Introduction

Porcine epidemic diarrhea (PED) is an acute intestinal infection caused by the porcine epidemic diarrhea virus (PEDV)1. It affects pigs of all ages, manifesting primarily through symptoms such as watery diarrhea, vomiting, anorexia, dehydration, and weight loss, particularly in piglets. The mortality rate can reach up to 100%, resulting in significant economic losses within the global pig-intensive farming industry2. Extensive research has been conducted on PEDV, focusing on its pathogenic mechanisms, diagnostic approaches, and detection techniques. However, the virus continues to evolve and mutate in nature, giving rise to new strains and frequently triggering fresh outbreaks2,3. Vaccination stands as a crucial measure in preventing and managing infectious diseases, providing protection for the body and enhancing specific immunity4. Presently, various vaccines are tailored to combat PEDV, including whole virus-inactivated vaccines and improved live vaccines5,6. While the inactivated vaccines boast safety and a straightforward production process, they solely induce humoral immunity. Conversely, live attenuated vaccines provoke robust cellular and humoral immune responses but pose the risk of virulence reversal7. Hence, it is imperative to vigorously advocate for the development and enhancement of novel vaccines to combat PEDV effectively.

PEDV is classified within the genus Alpha Coronavirus in the Coronaviridae family and falls under the order Nidovirales. It is characterized as a single-stranded RNA virus with a plus-stranded genome, measuring approximately 28 kb in size5. The virus contains seven open reading frameworks (ORFs) encoding four structural proteins: spiking (S), capsular (M), envelope (E) and nucleocapsid (N). Additionally, it encodes three non-structural proteins: open reading frame 1a/b and ORF36. The S protein, present as a trimeric glycoprotein on the viral surface, assumes a pivotal role in viral functionality4. It consists of two domains, S1 and S2, which facilitate virus entry, attachment to host cell receptors, and the induction of neutralizing antibody production. The S2 domain is responsible for membrane fusion and viral replication8. The S protein also determines the tissue tropism and host range of the virus, making it an attractive target for attenuated coronavirus and subunit vaccines9. However, employing a full-length S protein vaccine may elicit harmful immune responses and toxic side effects10. To mitigate this concern, the screening and design of dominant epitopes recognized by T and B cells can be effectively achieved through immunoinformatics11. Moreover, enhancing the antigenicity of these epitopes can be achieved through efficient delivery platforms. Ferritin, a self-assembled, symmetrical octahedral protein, is ideal for multivalently displaying antigens in highly ordered and repetitive arrays12. This study aims to control the sustained spread of PEDV by constructing a self-assembled nanoparticle vaccine targeting the dominant epitope of the S protein.

Materials and methods

Based on theoretical guidance, the study mainly described the prediction of PEDV S protein dominant epitopes (B cells, CD4+ and CD8+), vaccine design, 3D structural modeling, molecular docking, protein expression and immune simulation summary, and the results were shown in Fig. 1.

Fig. 1
figure 1

The flow chart was used to describe the screening of the PEDV S protein dominant epitope and the construction of a nanoparticle candidate vaccine (Biorender drawn).

Full size image

T cell and B cell epitope prediction

The PEDV-S protein (GenBankID: AAK38656) was retrieved in FASTA file format from the NCBI database (https://www.ncbi.nlm.nih.gov/protein) to serve as a template for cellular epitope prediction. For CD4+ epitope prediction, the NetMHCIIpan-4.0 server (https://services.healthtech.dtu.dk/service.php?NetMHCIIpan-4.0) was utilized, with adjustments made to the epitope length to ensure a strong binding epitope of 1513. Subsequently, the induction of IFN-γ and IL-4 was evaluated using the IFNepitope tool (https://webs.iiitd.edu.in/raghava/ifnepitope/index.php) and IL-4pred (http://crdd.osdd.net/raghava/il4pred/)14,15. CD8+ epitopes were predicted using the NetMHCpanEL4.1 server (http://tools.iedb.org/mhci/) and the NetMHCpan4.1 server (https://services.healthtech.dtu.dk/service.php?NetMHCpan-4.1)16,17. Epitope length was adjusted to 14, with the allele threshold set to approach 0, and binding sequences ranked in the top 5% based on their scores. B cell epitopes were predicted using the ABCpred server (https://webs.iiitd.edu.in/raghava/abcpred/ABC_submission.html) and the PEPTIDES server (http://imed.med.ucm.es/Tools/antigenic.html)18,19. For ABCpred, the length was adjusted to 16, and the threshold was set to 0.8. PEPTIDES performed predictive analysis on the entire amino acid sequence. All predicted epitopes were further evaluated for immunogenicity and toxicity using the IEDBMHC-I server (http://tools.iedb.org/immunogenicity/) and the ToxinPred server (https://webs.iiitd.edu.in/raghava/toxinpred/design.php)20,21. The final result consisted of non-toxic epitopes with immunogenicity scores greater than 0.25.

Screening and conservativeness analysis of dominant epitopes

The B cell epitopes, CD4+ epitopes, and CD8+ epitopes generated by the server underwent comparative screening analysis. Overlapping segments among these epitopes were identified and selected as dominant epitopes. The conservation of these dominant epitopes was evaluated using the IEDB server (http://tools.iedb.org/conservancy/) and 169 PDCoV-S protein amino acid sequences retrieved from the NCBI database22 (refer to Supplementary Table S1 for detailed information of GenBankID). Analysis parameters were set to ‘linear’ for the analysis type and ‘100%’ for the sequence identification threshold, with identical template sequences removed. The PEDV-S protein crystal structure (PDBID:7W6M) was selected from the PDB database (https://www.rcsb.org), and spatial mapping of the dominant epitopes was carried out using Pymol.

Vaccine design and evaluation

CD8+, CD4+, and B-cell dominant epitopes were tandemly linked using AAY, GPGPG and KK flexlinker, respectively, as previously described23. The C-terminus of the construct was ligated to ferritin (GenBank: WP_000949190) via (GGGGS)3 linker to present a multivalent display of the dominant epitope. The candidate vaccine was analyzed for antigenicity using the ANTIGENpro server (http://scratch.proteomics.ics.uci.edu/)24, while solubility during overexpression was assessed using the SOLpro server (http://scratch.proteomics.ics.uci.edu/index.html)24. Allergenicity was determined using the AllerTop Server (http://www.ddg-pharmfac.net/AllerTOP/)25, hydrophobicity was examined using the ProtScale server (https://web.expasy.org/cgi-bin/protscale/protscale.pl#opennewwindow)26, and transmembrane structural domains was performed using the TMHMM-2.0 server (https://services.healthtech.dtu.dk/services/TMHMM-2.0/)27. Additionally, physicochemical properties were assessed using the ExPASy server (https://web.expasy.org/protparam/)26, while the secondary structure was analyzed using the PSIPRED 4.0 server (http://bioinf.cs.ucl.ac.uk/psipred)28.

Vaccine tertiary structure prediction, refinement and validation

The candidate vaccine is modeled using the trRosetta server (https://yanglab.nankai.edu.cn/trRosetta/)29. Subsequently, the initial model was refined using the GalaxyRefine server (http://galaxy.seoklab.org/cgi-bin/submit.cgi?)30. Additionally, the quality of the model was assessed and compared using the PROCHECK server (https://saves.mbi.ucla.edu/) and authenticated using the ProSA-web server (https://prosa.services.came.sbg.ac.at/prosa.php)31,32.

Molecular docking, immunosimulation and computerized cloning

In this study, TLR-3 (PDBID:2A0Z) and the candidate vaccine were subjected to molecular docking using the pyDockWEB server (https://life.bsc.es/pid/pydockweb)33. The resulting complex was visualized using LigPlot+ to analyze the two-dimensional interworking interface. Immunization simulations were conducted using the C-ImmSim server (https://150.146.2.1/C-IMMSIM/index.php)34 with step-time settings of 1, 25 and 50, while other parameters were kept at their default values to assess the candidate vaccine's efficacy in inducing both humoral and cellular immunity. Codon optimization and reverse translation of the candidate vaccine amino acid sequences were accomplished using the JCat server (http://www.jcat.de/)35. The ExpOptimizer server (https://www.novopro.cn/tools/codon-optimization.html) (NOVOPRO, China) validated CAI values against GC content. Visualization of restriction endonuclease sites was facilitated using SnapGene, followed by their insertion into the adapted expression vector system.

Results

Screening of S protein dominant epitopes of PEDV

In order to enhance the reliability of the dominant epitope screening strategy, this study employed multiple immunological tools to analyze the same dominant epitope. The results revealed that all dominant epitopes were non-toxic and displayed high immunogenicity, as summarized in Table 1. A total of 5 B-cell dominant epitopes were obtained through ABCpred and PEPTIDES screening. Additionally, 3 CD8+ cell dominant epitopes were obtained using NetMHCpanEL4.1 and NetMHCpan4.1, while 3 CD4+ cell dominant epitopes were obtained through NetMHCIIpan-4.0 followed by IFNepitope and IL-4pred analysis. The B-cell epitopes, denoted as B1, B2, B3, B4 and B5, exhibited minimum identities of 93. 75%, 75.00%, 75.00%, 81.25% and 80.00%, respectively. Similarly, the CD4+ epitopes (CD4-1, CD4-2, and CD4-3) and CD8+ epitopes (CD8-1, CD8-2, and CD8-3) displayed minimum identities of 53.33%, 93.33%, 78.57%, 64.29% and 92.86%, respectively. These results indicate a relatively high level of identity for the dominant epitopes. Spatial mapping of the dominant epitopes in PyMOL revealed their presence on the surface of the S protein, as depicted in Fig. 2. These short surface peptides contribute to antigenic chimerism and stimulate the immune response. However, the exact spatial location of the CD4-3 and CD8-3 epitopes remains unclear due to the incomplete resolution of the crystal structure of the S protein (detailed data are shown in Supplementary Table S2).

Table 1 PEDV-S protein B-cell, CD4, CD8 dominant epitope screening results.
Full size table
Fig. 2
figure 2

Spatial mapping of dominant epitopes of PDDV-S protein. (A) PEDV structure; (B) S protein tertiary structure; (C) Red labeled B cell epitope, green labeled CD4+ epitope, purple labeled CD8+ epitope.

Full size image

Construction of PEDV-S protein vaccine

Based on the results of the screening analysis, we used KK, AAY, GPGPG, and (GGGGS)3 flexible linkers to connect B cell, CD8 and CD4 epitopes in a specific order to construct a candidate vaccine (detailed data are shown in Supplementary Table S3). To enhance the immunogenicity of the candidate vaccine, a double lysine KK linker was utilized for the B cell dominant epitope to maintain its independent immunogenic activity. Additionally, AAY and GPGPG linkers were used for the T cell dominant epitope to improve the recognition of candidate vaccine subunits. The (GGGGS)3 linker was utilized to connect epitopes to ferritin, ensuring stability and proper folding rate. Our analysis revealed that using a single GGGGS flexible linker resulted in low protein solubility during the construction of the candidate vaccine. However, when we introduced (GGGGS)3, protein solubility significantly increased (amino acid sequence in Fig. 3C).

Fig. 3
figure 3

Results of candidate vaccine evaluation. (A) Candidate vaccine hydrophobicity prediction; (B) Candidate vaccine transmembrane domain prediction; (C) Candidate vaccine secondary structure analysis.

Full size image

Evaluation of the candidate vaccine of PEDV-S protein

According to ANTIGENpro, the candidate vaccine exhibited a predicted immunogenicity of 0.606918. Additionally, SOLpro predicted an overexpression solubility of 0.584405 for the candidate. AllerTOPV2.0 classified the vaccine candidate as a non-sensitizing protein. ExPASy analysis revealed that the candidate vaccine has a full length of 374 amino acids, a molecular weight of 40.89 kDa, an isoelectric point of 5.48 and a molecular formula of C1855H2850N476O553S7. The antigenic stabilization index of the candidate was calculated to be 32.98, with a lipid index of 89.39, classifying the candidate vaccine as a stable protein. The results presented in Fig. 3A reveal that most of the amino acids in the hydrophobicity of the candidate vaccine exhibit negative values, indicative of hydrophilic characteristics. Figure 3B illustrates that none of the candidate vaccines possess transmembrane domains according to the TMHMM-2.0 prediction; instead, all are extracellular regions supporting soluble expression. The PSIPRED secondary structure prediction indicates that the candidate vaccine mainly consists of helical structures. The secondary structure prediction results are shown in Fig. 3C.

Modeling, refinement and evaluation of tertiary structures of vaccine candidates

TrRosetta utilizes homology modeling to generate 2D views (contact map, distance map, and orientation map) of the candidate vaccine (Fig. 4A–E) and the tertiary structure model (Fig. 4F). However, the initial model exhibited instability. To address this issue, the crude model underwent refinement using Galaxy-Refine, and a cartoon comparison diagram was generated using Pymol (Fig. 5B). PROCHECK was employed to generate Ramachandran diagrams for evaluating the model's quality before and after refinement. The Ramachandran score results for the initial model (Fig. 5A) revealed the most favored regions (A, B, L) at 89.8% and additional allowed regions (a, b, l, p) at 9.5%. After refinement, the Ramachandran scoring results (Fig. 5C) indicated an increase in most favored regions (A, B, L) to 93.3%, with additional allowed regions (a, b, l, p) at 6.0%. These scores indicate an improvement in the model's quality after refinement (Detailed data are shown in Supplementary Table S4). ProSAweb was utilized to enhance the quality of the refined model further. The model structures were scored and validated. The structural accuracy analysis reveals a z value of − 5.85 (Fig. 5D), indicating a more satisfactory native protein conformation. Additionally, predominantly negative energy scores for the local model quality (Fig. 5E) signify an ideal model suitable for further analysis.

Fig. 4
figure 4

Three-level structure modeling of the candidate vaccine. (A–E) are the Omega, Phi, Distance, Theta and Contact parameters of the two-dimensional structure; ferritin is red; B cell epitope is purple and green; CD4+ epitope is orange; CD8+ epitope is sea blue; TM-score: 0.488.

Full size image
Fig. 5
figure 5

Modeling and optimization of the three-level structure of candidate vaccine. (A) Ramachandran diagram of the crude model; (B) Comparison of the structure of the crude model and the refined model, the crude model and refined model are green and blue, respectively; (C) Ramachandran diagram of the refined model; (D) ProSA SEB refined model evaluation; (E) Local model quality assessment.

Full size image

Candidate vaccine docking with TLR3

TLR-3, a member of the TLR family known for its role in inducing the transcription of type I interferons, proinflammatory cytokines, and chemokines crucial for initiating the host's antiviral response, was investigated. Approximately 102 composite docking complexes were generated using pyDockWEB docking. The complex with the highest composite ranking was chosen from the docking results, as depicted in Fig. 6A. The composite scores for this complex were as follows: Electrostatics: − 13.194; Desolvation: − 32.294; Vdw: 40.519; Total: − 41.436. Utilizing Pymol, we zoomed in on the interacting amino acid residues, with corresponding results presented in Fig. 6B,C. LigPlot+ was employed to generate a two-dimensional interaction map based on the docking results of the candidate vaccine and TLR3, and the outcomes are displayed in Fig. 6D. This map allows clear observation of hydrogen-bonded amino acid residues and hydrophobic interacting amino acid residues.

Fig. 6
figure 6

Docking results of candidate vaccine with TLR3 molecules. (A) Docking of a candidate vaccine to TLR3 (candidate vaccine structure in red, TLR3 in green); (B) local zoom in to show amino acid residues; (C) amino acid residues for specific interactions; (D) 2D interactions schematic diagram.

Full size image

Immunosimulation and computerized cloning

Both specific and non-specific immune responses play a vital role in the process of vaccine immunity. Immune simulations on the candidate vaccine were conducted using C-ImmSim (Figs. 7, 8, 9). These simulations revealed that the candidate vaccine activated NK cells, giant cells and phages. Additionally, dendritic cells stimulated the differentiation and proliferation of B lymphocytes, leading to a significant increase in B cell populations and levels of IgG1 + IgG2, IgM and IgG + IgM antibodies. Similar trends were observed for CD4+ T cells (Th1 and Th2 cells) and CD8+ T cells (CTL), as well as immunoglobulins and immune complexes. The candidate vaccine also induced the production of IFN-γ by immune cells, forming three peaks. These findings indicate that after three rounds of immune stimulation, the simulation demonstrated a robust induction, thereby generating a potent immune response. Overall, the simulated immune response indicates that the vaccine can induce both cellular and humoral immunity, making it an excellent candidate.

Fig. 7
figure 7

Display of changes in the number of B cells and Th. (A) B lymphocytes; (B) Plasma B lymphocytes; (C) B lymphocyte population; (D) CD4 helper T lymphocytes; (E) CD4 helper T lymphocytes; (F) CD4T regulatory lymphocytes.

Full size image
Fig. 8
figure 8

Display of changes in the number of CTL, NK, MA, DC, and EP cells. (A) CD8 T cytotoxic lymphocytes; (B) Per CD8 T cytotoxic lymphocytes; (C) Natural killer cells; (D) Total macrophage count; (E) dendritic cells; (F) epithelial cells.

Full size image
Fig. 9
figure 9

Trends in changes in immunoglobulins, immune complexes, and concentrations of cytokines and interleukins. (A) Antigens and immunoglobulins; (B) Antigens and immunoglobulins (excluding ferritin); (C) Concentrations of cytokines and interleukins.

Full size image

Computer simulation clones of candidate vaccines

JCat reverse translation of the candidate vaccine amino acid sequence resulted in a candidate vaccine nucleotide sequence of 1122 bp. To enhance the expression of the candidate vaccine, codon optimization was performed in E. coli K-12. The optimized sequence showed a CAI improvement value of 1.0, which is considered optimal. With a GC content of 48.395%, falling within the recommended range of 30–70% for protein expression, the sequence proved suitable for expression in E. coli. Validation by the ExpOptimizer server revealed a value of 0.74, while the GC content remained consistent at 48.40% (Fig. 10A,B). SnapGene was used to insert the optimized sequence into the prokaryotic expression vector pET-32a(+) using BamHI and XhoI sites. The results of the constructed prokaryotic expression plasmid are depicted in Fig. 10C. The plasmids were subjected to PCR simulation, identified by double digestion and analyzed using 1% agarose gel electrophoresis. The results are shown in Fig. 10D,E. The candidate vaccine was successfully inserted, and the bands appeared in the correct position after double enzyme digestion (nucleic acid sequences and original images of simulated gel electrophoresis are shown in supplement).

Fig. 10
figure 10

Construction of candidate vaccine expression system. (A,B) ExpOptimizer server analyzed results of vaccine candidates; (C) Candidate vaccine (red) inserted into pET-32a(+); (D) PCR simulation analysis of candidate vaccines; (E) PET-32a-Vaccine plasmid double digested.

Full size image

Discussion

Coronaviruses are pathogens that originate from bats and have a wide range of hosts and tissue tropism2. The ongoing coronavirus disease 2019 (COVID-19) caused by SARS-CoV-2 poses a significant threat to the global economy and public health36. PEDV, another member of the coronavirus family, is notorious for its recurrent outbreaks in swine herds, causing widespread concern in the hog industry4. Prior to 2010, PEDV only had sporadic incidents and was limited to certain regions8. However, in October 2010, an epidemic with acute diarrhea symptoms broke out in southern provinces of China and quickly spread across the country37. The mutant strains of PEDV caused high mortality rates among piglets, reaching levels of 80–100%2. By 2013, the mutated virus strain had disseminated globally after being initially reported in the United States, exacerbating the detrimental impact of PEDV on the global pig industry and food safety38. The emergence of highly pathogenic PEDV strains underscores the limitations of the original vaccine strategy in providing cross-protection and managing virus escape39. Therefore, there is an urgent need to develop and improve vaccines to control and prevent PEDV. Ferritin possesses the capability to self-assemble and attach target antigens onto its surface via adsorption and covalent binding mechanisms40. This results in the formation of multivalent display nanoparticles, which play a crucial role in activating antigen-presenting cells (APCs)41. Additionally, it enhances antigen uptake by APCs, assists in antigen processing, fosters dendritic cell (DC) maturation, and augments B-cell activation42. Nanoparticle-based candidate vaccines are constructed through the integration of reverse vaccinology for comprehensive gene sequence screening and the incorporation of multiple B-lymphocyte or T-lymphocyte epitopes from viral proteins43. This approach addresses challenges associated with epitope shortness, poor stability, and low in vivo immunogenicity, thereby effectively eliciting both humoral and cellular immune responses.

The study was conducted to achieve immune balance, using multiple immunoinformatics tools to screen and construct the nanoparticle candidate vaccine for S-protein dominant epitopes. We identified and analyzed 5 B cell epitopes, 3 CD4+ and 3 CD8+ dominant epitopes. The B2 dominant epitope overlaps with the currently reported neutralizing epitope 499–638 aa44, suggesting strong potential for the candidate vaccine to neutralize the virus effectively. These dominant epitopes were then fused to ferritin in a specific order. Our analysis revealed that the candidate vaccine exhibited an immunogenicity score of 0.606918 and an overexpression solubility score of 0.584405, both exceeding the server's threshold for determination. The majority of the secondary structure of the vaccine consists of random coils, indicating a solid structural foundation and a loose spatial arrangement that facilitates epitope formation45. In the tertiary structure modeling, the refined model was increase in most favored regions (A, B, L) to 93.3%, with additional allowed regions (a, b, l, p) at 6.0%. Meanwhile, it was similar to the natural protein. The reliable structure of the energy score indicates that the structure of the model is stable, and it is helpful for the following research. Molecular docking investigations clearly demonstrate that the candidate vaccine forms a stable complex with TLR3. These findings imply that the candidate vaccine has the potential to activate intracellular signaling pathways, including NF-κB and cytokine release, thereby eliciting innate immune system activation and fostering durable adaptive immunity46. Although the C-IMMSIM immune simulation failed to simulate the self-assembly of ferritin into nanoparticles, it still illustrated significant antibody titer production against the antigen by the candidate vaccine. Furthermore, secondary and tertiary immune responses led to a gradual increase in active Th cell populations, total and memory TH cell counts, and cytotoxic T cells. Notably, after the three stages of immunity, IFN-γ exhibited a robust response among cytokines and interleukins, playing a crucial role in inhibiting viral replication, promoting T cell and natural killer cell production and aiding in viral clearance. Simulation data suggests that the candidate vaccine has the potential to induce both cellular and humoral immune responses. Additionally, these results are consistent with previous research on COVID-19 ferritin nanoparticle vaccines47, which have shown strong immune responses in animals. Therefore, it is anticipated that the actual immune response to the candidate vaccine will exceed expectations. Finally, the candidate vaccine was reverse-translated and molecularly cloned. The optimized sequence exhibited a CAI improvement value of 1.0 and a GC content of 48.395%, indicating high-level expression of the protein. However, to determine the practical application value of the candidate vaccine, further experiments involving protein expression and purification, as well as in vitro and in vivo studies, will be necessary to assess its ability to stimulate the immune response effectively.

Conclusions

This study successfully designed a recombinant multi-epitope nanoparticle vaccine targeting the PEDV S protein. The design process involved considering T/B cell dominant epitopes, spatial mapping, allele selection, homology modeling, molecular interactions and computer cloning. The innovative and logically structured design adopted in this study offers valuable insights for vaccine developers. The predictive evaluation and immune simulation enable the rapid design of vaccine molecules based on immunoinformatics. The constructed nanoparticle candidate vaccine provides a theoretical basis and offers data support for establishing a new vaccine development platform for PEDV.