Peptide vaccine against chikungunya virus: immuno-informatics combined with molecular docking approach

Primary sequence of the CHIKV structural polyprotein was obtained from the Uniprot (ID: Q1H8W5). Experimentally known crystal structure of CHIKV structural polyprotein was retrieved from Protein Data Bank (http://​www.​rcsb.​org/​pdb) (PDBId: 3N44). Protein sequence was analyzed for its physical and chemicals properties including half-life, GRAVY (Grand average of hydrophathicity), stability index, molecular weight and amino acid atomic composition through an online server Protparam [20]. Secondary structure of our CHIKV structural polyprotein was analyzed through PSIPRED [21].

Systematic strategy was used to design CHIKV structural polyprotein B and T cell epitopes. TMHMM was used for transmembrane topology prediction [22]. Freely available servers IEDB [23] and BCPREDS were recruited for B cell epitopes prediction [24]. Intracellular epitopes were excluded and only those epitopes were selected that were exposed on extracellular surface. Vaxijen 2.0 server was used for antigenicity analysis of selected epitopes [25]. Identification of B-cell epitopes was based on; antigenicity, flexibility, accessibility of surface, predictions of linear epitope and hydrophilicity [26]. Flexibility, accessibility of surface, isolation of linear epitope and hydrophilicity analysis were performed through Karplus and Schulz flexibility prediction tool, Emini surface accessibility prediction tool, Kolaskar and Tongaonkar antigenicity scale, Parker hydrophilicity prediction [26] and Bepipred linear epitope prediction algorithms. Prediction of beta turns in polyprotein was carried out by using Chou and Fasman beta turn prediction algorithm [27]. DiscoTope 2.0 server (http://​www.​cbs.​dtu.​dk/​services/​DiscoTope/​) was used to model B cell epitopes from 3D structure of CHIKV structural polyprotein. All predicted epitopes were given in plain format while the 3D structure of structural polyprotein was given in PDB format.

By using two different online available tools, CTL epitopes of targeted protein were predicted. Major Histocompatibility Complex (MHC) class I and II were predicted by ProPred-1 (http://​crdd.​osdd.​net/​raghava/​propred1/​) and Propred tool (http://​crdd.​osdd.​net/​raghava/​propred/​), respectively. The predicted results from both these tools are quite significant because they use huge number of alleles of HLAs (human leukocyte antigens) during calculations. For class I allele’s prediction, the threshold value of Proteasome and immune-proteasome were set to 5%.

Sequences of CHIKV structural polyprotein, belonging to 23 different countries, were taken from Genbank and Uniprot databases. Multiple sequence alignment (MSA) was performed to find out the conservation of selected epitopes by using CLC work bench (https://​www.​qiagenbioinforma​tics.​com/​products/​clc-main-workbench/​). The aligned files (.aln) were further used to construct phylogenetic tree through MEGA7 software (https://​www.​megasoftware.​net/​). Epitopes were selected based on their conserved regions. Moreover, Immune epitope database and analysis resource (IEDB) was used to cross-check the conservation of the selected epitopes.

3-D structure of all the predicted peptides were modeled through PEPFOLD server at RPBS MOBYL portal [28, 29], while 3D structure of human HLA-B7 allele crystallized at a resolution of 1.7 Å was taken from Protein databank (PDB ID: 3VCL), was used for molecular docking purpose. Peptide models (antigenic determinants) were docked against their respective HLA-B7 allele through Molecular Operating Environment (MOE) tool to analyze their inhibitory potential. Protocol for molecular docking using MOE has already been reported in various studies involving the antiviral discovery against dengue non-structural proteases [30‐32]. Docking protocol used in those studies involves removal of already bound peptide, protonation, energy minimization followed by removal of water molecules. Triangular matcher algorithm was applied as default peptide placement methods based on the receptor shape which quickly generates 1000 best poses of docked peptide without energy optimization. Energy estimation of the simulated poses was re-scored by applying London dG scoring function [33]. Functional form of London dG scoring function estimates ΔG of binding taking into account ligand flexibility, H-bonds, desolvation of each atom by applying equation given below:where, c is average energy loss/gain due to receptor–ligand movement, E_flex measures entropic loss because of the conformational flexibility of the ligand, c_HB estimates H-bond maximum energy whereas H-bond f_HB accounts for geometric imperfections. In case of metals c_M and f_M measures metal ligation energy. While, ΔD calculates desolvation energy of each atom by integrating London dispersion forces over solvent space through the formula given below:

For each peptide, top 10 ranked poses of London dG were further minimized by Force field refinement algorithm. Protein peptide interactions were than analyzed through LigX tool of MOE. Pymol and UCSF Chimera tools were used to generate figures of docked complexes [34].

Allergenicity analysis of predicted epitopes was done by Aller Hunter server (https://​omictools.​com/​allerhunter-tool). This server compares query sequences of peptides against the database of already reported allergens to give significant results.

The primary structure of the target sequence (Uniprot ID:Q1H8W5) was analyzed via Protparam. Physiochemical parameters of the CHIKV structural polyprotein computed through Protparam shows that protein is comprised of 1248 amino acid (aa) and have molecular mass of 138.31 kD. Theoretical iso-electric point (PI) of our subject protein was 8.90 which show its positive nature. Briefly, out of 1248 residues, 138aa were found as positively charged and 106 were calculated as negatively charged residues while Methionine (Met) was selected as an N-terminal residue in our subject structure. Protparam computed instability index, aliphatic index and GRAVY of target protein were 40.75, 74.93 and − 0.305 respectively.

Secondary as well 3D structure analysis of structural polyprotein through PSIPRED and UCSF Chimera respectively showed 50% beta sheet, 5% helixes and 45% loops (Additional file 1: Figure S1A). Crystal structure of structural polyprotein shown in Additional file 1: Figure S1B, C is a heteromer with chain A (blue) chain B (green) and chain F (red). Moreover, DiANNA 1.1 tool computed 24 disulfide (S–S) bond positions in target protein and assigned them score given in Additional file 1: Table S1. In the entire sequence, 49 cystine residues were calculated that are making 24 disulfide bonds at following sequence positions (106–416, 113–283, 128–287, 164–344, 269–741, 278–591, 286–887, 318–1068, 347–790, 353–714, 430–871, 450–1115, 478–550, 528–1080, 545–789, 716–1189, 721–781, 742–923, 783–903, 791–1242, 858–877, 872–1137, 905–1110, 1179–1185).

Vaxijen 2.0 server confirmed the antigenicity value of target protein which was 0.5106. Prediction of transmembrane topology of the target protein was accomplished by TMHMM, showing five transmembrane helices that have range from 690–712, 733–755, 765–787, 794–816 and 1223–1245 sequence positions. Residue ranging from 1–689, 756–764 and 817–1222 sequence positions were exposed on the surface whereas residues from 713–732, 788–793 and 1246–1248 were buried inside the protein structure.

Potential B-cell epitopes have distinct characteristics that guide B cells to identify and trigger robust immune response against different viral infections. Primary sequence of target protein was scanned through BCPRED and different tools of the IEDB were used to predict B cell epitopes. From all the BCPRED predicted epitopes, only eight epitopes were observed to be antigenic in nature through Vaxigen computed score (Table 1). BCPRED has predicted B cell epitopes with 75% accuracy. Fourteen residues long epitopes are considered to be sufficient to induce protective immune response. Epitopes PPFGAGRPGQFGDI, IPTGAGKPGDSGRP, NYPASHTTLGVQDI, PFHHDPPVIGREKF, TSAPCTITGTMGHF, TPYELTPGATVPFL, PEGYYNWHHGAVQY and WLKERGASLQHTAP were predicted with 1, 0.999, 0.973, 0.97, 0.92, 0.858, 0.745 and 0.711 score respectively. In addition to this it is necessary to find out the surface accessibility of potential B-cell epitopes. Kolaskar and Tongaonkar antigenicity measuring tools analyzed the structural polyprotein for prediction of B-cell epitopes by computing their physicochemical properties of its amino acid and their profusion in known B cell epitopes. Kolaskar and Tongaonkar analysis of highly antigenic peptides is given in Fig. 1a. The threshold value of calculation was set at 1.041 while window size was kept 7. It predicted the antigenic propensity values of protein, which are 1.041 (average), 0.860 (minimum) and 1.316 (maximum). Hydrophilic regions of proteins are mostly exposed on the surface area, which potentially play an immense role in evoking immune response. BCPRED score and computed antigenicity results of Vaxigen clearly demonstrates that except the first peptide highlighted in italic in Table 1, all of our predicted peptides are part of extracellular region of transmembrane protein and have the power to maximize an immune response within the host during CHIKV infection. In addition to the above mentioned analysis, identification of B cell epitopes was done by Disco Top 2.0 server by analyzing 3-D structure of CHIKV structural polyprotein. Threshold range was kept at − 2.0 and a total of seven epitopes were computed as distinct surface exposed regions, shown in Table 2. In Kolaskar and Tongaonkar analysis the threshold value of calculation was set at 1.041 and window size was kept 7. It predicted the antigenic propensity values of protein in range of 0.860–1.316 shown in Fig. 1a. Hydrophilic regions of proteins were mostly exposed on the surface area, which potentially play an immense role in evoking immune response. Therefore, it is necessary to find out the surface accessibility of potential B-cell epitopes. For surface accessibility analysis Parker hydrophilicity and Emini surface accessibility prediction tools were used. Emini surface accessibility analyzing tool’s results are given in Additional file 1: Table S2, while graphical representations of the outputs of both tools shown in Fig. 1b, c, respectively. These peptides were part of extracellular region of transmembrane protein and have the power to maximize an immune response. Higher antigenicity score has suggested that these peptides can play a crucial role in initiation of immune response.

To predict the beta turns in structural polyprotein, Chou and Fasman Beta turn analyzing algorithm measure surface exposed and hydrophilicity that play a critical role to initiate the immune response. Threshold of the tool was set at 0.988, it calculated the values which are 0.530 (minimum), 0.988 (average) and 1.443 (maximum). Graphical representation of Chou and Fasman’s output is shown in Fig. 1d. Results indicated that the regions from 60 to 100 amino acids and from 980 to 1020 amino acids are more inclined to induce B-turns in peptide structure. Experimental data has reported that those regions of epitope which interact with alleles or antibodies are mostly elastic in nature. To, detect the flexible regions of the target protein, Karplus and Schulz flexibility analyzing tool depicted that the region from amino acids from 60 to 100 sequence positions are highly flexible (Fig. 1e). Site of each predicted epitopes in its 3D structure is shown in Fig. 2.

T cell epitopes were predicted through the application of two different online servers. MHC class I alleles were predicted through ProPred while MHC class I alleles were predicted through Propred simply. Results of both MHC class I alleles and MHC class II alleles are given in Tables 3, 4. FASTA file of structural polyprotein sequences of the CHIKV was given as an input with 4% threshold level while keeping the proteasome filters and immune proteasome at on mode. Distinctive peptides against all MHC class I alleles were identified out of which we only retained peptides with antigenic properties. Antigenicity of ‘‘TAECKDKNL’’ peptide was higher amongst all peptides with the antigenicity score “1.6001” as shown italic in the Table 3. Similarly, for MHC class II alleles, sequence was also given in FASTA format to ProPred. Antigenicity score of ‘‘VRYKCNCGG’’ peptide predicted through Vaxijen 2.0 was 2.2127 which is greater (shown in italic) as compared to other MHC class II alleles binding peptides (Table 4).

Sequence of CHIKV structural polyprotein isolates from 23 different countries was subjected to multiple-sequence alignment to analyze the conservation of selected epitopes through CLC workbench. It was observed that all the selected epitopes are highly conserved in all sequences used for analysis (Additional file 1: Figure S2). A phylogenetic tree was constructed to show the evolutionary relation of CHIKV belonging to 23 different countries as shown in Fig. 3. Evolutionary analysis revealed that south east Asian CHIKV strains including Pakistan and India stains are less divergent and more closely related to Congo and African CHIKV strains.

Three dimensional structures of 4 MHC classes-I binding peptides were predicted through PEPFOLD. It produced five models of each peptide; one best model was selected for each peptide. Initially models were refined through energy minimization in MOE and peptide library comprised of four peptides was created to dock with solved structure of HLA-B7 allele.

Before docking, energy minimization was performed for a library of four potential peptides and then docked against HLA-B7. Crystal structure of HLA-B7 (PDBID: 3VCL) was already available with co-crystallized peptide. So focused docking was performed by using same active pocket to dock our peptide library. Ten confirmations for each epitope were generated and top ranked conformations based on their dock scores were selected (Table 5). Later on, interaction analysis using ligX tool was done which revealed that the peptide “VMHKKEVVL” with highest dock score (− 20.2086 kcal/mol) is interacting with key catalytic residues of chain A. Human HLA-B7 is a hetero-dimer structure, from the interaction analysis it was revealed that Lys-243, Glu-232 while Tyr-26 and Tyr63 from B chains were making stable hydrogen bonds with the aforementioned peptide. In addition to this four water molecules were observed to be involved in water mediated interaction between the peptide and Thr-233 (A chain), and Thr-10 (B chain) shown in Fig. 4a, b. Peptide “GTLKIQVSL” was docked (dock score − 19.6688 kcal/mol) within the catalytic pocket of receptor protein through four hydrogen bond interaction with Ser-4, Arg-6, Thr-31 and Asp102 (chain A) and one water mediated interaction with Gln-32 (Fig. 5a, b). Peptide ranked 3th in Table 5 has − 17.3569 kcal/mol of dock score with three stable hydrogen bonds between peptide and His-3, Thr-31, Asp-102 and Glu-180 shown in Fig. 6a, b. Similarly, peptide ranked at 4th with dock score (− 14.5594 kcal/mol) got 2 stable hydrogen bond interactions with Arg-6 and Glu-232. While one water mediated interaction between peptide and receptor (Thr-31) can also be observed in Fig. 7a, b.

Allergenicity assessment of query protein was performed via Aller Hunter to predict whether it is allergenic or not based on the FAO/WHO standards. Results showed that our query sequence was non-allergenic with a score of 0.0 (SP = 89.3%, SE = 91.6%).