infection prevalent in populations residing around agricultural areas in Argentina [1]. The first cases were reported in the 1950s during the unleashed propagation of Calomys musculus. Since then, hundreds of cases have been registered annually. Humans will become victims by the inhalation of rodent aerosols or excreta, generally in the harvest season, so it is a rodent-borne virus [2]. Common symptoms of AHF include flu-like signs, such as headache, malaise, and fever. Primary infection sites are the lungs and then circulate to parenchymal tissues [3]. The final stage of AHF shows neurologic and haemorrhagic complications. The World Health Organization (WHO) has classified AHF as an emerging disease, warranting immediate research to design antiviral agents and vaccine targeting the virus components.

JUNV’s genome consists of two single-stranded RNA segments: a small (S) segment of 3.4 kb and a large (L) segment of 7.2 kb. The S segment translates in the cytoplasm giving rise to nucleoproteins and glycoprotein precursors that mature into a glycoprotein complex after cleavage by cellular proteases. The glycoprotein is crucial for viral attachment to host cells, initiating the viral components' entry into the host cell [4,5,6]. Glycoprotein formation occurs in the endoplasmic reticulum and then infects parenchymal cells through circulation [4]. Glycoprotein also plays a role in cell pathogenicity [7].

In the late 1990s, an effective live-attenuated vaccine was developed, significantly reducing AHF incidence. However, there are no effective remedies to counter viral infection among victims [8]. Later, the application of immune plasma therapy to neutralize the antigens decreased the 1% fatality rate if treated before eight days of the onset of symptoms [9,10,11,12]. The FDA-approved nucleoside analogue ribavirin inhibits JUNV polymerase and serves as an antiviral, but its use is complicated due to side effects and less efficacy [12, 13]. Numerous small molecule (drugs) antiviral compounds have been reported as antagonists to JUNV in both in vitro [14,15,16,17] and in animal models [18, 19]. In silico approaches, such as molecular docking, have been employed to identify the potent drugs from the FDA against the glycoprotein of the Junin virus, revealing MK-3207 and dihydro-ergotamine [17]. Additionally, the live-attenuated vaccine strain Candid #1 (Can) has shown promise in inhibiting glycoprotein spread and reducing infections [18].

Until now, the experiments have primarily focussed on neutralized live-attenuated vaccines, but in this study, we aim to design a new vaccine specifically targeting the glycoprotein. Firstly, we have chosen the mature glycoprotein sequence in identifying and designing a new vaccine through immunoinformatics approaches such as predicting the B cell and T cell epitopes. Next, validating the peptide sequences to ensure the accuracy and reliability of epitopes. Subsequently, designing the vaccine by incorporating suitable adjuvants and linkers and analysing the physicochemical properties. Furthermore, detailed evaluations of the secondary and tertiary structural features of the designed vaccine sequence followed by refining the 3D structure. Furthermore, we checked the binding affinity score and molecular dynamic simulation of the designed vaccine. We presented the detailed methodology followed during this study in a pictorial format, making it visually accessible and facilitating a better understanding in Fig. 1.

Fig. 1
figure 1

The methodology followed in this study

Full size image

Methods

Sequence retrieval

The sequence of Junín virus glycoprotein precursor was retrieved from the National Centre for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/nuccore/MZ408914.1) [19] of 485 length peptide sequence. NCBI provides various tools and a repository database to access the sequence deposited in it.

Prediction of B cell and T cell [MHC I (TC/CD4+), MHC II (TH/CD8+)] epitopes

To predict B cell linear epitopes, we used the Immune Epitope Database (IEDB) (http://tools.immuneepitope.org/bcell/) and employed Bepipred Linear Epitope prediction 2.0 with a threshold value of 0.50 [20]. This prediction method is based on the input fasta format of the query protein sequence, and the server utilizes the random forest algorithm to distinguish between epitopes and non-epitope sequences found in crystal structures.

For predicting cytotoxic T cells/Class 1/CD4+ epitopes, we used artificial neural network 4.0 [21] and specified human MHC alleles as the source or reference to get the IC50 values of each predicted sequence in the IEDB Analysis Resource (http://tools.iedb.org/mhci/), which determines the subsequence’s binding ability with the specific MHC class I. Helper T cells/Class II/CD8+ epitopes predictions were carried out with the NN-align 2.3 (NetMHCII) method [22, 23] with the human HLA-DR data set and all MHC II allele sequences between 12 and 18 mers in length. The resulting output provided IC50 values for predicted epitopes in the IEDB Analysis Resource (http://tools.iedb.org/mhcii/).

Validation of epitopes

The predicted epitopes were subjected to validation for allergenicity, antigenicity, and toxicity. Allergenicity refers to the ability of an antigen to induce abnormal and hyperresponsive [23]. AllerTOP v. 2.0 (http://www.ddg-pharmfac.net/AllerTOP/) was utilized for allergenicity prediction, employing the k-nearest neighbour algorithm on a training set of 2427 allergens and 2427 allergens, based on ACC uniform length and QSAR with different lengths [24]. Antigenicity, on the other hand, triggers the immune response and describes the ability to bind paratopes [25]. VaxiJen 2.0 (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html) server used to predict the antigens and non-antigens [26]. This server was developed based on independent alignment methods with auto cross-covariance (ACC) of amino acid sequences transformation into principal vector properties to predict antigens and non-antigens. To access the peptides' toxicity, the ToxinPred server (https://webs.iiitd.edu.in/raghava/toxinpred/multi_submit.php) [27] was used. It utilizes the support vector machine (SVM) method with a threshold of 0.5 and an E value of 0.01 to predict the toxic and non-toxic peptides. Furthermore, population coverage analysis (http://tools.iedb.org/population/) [27] was performed to select the world population for MHC class I and MHC class II, which covers 3245 alleles from 16 geographical areas, 21 various ethnicities, and 115 countries.

Designing and characterization of structural vaccine

A serial arrangement of the vaccine was designed with beta-defensin 114 as an adjuvant and EAAK, AAY, GPGPG, and KK as flexible linkers [28, 29]. Epitope inclusion linkers and adjuvants were aimed to enhance protein stability and immunogenicity. To validate the designed vaccine, various physicochemical properties were calculated using ProtParam (https://web.expasy.org/protparam/) [30]. Antigenicity was determined using VaxiJen v 2.0 [26], while solubility with SolPro [31]. The presence of disulphide bonds was analysed through DLpro [32] through the SCRATCH web server [33] (https://scratch.proteomics.ics.uci.edu/). Allergenicity prediction was performed through AllerTOP v. 2.0 [24].

Secondary and tertiary structure assessment of antibody

Psipred [34] (http://bioinf.cs.ucl.ac.uk/psipred/) was designed based on the two feed-forward neural networks using the output from Position-Specific Iterated (PSI) – Blast. The SOPMA Server [35] (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html) uses amino acid sequences by the self-optimized prediction method to predict the secondary structure features of the protein.

For in silico studies requiring tertiary structure of the protein 3D protein models using were generated using the servers such as Iterative Threading ASSEmbly Refinement (I-TASSER) [36] (https://zhanggroup.org/I-TASSER/). The prediction is based on the multiple threading approaches from the protein data bank through Local Meta-Threading Server v 3.0. The complete model was built by iterative template-based fragment assembly. The generated models were refined using the ModRefiner [37] (https://zhanggroup.org/ModRefiner/) servers until they achieved Ramachandran-favoured regions greater than 90%.

Molecular docking and MM/GBSA evaluation

The glycoprotein (GP) complex with the Fab antibody of the Junin virus was obtained from the Protein Data Bank (PDB ID: 5NUZ) (https://www.rcsb.org/) and determined using the X-ray diffraction method. The raw protein structure was made by eliminating the extra bound ligands, water, and het atoms followed by saving them in.pdb format for docking. The GP and the designed antibody were docked in the ClusPro protein–protein docking server (https://cluspro.bu.edu/home.php) [38], which was different from the protein–ligand docking that we performed in our previous studies [39].

Subsequently, the docked complex underwent Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) analysis to determine various interaction energies such as binding affinity, Vander Waal, electrostatic, polar solvation-free energy, and solvation-free energy based on the empirical methodologies after minimizing the complex in the ff02 force field over the HawkDock server (http://cadd.zju.edu.cn/hawkdock/) [40].

Molecular dynamic (MD) simulations

The MD simulations were performed on the refined tertiary structure of the antibody in a water medium using GROMACS [41] (https://simlab.uams.edu/index.php) from the WebGRO sim lab UAMS. GROMOS96 43a1 force field [42] initially used the TIP4P solvation model, to build the triclinic box and maintained 0.15 molarity of NaCl at pH 7.4. Using the steepest descent type of energy minimization 800 kJ/mol at 10,000 steps with NVT and NPT equilibration at temperature 300 k and pressure at 1 atmosphere. The final MD was performed by a leap-frog integer method at 20 ns of time.

Results

Sequence retrieval

The NCBI Virus accession number for the glycoprotein sequences of the GenBank ID: MZ408914.1, represents the glycoprotein sequence of the S segment of 485 length peptide sequence linear RNA genome isolated from the Argentinian mammarenavirus on Feb 2022.

Prediction and validation of B cell and T cell epitopes

The B cell and T cell epitopes were predicted using different algorithms from the IEDB, and their respective antigenicity (A > 0.5), allergenicity (+), and toxicity (N) were manually analysed and curated (eliminated the non-antigenic (NA), allergenic (−), and toxic (T) peptides). Initially, fourteen B cell epitopes were retrieved as outputs, and they were curated based on properties such as antigenicity, allergenicity, and toxicity. Finally, four peptide sequences were selected for the vaccine construction in Table 1, after removing the duplicates. Based on the threshold values, numerous MHC class I and class II epitopes were obtained. Furthermore, filtered based on IC50 values (≤ 20), were selected followed by antigenic, non-allergen, and non-toxic peptides used in designing the antibody sequence in Table 2. Population coverage analysis for MHC I and MHC II epitopes was employed to estimate the potential target population for the predicted epitopes used in designing the vaccine, resulting in the best coverage value of 97.28% world population (Additional file 2: Table S1 and Figure S1).

Table 1 Predicted and selected B cell epitopes
Full size table
Table 2 Selected MHC class I and class II epitopes with antigenicity, allergenicity, and toxicity
Full size table

Designing and characterization of antibody structure

Before designing the vaccine, a suitable adjuvant, beta-defensin was selected, which is widely used. Additionally, appropriate linkers were exploited to connect intra B cells with KK, MHC I with AAY, and MHC II epitopes with GPGPG linkers and among them during vaccine designing. An EAAAK rigid linker, capable of forming an alpha helix at the amine terminal used to link the B cell epitopes and the adjuvant. The GPGPG linker was used between the MHC I epitopes and B cell epitopes. An AAY linker was employed to join the MHC I and MHC II epitopes followed by six histidine tags (6H) at the end in Fig. 2.

Fig. 2
figure 2

The sequential arrangement of epitopes in designing the vaccine

Full size image

The designed antibody underwent physicochemical analysis of the nine descriptor values determined through ExPASY ProtParam. The vaccine comprises 431 residues with 47,835.77 Daltons molecular weight (M.wt.). The isoelectric constant point (pI) of the designed antibody was found to be 9.65. The negatively (Asp + Glu) [−R] and positively (Arg + Lys) [+ R] charged residues were 29 and 57, respectively. The extinction coefficient (EC) value at 280 nm was 36,258 M−1 cm−1. The instability index (II) was 34.87, a value greater than 40 refers to unstable, and a value less than 40 stability. The aliphatic index (AI) was 81.25 and the grand average of hydropathicity (GRAVY) of − 0.254 in Table 3.

Table 3 Physicochemical properties of the designed vaccine sequence
Full size table

The descending order of the amino acid types in the antibody was: Polar > non-polar > basic > aromatic > acidic in Fig. 3. The constructed vaccine was validated as antigenic, non-allergenic, non-toxic, and had a solubility of 0.92. Moreover, it could form six disulphide bonds between the cysteine residues in Table 4 and Additional file 2: Table S2.

Fig. 3
figure 3

Different amino acid type content in the designed vaccine sequence

Full size image
Table 4 Vaccine antigenicity, allergenicity, toxicity, solubility, and disulphide bonds forming the number
Full size table

Secondary and tertiary structure assessment of antibody

The designed antibody fasta sequence was submitted to the SOPMA server for secondary structure prediction. The results indicated that the antibody comprises 26.68% of alpha helices, 41.07% of random coils, 25.99% of extended strands, and 6.26% of beta turns in Table 5. Figure 4, obtained from PSIPRED, exhibits the secondary structure confidence score and provides additional pictorial information about the secondary structure, as depicted in Additional file 2: Figure S2.

Table 5 Secondary features of the vaccine sequence
Full size table
Fig. 4
figure 4

The confidence level of prediction of the secondary structure via PSIPRED

Full size image

The tertiary model obtained from the I-TASSER, though constructed upon validating in the Ramachandran plot, does not have enough amino acids in the allowed regions, leading to a low confidence score (C-score) of − 2.17. The C-Score ranges from − 5 to 2, where a higher value indicates a better model. Based on the confidence score, template modelling (TM-score) and root-mean-square deviation (RMSD) values were calculated to assess the structural similarity between the model and the actual protein. That resulted in a TM-Score of 0.46 ± 0.15 and 12.2 ± 4.4 [43].

The RMSD quantifies the similarity between two superimposed atomic coordinates [44]. The model was sent for structure refinement in the Galaxy refiner, and the results are shown in Additional file 2: Table S3. Based on the Global distance test (GDT) score, RMSD value, molprobity, clash score, rotamers value, and Ramachandran-favoured regions, model 1 was selected as a vaccine which was furthermore analysed. The GDT-HA represents the high-resolution value of the structure, and the MolProbity value is a log-weighted value determined from scores such as clash score, Ramachandran-favoured percentage, and bad side-chain rotamers [45]. Ramachandran’s favoured percentage regions have residues of 90.9% in the most favoured regions of the built 3D antibody. Overall, the selected model was refined and met the criteria for a suitable vaccine candidate, based on the various evaluation structural parameters and properties.

Molecular docking and MM/GBSA evaluation

The interactions between Junín glycoprotein and the designed antibody were analysed through the ClusPro server, and a balanced complex is retrieved in Fig. 5. The interactions between the residues were visualized in the drug discovery studio. Before this, the vaccine formed a total of nine hundred fourteen interactions with the glycoprotein. Including four S–S bonds, twenty-five electrostatic bonds, one-hundred eighty-three hydrophobic bonds, and seven hundred-two hydrogen bonds. Hydrogen bond-forming interactions play a prominent role in drug availability and enhancing the binding chances. Furthermore, the distance criteria of less than two angstroms (< 2 Å) interaction distance among the hydrogen bonds were 280 (Additional file 1).

Fig. 5
figure 5

The docked complex of JUNV glycoprotein (Chain C and L) with the designed vaccine (Chain A)

Full size image

Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) was employed in determining the free energy between the glycoprotein and the designed vaccine (protein–protein interaction). Extensive methodologies in calculating the interaction energies include receptor’s grid identification, pre-processing the input files, T leap function, scoring various energies, and scoring. Briefly van der Waals (VDW) energy was − 151.33 kcal/mol, electrostatic energy (ELE) was − 272.63 kcal/mol, polar solvation-free energy (GB) was 424.39 kcal/mol, and solvation-free (SA) energy was − 15.93 kcal/mol. The total binding free energy of the complex was − 15.5 kcal/mol represented in Table 6. Overall, the analysis indicates strong and favourable interactions between the Junin glycoprotein and the designed vaccine, with a significant binding free energy, suggesting the potent vaccine’s effectiveness in combating the virus.

Table 6 MM/GBSA analysis representing various energies of the vaccine with glycoprotein
Full size table

Molecular dynamic simulations

Understanding the vaccine’s conformational relative stability of the backbone atoms using the RMSD measure provides valuable insights since it measures the deviations of the extent atoms from the starting point. Less deviation (low RSMD) value signifies a better stable protein structure. For the constructed vaccine, the average lowest RMSD value was 0.24 nm in Fig. 6A. However, there was a slight increase in atomic deviations to 0.6 nm at 0.3 ns. This increase in RMSD values between 0.6 to 0.8 nm from 0.3 to 15 ns can be attributed to the modelled vaccine structural interactions optimization.

Fig. 6
figure 6

The molecular dynamic simulations of the 3D vaccine in water. A Root-mean-square deviation (RMSD) computes the average distance between the backbone atoms of starting structure (reference structure) with simulated structures (frame by frame) when superimposed. B Root-mean-square fluctuation (RMSF) computes fluctuations (standard deviation) of atomic positions of each amino acid (residues) in the trajectory [X-axis = Time; Y-axis = Residue number]. C Hydrogen interactions forming residues

Full size image

Furthermore, the protein stability evaluated by root-mean-square fluctuations (RMSFs), which measures the protein residue’s elasticity and binding site adaptation phenomena, was determined. The vaccine, RMSF contains evident residue cluster fluctuations. On average, there were two fluctuations for every hundred residues. Residues 200–320 showed high fluctuations with significant up and down levels. Maximum fluctuations were observed in the 390–400 residues range, while minimum fluctuations between 280 and 389 in Fig. 6B.

Hydrogen bond trajectories retained during the MD simulation were studied to evaluate the strength and binding affinity of the complex. The designed vaccine was found to form several hydrogen bonds with an average of 140 hydrogen bonds when interacting with the antigenic protein. The ability to form multiple hydrogen bonds was crucial in inhibiting the target protein, thereby strengthening the complex’s free energy in Fig. 6C. Overall, analysis of RMSD, RMSF, and hydrogen bond trajectories provides valuable information about the stability and interactions of the designed vaccine, which were essential factors in determining its potential effectiveness as a vaccine.

Discussion

The general effective possibilities that are being examined are to control the dispense of Argentine haemorrhagic fever (AHF) by the design and development of suitable vaccines. The duration of the vaccine development from experimental studies to clinical trial studies takes an extensive period. In this modern era, due to the advancement of system applications in biology, numerous epitopes predicting epitopes are made available using various machine learning algorithms. Among them, multi-epitope-based vaccine designs employing predicting B cell and T cell epitopes have made a remarkable trend in bioinformatics. In addition, proper attention is required to develop a safe and viable vaccine.

To date, the primary approach to prevent the infection rate is neutralized monoclonal antibodies (MABs). Research on the MABs extracted from animal models (mouse) shows that the glycoprotein JUNV is involved in mimicking the human transferrin receptor 1 (hTfR1) during the binding process. This glycoprotein acts as a primary target and plays a predominant role in the prevention and treatment of infection [46]. The vaccine design according to the sequence-specific will act efficiently in inhibiting the antigen and will also be helpful in further mutational studies considering the evolution of the virus variants.

Three epitope types (B cell, TC, and TH) predictions were performed using multiple servers including experimentally based HLA class I and HLA class II alleles [47]. All predicted epitopes were validated based on antigenicity/non-antigenicity, allergenicity/non-allergenicity, and toxicity/non-toxicity and shared maximum population coverage across the world. The number of B cell epitopes obtained after screening was less, indicating fewer interactions. Considering the T cell epitope number obtained, the cell-mediated immunity and immune response generated were high and long-lasting. A commonly used adjuvant (beta-defensin) was selected in designing and the linkers were added appropriately in Fig. 2. To ensure safety and efficacy, the designed vaccine over-checked their non-allergenicity, antigenicity, and non-toxicity. Furthermore, physicochemical, solubility and disulphide bond parameters were checked to determine the stability of the vaccine in Table 4 and Additional file 2: Table S2. This comprehensive validation paves potential applications in immunization against the target antigen.

The physicochemical properties in Table 3 and Fig. 3 include the pI as the pH at which no electrical charge is present on the molecule or the total number of negative and positive charges are equal [30]. The isoelectric focusing technique is performed based on the pI values for separating the molecule from the complex [48]. These values help to isolate the respective protein of interest in the wet lab experiments upon digestion. The glycoprotein’s pI of 9.65 indicates that for wet-lab studies the experimental setup is to be maintained a basic environment during extraction. EC is defined as the amount of light absorbed per mole of protein at a specific wavelength of light. The protein’s EC value is calculated from the composition of tryptophan, tyrosine, and cysteine residues because these amino acids contribute significantly to measuring the protein’s optical density in the 276–282 nm range [30]. Protein–protein and protein-ligands quantitative study can be understood through the EC values [49]. The EC value exhibited was high representing a good sequence for further studies. II indicates the protein stability under both in vivo and in vitro conditions. Proteins with II greater than 40 are considered unstable proteins, while II less than 40 are stable [50]. II of the designed vaccine was 34.87 inferring the stability. AI is another parameter that describes protein stability at temperatures. AI is defined as the relative volume occupied by aliphatic side chains like alanine (Ala), valine (Val), leucine (Leu), and isoleucine (Ile) [30, 48]. The high value of AI indicates the increase in the thermostability nature of the protein which is an additive factor for wet lab studies. The Aliphatic index (AI) of the vaccine sequence is 81.25 indicating the thermostability character. GRAVY value ranges from − 4 to + 4 indicating the hydrophilic and hydrophobic nature of the proteins [51]. The low GRAVY range indicates the possibility of being a globular (hydrophobic) protein rather than membranous (hydrophilic). Thus, the vaccine contains a globular protein residue with a value of − 0.254.

The tertiary structure of the vaccine has been generated from the I-TASSER and further processed in ModRefiner to achieve optimum descriptor values such as GDT-HA, RMSD, Molprobity, C-score, poor rotamers, and Ramachandran-favoured regions greater than or equal to 90% in Additional file 2: Table S3, aids to validate the designed structure all of which are crucial in validating the structural integrity [52, 53]. Subsequently, molecular docking of JUNV glycoprotein and the construct was performed to achieve a balanced complex model from the ClusPro server in Fig. 5. The post-docking analysis, specifically the MM/GBSA method in Table 6, of the docked complex resulted in favourable interaction energies within the docked complex. Moreover, MD simulations on the vaccine in water in Fig. 6 have indicated that the designed construct exhibited notable stability without higher deviations and fluctuations with maximum hydrogen bond forming ability that emphasizes the vaccine stability. These findings collectively underscore the designed vaccine’s reliability and robustness for potential application in combating the JUNV.

Conclusion

In this study, we focussed on vaccine design based on the glycoprotein of the JUNV on employed immunoinformatics approaches to identify antigenic peptides because of the novelty of identifying the peptides using machine learning tools. The vaccine was constructed after selecting the antigenic, nonallergic, and non-toxic and deleting the duplicate and overlapping sequences. The vaccine was designed using 4 B cell, 9 MHC class I, and 8 MHC class II T cell epitopes with appropriate linkers and adjuvant. The tertiary model was generated and refined to meet suitable specific criteria especially Ramachandran-favoured regions greater than 90%. After the physicochemical properties’ determination, molecular docking of JUNV glycoprotein with the designed vaccine performed a balanced complex model chosen for analysing the interactions among them showed 280 hydrogen bond forming residues with radii less than 2 angstroms. MM/GBSA methodology exhibited − 15.5 kcal/mol of binding free energy of the complex. The MD simulations of the vaccine that were performed exhibited fewer values of fluctuations and deviations that signify the vaccine's stability. However, the present work was based on in silico methodologies with promising results that will be evident in the wet lab works to justify the findings.