Design and computational evaluation of a prophylactic and therapeutic multi-epitope vaccine candidate against cervical cancer

Abstract

Cervical cancer is the fourth most common cancer among women worldwide, caused by the human papillomavirus (HPV). HPV16 /18 are strongly associated with the development of cervical cancer. HPV vaccines are not widely available in economically underdeveloped areas. They also have limited efficacy against pre-existing HPV infections and cervical lesions. Therefore, the study aims to computationally design a vaccine that induces both prophylactic and therapeutic immunity against cervical cancer. Using computational approaches, we designed a multi-epitope vaccine incorporating 62 cytotoxic T lymphocyte (CTL) epitopes, 7 helper T lymphocyte (HTL) epitopes, and 3 linear B-cell epitopes from conserved regions of HPV16/18 E6, E7, and L2 proteins. To increase immunogenicity, the adjuvant RS-09 (APPHALS) was added to the N-terminal of the vaccine. The chosen CTL and HTL epitopes have the potential to achieve 100% worldwide population coverage. The vaccine candidate has appropriate physicochemical properties. The vaccine’s secondary and tertiary structures were predicted, followed by refinement and validation of the models. The vaccine has a high affinity for Toll-like receptor 4 (TLR4), as indicated by its molecular docking score of -1164.4 kcal/mol and RMSD fluctuation range of 17 to 22.5 angstroms (Å) at the end of the MD simulation period. Immune response simulation showed that the vaccine candidate could induce strong humoral and cellular immune responses. The vaccine sequence’s GC content was calculated to be 48.99%, and subsequently, in silico cloning of the vaccine was performed in the pET28a (+) vector. The results reveal that this vaccine is highly immunogenic; however, experimental testing is required to confirm its efficacy and safety.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12985-026-03072-x.

Introduction

Cervical cancer is the fourth most common cancer among women worldwide, with around 660,000 new cases recorded in 2022. In that year, low- and middle-income countries accounted for more than 94% of the 350,000 cervical cancer deaths [1]. The etiological agent of cervical cancer is the human papillomavirus (HPV), consisting of approximately 200 types, classified as low-risk HPVs (LR-HPVs) and high-risk HPVs (HR-HPVs) based on their carcinogenic potential, principally determined by differences in the E6 and E7 viral proteins [2]. Almost 95% of cervical cancer is associated with chronic HR-HPV infection [3]. The most common HR-HPV types in cervical cancer are HPV16 and HPV18, accounting for 55.2% and 14.2% of cases, respectively [4].

HPVs belonging to the Papillomaviridae family measure 50–60 nanometers (nm) in diameter and are non-enveloped [5]. This virus possesses an approximately 8,000 base pairs (bp) circular double-stranded DNA genome encased in a protein capsid comprised of L1 and L2 proteins [6]. The genome is structured into three functional segments: the early region (E1-E8), the late region (L1, L2), and the long control region (LCR) [7]. The E region encodes seven viral nonstructural proteins: E1, E2, E1ˆE4, E5, E6, E7, and E8ˆE2. These proteins are involved in HPV replication, transcription, translation, and transformation. The L region generates two viral capsid proteins: L1 and L2. The LCR region, or upstream regulatory region (URR), lacks protein-coding genes [8].

The main function of the E6 protein in the HR-HPV types is to induce the ubiquitination of the p53 proteins. The interaction of E6 with p53 and its ubiquitination are dependent on E6-AP (E6-associated protein) [9–11]. The p53 protein is a transcription factor that regulates the expression of genes associated with cell cycle arrest and apoptosis, including the cyclin-dependent kinase inhibitor p21 [12]. The HR-HPV E7 oncoprotein inactivates the tumor suppressor protein retinoblastoma (Rb) [13]. The Rb frequently inhibits G1-to-S phase cell cycle progression. Rb inactivation by E7 disrupts cell cycle regulation and promotes uncontrolled cell proliferation [14–18]. The synergistic effects of E6 and E7 oncoproteins can induce unregulated cellular proliferation and precancerous lesions. Infected cells exhibit loss of differentiation and suffer genetic and epigenetic alterations, which promote cancer development [14, 17, 19, 20].

It has recently been proposed that L2-based vaccination, rather than virus-like particles (VLPs) based on L1 immunization, could elicit broader cross-type protective immunity at a lower cost [21]. L2 induces immunological responses in the same manner as L1. However, in contrast to L1, the L2 capsid protein has a highly conserved sequence across different types of HPV and can neutralize antibodies produced during multiple types of HPV infections, offering broader protection [22]. The main challenge is the relatively low immunogenicity of L2 [21, 23]; to overcome this weakness, it may be formulated in diverse ways, including multivalent L2 epitopes (peptide vaccine), fusions with L1 and other immunogenic proteins, and multi-epitope vaccines [24, 25].

HPV is transmitted through sexual contact and impacts sexually active individuals with genital infections [26]; however, these infections seldom lead to malignancy [27]. Data indicate that 60% of HPV infections are spontaneously resolved within one year, 90% within two years, and just a minuscule percentage advance to precancerous lesions and cancer [28].

HPV infection occurs via microscopic lesions in the epithelium, hence exposing the basal cells to the virus [29]. The virus initially infects epithelial cells in the basal layer [30]. The α6β4 integrin complex serves as the receptor for HPV entrance into epithelial cells [31]. Following viral penetration into basal layer keratinocytes, the HPV genome replicates as an episome, with approximately 50–100 copies per cell, and is subsequently stably maintained in these undifferentiated cells [32]. Upon division of infected cells, viral DNA is disseminated among all infected cells. Viral particles are subsequently expelled from the superficial layers of the stratified epithelium [33].

Reverse vaccinology is a new vaccine development technique that uses pathogen genomic data to identify potential vaccine candidates [34]. This approach differs from conventional methods in that it uses computational analysis of the pathogen’s genome rather than solely cultivating and examining the pathogen in a laboratory, significantly reducing the time and costs associated with vaccine development when compared to traditional techniques [34]. It allows for the in silico discovery of more immunogenic antigens while simultaneously analyzing their resemblance to host proteins, hence increasing reliability against undesired effects [35]. This involves using computational models to predict which proteins are likely to be surface-exposed, secreted, or otherwise available to the immune system [34]. Chosen proteins can be used to predict epitopes, increasing the likelihood of detection by populations other than the target populations and allowing for more focused preventative interventions [36]. Reverse vaccinology has been extensively employed to develop multi-epitope vaccines targeting many pathogens, including human immunodeficiency virus (HIV) [37–41], coxsackievirus (CV) [42–44], SARS-CoV-2 [45–49], influenza A virus [50–52], Mycobacterium tuberculosis [53, 54], Enterobacter cloacae complex [55, 56], Helicobacter pylori [57–60], Leishmania infantum [61, 62], Toxoplasma gondii [63, 64], and others.

The US Food and Drug Administration (FDA) has approved three L1-based HPV vaccines: Cervarix^®, Gardasil^®, and Gardasil^®9 [65]. These vaccines have demonstrated exceptional clinical efficacy. Furthermore, three new prophylactic vaccines—Cecolin^®, Walrinvax^®, and Cervavac^®—have just been licensed in India and China, and they are significantly less expensive than the originals [66]. The development of HPV prophylactic vaccines is a major step forward in cervical cancer prevention, but these vaccines are still expensive and not widely available in economically underdeveloped areas. Moreover, prophylactic vaccines exhibit restricted efficacy against pre-existing HPV infections and cervical lesions [67]. Currently, there is no definitive treatment for HPV infection, and the conventional approach for cervical cancer involves a combination of surgical intervention and chemoradiotherapy [68]. Nonetheless, individuals with advanced, metastatic cervical cancer continue to encounter a bleak clinical prognosis while receiving routine treatment [69]. Therefore, the development of a vaccine capable of inducing both prophylactic and therapeutic immunity is imperative.

This study aims to develop a multi-epitope vaccine that can have both prophylactic and therapeutic effects against cervical cancer. This study identified the best cytotoxic T lymphocyte (CTL), helper T lymphocytes (HTL), and linear B-cell epitopes from the conserved regions of the E6, E7, and L2 proteins of HPV16/18 types. To develop the multi-epitope vaccine construct, the identified epitopes were joined together using linkers and adjuvants. The vaccine’s secondary and tertiary structures, as well as its physicochemical properties, were then predicted. The vaccine’s interaction with Toll-like receptor 4 (TLR4) was examined using molecular docking, and its stability was determined using molecular dynamics (MD) simulation to ensure immune response induction. The vaccine’s ability to produce an immune response was evaluated using immune response simulation. Finally, codon optimization was conducted for expression in the Escherichia coli (E. coli) K12 strain, and the vaccine was in silico cloned into the pET28a (+) vector to confirm expression.

Materials and methods

Protein sequences retrieval and multiple sequence alignment

The sequences for the E6, E7, and L2 proteins of various HPV16/18 types were acquired in FASTA format from the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/). To develop a universal vaccine that successfully targets all HPV16/18 strains, it was essential to find variations or mutations among the sequences of different strains; hence, multiple sequence alignment (MSA) was performed using the Clustal Omega tool (https://www.ebi.ac.uk/jdispatcher/msa/clustalo). Clustal Omega is a new MSA software that utilizes seeded guide trees and HMM profile-profile algorithms to generate alignments among three or more sequences [70].

Epitope prediction and screening

The identified conserved sequences of target proteins across different strains were subjected to epitope prediction. To predict CTL epitopes, we used the MHC Class I Tools Suite (Immunogenicity, Processing, Binding/Elution) of the Immune Epitope Database (IEDB) (https://nextgen-tools.iedb.org/pipeline?tool=tc1) [71]. CTL epitopes with binding affinity for 27 alleles (the default in IEDB) were predicted using the NetMHCpan 4.1 EL (recommended epitope predictor-2023.09) method [72]. HTL epitopes were predicted using IEDB’s MHC Class II Tools Suite (Immunogenicity, Binding/Elution) (https://nextgen-tools.iedb.org/pipeline?tool=tc2) [71]. HTL epitopes that showed binding affinity to the 27 alleles (the default in IEDB) were predicted using the NetMHCII pan 4.1 EL (recommended epitope predictor-2023.09) method [72]. We employed the Bepipred Linear Epitope Prediction 2.0 method [73] to predict linear B-cell epitopes using the IEDB’s antibody epitope prediction tool (https://tools.iedb.org/bcell/) [71]. The selected CTL and HTL epitopes with a median binding percentile value of ≤ 20, along with linear B-cell epitopes to be at least 6 amino acids long, were subsequently assessed for their antigenicity, allergenicity, and toxicity profiles. HTL epitopes were also tested for the potential to induce interferon gamma (IFN-γ) production.

The antigenicity of epitopes was determined with the VaxiJen v2.0 server (https://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html). This is the first server that can predict protective antigens autonomously, without the need for alignment. It was designed to enable antigen classification based simply on protein physicochemical properties, rather than sequence alignment [74–76]. The target organism for this test was “virus”, with an antigenicity threshold set at 0.4. To assess allergenicity, epitopes were assessed using the AllerTOP v2.1 server (https://www.ddg-pharmfac.net/allertop_test/). This server employs an auto cross-covariance (ACC) transformation to convert protein sequences into uniform, equal-length vectors. It has been utilized in quantitative structure-activity relationship (QSAR) analyses of peptides of varying lengths [77]. The CSM-Toxin server (https://biosig.lab.uq.edu.au/csm_toxin/) was employed to assess the toxicity of the epitopes. CSM-Toxin is a powerful in silico classifier for protein toxicity that relies solely on the protein’s basic sequence. The method encodes protein sequence data utilizing a deep learning natural language model to interpret the “biological” language, wherein residues are seen as words and protein sequences as sentences. The CSM-Toxin methodology effectively detects potentially harmful peptides and proteins, achieving MCC values of up to 0.66 in both cross-validation and various non-redundant blind assessments [78]. IFN-γ is a cytokine that stimulates T helper type 1 (Th1) responses [79]. It boosts the antigen presentation ability of antigen-presenting cells (APCs) and promotes the development of CD4⁺ Th1 cells [80, 81]. Therefore, the HTL epitopes were assessed for IFN-γ induction using the IFNepitope server (http://crdd.osdd.net/raghava/ifnepitope/design.php) [82].

Population coverage analysis

Geographic and ethnic differences can cause variations in the distribution and expression of human leukocyte antigen (HLA) alleles around the world. As a result, evaluating epitope population coverage is critical for determining the efficacy of the proposed vaccine across varied worldwide populations [83]. We employed the IEDB population coverage tool (http://tools.iedb.org/population/) to conduct assessments of population coverage [84]. The chosen CTL and HTL epitopes were assessed individually and in combination to ascertain their population coverage throughout 16 geographical areas and the world.d.

Multi-epitope vaccine construction

To avoid duplication, epitopes having sequences embedded in other epitopes were removed at this step. Epitopes that passed all of the predefined criteria were then concatenated using appropriate linkers to form the final vaccine construct. Linear HTL epitopes were connected using GPGPG (Gly-Pro-Gly-Pro-Gly) linkers, while CTL and linear B-cell epitopes were linked using KK (Lys-Lys) linkers. Furthermore, to improve the vaccine candidate’s immunogenicity, the TLR4 agonist (RS-09; Sequence: APPHALS) was introduced as an adjuvant to the N-terminal end of the vaccine construct via an EAAAK (Glu-Ala-Ala-Ala-Lys) linker. The employment of synthetic adjuvants (RS09) is regarded as a safer strategy and an enhancement over conventional vaccination approaches [85]. A 6xHis tag was ultimately attached to the C-terminal end of the vaccine to enhance its purification and identification [86]. The GPGPG linker is a glycine-rich linker that enhances construct solubility while improving activity, accessibility, and flexibility in surrounding domains [87]. The KK linker is the target sequence for cathepsin B, a lysosomal protease crucial for antigen processing. Linking two epitopes with a KK linker mitigates the induction of antibodies against the peptide sequences formed by linear concatenation of individual epitopes [88]. The EAAAK linker is a stiff and helical peptide linker employed in protein engineering, especially for the assembly of fusion proteins. It is engineered to preserve a distance and minimize interference among various protein domains inside a fusion protein. The EAAAK sequence facilitates the development of an alpha-helix, enhancing the linker’s stiffness and its capacity to regulate the spatial configuration of fused protein domains [89, 90].

Assessment of antigenicity, allergenicity, toxicity, solubility, and physicochemical characteristics of the vaccine

The antigenicity of the proposed multi-epitope vaccine was evaluated using the VaxiJen v2.0 server, with a threshold value of 0.4; furthermore, ANTIGENpro (https://scratch.proteomics.ics.uci.edu/) was employed to check the vaccine’s antigenicity. ANTIGENpro is the first whole protein antigenicity predictor developed using reactivity data derived from protein microarray analysis for five pathogens. The predictor is based on protein sequences, does not require alignment, and is independent of pathogens [91]. The AllerTOP v2.1 and CSM-Toxin servers were employed to evaluate the allergenicity and toxicity of the vaccine candidate, respectively. The solubility of the multi-epitope vaccine was assessed via the Protein-Sol (https://protein-sol.manchester.ac.uk/) and SOLpro (https://scratch.proteomics.ics.uci.edu/) servers. According to the Protein-Sol server, the population average for the experimental dataset (PopAvrSol) is 0.45; hence, any scaled solubility value more than 0.45 is considered more soluble than the average soluble E. coli protein [92]. SOLpro server uses a two-stage SVM architecture based on multiple representations of the primary sequence to predict a protein’s tendency for solubility upon overexpression in E. coli [93]. The Expasy ProtParam tool (https://web.expasy.org/protparam/) was used to predict the vaccine candidate’s physicochemical properties [94], which include molecular weight, theoretical pI, total number of negatively and positively charged residues, formula, total number of atoms, half-life, instability index, aliphatic index, and grand average of hydropathicity (GRAVY).

Secondary structure prediction of the vaccine

The percentage of secondary structural elements in the multi-epitope vaccine, comprising alpha helix, extended strand, and random coil, was calculated using the Prabi server (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_gor4.html) [95]. It predicts protein secondary structure using the GOR IV approach. GOR IV is the fourth version of the Garnier-Osguthorpe-Robson (GOR) method, which is a computational approach for predicting protein secondary structure with a mean accuracy of 64.4%. It uses information theory to assess the probability of amino acids adopting various secondary structural conformations and considers all probable pair frequencies within a window of 17 amino acid residues [96].

Tertiary structure prediction, refinement, and validation

The GalaxyTBM server (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=TBM) was used to predict the vaccine’s three-dimensional structure. GalaxyTBM is a new template-based modeling (TBM) approach that initially constructs the more reliable core region using several templates, followed by the identification and re-modeling of the less trustworthy, variable local regions, such as loops or termini, through an ab initio process. This TBM approach is based on “Seok-server”, which was evaluated in CASP9 and found to be one of the best TBM servers [97]. The GalaxyTBM server generated five 3D models of the vaccine, which were subsequently evaluated for quality using the PROCHECK tool from the SAVES v6.1 server (https://saves.mbi.ucla.edu/) and the ProSA-web server (https://prosa.services.came.sbg.ac.at/prosa.php). The best model was then refined using the GalaxyRefine server (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=REFINE). The GalaxyRefine server employs a refining approach that has been successfully validated in CASP10. The approach initially reconstructs side chains, followed by sidechain repacking and then total structural relaxation by molecular dynamics simulation. The CASP10 assessment indicated that this approach had the highest efficacy in enhancing local structural quality [98]. We assessed the quality of the vaccine’s refined structure compared to the initial structure using the PROCHECK tool from the SAVES v6.1 server and the ProSA-web server. PROCHECK is a software tool that evaluates the stereochemical quality of protein structures, with particular emphasis on their conformation and overall quality [99]. ProSA assigns an overall quality score to each protein 3D structure it analyses. If this score falls outside of the range typically observed in natural proteins, the structure probably contains errors [100, 101].

Discontinuous B‑cell epitopes prediction

Using the IEDB’s ElliPro tool (http://tools.iedb.org/ellipro/), we analyzed the vaccine’s refined 3D structure to identify discontinuous B-cell epitopes. ElliPro assigns a score to each predicted epitope, which is presented as a Protrusion Index (PI) value calculated from the average of the epitope residues [102]. For this prediction, we set a minimum score of 0.5 and a maximum distance of 6 angstroms (Å).

Disulfide engineering of the vaccine construct

The addition of new disulfide bonds into protein structures is widely employed to enhance protein stability, alter functional properties, and facilitate the investigation of protein dynamics [103]. We employed the Disulfide by Design 2 v2.13 server (http://cptweb.cpt.wayne.edu/DbD2/) to identify possible residue pairs in the vaccine’s refined 3D structure that could be mutated to cysteine for disulfide bond formation. This server facilitates the systematic prediction of disulfide bonds, hence significantly improving protein stability [103, 104].

Molecular Docking of the vaccine with TLR4

The ClusPro 2.0 server (https://cluspro.org/help.php) was used to perform molecular docking between the vaccine candidate and the TLR4 [105–109]. TLR4’s crystal structure (PDB ID: 4G8A) was retrieved in PDB format from the RCSB Protein Data Bank (RCSB PDB) (https://www.rcsb.org/). Before docking, the TLR4 was prepared in UCSF Chimera software v 1.10.2 [110]. by removing co-crystallized ligands and water molecules, docking was then performed using the default parameters. PDBsum server (http://www.ebi.ac.uk/thornton-srv/databases/pdbsum/) was used for mapping the protein-protein interactions of the best docked vaccine-TLR4 complex [111].

Molecular dynamics simulation

The chosen docked complex underwent MD simulation utilizing the Assisted Model Building with Energy Refinement (AMBER) version 20 software [112]. The antechamber program was employed for the preparation of the complexes [113]. The Leap software was utilized to immerse complexes in the TIP3P solvation box. The ff14SB force field was utilized to elucidate the system’s intermolecular interactions. Na⁺ and Cl⁻ ions were supplied to the protein surface in optimal proportions to neutralize the system’s overall charges. The energy minimization parameters were optimized via the steepest descent integrator over 5000 steps. The system was thereafter incrementally heated to 300 K, with temperature regulation maintained throughout the experiment via the Langevin algorithm. The SHAKE method was utilized to restrict all hydrogen bonding (H-bonds) [114]. The system was equilibrated at constant pressure using the NPT ensemble. The simulation was conducted for 200 nanoseconds (ns), and critical metrics, including root mean square deviation (RMSD) for stability, root mean square fluctuation (RMSF) for flexibility, and radius of gyration (Rg) for the compactness of the vaccine-receptor complex, were graphically represented. The binding free energies for the TLR4-vaccine complex were determined via Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) and Molecular Mechanics-Generalized Born Surface Area (MM-GBSA) approaches, employing the AMBER MMPBSA.py program [115]. The binding energies were computed across the simulation system using 1000 frames chosen at consistent intervals from the simulation trajectories.

Immune response simulation of the vaccine candidate

To assess the immunogenic potential of the multi-epitope vaccine, we run an in silico immunological simulation on the C-ImmSim server (https://kraken.iac.rm.cnr.it/C-IMMSIM/index.php?page=1). This computational system simulates the human immune response to antigen exposure, offering valuable insights into vaccine efficacy. Because C-ImmSim is an agent-based model, each agent (e.g., a cell) and its interactions are individually simulated, resulting in millions of bindings in a typical simulation. As a result, a new, fast Position Specific Scoring Matrix (PSSM)-based method was developed for this server, with minimal sacrifices in terms of performance prediction [116]. To investigate the immune system’s cellular and humoral responses to the proposed vaccine, three doses were administered at four-week intervals. The simulation ran for 1050 steps, with time steps of 1, 84, and 168. The remaining simulation parameters were left at their default settings [117].

Codon optimization and in Silico cloning

The multi-epitope vaccine was reverse translated from protein to nucleotide and codon optimized for expression in the E. coli K12 strain, using the Java Codon Adaptation Tool (JCat) (http://www.jcat.de/) [118]. The tool outputs guanine-cytosine (GC) content and the Codon Adaptation Index (CAI) to measure vaccine expression. The restriction sites for BamHI (5’-GGATCC-3’) and XbaI (5’-TCTAGA-3’) were introduced at the N- and C-terminals of the optimized nucleotide sequence, respectively. The modified sequence of the vaccine construct was then inserted into multiple cloning sites (MCS) of the pET28a (+) expression vector using SnapGene 3.2.1 software.

Results

Protein sequences retrieval and multiple sequence alignment

A total of 93, 49, 37, 18, 98, and 51 appropriate sequences were identified for the E6 HPV16, E6 HPV18, E7 HPV16, E7 HPV18, L2 HPV16, and L2 HPV18 proteins from various strains, respectively. Clustal Omega subsequently performed MSA to identify the conserved regions of the target proteins (Supplementary Files 1–6). The conserved regions were found by ensuring that there were no gaps in the protein sequence and that there was the largest number of identical amino acids.

Epitope prediction and screening

The IEDB predicted a total of 247 CTL epitopes, 26 HTL epitopes, and 13 linear B-cell epitopes derived from target proteins. Of those, 91 CTL epitopes (Supplementary Table 1), 7 HTL epitopes (Supplementary Table 2), and 5 linear B-cell epitopes (Supplementary Table 3) were identified as antigenic, non-allergenic, non-toxic, and IFN-γ inducers (in the case of HTL epitopes).

Population coverage analysis

The analysis of population coverage indicated that the chosen CTL and HTL epitopes individually covered approximately 98.72% and 99.79% of the world population, respectively. However, when these epitopes were used together, the population coverage increased to 100.0%. The population coverage of the chosen CTL epitopes is highest in Europe (99.76%) and lowest in Central America (8.33%). The chosen HTL epitopes have the highest population coverage in North America (100.0%), but the lowest coverage in South Africa (7.65%). When the CTL and HTL epitopes were combined, population coverage above 90% in all geographic regions, with 100% coverage in Europe, North America, Oceania, South America, South Asia, and West Africa. Figure 1 and Supplementary Tables 4–6 represent the population coverage of CTL, HTL, and combined CTL and HTL epitopes throughout 16 geographical areas and the world.

Fig. 1.

Population coverage (%) of the selected CTL and HTL epitopes—analyzed individually and in combination—across 16 geographical areas and the world. The combination of CTL and HTL epitopes show population coverage over 90% in all regions, with complete (100%) coverage in Europe, North America, Oceania, South America, South Asia, and West Africa

Multi-epitope vaccine construction

The final vaccine construct has 888 amino acids organized into 7 HTL epitopes, 62 CTL epitopes, 3 linear B-cell epitopes, 1 adjuvant (RS09), 1 6xHis tag, 1 EAAAK linker, 6 GPGPG linkers, and 65 KK linkers (Supplementary Files 7, Fig. 2A).

Fig. 2.

Types of vaccine structures. (A) Linear structure of the vaccine candidate (amino acid composition). (B) Secondary structure of the vaccine construct, which consists of 19.37% alpha helix (blue), 21.40% extended strands (red), and 59.23% random coil (purple). (C) Refined tertiary structure of the multi-epitope vaccine

Assessment of antigenicity, allergenicity, toxicity, solubility, and physicochemical characteristics of the vaccine

To evaluate the vaccine’s antigenicity, we used two servers: VaxiJen v2.0, which predicted an antigenicity score of 0.7949 (> 0.4), and ANTIGENpro, which reported a score of 0.65. The AllerTOP v2.1 and CSM-Toxin servers predicted the vaccine to be non-allergenic and non-toxic, respectively. The Protein-Sol server assigned a solubility score of 0.682 (> 0.45) to the vaccine design, while the SOLpro server reported a score of 0.934103. The Expasy ProtParam tool was used to estimate the physicochemical properties of the multi-epitope vaccine, as detailed in Table 1.

Table 1.

The predicted physicochemical features of the vaccine candidate

Parameters	Value
Molecular weight	97.606 kDa
Theoretical pI	10.89
Total number of negatively charged residues (Asp + Glu)	41
Total number of positively charged residues (Arg + Lys)	220
Formula	C4379H7277N1295O1185S18
Total number of atoms	14,154
Half-life	4.4 h (mammalian reticulocytes, in vitro) > 20 h (yeast, in vivo) > 10 h (E. coli, in vivo)
Instability index	38.66
Aliphatic index	57.31
GRAVY	−0.883

Secondary structure prediction of the vaccine

The Prabi server predicted that the secondary structure comprises 19.37% (172 aa) alpha helix, 21.40% (190 aa) extended strand, and 59.23% (526 aa) random coil (Fig. 2B).

Tertiary structure prediction, refinement, and validation

The GalaxyTBM server predicted the five models; all of the models were analyzed using the PROCHECK tool from the SAVES v6.1 server and the ProSA-web server, with model 5 selected as the best model and uploaded to the GalaxyRefine server for refinement. We assessed the five refined models provided by the GalaxyRefine server using a variety of quality evaluation metrics, including GDT-HA, RMSD, MolProbity, Clash score, Poor rotamers, and Rama favored (Supplementary Table S7). Based on its high GDT-HA (0.9927) and Rama favored (93.7), as well as its low RMSD (0.279), MolProbity (2.085), clash score (13.8), and poor rotamers (0.3), the refined model 1 (Fig. 2C) was chosen for further analysis. The Ramachandran plot of the chosen initial model revealed that 87.4%, 9.7%, 0.8%, and 2.1% of the residues were found in the most favoured, additional allowed, generously allowed, and disallowed regions, respectively (Fig. 3A). After refinement, the proportion of residues in the most favored, additionally allowed, generously allowed, and disallowed regions was 89.9%, 7.6%, 0.7%, and 1.8%, respectively (Fig. 3B). The initial model’s ProSA Z-score was − 4.68 (Fig. 3C), but following refinement, it changed to −4.78 (Fig. 3D).

Fig. 3.

Validation of the multi-epitope vaccine. (A) Ramachandran plot shows that 87.4%, 9.7%, 0.8%, and 2.1% of residues of the initial 3D structure of the vaccine were found in the most favoured, additional allowed, generously allowed, and disallowed regions, respectively (B). Ramachandran plot shows that 89.9%, 7.6%, 0.7%, and 1.8% of residues of the refined 3D structure of the vaccine were located in the most favoured, additional allowed, generously allowed, and disallowed regions, respectively. (C) ProSA Z-score of the initial and (D) refined 3D structure of the vaccine candidate are − 4.68 and − 4.78, respectively

Discontinuous B‑cell epitopes prediction

The tool identified 14 discontinuous B‑cell epitopes in the 3D refined structure of the vaccine. The number of residues in these epitopes varies from 3 to 132 amino acids, as does their score, which ranges from 0.501 to 0.875 (Fig. 4, Supplementary Table 8).

Fig. 4.

The 3D representation of discontinuous B-cell epitopes in the 3D refined structure of the multi-epitope vaccine. (A) 8 residues with a score of 0.875. (B) 3 residues with a score of 0.866. (C) 13 residues with a score of 0.854. (D) 110 residues with a score of 0.823. (E) 95 residues with a score of 0.819. (F) 132 residues with a score of 0.656. (G) 15 residues with a score of 0.614. (H) 35 residues with a score of 0.593. (I) 16 residues with a score of 0.587. (J) 7 residues with a score of 0.578. (K) 24 residues with a score of 0.555. (L) 4 residues with a score of 0.535. (M) 4 residues with a score of 0.517. (N) 4 residues with a score of 0.501

Disulfide engineering of the vaccine construct

The Disulfide by Design 2 v2.13 server identified 62 possible residue pairings in the vaccine’s revised 3D structure that might form disulfide bonds (Supplementary Table S9). The χ3 angle peaks at −87° and + 97°, and approximately 90% of disulfide bonds have an energy value < 2.2 kcal/mol. Based on these parameters, we chose 9 residue pairs for disulfide bond formation: GLY150-GLN155, LYS170-GLY205, GLU211-LYS280, PRO351-SER354, LYS677-THR680, GLN685-SER691, ALA734-SER782, TYR735-VAL738, and THR789-PRO841 (Fig. 5).

Fig. 5.

The disulfide engineering of the refined 3D model of the multi-epitope vaccine. (A) The wild type. (B) The mutant type, with the 9 disulfide bonds depicted as yellow sticks enclosed in yellow rings

Molecular Docking of the vaccine with TLR4

The ClusPro 2.0 server carried out the molecular docking analysis between TLR4 and the vaccine’s refined 3D structure. For the vaccine-TLR4 docked complex, it generated 27 clusters (Supplementary Table S10). The best model for MD simulation was thought to be cluster 0, which had the most members (40) and the most negative energy (−1164.4 kcal/mol) (Fig. 6A). Additionally, the PDBsum server was used to conduct the interaction analysis. The results indicated that in the docked vaccine-TLR4 complex, 9 residues from TLR4 chain A interacted with 10 residues from the vaccine, 6 residues from TLR4 chain B interacted with 6 residues from the vaccine, and 20 residues from TLR4 chain C interacted with 17 residues from the vaccine (Fig. 6B). No H-bonds were established between TLR4 chain B and the vaccine; however, 5 H-bonds were formed between TLR4 chain A and the vaccine, and 7 H-bonds were formed between TLR4 chain C and the vaccine (Table 2).

Fig. 6.

Molecular docking of the proposed vaccine construct with TLR4. (A) The vaccine-TLR4 docked complex with energy score of −1164.4 kcal/mol. The vaccine was depicted as a sphere, while the TLR4 was represented as a ribbon. (B) Interaction pattern of the vaccine and TLR4 chains A, B, and C

Table 2.

A list of residues involved in hydrogen bond interactions within the docked complex of the vaccine candidate and chains A and C of TLR4, along with their corresponding distances

TLR4 (Chain A)-Vaccine
TLR4 residues	Vaccine residues	H-bond distances (Å)
ASP294	SER868	2.85
ASP294	LYS849	2.59
TYR296	THR857	2.86
LYS341	GLN854	2.63
LYS 341	GLN854	2.55

TLR4 (Chain C)-Vaccine
LEU54	HIS884	2.71
GLU92	GLY881	2.96
LYS122	THR871	2.72
LYS122	SER872	2.59
LYS122	GLY873	2.93
LYS122	ALA874	2.52
LYS122	PRO875	3.21

Molecular dynamics simulation

To examine the stability of the docked vaccine-TLR4 complex, an MD simulation run of 200 ns was performed using AMBER version 20 software. During the initial phase of the simulation, up to 120 ns, the RMSD value escalated from zero to 18 Å, subsequently fluctuating between 17 and 22.5 Å until the end of the simulation (Fig. 7A). The vaccine residues show a high RMSF value attributable to the significant proportion of coils in the vaccine structure, resulting in flexibility, whereas the TLR4 residues show a low RMSF value due to their predominantly beta-sheet configuration (Fig. 7B), which is more rigid and stable owing to robust interstrand hydrogen bonding that reduces structural fluctuations [119]. The Rg plot showed an initial increase for the first 20 ns of the simulation, followed by a decline until 100 ns, after which it stabilized until the end of the simulation (Fig. 7C). The binding free energy of the docked TLR4-vaccine complex was calculated using the MM-PBSA and MM-GBSA methods to assess the strength and stability of its interactions. The total binding free energies, calculated using the MM-PBSA and MM-GBSA methods, were − 284.99 and − 283.53 kcal/mol, respectively. The contribution of van der Waals energy is significant in both methods (Table 3); these interactions are essential for facilitating molecular recognition between the TLR4 receptor and the vaccine, so ensuring effective activation of immune responses.

Fig. 7.

The MD simulation of the selected docked vaccine-TLR4 complex during a 200 ns period. (A) RMSD plot. The RMSD increased from 0 to 18 Å in the first 120 ns, subsequently fluctuated between 17 and 22.5 Å until the end of the run. (B) RMSF plot. The peaks contain residues with a high level of flexibility. (C) Rg plot. The Rg value shows little variation from the start to the end of the simulation

Table 3.

Binding free energy values estimated for the selected docked vaccine-TLR4 complex

Energy parameters	Value (kcal/mol)
MM-PBSA
Van der Waals energy	−256.10
Energy electrostatic	−55.69
Total gas phase energy	−311.79
Total salvation energy	26.80
Net energy	−284.99
MM-GBSA
Energy van der Waals	−256.10
Energy electrostatic	−55.69
Total gas phase energy	−311.79
Total energy salvation	28.26
Net energy	−283.53

Immune response simulation of the vaccine candidate

The C-ImmSim server precisely predicted immunological responses following the administration of three vaccine doses at four-week intervals. The initial vaccine injection triggered a primary immune response, resulting in the production of IgM and IgG (IgG1 + IgG2). The second and third vaccine injections dramatically boosted IgM and IgG1 + IgG2 levels, indicating that secondary and tertiary responses caused strong antibody reactions as antigen concentrations decreased. During the initial two injections, the elevation of IgM levels surpassed those of IgG1 + IgG2; conversely, in the third injection, the rise in IgG1 + IgG2 exceeded that of IgM (Fig. 8A). Following each of the three injections, the population of B-cell memory, B cell isotypes, and active B cells increased. After three injections, the B isotype IgG2 population increased very slightly, whereas the B isotype IgG1 and B isotype IgM populations increased dramatically (Fig. 8B, C). Following the administration of the second and third doses, the number of TH memory cells, TH active cells, and TH resting cells increased; a similar pattern was observed for the population growth of TC memory cells and TC active cells, but TC resting cells showed the inverse trend (Fig. 8D, E, F, G). The concentrations of cytokines, including IFN-γ, transforming growth factor-beta (TGF-β), IL-10, IL-12, and IL-2, increased after each of the three injections. However, the levels of IFN-γ, TGF-β, IL-10, and IL-12 decreased from the first to the third response, although IL-2 showed a rise during the same interval (Fig. 8H).

Fig. 8.

Characterization of the immunological response elicited by the vaccine candidate. (A)Immunoglobulin responses. (B) The B cell population. (C) The B cell population per state. (D)The population of total and memory TH cells. (E) The TH cell population per state. (F) The population of total and memory TC cells. (G) The TC cell population per state. (H) The cytokine concentration

Codon optimization and in Silico cloning

The multi-epitope vaccine was back-translated into a nucleotide sequence of 2664 bp and optimized for expression in the E. coli K12 strain with JCat. The GC content of the cDNA sequence was determined to be 48.99%, and the CAI was calculated to be 1. The vaccine sequence was then inserted into the pET28a (+) expression vector, between the BamHI (198) and XbaI (2868) restriction sites. After cloning, the cloned vector had a total length of 7902 bp (Fig. 9).

Fig. 9.

In silico cloning of a multi-epitope vaccine into the pET28a (+) expression vector. The vaccine is shown by the yellow region surrounded by the BamHI (198) and XbaI (2868) restriction sites, while the black region displays the pET28a (+) expression vector

Discussion

The World Health Organization (WHO) estimates that almost all individuals who are sexually active will develop HPV at some stage in their lives. This makes it one of the most frequent sexually transmitted viruses worldwide [1]. While most HPV infections self-resolve, persistent infection with HR-HPV types can result in many cancers, including cervical cancer, which remains the leading cause of cancer-related mortality among women worldwide [120].

Despite the notable success of prophylactic HPV vaccines introduced over the last 18 years, cervical cancer remains a significant global challenge. The nearly 10% increase in new cervical cancer cases worldwide between 2020 and 2022 emphasizes the critical need for the development of novel vaccines with both prophylactic and therapeutic properties that provide broader coverage, greater affordability, and better transport and storage conditions [66]. Multi-epitope vaccines may serve as an appropriate solution to these challenges. These vaccines, by targeting several epitopes, can elicit both humoral (antibody-mediated) and cellular (T-cell-mediated) immunity, resulting in more robust and long-lasting protection [121]. They also offer additional benefits, such as the potential for broad protection against multiple pathogen strains and mutations, enhanced immune responses, and a lower risk of side effects. Multi-epitope vaccines are usually engineered to be more stable and resistant to degradation than traditional vaccines, making storage and transportation easier [122]. The fundamental drawback of multi-epitope vaccines is that they are less effective in eliciting a strong immune response for long-term immunity and frequently require adjuvants and boosters to establish and maintain protective immunity [123].

Several studies have been conducted on the in silico design of multi-epitope vaccines to develop effective immunotherapies against cervical cancer. Kayyal et al.. used in silico analysis to design two multi-epitope vaccine candidates that included immunogenic and conserved epitopes from HPV16/18’s L1, L2, and E7 proteins (DNA fusion L1-L2-E7). The presence or lack of heat shock protein 70 (HSP70) regions differentiated these constructs [124]. In the study conducted by Kumar et al.. the L1, E5, E6, and E7 oncoproteins from HPV 16/18 were used for epitope prediction; their recombinant chimeric vaccine construct included selected HTL and CTL epitopes, along with the adjuvant β-defensin [125]. Shahab et al.. designed and evaluated a chimeric protein vaccine that included CTL, HTL, and linear B-cell epitopes derived from the L2 protein from HPV16, as well as a 50 S ribosomal protein L7/L12 adjuvant [126]. The vaccine construct in Zhu et al..‘s study contained CTL and HTL epitopes from the E6 and E7 oncoproteins of HPV16, together with β-defensin 3 as an adjuvant [127]. Ehsasatvatan et al.. designed a multi-epitope vaccine that included CTL, HTL, and linear B-cell epitopes from L1 protein for five LR-HPV types (HPV6, 11, 42, 43, 44) and 12 h-HPV types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59); in this construct, cholera toxin subunit B (CTB) and TLR4 agonist RS09 were added as adjuvants to the C- and N-terminal of the multi-epitope vaccine candidate, respectively [128]. These works show how immunoinformatics and computational approaches can be used to develop successful vaccines against cervical cancer, laying the groundwork for future experimental validation and clinical applications.

In the present study, potential epitopes that elicit humoral and cellular immune responses have been identified from conserved regions of target proteins using a suite of bioinformatics tools. The epitopes are antigenic, non-toxic, non-allergenic, and capable of inducing IFN-γ (for HTL epitopes). Conservancy analysis revealed that all chosen epitopes were significantly conserved among HPV16/18 strains, indicating a probability of cross-protection against these viral strains.

Although real-world immune responses might vary, the population coverage findings from the study indicate that the epitopes in the designed vaccine construct can probably cover 100% of the world’s population, implying that the proposed vaccine is likely to be successful for the whole worldwide population. Clinical trials are required to determine the efficacy of the vaccine in a variety of populations, as well as the effect of genetic and environmental factors on immune responses [129].

Following the overlap analysis, some identified epitopes were removed, and the remaining epitopes constituted the main structure of the vaccine construct. The final vaccine composition included 7 HTL epitopes, 62 CTL epitopes, 3 linear B-cell epitopes, RS09 as an adjuvant, linkers such as EAAAK, GPGPG, and KK, along with a 6xHis tag. TLR4 is a key innate immune system component that recognizes pathogen-associated molecular patterns (PAMPs), such as lipopolysaccharides (LPS) from Gram-negative bacteria, as well as damage-associated molecular patterns (DAMPs) released by stressed or dying cells [130]. When TLR4 is activated, it triggers signaling cascades via myeloid differentiation factor 88 (MyD88)-dependent and TIR-domain-containing adapter-inducing interferon-β (TRIF)-dependent pathways, activating nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and interferon regulatory factor 3 (IRF-3). This process results in the generation of proinflammatory cytokines and type I interferons, which act as a bridge between innate and adaptive immunity [131, 132]. TLR4 promotes the proliferation and apoptosis resistance of HPV-related cervical cancer cells [133]. When the TLR4/MyD88/NF-κB pathway is activated, HPV-related cervical cancer cells produce large quantities of proinflammatory cytokines, suggesting that TLR4 promotes cervical cancer growth through establishing an immunosuppressive microenvironment [133]. Because TLR4 is expressed in a variety of cell types (tumour cells, dendritic/immune cells, and stromal cells) [134], and downstream signaling can produce quite varied results depending on context, there are two therapy scenarios: (1) Blocking TLR4 (or downstream MyD88/NF-κB) reduces pro-tumor cytokines, reducing HPV-related cervical cancer cell proliferation and invasion [135]. (2) TLR4 activation of antigen-presenting cells can enhance adaptive anti-tumor immunity. TLR4 agonists significantly increase dendritic-cell maturation and CD8⁺ T-cell responses to HPV antigens (E6/E7), hence boosting vaccine-driven tumor immunity [136] which is why TLR4 ligands are used as vaccine adjuvants [137]. For this reason, RS09 (TLR4 agonist) was employed as an adjuvant in this study.

Considering that the amino acid composition of our vaccine design differs from those developed in prior studies, it is expected that vaccine features associated with this composition will likewise vary. Computational analysis indicates that the proposed vaccine may be antigenic, non-allergenic, non-toxic, and soluble. The vaccine has a molecular weight of 97.606 kDa, which is below 110 kDa, facilitating purification and processing during the experiment [138]. The vaccine’s theoretical pI was determined to be 10.89, indicating an alkaline nature. The predicted theoretical pI for the vaccines designed by Kumar et al. [125]., Shahab et al. [126]., and Ehsasatvatan et al. [128]. was likewise greater than 8, indicating that these vaccinations are alkaline, similar to our vaccine. Our vaccine candidate has a predicted half-life of 4.4 h in mammalian reticulocytes, > 20 h in yeast, and > 10 h in E. coli, whereas the vaccine designed by Ehsasatvatan et al. [128]. has a predicted half-life of 3.5 h in mammalian reticulocytes, 10 min in yeast, and > 10 h in E. coli, indicating that our vaccine is likely to be exposed to the immune system for a longer period of time. The vaccine’s instability index, essential for maintaining structural integrity during storage and delivery [139], was determined to be below 40 (38.66), implying potential stability of the vaccine protein structure [140]; however, this parameter value was greater (41.53 and 43.63) for both vaccine constructs designed by Kayyal et al. [124]. The vaccine’s aliphatic index was 57.31, which is typically in the range of 50 to 160. A higher aliphatic index has been related to greater thermostability in proteins [141]. It indicates that our vaccine can endure various temperature conditions without compromising its immunogenicity, thereby ensuring the vaccine’s efficacy during production and delivery [142]. A negative GRAVY value (−0.883) for the vaccine candidate signifies protein solubility in water (hydrophilicity), essential for efficient antigen delivery [143]; conversely, the parameter is positive (0.186) in the vaccine designed by Kumar et al. [125]., implying it is probably less soluble than our vaccine.

A three-dimensional structure of the multi-epitope vaccine was constructed, and the refining procedure enhanced the model’s quality. A quality assessment indicated that in the improved vaccine structure model, the majority of residues are located in the most favored region, and its ProSA Z-score falls within the range typical of native proteins.

Molecular docking analysis predicted a strong binding affinity between the vaccine and TLR4. The MD simulation was performed for 200 ns to confirm the vaccine’s docking with TLR4. The RMSD plot revealed high values, which could be attributable to the vaccine’s multiple coil regions. Coils are naturally flexible and dynamic structural components that lack stable hydrogen-bonding networks. This intrinsic flexibility enables the vaccine structure to assume a variety of conformations during MD simulation [144, 145], which are frequently required for optimal interaction with immune receptors such as TLRs, improved antigen presentation, and immunological activation. This flexibility may help the vaccine adjust to the various conformational states required for optimal epitope presentation [146]. Our vaccine candidate could interact strongly with TLR4 in the docked complex, as indicated by negative binding free energy values determined using MM-PBSA and MM-GBSA methods.

The immune response simulation revealed that administering our candidate vaccine could increase the amount of memory B and T cells, as well as the levels of antibodies in secondary and tertiary immunological responses, when compared to the primary immune response. The cytokine profiles, characterized by an initial peak of IFN-γ and prolonged IL-2 activity, highlight the specialized immunological signals required for effective immune regulation. Several experimental studies have indicated the immunogenicity of different identified epitopes derived from HPV16/18 E6 and E7 proteins. For example, Bahmani et al.. developed a multi-epitope vaccine including epitopes derived from HPV16 E7 protein, and to assess its immunogenicity, C57BL/6 mice were administered with three doses of the vaccine subcutaneously at two-week intervals. The level of antibodies in immunized mice serum was measured using an enzyme-linked immunosorbent assay (ELISA) and the study clearly showed that antibody production was significantly increased after three doses of immunization [147]. In another study, Oliveira et al.. described the development and assessment of a multi-epitope vaccine containing the immunogenic HPV16 E6 and E7 epitopes. C57BL/6 mice received an initial subcutaneous injection of 7.5 × 104 TC-1 cells. After three days, the immunization schedule began, which included three subcutaneous injections of E6E7. The vaccination stimulated strong activation in E6/E7-specific T-cells [148]. The results of the aforementioned studies are consistent with our computational findings.

The estimated GC content percentage in this study is 48.99%; a GC content of 30–70% is considered ideal [149]. The CAI value for the vaccine sequence is 1, which falls within the optimal range of 0.8–1.8 [150], indicating that our vaccine candidate has promising expression potential.

While our computational findings are intriguing, we note that the study is primarily based on computational predictions, which may not fully capture the complexities of immune responses in vivo. As a result, experimental validation using immunological tests and animal models is critical for determining the anticipated antigenicity and safety of the new multi-epitope vaccine. Furthermore, we have identified future directions, which include the incorporation of more comprehensive population coverage assessments, the refinement of epitope selection using novel structural bioinformatics tools, and the research of delivery routes to improve vaccine efficacy. These improvements clarify our findings and provide a road map for translating this bioinformatic framework into actual vaccine development.

Conclusion

This study employed immunoinformatics approaches to design a multi-epitope vaccine for cervical cancer that could have both prophylactic and therapeutic potential. To identify T- and B-cell epitopes, we focused on HPV16/18 E6, E7, and L2 proteins. The proposed multi-epitope vaccine has potentially desirable physicochemical features, including the potential to be soluble and stable while being free of toxicity and allergenicity. This vaccine candidate could provide broad coverage across the worldwide population, and induce strong cellular and humoral immune responses. The molecular docking analysis between the vaccine and TLR4 revealed a stable vaccine-receptor complex capable of eliciting a strong immune response. This study provides impressive results; however, experimental validation is essential for confirming the candidate vaccine’s efficacy and safety.

Author contributions

SHE: Writing – original draft, Methodology. SA: Methodology, Visualization. MR: Data curation. RB: Visualization, Validation. FA: Software. ShA: Supervision, Project administration. SS: Supervision, Project administration, Writing – review & editing. All authors reviewed the manuscript.

Funding

This study was supported by a grant from Semnan University of Medical Sciences (Grant Number: 4142).

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files.

Declarations

Ethics approval and consent to participate

The ethical committee of Semnan University of Medical Sciences approved this study with the number: IR.SEMUMS.REC.1403.180.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.