Prediction of an Epitope-based Computational Vaccine Strategy for Gaining Concurrent Immunization Against the Venom Proteins of Australian Box Jellyfish.

BACKGROUND: Australian Box Jellyfish (C. fleckeri) has the most rapid acting venom known to in the arena of toxicological research and is capable enough of killing a person in less than 5 minutes inflicting painful, debilitating and potentially life-threatening stings in humans. It has been understood that C. fleckeri venom proteins CfTX-1, 2 and HSP70-1 contain cardiotoxic, neurotoxic and highly dermatonecrotic components that can cause itchy bumpy rash and cardiac arrest. SUBJECTS AND METHODS: As there is no effective drug available, novel approaches regarding epitope prediction for vaccine development were performed in this study. Peptide fragments as nonamers of these antigenic venom proteins were analyzed by using computational tools that would elicit humoral and cell mediated immunity, were focused for attempting vaccine design. By ranking the peptides according to their proteasomal cleavage sites, TAP scores and IC50<250 nM, the predictions were scrutinized. Furthermore, the epitope sequences were examined by in silico docking simulation with different specific HLA receptors.
RESULTS: Interestingly, to our knowledge, this is the maiden hypothetical immunization that predicts the promiscuous epitopes with potential contributions to the tailored design of improved safe and effective vaccines against antigenic venom proteins of C. fleckeri which would be effective especially for the Australian population.
CONCLUSION: Although the computational approaches executed here are based on concrete confidence which demands more validation and in vivo experiments to validate such in silico approach.

INTRODUCTION

Chironex fleckeri (Australian box jellyfish) commonly known as sea wasp or marine stingers that inhabit and roam coastal water from northern Australia, New Guinea north to the Philippines and Vietnam.[1] It is considered the most venomous marine animal and most dangerous cubozoan jellyfish to humans and its occurrence in the tropical coastal waters of Australia is primarily a problem, particularly in summer. At least 70 deaths have been reported due to C. fleckeri envenoming occurred in Australia. Apart from these, numerous deaths from related species also have been reported in the South India, Malaysia, Japan, Philippines, Maldives islands, Papua New Guinea, Java, and Gulf of Thailand, but most encounters appear to result only in mild envenomation. Australian box jellyfish produces exceptionally potent and rapid-acting venom and its stings to humans cause severe localized and systemic effects that are potentially life-threatening to humans. The venom of C. fleckeri contains a variety of bioactive and complex mixture of venom proteins which are stored and discharged by small, highly pressurized stinging capsules called nematocysts. Recent studies reveal that the venom proteins have cytolytic, cytotoxic, cardiotoxic (attacks the heart), and highly dermatonecrotic (destroys skin) components which are rapidly absorbed into the circulation after injection.[2] Surprisingly, the onset of symptoms has been reported extremely rapid in several case studies.[3] For its notorious sting, a massive dose of venom can cause cardiac dysfunction, cardiac arrest (arrhythmias), resulting in loss of consciousness and death within 5 min of being stung in severe cases. Children are closely vulnerable to this life-threatening venom proteins because of their smaller body mass.[4] Moreover, the box jellyfish venom has multiple effects attacking the nervous system, skin, and heart simultaneously. Symptoms of major C. fleckeri stings include rapid acute cutaneous inflammation, dermonecrosis, excruciating pain, permanent scarring, hypotension, shock, dyspnoea, hypertension, impaired consciousness, pulmonary edema, and cardiac dysfunction.[5] To date only two C. fleckeri venom proteins, CfTX-1 and -2 have been significantly identified which have potent hemolytic activity with cutaneous inflammation and may function as a pore-forming toxin. Hence, it can disrupt normal transmembrane ion concentration gradients in susceptible cells.[6] Moreover, heat shock proteins of C. fleckeri (HSP70-1) may play a crucial role in antigen and cross presentation.[7] In addition, heat shock protein-derived immunodominant epitopes are exploitable as therapeutic peptides in allergies have been recently reported.[8] Domestic vinegar and antivenom have been widely used as first aid treatment to neutralize against this rapid acting venom, but some people still die despite its administration. Although life-saving antivenoms have an immunoglobulin pool of unknown antigen specificity and known redundancy that necessitates the delivery of large volumes of heterologous immunoglobulin to the envenomed victim. Consequently, it increases the risk of serum sickness and anaphylactoid which has a strong adverse effect.[9] With prospects, recent developments in computational tools have paved the way to predict and to prosecute further assay of B cell and t cell epitopes from antigenic proteins in specialized tasks. This has led to peptide-based vaccines design planning that is more specific, secured, optimized, and easy to predict the peptide binding to human leuckocyte antigen (HLA) alleles using structural and modelling methodologies. Surprisingly, it has gained momentum in recent years in alleviating to some crucial immunological infections.

SUBJECTS AND METHODS

Protein sequence retrieval

The toxin protein sequences of C. fleckeri were retrieved from protein database of National Center for Biotechnology Information (NCBI, http://www.ncbi.nlm.nih.gov/protein/) by GenBank accession no. ABS30940.1 for Toxin-1, ABS30941.1 for Toxin-2 and ACS12895.1 for HSP70-1. The sequences were analyzed with a view to recognizing the immunologically relevant regions, which was done by studying antigenecity, solvent accessible regions, and Major Histocompatibility Complex (MHC) class I and II binding sites.

Antigenic peptide prediction

In this method, the potential hydrophilic regions of the proteins were found out in order to identify the antigenic determinants. Antibody epitope prediction of Immune Epitope Database (IEDB) analysis resource server (http://tools.immuneepitope.org/tools/bcell/tutorial.jsp) was used which predicted the sites that produce antigenic response against an antigenic protein. Antigenic epitopes are determined using several prediction methods, for example, Kolaskar and Tongaonkar antigenicity,[10] and Hopp and Woods hydrophobicity[11] methods.

Secondary structure prediction

We used ExPASy's secondary structure prediction server (http://web.expasy.org/protparam/)[12] to get an idea about the secondary structure of the venom proteins. Several parameters given by ProtParam tool were studied, for example, molecular weight, theoretical pI, amino acid composition, atomic composition, extinction coefficient, estimated half-life, instability index, aliphatic index, and grand average of hydropathicity.[131415]

Solvent accessible regions and hydrophilicity estimation

Hydrophilicity of the proteins was estimated by using ProtScale (http://web.expasy.org/protscale/) server of ExPASy. By using Wolfenden et al., and Eisenberg et al.,[161718] hydrophobicity scales, we found different hydrophobicity plots, which were then analyzed to predict the solvent accessible regions and to estimate the hydrophobic sites on the protein.

Beta turn prediction

Beta turn in the selected protein structures for epitope prediction was determined by using Levitt scale[19] by ExPASy's ProtScale server.

Prediction of major histocompatibility complex binding peptide

To predict the MHC binding peptides for the venom proteins of C. fleckeri, we used two options provided by IEDB analysis resource. For MHC class I peptide prediction, we used Proteasomal cleavage/transporter of antigenic peptides transporter/MHC class I combined prediction server (http://tools.immuneepitope.org/processing/) and for MHC class II peptide, we used MHC II-binding prediction (http://tools.immuneepitope.org/mhcii). In both the prediction servers, we used the artificial neural network prediction method to predict the potential nonamers that may efficiently bind to the binding groove of the MHC molecules.

Allergenicity assessment

In order to assay the degrees of allergenicity we operated AllerHunter (http://tiger.dbs.nus.edu.sg/AllerHunter/index.html). A combinational prediction by using both Support Vector Machine (SVM) and pair-wise sequence similarity makes AllerHunter a very useful program for cross-reactive allergen prediction. Cross-reactivity is a phenomenon which is based on similarity among proteins and allergens, whereas allergenecity means the ability of an allergen to induce immunoglobulin E antibody production. AllerHunter predicts allergens as well as nonallergens with high specificity. Moreover, it does not compromise its efficiency while classifying proteins with similar sequence to known allergens.

Docking simulation

We performed in silico docking simulation to find out whether or not the predicted peptides will bind to the MHC molecules when these will be applied for further in vivo experiments. For docking simulation study, we used Autodock Vina[20] developed by The Scripps research Institute. To run the docking simulations, three MHC I molecules (PDB ID: 1A1O, 1DUZ and 1JHT) and three MHC II molecules (PDB ID: 1AQD, 1DLH and 1H15) were selected. Protein Data Bank (PDB) files for the predicted epitopes were prepared by using HHPred to use them as ligands. Autodock tools were used for preparation of receptor and ligand molecules for docking simulations at the binding groove of the MHC molecules.

RESULTS

Secondary structure analysis

The secondary structural features of C. fleckeri toxin-1, toxin-2, and HSP70-1 are summarized in Table 1. All of these proteins were found to be rich in leucine, isoleucine, and glycine residues. Toxin-1 and toxin-2 were found to be alkaline in nature, while HSP70-1 was found to be acidic.

Table 1

Secondary structural analysis of Chironex fleckeri dermatonecrotic proteins by ProtParam tool

Solvent accessible regions

For C. fleckeri toxin-1, the minimal value in Eisenberg hydrophobicity scale was -1.154 and maximal value was 0.993. Eisenberg scale puts negative values for hydropathic residues in protein. According to Eisenberg's scale, the most hydrophilic regions of toxin-1 were 70-83, 88-128, 134-141, 164-170, 181-204, 223-231, 235-260, 305-311, 325-341, 345-352, 399-404, and 431-444 [Figure 1]. For toxin-2, the predicted hydrophilic regions were 67-80, 85-125, 132-138, 181-201, 220-260, 282-289, 303-310, 319-337, 361-367, 395-402 and 444-454 [Figure 1]. Again, for HSP70-1 the hydrophilic regions were 26-31, 60-67, 112-125, 141-150, 162-184, 194-227, 281-296, 332-340, 346-353, 366-381, 389-412, 436-443, 456-469, 473-489, and 601-609 [Figure 1]. From the analysis, it was found that toxin-1, toxin-2, and HSP70-1 of C. fleckeri were hydrophilic in nature with high flexibility and low complexity segments. In a vaccine design program, it is the first step to make sure that the predicted antigenic fragments can bind to MHC molecules.

Figure 1

Graphical representation of solvent accessible region (Left - eisenberg) and beta turn region (Right - levitt) analysis of Chironex fleckeri venoms

Antigenic peptide evaluation

By analyzing numerical and graphical data, it was found that according to Hopp and Woods scale the regions 19-24, 31-42, 48-69, 84-91, 105-111, 129-135, 138-152, 158-164, 171-181, 208-220, 258-285, 294-303, 310-324, 357-364, 371-380, 390-398, and 415-421 contained the potential hydrophilic regions for toxin-1 [Figure 2]. For toxin-2, the predicted hydrophilic regions were 28-66, 81-87, 102-108, 126-132, 139-149, 168-180, 205-217, 255-282, 291-301, 308-320, 348-360, 381-395, and 403-416 [Figure 2]. Again, for HSP70-1, the predicted regions were 24-38, 45-57, 68-88, 94-111, 126-139, 152-162, 185-194, 211-217, 227-237, 241-289, 298-310, 313-332, 354-369, 381-389, 413-424, 443-455, 490-501, and 551-603 [Figure 2]. According to Kolaskar and Tangaonkar antigenicity scale, at 1.0 as the threshold level, the most likely antigenic determinants for toxin-1 were at 89-109, 134-146, 181-211, 224-250, 285-294, 342-360, 399-407, 409-417, and 422-429 [Figure 2]. For toxin-2, the antigenicity plot was predicted as 85-106, 129-139, 178-208, 221-247, 283-291, 344-359, 374-386, and 395-413 were the potential antigenic sites [Figure 2]. Antigenic sites for HSP70-1 were predicted at the regions 136-150, 192-199, 205-213, 332-340, 366-382, 388-403, 423-430, 435-444, and 485-493 [Figure 2]. It was found that according to Levitt scale, amino acid residues 23-30, 38-45, 60-65, 83-91, 126-135, 172-184, 247-256, 260-269, 278-284, 291-300, 315-324, 352-368, 373-383, and 416-432 fall inside the beta turn region for toxin-1 by considering 1.000 as the threshold level and the upper section of the graph was analyzed as beta turn region [Figure 1]. For toxin-2 and HSP70-1 the beta turn regions are shown in [Figure 1]. These predicted fragments are assumed to bind with MHC molecules of immune system, which is the first step toward vaccine design.

Figure 2

Graphical representation of antigenic peptide evaluation by Hopp and Woods (left) and Kolaskar and Tangaonkar (right) of Chironex fleckeri venoms

Allergenecity evaluation

The query sequences did not meet the criteria set by the Food and Agriculture Organization (FAO)/World Health Organization (WHO) evaluation scheme for cross-reactive allergen prediction. So, the query sequences were classified as a nonallergen by the FAO/WHO evaluation scheme. Both toxin-1 and toxin-2 were predicted as a potential nonallergen with a prediction score of 0.0 (Sensitivity, SE = 91.6%; Specificity, SP = 89.3%). HSP70-1 was also a nonallergen with a score of 0.04 (Sensitivity, SE = 96.0%; Specificity, SP = 45.9%).

Prediction of major histocompatibility complex-binding peptide

A total of 58 alleles were analyzed for MHC class I peptide prediction by using artificial neural network method.[2122] Again 26 MHC class II alleles were analyzed for prediction of MHC II-binding peptides from the selected venom proteins. We predicted three nonamers which showed sufficiently high results in the prediction methods that were used in this study. The predicted nonamers were “ILLDLYQLV” for toxin-1, “FIAMVVQRI” for toxin-2, and “FQHGKVEII” for HSP70-1. Theses peptides showed interaction with multiple MHC class I and MHC class II alleles. Interaction among different alleles with these peptides is summarized in Table 2 [Supplementary materials 1–3] and Table 3 [Supplementary materials 4–6]. In case of MHC class II prediction, artificial neural network method was used.[23] For selection of all the MHC-binding peptides, MHC IC50 score was below 250 nM. The predicted epitope for toxin-1 interacted with three MHC I alleles (belong to two supertypes A, C) and 12 MHC II alleles (belong to three supertypes and six complexes). The epitope FIAMVVQRI interacted with five MHC I alleles (belong to two supertypes) and 15 MHC II alleles (belong to two supertypes and six complexes). Epitope for HSP70-1 interacted with five MHC I alleles (belong to three supertypes A, B, and C) and five MHC II alleles.

Table 2

Prediction of MHC class I peptides of Chironex fleckeri venom by using proteasome/transporter of antigenic peptides/MHC-combined method

Supplementary material 1

Toxin 1 MHC 1 allele interaction

Supplementary material 3

HSP70-1 MHC 1 allele interaction

Table 3

Prediction of MHC class II peptides of Chironex fleckeri venom by using artificial neural network method

Supplementary material 4

Toxin 1 MHC II allele interaction

Supplementary material 6

HSP70-1 MHC II allele interaction

Docking simulation results

The area that were selected on the receptor molecules for docking with the epitopes are summarized in Table 4. One angstrom, spacing was used to select the binding site. The center box area was positioned carefully to make the docking of ligands at the binding groove of the receptors. The three predicted peptides showed significant binding affinity to the MHC receptors [Table 5], except for a few, compared to the Epstein-Barr virus epitope-binding energy with 1H15 (5.9 Kcal/moL). Strong binding affinity gives a clear idea that peptide vaccine designed by using these epitopes may efficiently work in vivo to elicit humoral and cell-mediated immunity [Figures 3 and 4].

Table 4

Binding site coordinates for protein-ligand docking between MHC molecules and peptides prepared by autodock tools

Table 5

Docking simulation results prepared by autodock vina

Figure 3

Visualization of best docking results for predicted peptides with MHC class I receptors by using Autodock Tools. (a-c) represents docking images with 1A1O, (d-f) represents docking images with 1DUZ, (g-i) represents docking images with 1JHT

Figure 4

Visualization of best docking results for predicted peptides with MHC class II receptors by using Autodock tools. (a-c) represents docking images with 1AQD (d-f) represents docking images with 1DLH, (g-i) represents docking images with 1H15

DISCUSSION

Prediction of epitope and mapping theses on the protein surface is a vital step for epitope-based vaccine design. A number of ways were attempted in earlier studies but here tried to predict the epitopes more accurately by starting from the very basic step- like finding the hydrophilic regions of the proteins and ending by docking of epitopes to their respective receptors. To find out the hydrophobicity scores, which actually give us the index for hydrophilic regions, we used Eisenberg hydrophobicity and Wolfenden hydrophobicity scales. These scales put a positive score for the nonpolar residues and negative score for polar residues of a given protein. From these data, we measured the hydrophilic regions of the selected proteins, which are supposed to be antigenic in nature and more exposed to the surface of the protein. The regions with maximum hydrophilic scores are analyzed as antigenic sites, because these regions are unstructured and solvent accessible which make it easier for antibodies to recognize the native proteins.[24] Wolfenden hydrophobicity plot puts an overall impression on hydrophilicity of a protein molecule. Numerical and graphical analysis showed that the proteins were hydrophilic in nature. According to Wolfenden hydrophobicity plot, glycine, leucine, and isoleucine are the most hydrophobic residues.

The local hydrophilic region of the protein which is typically more exposed to the surface is detected as the antigenic site and the corresponding amino acids of these sites are detected as the antigenic peptides. Hopp and Woods hydrophilicity scale and Kolaskar and Tangaonkar antigenecity scale were used to predict the antigenic peptides of the selected toxin proteins. Hopp and Woods hydrophobicity scale is actually a hydrophilicity scale in which window size 7 gives the ideal values for a protein hydrophilicity nature. Hopp and Woods scale assigns nonpolar residues with a negative value. Kolaskar and Tangaonkar antigenecity scale is the simplest method for determining antigenic determinants. This method is based on the occurrence of amino acid residues in experimentally determined epitopes. In several experimental studies, it was found that the antigenic parts of a protein belong to the beta turn regions.[2526] To validate the predicted hydrophilic and antigenic parts of the protein we analyzed Levitt and Deleage and Roux beta turn scale.

The AllerHunter score value is the probability that a particular sequence is a cross-reactive allergen. However, the threshold for prediction of cross-reactive allergen is adjusted such that a sequence is predicted as a cross-reactive allergen if its probability is >= 0.06. The probability threshold was determined during the fine-tuning of prediction model. AllerHunter has optimum prediction result at that particular threshold. The FAO and WHO evaluation scheme is a guideline by the FAO and WHO for sequence-based alelrgenicity prediction. This guideline clearly states that a sequence can be a potentially allergenic if it either has an approximated identity of at least 6 contiguous amino acids or >35% sequence identity over a window of 80 amino acid chains when compared to known allergens.[27] So, if a vaccine was developed by using the venom peptides, it will not create allergic reactions.

MHC class II molecules are highly polymorphic in nature, and this polymorphism exclusively corresponds with a few differences along the peptide-binding groove in antigenic fragments.[28] The binding between antigenic peptides (epitopes) and the MHC molecule is a crucial step in the cellular immune response. For the prediction of MHC binding molecules in both cases (MHC I and II) artificial neural network method was used. For t cell class I epitope prediction, the neural network method was designed by combining sparse encoding, blosum encoding, and input derived from hidden markov models.[21] In this study, for MHC class II peptide prediction, we used artificial neural network-based method NN-align which was evaluated by 14 human MHC class II alleles.[23]

In this study, we tried to minimize the predicted promiscuous epitopes and pinpoint the efficient epitope sequences that have the greatest chance for eliciting cell-mediated immunity in human body against box jellyfish venom. As it is a concern that the prediction-based epitope design might not work in reality, we docked the predicted MHC peptides with HLA molecules to find out whether or not the vaccine designed by using the predicted epitopes will elicit sufficient immunological response in vivo. Lower energy scores represent better binding between receptor and ligand.[29]

More interestingly, sequence similarity search between toxin-1 and toxin-2 by using CLUSTALW multiple sequence alignment web server [Figure 5] showed that the epitope that was predicted for toxin-2 (FIAMVVQRI) would also elicit immune response against toxin-1. Again, we analyzed the B-cell epitope prediction method by Kolaskar and Tangaonkar method [Table 6] and found out that the selected epitopes fall inside the sequences predicted for B-cell immunity. So, we summarize that designing of a trivalent vaccine by using these three epitopes may elicit both humoral and cell-mediated immunity against box jellyfish venom.

Figure 5

Multiple sequence alignment between toxin-1 and toxin-2 by using CLUSTALW web server [* (star mark) designates similarity between two proteins]

Table 6

B cell epitope regions of Chironex fleckeri venom proteins predicted by Kolaskar and Tangaonkar antigenecity scale

CONCLUSION

All of these computational approaches demonstrate the importance of C. fleckeri venom proteins as valuable immunodiagnostic tool for initial research methodologies with a view to future disease diagnosis and drug design against this fatal venom. The findings of this study yet need to be validated in future by experimental procedures; however, the given information and approaches in this study will be more blissful for researchers to investigate novel human therapeutics-like design of subunit and synthetic peptide vaccine from this world's most venomous marine creature C. fleckri. This superficial concept can be implemented to design synthetic and subunit peptide vaccine against these lethal venom proteins that may save thousand lives especially in Australia where it poses a major problem.

Supplementary material 2

Toxin 2 MHC 1 allele interaction

Supplementary material 5

Toxin 2 MHC II allele interaction