Abstract

Introduction: African swine fever virus (ASFV) causes African swine fever, a highly fatal hemorrhagic viral disease of domestic swine, severely impacting the development of swine industries in affected countries. The management of this disease is significantly hindered by the absence of protective vaccines against the virus. A cost-effective approach to developing potent vaccines is by employing immunoinformatic tools to identify highly conserved IFN-g , CD4+, CD8+, and B-cell epitopes. These can then be combined to create a multi-epitope broad-spectrum vaccine.

Methods: Therefore, in this study, immunoinformatic tools were utilized to identify CD8+ T-cell, CD4+ T-cell, B-cell, and IFN-g epitopes of the ASFV major coat protein p72, CD2 homologue (CD2v), and C-type lectin-like proteins, which are promising vaccine candidates capable of eliciting a protective immune response against the virus. The epitopes were computationally assembled to generate a multi-epitope subunit vaccine against ASFV. Molecular docking was employed to assess the interaction between the vaccine construct and immune receptors Toll-like receptor 9 (TLR-9) and Swine Leukocyte Antigen-1 (SLA-1).

Results: Molecular dynamic simulation revealed stable interactions between the vaccine construct and the immune receptors. In silico cloning and codon optimization were utilized to enhance the efficient expression of the vaccine in an E. coli expression system. The potential of the vaccine to provoke effective immune responses was evaluated using in silico immune simulation.

Conclusion: These computational approaches have shown that the vaccine is structurally stable and capable of inducing both humoral and cell-mediated immune responses against ASFV. However, this study needs to be validated further experimentally.

Introduction

African swine fever (ASF) is a highly contagious, fatal hemorrhagic disease that affects domestic pigs and wild boars (Sus scrofa), negatively impacting the development of the swine industry in affected countries. The disease was first described in Kenya in 19211 and is now enzootic in Sub-Saharan Africa, the Russian Federation, the Caucasus, Eastern Europe, and, most recently, China2, 3. Currently, there is no vaccine or effective therapeutics for proper management of this malady, and as such, culling infected pigs is the only method of disease control. The African swine fever pandemic has resulted in about a $2 billion loss to the swine industry worldwide. The causative agent, ASFV, belongs to the family Asfarviridae, the genus Asfarvirus, and it is the only known DNA-containing arbovirus4. The 200 nm-diameter African swine fever virus is a nucleocytoplasmic large double-stranded DNA virus. The virion's structure includes an enclosed nucleoid and core shell, icosahedral inner capsid, icosahedral membrane, outer capsid, and outer envelope5. A plethora of proteins contribute to the production of the outer capsid, including p72, p49, and p17, but p72, encoded by the B646L gene, is the main capsid protein of ASFV. It has been utilized to characterize the virus into genotypes and is a crucial target for the development of an ASF vaccine5. The outer envelope harbors the virus attachment protein p12, and the EP402R and EP153R gene products, CD2v and C-type lectin, respectively. CD2v is a homologue of CD2, a cell adhesion molecule expressed by T-lymphocytes (T-cells) and natural killer (NK) cells6. CD2v and C-type lectin proteins have been demonstrated to be responsible for the hemadsorption propensity of ASFV, and this has been employed in hemadsorption inhibition (HAI) serological typing of ASFVs into eight serogroups, although more serogroups might still be undetected. These serotype-specific proteins, CD2v and/or C-type lectin, have also been revealed to be crucial for defense against homologous ASF infection, indicating that ASF protective immunity may be serotype-specific. Since there is no effective vaccine against ASF currently, and taking into account the immunogenicity of p72, CD2v, and C-type lectin proteins, a pragmatic approach is the development of a multi-epitope vaccine against the virus. Multi-epitope vaccines are advantageous over classical vaccines for the following reasons: (I) They incorporate certain IFN-epitopes and other adjuvant-capable innate immune response components, which can improve immunogenicity and long-lasting immune responses; (II) they consist of cytotoxic T-lymphocytes (CTL), T-helper (Th), and B-lymphocyte (B-cell) epitopes, which could resolve the issue of various ASFV serotypes/genotypes in vaccine development; (III) they reduce unwanted components that can provoke hypersensitivity and immunopathologic responses, and very importantly; (IV) they contain several MHC-restricted epitopes that can be detected by T-cell receptors (TCRs) from various clones from different T-cell subsets. The presentations of viral peptides in the context of MHC class I and class II are essential for anti-viral T-cell responses. Unlike B-cell epitopes responsible for eliciting neutralizing antibodies, T-cell epitopes can be located anywhere in a viral protein since cells can process and present both intracellular and extracellular viral peptides. The majority of MHC class I-restricted T-cell epitopes originate from intracellular antigens that are cleaved by the proteasome and transported into the endoplasmic reticulum by the transporter associated with antigen processing proteins (TAP), where they can bind to MHC class I molecules, which are then transported to the cell surface for recognition by CD8+ T-cells. However, extracellular proteins picked up by professional antigen-presenting cells and cleaved in lysosomes are the main source of MHC class II-restricted T-cell epitopes. These proteins can attach to MHC class II molecules, be transported to the cell surface, and then be recognized by CD4+ T cells7. A vaccine-induced CD8+ T-cell-mediated immune response might be long-lived and cross-serotype and thus deserves further attention. The swine MHC, designated swine leukocyte antigen (SLA), consisting of SLA class I loci (SLA-1, SLA-2, SLA-3) and SLA class II loci (DRB1, DQB1, DQA)8, 9, has been shown to be involved in swine immune response to various viral infections such as Foot-and-mouth disease virus and influenza-A viruses. This is possible because of the presence of six pockets (A to F) in the peptide-binding groove (PBG) of SLA-1 that the viral peptides could bind10. Since the mainstay of developing an efficacious multi-epitope vaccine is the selection of appropriate candidate antigens and their immunodominant epitopes, these sites can be easily selected using immunoinformatics. Therefore, the immunoinformatics approach is a good alternative for streamlining a cheap and straightforward experimental approach, since the experimental selection of B- and T-cell epitopes is time-consuming and expensive. Immunoinformatics employs computational methods such as machine learning, artificial neural networks, and deep learning to predict B- and T-cell epitopes. Herein, B- and T-cells of ASFV CD2v, C-type lectin, and p72 epitopes were predicted and validated using immunoinformatic methods. These epitopes were employed to construct and validate the ASFV multi-epitope vaccine. These findings may provide a basis for the development of other multi-epitope vaccines.

Methods Protein sequences

The amino acid sequences of CD2v, C-type lectin, and p72 of all serotypes/genotypes of ASFV were retrieved from GenBank. The retrieved sequences for each protein were aligned within serotypes using Geneious software. The consensus sequences of each protein were subjected to Blast-p analysis (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The best matches for each protein sequence, CD2v (AJB28366), C-type-like protein (AAC28412), and p72 (QID21249), were selected for epitope screening. The antigenicity (at a threshold of 0.4) of selected proteins was evaluated via VaxiJen v2.0 (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html)11.

T-cell and B-cell epitope prediction

An array of tools exists for predicting epitopes that could be presented by MHC on host cells to T-cells. The Immune Epitope Database and Analysis Resource (IEDB), a web-based epitope analysis resource that includes tools for T cell and B cell epitope prediction, was utilized. This resource employs artificial neural network (ANN), stabilized matrix method (SMM), and combinatorial peptide libraries (CombLib) to predict the natural process and presentation of peptides by MHC I. IEDB MHC-I NetMHCpan EL 4.1 has primarily been trained on existing MHC–peptide binding data from chimpanzee, cow, gorilla, human, macaque, mouse, pig, dog, and horse to predict the peptide-binding specificity of MHC molecules. It was used to screen the extracellular domains of CD2v (amino acid positions 17-204), C-type lectin (amino acid positions 52-161), and p72 as epitopes for SLAs that would be presented to CD8+ T-cells (CTLs). For SLA-II (HLA-II) alleles, which were employed to select peptides for presentation to CD4+ T-cells since SLA-II is not present in IEDB, it has been demonstrated that HLA-II can serve as a proxy for other mammals12. Furthermore, SLA-II genes have shown much stronger sequence similarity with their HLA-II counterparts than with each other13. The IEDB-recommended method2.22 was used to screen for the aforementioned proteins for potential peptides as epitopes for MHC-II (HLA-II) that would be presented to CD4+ T-cells (HTLs). The ABCpred server (http://crdd.osdd.net/raghava/abcpred/) was utilized for B-cell epitope prediction, which is based on ANN for predicting linear epitopes.

Conformational/discontinuous B-cell epitopes were predicted from the proteins using IEDB Discotope 2.0 (http://tools.iedb.org/discotope/) with default parameters14. The discontinuous epitopes were selected against the resolved structure of ASFV p72 (PDB:6KU9). Epitopes were chosen based on promiscuity, antigenicity, and non-toxicity criteria. The antigenicity of selected epitopes was verified by VaxiJen v2.0 (at a threshold of 0.4)11. ToxinPred (http://crdd.osdd.net/raghava/toxinpred/) was used to predict the nontoxic/toxic properties of epitopes15.

IFN-γ epitope prediction

IFN-γ plays a critical role in antiviral, anti-tumor, and immune regulatory activities. It triggers both innate and adaptive immune responses by activating macrophages and NK cells16. Thus, identifying IFN-γ inducing epitopes is crucial for designing a potential multi-epitope vaccine. The IFNepitope server (https://crdd.osdd.net/raghava/ifnepitope/) was used for predicting IFN-γ epitopes in the target proteins.

Multi-epitope subunit vaccine construction, structural modeling, refinement, and validation

The vaccine was constructed by linking Suis scrofa β-defensin, the selected CTL, HTL, B-cells, and IFN-γ inducing epitopes, along with the TAT protein together via linkers such as EAAAK, GPGPG, KK, and AAY. β-defensins, endogenous alarmins/antimicrobial peptides, are responsible for promoting both local innate and adaptive systemic immune responses against pathogens. They play roles in immune responses by inducing innate immune cells and recruiting naïve T-cells through the chemokine receptor-6 (CCR-6)17. This antimicrobial peptide was appended at the N-terminal of the vaccine construct to enhance the immune response provoked by the vaccine. At the C-terminal of the vaccine construct, a 12-mer cell-penetrating peptide with residues GRKKRRQRRRPQ, derived from the Human immunodeficiency virus-1 (HIV-1) transactivator of transcription (TAT) protein18, was grafted. This peptide, with cell-penetrating properties, will assist the vaccine construct in intracellular delivery and sub-cellular localization. The amino acid residues of β-defensin at the N-terminal of the vaccine construct were linked with CTL epitopes via EAAAK linkers, and the CTL epitopes were linked through AAY linkers. CTL epitopes were joined with HTL epitopes by GPGPG, while B cells, IFN-γ, and the TAT protein were linked via KK linkers. These linkers are essential for protein folding, flexibility, and separation of functional domains of the vaccine construct. A Protein BLAST search was employed using default parameters to ensure that the vaccine construct is not homologous to the pig proteome. Proteins with less than 37% identity with 100% coverage are regarded as non-homologous. The physicochemical parameters of the vaccine construct were determined using the ProtParam tool (http://web.expasy.org/protparam/)19. ProtParam predicts various physicochemical parameters such as the number of amino acids, molecular weight, theoretical isoelectric point (pI), amino acid composition, atomic composition, formula, extinction coefficients, estimated half-life, instability index, aliphatic index, and the GRand AVerage of hydropathY (GRAVY). The VaxiJen v2.0 server (http://www.ddgpharmfac.net/vaxijen/VaxiJen/VaxiJen.html) was used to predict the antigenicity of the vaccine construct through an alignment-free approach based on autoregressive moving average_cross_mean absolute deviation (ARMA-MAD) transformation of protein sequences into uniform vectors of principal amino acid properties11. A key feature of the vaccine is that it should be non-allergenic, as allergenic proteins can induce hypersensitivity in the host. Therefore, the vaccine must be non-allergenic. The AlgPred server (http://www.imtech.res.in/raghava/algpred/) was used to determine the allergenicity of the vaccine construct. A hybrid approach (SVMc+IgE epitope+ARPs BLAST+MAST) in AlgPred was utilized to predict the allergenicity of the vaccine construct based on its high accuracy and sensitivity20. The tertiary or three-dimensional (3D) structure of the vaccine construct from its linear structure was generated using a transform-restrained Rosetta (trRosetta) algorithm, which predicts protein 3D structures based on predicted inter-residue orientations through a deep learning-based method21. The predicted 3D structure was refined using ModRefiner, an algorithm used for atomic-level, high-resolution protein structure refinement, which can start from either a C-alpha trace, a main-chain model, or a full-atomic model22. Thereafter, the 3D structure was validated using the ERRAT score at the ERRAT server (https://servicesn.mbi.ucla.edu/ERRAT/), which evaluates the calculation of non-bonded interactions in the vaccine construct structure. The overall quality of the generated model of the vaccine construct was also validated by PDBSum Generate (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/Generate.html) for Ramachandran plot analysis and G-factor determination. The Ramachandran plot reveals favored regions for backbone dihedral angles Phi (ϕ) and Psi (ψ) against amino acid residues in protein structure23. This was followed by ProSA-web analysis to validate the structure based on the Z-score prediction.

Molecular docking of the vaccine construct with swine immune receptors

Germline-encoded pattern recognition receptors (PRRs) such as toll-like receptors (TLRs) and inflammasomes play a pivotal role in inducing an anti-viral state in the host through the innate immune response via distinct signaling pathways, resulting in the production of pro-inflammatory cytokines, chemokines, and interferons24. Moreover, TLRs activate antigen-presenting cells (APCs) to work in concert with adaptive immunity for pathogen elimination and induction of long-term immunity25. These receptors recognize conserved viral molecular patterns (PAMPs) and subsequently provoke an immune response in the host if an antigen or vaccine interacts appropriately with the target immune cells. Toll-like receptors are type 1 transmembrane horseshoe-shaped proteins that recognize PAMPs and signal via MyD88-dependent or TRIF-dependent pathways. TLRs 1, 2, 4, 5, and 6 are located on the cell surface, whereas TLRs 3, 7, 8, and 9 are situated on endosomal membranes and detect viral nucleic acids26. TLR9, in particular, recognizes double-stranded DNA containing unmethylated CpG motifs in endosomes and has been found to induce innate immunity in herpesvirus infections. Therefore, molecular docking analysis was carried out to demonstrate the binding affinity between the vaccine construct and immune receptors TLR-9 (TLR-9, PDB:3wpb) and SLA-1 (SLA-1*0401, PDB:3qq4). The docking of the vaccine construct with TLR-9 and SLA-I receptors was performed using High Ambiguity Driven protein-protein DOCKing (HADDOCK) 2.4 (https://www.bonvinlab.org/software/haddock2.4/)27. The best cluster was selected from the docked clusters based on the lowest HADDOCK score. Thereafter, the HADDOCK Refinement Interface was employed to refine the chosen cluster. The best structure after refinement from each docked complex was chosen, and their binding affinity was determined using the PROtein binDIng enerGY prediction (PRODIGY) web server28. PRODIGY employs a simple but powerful predictive model based on intermolecular contacts and properties derived from non-interface surfaces to predict the binding affinity of protein-protein complexes through their 3D structure29. Finally, the interacting residues between the vaccine construct and the immune receptors were mapped using PDBsum (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/Generate.html)30.

Molecular dynamics simulation

Molecular dynamics simulation was performed using the inbuilt commands of GROMACS (GROningen MAchine for Chemical Simulations) to evaluate the structural properties and interactions in vaccine construct-TLR-9 and vaccine construct-SLA-1 complexes. The Assisted Model Building and Energy Refinement 03 (AMBER03) force field was used to create the topology file essential for energy minimization and equilibration. The spc/e water model was employed as the solvent to simulate the complex with periodic boundary conditions. The net charge of the complex was determined, and charged ions were added to neutralize the system. Energy minimization was carried out before simulation to ensure the adequate geometry of the system and the absence of steric clashes using the steepest descent algorithm approach. The trajectories generated from the simulation (10 ns) were analyzed for the stability of the energy-minimized complex in terms of root mean square deviation (RMSD) and root mean square fluctuation (RMSF) of backbone atoms and side chains, respectively.

Codon optimization and in silico cloning

To generate an appropriate plasmid construct harboring the vaccine construct sequence, codon optimization was embarked upon. Codon optimization is a method to achieve translation efficiency of foreign genes in the host if the use of codons differs between the two organisms. The Java Codon Adaptation Tool (JCat) (https://www.jcat.de/) was used for codon optimization to generate a DNA sequence of the vaccine construct that can be efficiently expressed in the E. coli K-12 strain31. The Codon Adaptation Index (CAI), a technique for analyzing codon usage bias32, and the guanine and cytosine (GC) contents were evaluated. To ensure translation efficiency, the RNAfold web server was employed to predict the secondary structure of the optimized codon sequences. Sticky ends of the restriction sites for BamHI and EcoRI restriction enzymes were grafted to allow restriction and cloning at the start (N-terminal) and end (C-terminal) of the modified vaccine sequence, respectively. The modified nucleotide sequence of the vaccine was cloned into the E. coli pET-28a (+) vector using the SnapGene tool (https://www.snapgene.com/) to ensure its in vitro expression.

Immune Response Simulations

To predict and characterize the immune response profile of the vaccine construct in the host, computational immune simulations were employed by the C-IMMSIM server (http://kraken.iac.rm.cnr.it/C-IMMSIM/)33. The C-ImmSim uses the Celada-Seiden model to describe both humoral and cell-mediated immune profiles of a mammalian immune system against a designed vaccine. C-IMMSIM is an agent-based model that employs position-specific scoring matrices (PSSM) derived from machine learning techniques for predicting immune interactions. The summary of the pipeline for vaccine development is described (Figure 1).

× Figure 1 . The schematic workflow of the designed study . The strategy employed in the study entails several steps involved in identifying the target proteins. CTL, HTL, IFN-γ and B cell epitopes predictions from the target proteins. Thereafter, vaccine construction and its quality checks. Molecular docking with immune cell receptors TLR-9 and SLA-1. This was proceeded by molecular dynamic simulation to assess vaccine construct stability. Finally, codon adaptation and immune simulation to assess vaccine induction of immune response. Figure 1 . The schematic workflow of the designed study . The strategy employed in the study entails several steps involved in identifying the target proteins. CTL, HTL, IFN-γ and B cell epitopes predictions from the target proteins. Thereafter, vaccine construction and its quality checks. Molecular docking with immune cell receptors TLR-9 and SLA-1. This was proceeded by molecular dynamic simulation to assess vaccine construct stability. Finally, codon adaptation and immune simulation to assess vaccine induction of immune response.

Table 1.

CTL epitopes predicted using NetMHCpan 1.2 revealing promiscuity

Protein SLA I CD2v Peptides (positions) Alleles Antigenicity Toxicity TNKSFLNYYW (151-160) SLA-2*0402, SLA-1*0501, SLA-2*0401 0.5346 Non-toxic C-type lectin KYTGLIDKNY (6-15) SLA-2*0102, SLA-1*0501 1.1132 Non-toxic FSNNIDEKNY (81-90) SLA-2*1002 0.4413 Non-toxic KKYNYESGYW (116-125) SLA-1*0501, SLA-3*0701, SLA-3*0301, SLA-3*0303, SLA-3*0401 1.0042 Non-toxic KKVNYTGLLF (145-154) SLA-1*0401, SLA-3*0601, SLA-3*0301, SLA-3*0303, SLA-3*0304, SLA-1*0801, SLA-3*0701 1.2916 Non-toxic p72 YGKPDPEPTL (40-49) SLA-1*1101, SLA-3*0101, SLA-2*0701, SLA-2*0501, SLA-3*0501, SLA-3*0502, SLA-3*0503, SLA-6*0101, SLA-6*0102, SLA-6*0103, SLA-6*0104, SLA-6*0105 1.4480 Non-toxic KPYVPVGFEY (65-74) SLA-1*0701, SLA-1*0702, SLA-2*0102, SLA-2*0101, SLA-1*0501, SLA-2*0302, SLA-2*0401, SLA-2*0402, SLA-3*0601, SLA-1*0201, SLA-1*0202, SLA-1*0801, SLA-1*0601, SLA-1*1301, SLA-3*0701, SLA-2*1001, SLA-2*1002, SLA-2*0601,