Integrative immunoinformatics and structural modeling for the rational design of a multi-epitope vaccine candidate against human cytomegalovirus

Abstract

Human cytomegalovirus (CMV) is a globally widespread pathogen associated with significant morbidity in immunocompromised individuals. Despite its clinical importance, no licensed vaccine is currently available. This study aimed to design a rational multi-epitope vaccine candidate targeting CMV using an integrative approach combining immunoinformatics and structural biology. Viral proteins were screened to identify epitopes with high affinity for B cells, cytotoxic T cells (CTLs), and helper T cells (HTLs) using the Immune Epitope Database (IEDB). Selected epitopes were filtered according to their antigenicity and toxicity and then assembled into a chimeric construct incorporating an immunostimulatory adjuvant. The designed vaccine was evaluated for its physicochemical properties, validated by Ramchandran and ERRAT analyses. Molecular modeling demonstrated strong and stable interactions with key innate immunity receptors, including TLR7 and TLR9, interactions confirmed by molecular dynamics simulations. In silico immune simulation predicted a robust and durable immune response, characterized by high levels of IgM and IgG, as well as significant activation of CD4 + and CD8 + lymphocytes and innate immunity components. These results highlight the potential of the proposed multi-epitope construct as a promising vaccine candidate against HCMV. However, experimental validation is essential to confirm its immunogenicity, safety, and translational applicability.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-61161-x.

Introduction

Human cytomegalovirus (CMV) is a virus carried by a large majority of individuals worldwide. It is a large DNA virus that lies dormant in the herpes family, often without their knowledge^1,2. Although it goes unnoticed in healthy bodies, trouble manifests itself in babies immediately after their birth, in people with organ donations, in people with HIV/AIDS, or in any other person whose immunity has been weakened by medications³. CMV exposure increases with age and is widespread globally⁴. Estimates vary across continents, generally reflecting socioeconomic development and access to healthcare⁴. In Africa, it exceeds 90–95% in many areas⁵. In sub-Saharan Africa, seroprevalence among adults approaches 100%^5,6. In Asia, CMV exposure is approximately 81–95%⁷. The highest rates are typically observed in the Eastern Mediterranean region and South Asia. In South America, exposure is estimated to be over 95% in most adult populations^4,8. This infection is responsible for 10% of paralysis and 25% of deafness in children⁹. A retrospective study conducted in Asia indicated that human cytomegalovirus caused 75% of deaths in immunocompromised individuals¹⁰. Being inherited at birth, this virus has been identified as the leading cause of childhood hearing and brain growth retardation compared to any other non-inherited factor¹¹. It stays silent in blood related cells and comes out as an unwanted guest once the defenses have been compromised¹².

At this moment, the majority of individuals with HCMV receive assistance with the use of medications such as Ganciclovir, Valganciclovir, Foscarnet, or Cidofovir¹³. Obstructing the way, the virus multiplies itself occurs when these medications sluggish its polymerase activity albeit latent residues remain, which can resurface at a later time. Despite their effectiveness in acting on the virus, they should not be used for too long as they damage blood cells or even the kidneys, making it difficult to treat a baby and people with a weak immune system^14,15. To escape the action of antiviral drugs and ensure their replication, some viruses like HCMV have developed bypass mechanisms, thus compromising the effectiveness of therapies over time^16,17. Recent drugs such as Letermovir, which is a viral terminase inhibitor, have provided alternatives although they remain incomplete, effective only at different stages of infection^18,19. After every unsuccessful therapy, it is obvious that there is a necessity to have a reliable vaccine to prevent the spread of HCMV and its negative outcomes^18,19.

Glycoprotein B is one of the viral proteins expressed on the surface, which is produced by UL55 gene and is active during infection^19,20. The combination of the outer layer of the virus and the cell it infects depends on viral entry through this protein²¹. Among all the proteins, gB is an especially good one abundant and capable of provoking both antibody and T-cell responses^22,23. There is currently no licensed vaccine to prevent human cytomegalovirus (CMV) infection²⁴. Due to the virus’s complex mechanisms of evasion from the immune system, combating CMV remains a major public health priority. Several promising candidates are currently undergoing clinical trials. Candidates such as the gB/MF59 vaccine target the surface antigen glycoprotein B (gB) to trigger the production of neutralizing antibodies. Although it has shown partial protection (approximately 43% to 50%) against viral acquisition, its use and efficacy have several major limitations, including limited and incomplete efficacy, a weak neutralizing response, and unblocked acquisition²⁵. In immunocompromised patients (such as transplant recipients) or seronegative women, it reduces viral load and the duration of viremia but does not provide sterilizing immunity²⁶. The protection MF59 booster plus a lab-produced form of gB reduced the odds of infection by approximately half in the meantime, but the protection failed quickly²⁵. Although these results indicate that flooding the system with full-length protein does not create any broad immunity, it may be necessary to design smarter forms which target dominant immune targets instead²⁵.

In spite of advances that have been made in vaccine research for Human Cytomegalovirus infection, the previous vaccines that were created, especially the subunit vaccines that utilize glycoprotein B (gB) in their formulations, have shown little success when used in clinical settings^25,26. This is because even though these vaccines elicit some form of response from the body’s immune system, the effectiveness of these responses has been found to be partial and temporary. Moreover, these vaccines have also failed to induce a strong and sustainable cellular immune response which is important for fighting HCMV infections, especially among high-risk individuals like those who have compromised immune systems^25,26.

Currently, the immunoinformatics approaches help in designing vaccines rationally by using computational approach and minimizing the use of empirical approach^8,27–29. The idea of creating a vaccine from several key components of HCMV glycoprotein B thus appears particularly promising^30,31. Rather than whole proteins, this vaccine would consist of robust and stable fragments, recognized by B and T lymphocytes, which could induce broad immunity without undesirable side effects. Since the structure of gB is more resistant to degradation and better recognized by the immune system thanks to the addition of specific helper molecules and spacers^30,31, the development of a vaccine targeting this protein would be a significant advantage in the vaccination of human cytomegalovirus infection. Furthermore, the use of computer motion models allows observation of how the construct moves and behaves within the body, thus ensuring the integrity of its structure before laboratory testing^32,33. Therefore, this work relies on immunoinformatic modeling, pattern prediction, and interaction analysis to design a multiepitope vaccine against human cytomegalovirus gB. This is therefore a computational study which will require experimental validation for the approval of the predicted vaccine sequence.

Materials and methods

Material

Webservers and softwares

Table S1 represents Web servers and oftwares used in this study. These databases/softwares were UniProtKB, ExPASy, PSIPRED, Swiss-Model, DeepTMHMM, VaxiJen v2.0, DiANNA 1.1, C-ImmSim, JCat. DNA, Hdock, iMODS, MHC-I Binding Predictions, MHC-II Binding Prediction, SnapGene, LigPlot+, algpred2, ToxinPred2, PRRpred, and Alphafold server.

HCMV proteins

Viral proteins from the HCMV genome were studied for the preliminary evaluation of their antigenic properties and their toxicological and allergic profiles. This included envelope glycoprotein-B (gB, UniProt ID: UL55), UL96 (UniProt ID: P16787), CCAAT/enhancer-binding protein (CEACAM1-like UL7) (UniProt ID: Q6SWC3), alkaline nucleases (UniProt ID: F5HF49), UL13 (UniProt ID: Q6SWB8), US11 (glycoprotein US11, UniProt ID: P09727), UL10 (UniProt ID: R4SHH4), envelope glycoprotein-H (gH, UniProt ID: Q6SW67), envelope glycoprotein-L (gL, UniProt ID: Q68674), UL22A(glycoprotein UL22A, UniProt ID: P16845), and glycoprotein-O (gO, UniProt ID: F5HGP1). These protein sequences were obtained through the UniProt database (https://www.uniprot.org).

Methods

This study was conducted according to the protocol described in Fig. 1.

Fig. 1.

Study design workflow in the vaccine construct.

Antigenic and safety properties evaluation of HCMV proteins

Antigenicity was predicted using VaxiJen v2.0 (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/, Threshold value: 0.5), which estimates the probability that a protein is an antigen^34–36. Allergenicity was assessed using Alerpred2 (https://webs.iiitd.edu.in/raghava/alergpred/, Threshold value: 0.3), which classifies proteins as allergenic or non-allergenic³⁷. Toxicity prediction was performed using ToxinPred (https://webs.iiitd.edu.in/raghava/toxinpred/, Threshold value:0.6), which evaluates the toxic potential of peptides and proteins. Neurotoxicity screening was included as an additional safety filter in Ntxpred2 (https://webs.iiitd.edu.in/raghava/ntxpred2/disp_c1.php?ran=40271, Threshold value: 0.3) to identify proteins with potential neurotoxic risk³⁸.

Prediction of protein topology and transmembrane domains

The membrane topology of each of the selected viral proteins (gB, UL7, Alk Nuclease, UL13, US11, UL10, gH, gL, and UL22A) was predicted using the TMHMM server. For each protein, the amino acid sequence in FASTA format was analyzed separately by submitting the sequences to the TMHMM server (version 2.0) (https://services.healthtech.dtu.dk/services/TMHMM-2.0/)^39,40. The algorithm produced a list of probabilities for each amino acid residue being inside (cytoplasmic side) the membrane, outside (non-cytoplasmic/extracellular) the membrane, or within a transmembrane helix. The predicted transmembrane regions were those that fall into areas of high probability for being transmembrane and have characteristics indicative of a membrane spanning structure^39,40.

Target proteins selection

In order to establish priorities in selecting the HCMV proteins for further investigation as prospective vaccine candidates, a scoring system had been developed, which included four principal criteria: antigenicity, allergenicity, toxicity/neurotoxicity, and protein topology. The main criterion used in the study was antigenicity and thus was allocated four points. Proteins with antigenicities ≥ 0.50 got all four points, with 0.45–0.49 having three points, with 0.40–0.44 awarded with two points, with 0.35–0.39 obtaining one point, and proteins below that threshold – zero points. Allergenicity was awarded two points for non-allergens and zero points for allergens. Toxicity and neurotoxicity were assessed at once with proteins being either non-toxic and non-neurotoxic and receiving two points, exhibiting at least one negative property receiving one point and both characteristics being negative giving zero points. Finally, protein topology and cellular accessibility were awarded two points each. Proteins that are mostly found extracellularly (> 70%) got two points, proteins with partial exposure (30–70%) one point, while intracellular ones – zero points. This method was initiated in light of the work of Zahraei et al.⁴¹.

Phylogeny tree analysis

To evaluate the evolutionary relationships and conservation of cytomegalovirus glycoprotein B among different viral strains, a phylogenetic analysis was performed to ensure the design of a more suitable and broad-spectrum vaccine. To perform this analysis, glycoprotein B (gB) protein sequences obtained from the UniProt and NCBI protein sequence databases for Human betaherpesvirus 5 and its associated glycoproteins⁴². The amino acid sequence entries that were analysed were sorted by their NCBI accession number which included: ACS92156.1 (gB of Human cytomegalovirus (HCMV)), XOL08432.1, AAA45932.1, 5CXF_A, XOL08435.1, XUA97349.1, CAA07368.2, YCN21674.1, YCN21491.1, AAA45927.1, XCZ56811.1, ALJ56180.1, ALJ56184.1, XCZ56814.1, AJF36318.1. NCBI/BLASTP were used to confirm homologous relationships between the selected sequences based on their similarity and identity of amino acid sequences. The FASTA sequences that were retrieved were aligned together using ClustalW, into MEGA Software, version X^43,44. The construction of phylogenetic tree were performed using MEGA’s Maximum Likelihood (ML) methods, and to determine whether the inferred phylogenetic relationships represented true phylogenetic relationships, or just coincidental associations, by performing a bootstrap analysis (1000 replicates)^43,44.

Antigenic sequence retrieval

The human cytomegalovirus gB gene (accession number of the AD169 strain is P06473) was retrieved into the UniProtKB database (https://www.uniprot.org/)^45,46.

Prediction of T and B cell epitopes

The selected sequence of the glycoprotein B was retrieved in FASTA format and submitted for epitope prediction through immunoinformatics tools. The prediction of CTL epitopes was conducted through the IEDB MHC-I binding prediction tool by using the recommended approach of NetMHCpan EL 4.1. The epitopes having high binding affinity to a range of alleles of HLA class I having a percentile rank ≤ 1 was considered. HTL epitopes were predicted through the IEDB MHC-II binding prediction tool by using the recommended approach of consensus. Those peptides showing a high binding affinity to the major alleles of HLA class II having a percentile rank ≤ 2 was considered for epitope prediction. Linear B-cell epitopes were predicted by using the BepiPred 2.0 server having the default threshold value of 0.5^29,47.

Epitope screening of immunological and physicochemical properties

Antigenicity of the candidate epitopes was determined on the VaxiJen v2.0 server, and their allergenicity in AlgPred2 (https://webs.iiitd.edu.in/raghava/algpred2/batch.html, Threshold value: 0.3)³⁷. Toxicity of these cadidate epitopes was predict using ToxinPred2 (https://webs.iiitd.edu.in/raghava/toxinpred2/index.html, Threshold value:0.6)⁴⁸. The stability of gB and constructed vaccine was predict using ProtParam (Expasy). PRRpred database was used to predict the pattern recognition receptors of predited epitopes (PRRpred, https://webs.iiitd.edu.in/raghava/prrpred, Threshold value: 0.5)⁴⁹. Peptides with a desirable immunogenic profile and which were non-allergenic, non-toxic, and stable were only selected to build the vaccine. In all the programs, the defaults of the tools were used.

Design of the multi-epitope vaccine (MEV)

To design the vaccine sequence, the different types of epitopes were combined, starting with the adjuvant sequence, which was linked to B-cell epitopes via the EAAAK sequence. These were then linked to MHC-I epitopes via the GGGS sequence, which in turn were linked to MHC-II epitopes via the AAY sequence. Finally, a 6X histadine tag sequence (HHHHHH) was added^50–53.

Discontinuous B-cell, concervancy and cluster epitope prediction

Discontinuous B-cell epitope were predicted as discontinuous epitopes by a structural computation approach based on the previously described methods^39,50. Vaccine sequences in the PDB format were uploaded to the IEDB database (http://tools.iedb.org/ellipro/). Each predicted epitope was analyzed according to its length, location on the protein chain, and the conservation score. For evaluating the level of conservation of predicted vaccine epitopes, comparative sequence analysis was performed of five human cytomegalovirus envelope glycoprotein B protein sequences. Proteins used for analysis were obtained from the UniProt database (https://www.uniprot.org/)^54–56. The sequences of epitopes of the developed vaccine and strain proteins were uploaded to the database for conservation evaluation. The protein sequence match percentage was determined under the conditions of 100% identity (https://tools.iedb.org/conservancy/). Similarly, for the epitope clustering analysis, the vaccine epitopes were entered into the IEDB database (https://tools.iedb.org/cluster/), then a selection sequence identity threshold and a minimum and maximum peptide length were set at 7 and 25 respectively, and the analysis was launched⁵⁷.

2D, 3D modelling, refinement, and validation

Three-dimensional structure prediction of the predicted multiepitope vaccine was performed using AlphaFold2. The entire sequence of amino acids of the vaccine was entered into the AlphaFold2 database, and a structure with the highest confidence score was considered for further analysis. The accuracy of the predicted structure was analyzed based on the pLDDT scores generated by AlphaFold2⁵⁸. Stereochemical quality of the predicted 3D structure was analyzed using the Ramachandran plot obtained through the PROCHECK server according to Owona et al.⁵⁹

Antigenicity prediction of vaccine sequence

To determine the antigenicity of the vaccine sequence, the method of Kolaskar and Tongaonkar⁶⁰. was used. Predictions were based on a table listing the frequency of occurrence of amino acid residues in experimentally known segmental epitopes. The accuracy of the method is estimated at approximately 75%. The vaccine sequence was then entered into the Immunomedicine Group tools Predicting Antigenic Peptides database (http://imed.med.ucm.es/Tools/antigenic.pl) to initiate prediction and visualize the results⁶⁰.

Interaction study by molecular docking

Docking of the multi-epitope vaccine with the TLR7, TLR8, and TLR9 immune receptors was carried out in order to determine the binding properties of the vaccine-receptor interactions. Docking simulation was done on the HDOCK webserver (http://hdock.phys.hust.edu.cn/)^61,62. The three-dimensional structure of the vaccine along with that of the respective immune receptors TLR7, TLR8, and TLR9 was submitted to the server in PDB format. Docking was done using the default parameters of the HDOCK protocol. The docking complexes obtained from the simulation were scored depending on the binding energy and the stability of the receptor-ligand interaction. The docking complex having the minimum score for each receptor was chosen for further visualization using LigPlot +⁶³.

Normal mode and MD analysis analysis

The stability of docked complexe between vaccine and protein TLR7, TLR8, and TLR9 was performed usind the IMODS server (https://imods.iqf.csic.es/) according to the study of Emmanuel et al.⁶⁴. Dynamic simulation at 100ns was performed according to Asim et al.³⁹. MD simulations were carried out for evaluating the conformational dynamics and stability of the vaccine-TLR protein complexes using Desmond software³⁹. To check the stability of the structure of the complex, the RMSD parameter was calculated, while the RMSF was used to examine the flexibility of the residues. The trajectories generated from the Desmond simulation were converted to XTC format using MDtraj and analyzed using the Gromacs tools^28,39.

Simulation of the immune response

An agent-based simulation of the immune response to the vaccine construct FASTA sequence was carried out using the C-ImmSim platform (https://kraken.iac.rm.cnr.it/C-IMMSIM/index.php)^39,65. This was performed by using a random seed of 12,345, a simulation volume of 10, and 100 simulation steps. The HLA profile of the virtual host was defined by the alleles HLA-A01:01/A01:01 and HLA-B07:02/B07:02 for the MHC class I molecules, while for the MHC class II molecules, HLA-DRB101:01/DRB101:01 was specified. One dose of antigen injection was delivered to the host at simulation time step 4 with the delivered antigen agent specified as a virus. For the virulence parameter, a multiplication factor of 0.2, infectivity 0.6, and a maximum number of viral particles of 100 were specified. The amount of antigen to be delivered was 1000 units^39,65.

Codon optimisation and in silico cloning

Codon optimization is essential for protein expression. This optimization was performed using the Java Codon Adaptation (JCat: http://www.jcat.de/) web server to maximize gene expression using the Escherichia coli codon system. The codon adaptation index (CAI) and the GC content of the constructed vaccine are crucial factors for assessing gene expression levels. Optimal CAI and GC values are between 0.8 and 1.0 and between 30% and 70%, respectively⁶⁶. The codon-optimized nucleotide sequence was checked for RNA secondary structure using the RNAfold server (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi). The optimized gene inserted into a pET-28a(+)-ring-shaped using SnapGene^51,67,68.

Results

Allergenicity, toxicity and antigenicity profile of HCMV proteins

Table 1 presents the allergenicity, antigenicity, and toxicity scores of various human cytomegalovirus proteins. According to this table, the B envelope glycoprotein has an allergenicity score of 0.14 (non-allergenic), a toxicity score of 0.49 (non-toxic), an antigenicity score of 0.5353 (antigen), and a neurotoxicity score of 0.447 (non-toxic); the UL96 protein has scores of 0.23 (non-allergenic), 0.52 (non-toxic), 0.4055 (antigen), and 0.03 (non-toxic); the UL7 protein, related to CEACAM1, has scores of 0.06 (non-allergenic), 0.30 (non-toxic), 0.3165 (non-antigen), and 0.18 (non-toxic). Alkaline nuclease has the following scores: 0.01 (non-allergenic), 0.14 (non-toxic), 0.3935 (non-antigenic) and 0.165 (non-toxic); the uncharacterized protein UL13 has the following scores: 0.14 (allergenic), 0.28 (non-toxic), 0.4226 (antigenic) and 0.293 (non-toxic); UL10 has the following scores: 0.28 (non-allergenic), 0.28 (non-toxic), 0.4607 (antigenic) and 0.076 (non-toxic); US11 has the following scores: 0.08 (non-allergenic), 0.19 (non-toxic), 0.3960 (non-antigenic) and 0.211 (non-toxic). Envelope glycoprotein H has the following scores: 0.04 (non-allergenic), −0.28 (non-toxin), 0.4522 (non-antigen) and 0.618 (neurotoxin); envelope glycoprotein L has the following scores: 0.08 (non-allergenic), 0.32 (non-toxin), 0.3492 (non-antigen) and 0.063 (non-toxic); envelope glycoprotein O has the following scores: 0.18 (non-allergenic), −0.35 (non-toxin), 0.3949 (non-antigen) and 0.242 (non-toxic); and glycoprotein UL22A has the following scores: 0.07 (non-allergenic), 0.29 (non-toxin), 0.3835 (non-antigen) and 0.618 (neurotoxin). It thus appears that the envelope glycoprotein B is the most promising vaccine candidate due to its highest antigenicity combined with favorable non-allergenic, non-toxic and non-neurotoxic properties.

Table 1.

Allergenicity, toxicity and antigenicity profile of HCMV proteins.

Proteins	Allergenicity		Toxicity		Antigenicity		Neurotoxin
	Score	Prediction	Score	Prediction	Score	Prediction	Score	Prediction
Envelope glycoprotein B	0.14	Non-allergen	0.49	Non-toxin	0.5353	Antigen	0.447	Non-Toxic
Protein UL96	0.23	Non-allergen	0.52	Non-toxin	0.4055	Antigen	0.035	Non-Toxic
CEACAM1-like protein UL7	0.06	Non-allergen	0.3	Non-toxin	0.3165	Non-antigen	0.18	Non-Toxic
Alkaline nuclease	0.01	Non-allergen	0.14	Non-toxin	0.3935	Non-antigen	0.165	Non-Toxic
Uncharacterized protein UL13	0.14	Allergen	0.28	Non-toxin	0.4226	Antigen	0.293	Non-Toxic
UL10	0.28	Non-allergen	0.28	Non-toxin	0.4607	Antigen	0.076	Non-Toxic
US11	0.08	Non-allergen	0.19	Non-toxin	0.3960	Non-antigen	0.211	Non-Toxic
Envelope glycoprotein H	0.04	Non-allergen	−0.28	Non-toxin	0.4522	Non-antigen	0.618	Neurotoxic
Envelope glycoprotein L	0.08	Non-allergen	0.32	Non-toxin	0.3492	Non-antigen	0.063	Non-Toxic
Envelope glycoprotein O	0.18	Non-allergen	−0.35	Non-toxin	0.3949	Non-antigen	0.242	Non-Toxic
Glycoprotein UL22A	0.07	Non-allergen	0.29	Non-toxin	0.3835	Non-antigen	0.618	Neurotoxic

Transmembrane domains and protein topology

Figure 2 presents the transmembrane domains and protein topology of HCMV selected proteins. According to this figure, the majority of envelope glycoprotein B is predicted to be exposed to the extracellular environment for most of its length (position 30 to position 720); with the probability of being exposed to the extracellular environment consistently greater than 0.900. The C-terminal end of the protein contains a region that is predicted to contain a single transmembrane helix (approximately residues 730–770). Unlike UL7, which has type I membrane protein topology, UL7 has the greatest degree of predicted transmembrane region at its N-terminal end (approximately residues 15–40). While alkaline nuclease has very few predicted significant transmembrane helices across its entire length, the probability of being transmembrane and intracellular for alkaline nuclease was also consistently close to 0.000, while it showed consistently high probabilities for being exposed to the extracellular environment (about 0.990). The UL13 protein has a predominantly non-transmembrane topology. Overall, UL13 has very low probabilities of being transmembrane, and no evidence exists for stable transmembrane helices at any position throughout the protein; however, there is a very small transmembrane signal at the N-terminal region.

Fig. 2.

Transmembrane domains and protein topology.

US11 glycoprotein exhibited topological features different from other contributions for the proteins considered in this study, as it appeared to be a membrane-associated viral glycoprotein. Almost all of the predicted sequence of US11 was found to be located outside the membrane (extracellular) region. The highest probability of being an extra membrane protein was found at the N-terminus and mid-region of the protein. UL10 glycoprotein exhibited a well-defined membrane-associated topology, especially at the C-terminus of the protein. A transmembrane region was identified with a high likelihood of being transmembrane (approximately amino acids 205–225). H envelope glycoprotein showed a typical topology for conserved glycoproteins known to be associated with a herpesvirus fusion process and was predicted to remain outside the membrane (extracellular) for almost its entire length, but had a probability of being an extra membrane protein approaching one for all segment of the protein. A TMHMM analysis suggested that UL22A glycoprotein was composed of a single transmembrane domain located in the N-terminal portion of the protein. Amino acid residues corresponding to positions 5–25 had the highest probabilities of being transmembrane, while the balance of the protein was expected to be largely extracellular. However, it can be noted that a transmembrane portion of the HCMV gB located between residues 750–800 can be difficult to extract and purify; which can be overcome by producing recombinant soluble versions of the glycoprotein or by using modern vaccine platforms and mRNA vaccines^62,63. Removal of the hydrophobic domain would also facilitate the extraction, purification and stability of the vaccine protein. Recombinant expression systems would also make it possible to conserve important immunogenic epitopes to induce an effective immune response.

Selected protein features and scores

According to the suggested method of classification (antigenicity: 4 points; allergenicity: 2 points; toxicity/neurotoxicity: 2 points; extracellular location: 2 points; highest possible score: 10 points), the HCMV proteins may be prioritized in the following manner (Table 2):

Table 2.

Selected protein features and scores (by our scoring system).

Rank	Protein	Antigenicity	Allergenicity	Toxicity/neurotoxicity	Topology	Total score (/10)
1	Envelope glycoprotein B	4	2	2	2	10
2	UL10	3	2	2	2	9
3	UL96	2	2	2	2	8
4	Envelope glycoprotein H	3	2	0	2	7
5	US11	1	2	2	2	7
6	Glycoprotein O	1	2	2	2	7
7	Alkaline nuclease	1	2	2	2	7
8	Glycoprotein L	0	2	2	2	6
9	UL13	2	0	2	1	5
10	UL22A	1	2	0	2	5
11	UL7	0	2	2	1	5

Structure and functional role of HCMV gB

Table S2 presents the order of the envelope glycoprotein B (gB) of HCMV. This protein is composed of 906 amino acids. The table shows regions rich in serine and threonine residues. One end of the molecule includes a hydrophobic segment that traps the protein in the outer layer, followed by a bond with a small inner segment that determines its movement and positioning within the viral particle.

Biochemical characteristics and stability of GB_HCMVA

Table S3 presents the composition and stability of GB_HCMVA. The protein is robust due to its uniform chemical structure. Composed of 906 amino acids, its mass is 102,003.89 daltons. Its pH is close to 6.86, indicating slight acidity. Its complete formula is C₄₅₁₂H₇₀₂₇N₁₂₄₁O₁₃₉₃S₃₃, for a total of 14,206 atoms. Several polar groups are present: serine (9.6%), threonine (7.8%), and tyrosine (5.6%). These groups promote protein hydrolysis, which enhances its affinity for surfaces. Valine forms strong bonds with leucine and isoleucine, giving the structure good heat resistance. Cysteine is present in small quantities (1.8%), while methionine is slightly more abundant. Disulfide bonds are therefore infrequent. The figures attest to its durability: one measurement, at 37.07, is below the typical degradation risk threshold. The value of 80.36 indicates a high density of side chains, a sign of a rigid internal structure. Water absorption is low (−0.326), moderate, and distinct, with a strong affinity for moist environments. A key wavelength exhibits light absorption of approximately 120,000 per molar centimeter, resulting in accurate measurements without stress. Its lifespan is a full day outside cells, over twenty hours when transported by yeast, and over ten hours inside the cell walls of E. coli. Designed for robustness, it withstands shocks and demanding environments.

Protein structure prediction

Tertiary structure prediction of GB_HCMV

Tertiary structure prediction of GB_HCMV

Figure 3 shows the ramachandran plot of the gB protein. This figure reveals that 83.1% of the residues are in the favorable region. This trend indicates the structural model is well-folded, stable, and highly conserved.

Fig. 3.

Ramachandran diagram of GB_HCMV.

Analysis of favorable and critical conformations of the vaccine

Figure 4 presents the 3D structure of the gB protein. This figure shows a heat map indicating the position of potentially misaligned atoms. It is dark green at zero and light green around thirty, mapping the stability of each point during alignment. The shape prediction is represented by a ribbon, the color of which is determined not randomly, but by the level of confidence in the prediction. Dark blue indicates near certainty (greater than 90 on a specific scale constructed for this object). Light blues correspond to values between 70 and 90, and we are relatively confident in what we observe. Yellow appears when values reach 50 to 70, meaning the situation becomes unclear. Orange appears at values between 50 and 50, which is difficult to interpret.

Fig. 4.

Visualization of the predicted three-dimensional structure of the protein GB_HCMVA.

Secondary structure prediction of GB_HCMV

A two-dimensional structural analysis was performed to visualize the organization of the selected glycoprotein antigen B. The resulting structure is illustrated in Fig. 5. It reveals a pattern typical of membrane or extracellular proteins. At the N-terminus, a sequence of leucine, valine, isoleucine, and cysteine is observed, likely highly hydrophobic perhaps a signal or a bridge across the lipid bilayer. Yellow areas indicate the formation of neutral amino acid clusters, suggesting the possibility of alpha helix formation. Proline-rich and polar regions exhibit pink areas, indicating loose loops or bends connecting solid areas. Rather than following one another, hydrophobic and hydrophilic motifs alternate between the deep beta sheets. Based on the visual distribution of the secondary structure prediction, the protein appears to contain predominantly alpha helices and beta sheets, with smaller proportions of random coils/turns and other elements. The approximate composition is summarized in Table 3.

Fig. 5.

Secondary structure prediction of GB_HCMV.

Table 3.

Helice composition of GB_HCMV.

Secondary structure element	Approximate percentage
Alpha helices	40%
Beta sheets (extended strands)	35%
Random coils/loops	20%
Beta turns and other structures	5%

Disulfide bridges and structural stabilization of gB of HCMV

Table S4 illustrates how disulfide bridges maintain the structure of the HCMV gB protein. The chain contains six cysteines, forming three potentially strong bonds. It is strongly predicted that there are only two pairs of bonds: the first links PKFTKCRSPER to RETFSCHWTIQ (score of 0.995444), and the second links IWIPYCIKLTS to TVDEKCFSVDE (0.99584). These near-perfect scores indicate the reliability of these relationships. All other matches are negligible, with weights ranging from 0.01 to 0.3, which confirms the hypothesis that only these three bonds are dominant.

Phylogeny tree

Figure 6 presents the phylogeny tree of gB sequences. According to this sequences, ACS92156.1, ALJ56180.1 and 5CXF_A formed a strongly cohesive, closely related cluster of sequences, supported by a bootstrap value of 81–82%, suggesting that they are highly homologous and share a close evolutionary relationship. Similarly, the gB sequences of XOL08432.1 and XUA97349.1 also formed a closely related pair of gB sequences with very high bootstrap support (99%), suggesting that these gB proteins are highly homologous and are likely derived from closely related viral strains. Despite differences in length, the grouping of AAA45932.1, AAA45927.1 and AJF36318.1 produces a strong evolutionary relationship, confirming gB proteins, as most of these entries represent partial proteins. The gB sequences of YCN21674.1 and YCN21491.1 also form a unique and relatively divergent branch of gB sequences, which is separate from the main HCMV cluster. Consequently, these gB proteins, as their annotation of Herpesviridae sp. suggests, may belong to more distantly related lineages within the herpesviruses or represent highly divergent variants of gB proteins.

Fig. 6.

Phylogeny tree of gB sequences.

Identification of HCMV gB B-cell epitopes

Table S5 presents the B-cell predicted epitopes of HCMV gB. This table reveals the presence of 31 potential straight segments, detectable by B lymphocytes and distributed along the chain (Table S5). Extending towards the C-terminus, two segments protrude: one between 792 and 843, the other between 856 and 882, the latter being longer than the majority. Smaller fragments such as TEEC (607–610), PLEN (654–657), or ELR (670–672) contain less material but can also serve as anchors and specify antibody binding specificity. The overall minimum score is 0.029, but a peak of 3.097 is observed, indicating areas of intense interest where immune attention is particularly focused, notably on the C-terminus and on serine- and threonine-rich regions of the protein.

Prediction of CD8⁺ epitopes of gB from HCMV

Table S6 presents the CD8 + HCMV gB epitopes. The table shows 20 target epitopes, with a majority of scores exceeding 0.97. The significance of high scores is reaffirmed, suggesting strong adhesion. The RVYQKVLTF peptide exhibits a near-perfect score of 0.994745 for HLA-A32:01. It is followed by FANSSYVQY, with a score of 0.990261 for HLA-B35:01.

Prediction of gB CD4⁺ epitopes for MHC-II

Table S7 presents the predicted CD4 + cell specific HCMV gB epitopes. Twenty-five epitopes were predicted, with a uniform size of 15. This table indicates that scores below 0.1 are particularly significant when they are high at the other end of the spectrum, notably in association with variants such as HLA-DRB115:01 or in combination with pairs like HLA-DQA105:01 and DQB103:01. The scores for these epitopes range from 0.2003 to 0.9446, with a trend toward 0.8, suggesting a specific action of these epitopes on CD4 + cells.

Immunogenic profile and safety of gB MHC-I, MHC-II, and B-cell epitopes

Table 4 presents the evaluation of MHC-I epitopes using various biological and immunological parameters, such as toxicity, allergenic potential, receptor recognition, and the ability to elicit an immune response. The majority of epitopes are non-toxic, although some, such as SEAEDSYHF, NTDFRVLELY, and TSGGLVVFW, are. Others, such as NTDFRVLELY and ETIYNTTLK, are not, while some, such as FANSSYVQY and YVAPPMWEI, elicit allergic reactions. Regarding their detection by cells, the probability is estimated as zero or variable, meaning there is a very low probability of activation of the integrated receptors. Their ability to elicit an immune response appears to be classified as “immunogenic,” with a specific probability score. The epitopes SEAEDSYHF, YVAPPMWEI, and TSGGLVVFW are indicators of an immune response with a probability of 66%. RVYQKVLTF and FANSSYVQY both have a rate of 100%, but are considered non-immunogenic. Table 5 presents an evaluation of the toxicity, immunogenicity, allergenicity, and receptor recognition profile of certain B epitopes. The sequences IQDEKN, ERTIRSEAEDSYHF, and YNQTYEKY show positive results: no signs of danger, no allergic reaction, not recognized by certain receptors, but perfectly capable of triggering an immune response. In summary, this table shows that the epitopes GVNTTKY, IQDEKN, GAAGKA, and ADGTTVTSGSTKDTSLQAPPSYEESV are non-allergenic, non-toxic, and immunogenic, with scores ranging from 0.51 to 0.71, 0.28 to 0.51, and 66% to 100% for toxicity, allergenicity, and immunogenicity, respectively. Table 6 presents an evaluation of the toxicity, immunogenicity, allergenicity, and receptor recognition profile of CD4⁺ epitopes. The sequences YDTLRGYINRALAQI, AIGAVGGAVASVVEG, and KNGYRHLKDSDEEEN exhibit allergenicity. Most epitopes show low immunogenic potential, with the exception of certain sequences, such as VDQRRTLEVFKELSK, which induce immune responses. Only the epitope sequences VDQRRTLEVFKELSK, KNGYRHLKDSDEEEN, QLTFWEASERTIRSE, and DQRRTLEVFKELSKI showed no toxicity or allergenicity but demonstrated immunogenicity.

Table 4.

Toxicity and immunogenicity profile of MHC-I epitopes.

Epitopes MHC-I	Toxicity			Allergenicity			Pattern/non-recognition receptor			Immunogen
	ML score	Hybrid score	Prediction	ML score	Hybrid score	Prediction	Score	Hybrid-score	Prediction	Yes or no	Probability
SEAEDSYHF	0.61	0.61	Non-toxin	0.36	0.36	Non-allergen	0.34	0.34	Non-	Yes	66%
YVAPPMWEI	0.57	0.57	Non-toxin	0.33	0.33	Non-allergen	0.4	0.4	Non	Yes	66%
TSGGLVVFW	0.62	0.62	Non-toxin	0.31	0.31	Non-allergen	0.49	0.49	Non	Yes	66%
ETIYNTTLK	0.62	0.62	Non-toxin	0.24	0.24	Non-allergen	0.25	0.25	Non	Yes	66%
YVAPPMWEI	0.57	0.57	Non-toxin	0.33	0.33	Non-allergen	0.4	0.4	Non	Yes	66%
TSGGLVVFW	0.62	0.62	Non-toxin	0.31	0.31	Non-allergen	0.49	0.49	Non	Yes	66%
SEAEDSYHF	0.61	0.61	Non-toxin	0.36	0.36	Non-allergen	0.34	0.34	Non	Yes	66%
FTYDTLRGY	0.62	0.62	Non-toxin	0.32	0.32	Non-allergen	0.43	0.43	Non	Yes	66%
DVVGVNTTK	0.65	0.65	Non-toxin	0.36	0.36	Non-allergen	0.42	0.42	Non	Yes	66%
DVVGVNTTKY	0.62	0.62	Non-toxin	0.34	0.34	Non-allergen	0.42	0.42	Non	Yes	66%
ETIYNTTLKY	0.62	0.62	Non-toxin	0.25	0.25	Non-allergen	0.24	0.24	Non	Yes	66%

Table 5.

Toxicity and immunogenicity profile of B-cell epitopes.

B Cell Epitope	Toxicity			Allergenicity			Pattern/non-recognition receptor			Immunogen
	ML score	Hybrid score	Prediction	ML score	Hybrid score	Prediction	Score	Hybrid-score	Prediction	Yes or no	Probability
GVNTTKY	0.61	0.61	Non-toxin	0.34	0.34	Non-allergen	0.52	0.52	Pattern	Yes	66%
IQDEKN	0.6	0.6	Non-toxin	0.28	0.28	Non-allergen	0.36	0.36	No	Yes	100%
GAAGKA	0.71	0.71	Non-toxin	0.51	0.51	Non-allergen	0.56	0.56	Pattern	Yes	100%
ADGTTVTSGSTKDTSL QAPPSYEESV	0.51	0.51	Non-toxin	0.34	0.34	Non-allergen	0.31	0.31	No	Yes	66%

Table 6.

Toxicity and immunogenicity profile of MHC-II epitopes.

Epitopes MHC-II	Toxicity			Allergenicity			Pattern/non-recognition receptor			Immunogen
	ML score	Hybrid score	Prediction	ML score	Hybrid score	Prediction	Score	Hybrid-score	Prediction	Yes or no	Probability
VDQRRTLEVFKELSK	0.66	0.66	Non-toxin	0.35	0.35	Non-allergen	0.45	0.45	No	Yes	66%
KNGYRHLKDSDEEEN	0.61	0.61	Non-toxin	0.29	0.29	Non-allergen	0.41	0.41	No	Yes	66%
QLTFWEASERTIRSE	0.59	0.59	Non-toxin	0.28	0.28	Non-allergen	0.51	0.51	Pattern	Yes	66%
DQRRTLEVFKELSKI	0.63	0.63	Non-toxin	0.34	0.34	Non-allergen	0.48	0.48	No	Yes	66%

Modular design of a multi-epitope vaccine

Figure 7 shows the sequence of the vaccine constructed from the selected epitopes. This figure illustrates that the vaccine is a multi-epitope vaccine. It begins with an adjuvant region (in green), which promotes antigen presentation by acting as an immune stimulant. A rigid EAAAK linker (shown in cream) separates the adjuvant from the B-cell epitopes, shown in purple. These epitopes trigger the activation of antibody-mediated immunity and are separated by flexible CPGPG linkers to allow structural accessibility. The golden-yellow boxes then mark the MHC-I epitopes, which are essential for the activation of cytotoxic T lymphocytes (CTLs) and separated by AAY linkers that enhance proteasomal processing. The construct is then completed with blue MHC-II epitopes, which activate helper T lymphocytes and lead to long-term immunity. Expression is completed by a red 6x His tag, which allows for easy purification during protein production.

Fig. 7.

Modular design of a multi-epitope vaccine. Adjuvant = 50 s ribosomal46 PROTEIN L7/L12.

Epitope concervancy of vaccine

The epitope conservancy analysis as shown in the Table 7. This table indicates the presence of considerable heterogeneity in terms of the level of conservancy found within the different sequences. The longer epitope ADGTTVTSGSTKDTSLQAPPSYEESV do not appear to have matches among the different proteins, and they tend to possess fairly low maximum identities (often less than 33%). This implies great variation and reduced chances of developing cross-immunity in the population. On the other hand, there are several smaller epitopes (from 6 to 15 amino acid lengths), GVNTTKY, IQDEKN, GAAGKA, SEAEDSYHF, YVAPPMWEI, ETIYNTTLK, FTYDTLRGY, and DVVGVNTTKY, that seem highly conserved (over 60% of the proteins tested, identity percentage as high as 100%), which is highly encouraging for vaccine or antibody development efforts.

Table 7.

Epitope concervancy of predicted vaccine.

Epitope	Epitope sequence	Epitope length	Percent of protein sequence matches at identity ≤ 100%	Minimum identity	Maximum identity
EP1	GVNTTKY	7	60.00% (3/5)	42.86%	100.00%
EP2	IQDEKN	6	60.00% (3/5)	50.00%	100.00%
EP3	GAAGKA	6	60.00% (3/5)	33.33%	100.00%
EP4	ADGTTVTSGSTKDTSLQAPPSYEESV	26	40.00% (2/5)	26.92%	100.00%
EP5	SEAEDSYHF	9	60.00% (3/5)	44.44%	100.00%
EP6	YVAPPMWEI	9	60.00% (3/5)	33.33%	100.00%
EP7	TSGGLVVFW	9	20.00% (1/5)	33.33%	100.00%
EP8	ETIYNTTLK	9	60.00% (3/5)	44.44%	100.00%
EP9	YVAPPMWEI	9	60.00% (3/5)	33.33%	100.00%
EP10	TSGGLVVFW	9	20.00% (1/5)	33.33%	100.00%
EP11	SEAEDSYHF	9	60.00% (3/5)	44.44%	100.00%
EP12	FTYDTLRGY	9	60.00% (3/5)	33.33%	100.00%
EP13	DVVGVNTTK	9	60.00% (3/5)	44.44%	100.00%
EP14	DVVGVNTTKY	10	60.00% (3/5)	40.00%	100.00%
EP15	ETIYNTTLKY	10	60.00% (3/5)	40.00%	100.00%
EP16	VDQRRTLEVFKELSK	15	60.00% (3/5)	40.00%	100.00%
EP17	KNGYRHLKDSDEEEN	15	60.00% (3/5)	26.67%	100.00%
EP18	QLTFWEASERTIRSE	15	60.00% (3/5)	26.67%	100.00%
EP19	DQRRTLEVFKELSKI	15	60.00% (3/5)	40.00%	100.00%

Epitope cluster of constructed vaccine

Data from the clustering analysis of the epitopes present in the engineered vaccine are shown in Table 8. The figure reveals the number of clusters found to have 60% similarity among 23 epitopes is 15. The four groups can be considered as conserved, whereas the rest of the epitopes form individual clusters. The conserved sequences can be the immunogenic regions and include DVVGVNTTKY and VDQRTLEVFKELSKI.

Table 8.

Epitope cluster of constructed vaccine.

Cluster.sub-cluster number	Peptide number	Alignment	Position	Description	Peptide
1.1	Consensus	KPLLKXXXXXXXXAAAKXXXAAGAAPAAGAAEA AEEQSEFDVILEAAGDKKI	–	–	–
1.1	1	KPLLEKVAKAEAEDAEAKKLEAAGA	1	seq5	KPLLEKVAKAEAEDAEAK KLEAAGA
1.1	2	---KKFETFVTAAAPVAVAAAGAAPA-	5	seq2	KKFETFVTAAAPVAVAA AGAAPA
1.1	3	-----------GAAGKA-----------	14	seq10	GAAGKA
1.1	4	--------------EAAAK---------	14	seq7	EAAAK
1.1	5	----------------------GAAVEAAEEQSEFDVILEAAGDKKI	30	seq3	GAAVEAAEEQSEFDVILEAA GDKKI
2.1	Consensus	DWGVNTTKY	–	–	–
2.1	1	DWGVNTTKY	1	seq22	DWGVNTTKY
2.1	2	DWGVNTTK-	1	seq21	DWGVNTTK
2.1	3	---GVNTTKY	4	seq8	GVNTTKY
3.1	Consensus	VDQRRTLEVFKELSKI	–	–	–
3.1	1	VDQRRTLEVFKELSK-	1	seq25	VDQRRTLEVFKELSK
3.1	2	-DQRRTLEVFKELSKI	2	seq28	DQRRTLEVFKELSKI
4.1	Consensus	ETIYNTTLKY	–	–	–
4.1	1	ETIYNTTLKY	1	seq23	ETIYNTTLKY
4.1	2	ETIYNTTLK-	1	seq16	ETIYNTTLK
5.1	Singleton	MAKLSTDELLDAFKEMTLLLSDDFV	–	seq1	MAKLSTDELLDAFKEMTL LLSDDFV
6.1	Singleton	SEAEDSYHF	–	seq13, seq19	SEAEDSYHF
7.1	Singleton	KNGYRHLDKSDEEEEN	–	seq26	KNGYRHLDKSDEEEEN
8.1	Singleton	QLTWEASERTIRSE	–	seq27	QLTWEASERTIRSE
9.1	Singleton	CPGPG	–	seq12	CPGPG
10.1	Singleton	GVKIVWREIVSGLGKEAKDLVGAP	–	seq4	GVKIVWREIVSGLGKEAKD LVGAP
11.1	Singleton	IQDEKN	–	seq9	IQDEKN
12.1	Singleton	TSGGLVFW	–	seq15, seq18	TSGGLVFW
13.1	Singleton	YVAPPMWEI	–	seq14, seq17	YVAPPMWEI
14.1	Singleton	FTYDTLRGY	–	seq20	FTYDTLRGY
15.1	Singleton	HHHHH	–	seq29	HHHHH

Predicted discontinuous epitopes of constructed vaccine

Discontinuous epitopes of the constructed vaccine are provided in Table 9. The prediction of the discontinuous epitopes provides information concerning three structurally different areas, each with diverse immunological scores. Epitope 1 is made up of 23 residues and has a higher score of 0.86. This score implies that there is a higher probability of the epitope being recognized. Epitope 2 is relatively shorter than epitope 1; it comprises 15 residues and a medium score of 0.712. Epitope 3 is relatively long (51 residues) and has a low score of 0.641.

Table 9.

Predicted discontinuous epitopes.

No.	Residues	Number of residues	Score	3D structure
3	A: D177, A: G178, A: T179, A: T180, A: D188, A: T189, A: S190, A: L191, A: A193, A: P194 A: P195, A: S196, A: E198, A: E199, A: S200, A: V201, A: C202, A: P203, A: P205, A: G206, A: A209, A: E210, A: D211 A: S212, A: Y213, A: H214, A: F215, A: T225, A: S226, A: G227, A: G228, A: L229, A: E250, A: I251, A: T252, A: S253, A: G254 A: G255, A: L256, A: V257, A: V258, A: F259, A: W260, A: S261, A: E262, A: A263, A: E264, A: D265, A: S266, A: Y267, A: H268	51	0.641
2	A: Y216, A: V217, A: A218, A: Y237, A: N238, A: T239, A: T240, A: L241, A: K242 A: Y243, A: V244, A: A245, A: P246, A: P247, A: W249	15	0.712
1	A: D312, A: Q313, A: R314, A: R315, A: T316, A: L317, A: E318, A: V319, A: F320 A: K321, A: E322, A: L323, A: S324, A: K325, A: K326, A: N327, A: G328, A: Y329, A: R330, A: H331, A: L332, A: K333, A: D334	23	0.86

Molecular profile and stability of the multi-epitope vaccine

Table S11 presents the physicochemical properties of the designed vaccine. According to this figure, the designed vaccine is composed of 225 amino acids. Its average molecular mass is 25,512.14 daltons. Its isoelectric point (pI) is 5.28, indicating a tendency towards acidity. Its structure contains only 3,521 atoms. Its chemical formula is C₁₁₄₀H₁₇₁₆N₃₀₀O₃₆₂S₃. Its water solubility is excellent due to the presence of polar groups. Glutamate is present in a large quantity (9.8%). Threonine represents 11.1% of the composition, while serine is present at 7.1%. Valine is present at 8%, and leucine at 5.3%. However, cysteine is present in very small quantities (0.4), and disulfide bond formation is difficult. Stability remains significant, but it is short-lived: less than one hour in cultured mammalian cells. The instability index, at 35.66, indicates good stress resistance. With an aliphatic absorption coefficient of 62.80, heat resistance is satisfactory. Attraction to water is high, as indicated by the GRAVY value of −0.672, which suggests easy dissolution.

Tertiary structure predicted for HCMV-vaccine

Analysis of favorable and critical conformations of the vaccine

Figure 8 shows the Ramachandran plot of vaccine. The Ramachandran plot analysis of HCMV gB vaccine showed that there was a good stereochemical quality and stability of the structure. In total, 119 residues (79.9%) were situated in most favored regions, 20 residues (13.4%) and 6 residues (4.0%) were in additionally and generously allowed regions, correspondingly. Only 4 residues (2.7%) resided in disallowed regions; hence, the amount of unfavorable conformations was low. Residues predominantly occupied typical α-helix and β-sheet regions; thus, it is evident that predicted secondary structure elements are correctly folded. Even though the percentage of residues in most favored regions was somewhat less than the optimal 90% that is usually accepted for protein high-resolution models, the residue distribution indicates that the vaccine construct has reliable three-dimensional conformation.

Fig. 8.

Ramachandran plot of the GB_HCMV constructed vaccine.

Predicted three-dimensional architecture of the vaccine

Figure 9 shows the 3D structure of the constructed vaccine. The 3D structure of the vaccin as predicted shows an overall reliable folding with a high-confidence structured core (pLDDT > 90) identified in the dark blue parts, and the ends, as well as certain loops display increased flexibility or intrinsic disorder (pLDDT < 70), as the segments colored with yellow and orange indicate; this distribution is supporting the position error map where low error values (dark green areas) indicate the goodness of the model in the central areas, implying a good functional architecture.

Fig. 9.

Predicted three-dimensional architecture of the vaccine.

Secondary structure predicted for the HCMV vaccine

Figure 10 presents the 2D structure of the constructed vaccine sequence. This figure shows that the sequence contains hydrophobic regions, represented in green in the lower part of the map. These are visible at positions 14–18, 42–44, 150–160, 240–260, and 330–345. These segments are also surrounded by orange areas, indicating the presence of small, nonpolar amino acids, which highlights their low hydrophilicity. Cysteine residues, recognizable by their light blue color, are few in number but uniformly distributed. They are found mainly in the 190–205 region, where they appear as isolated dots surrounded by larger clusters of orange and green residues. Other groups of cysteines are located between positions 245 and 255, and the last one at the end of the sequence, between positions 310 and 320. The α helices, marked in red, extend over long distances, notably between positions 10 and 40, 100 and 150, 170 and 200, and finally between positions 300 and 360 (Table 10).

Fig. 10.

Secondary structure predicted of HCMV-vaccine.

Table 10.

Helice composition of constructed vaccine.

Secondary structure element	Approximate percentage
Alpha helices	48%
Beta sheets (extended strands)	18%
Random coils/loops	28%
Beta turns and other structures	6%

N-glycosylation sites in the vaccine sequence

Figure 11 shows the probability of N-glycosylation sites in a protein chain of amino acids. On the bottom, there are numbers, starting with zero up to three hundred and fifty, indicating the location of each unit. The side scale ranges between one and zero five points, which indicates the magnitude of strength of the signal. On the half point of zero the line is a deep violet colouring transversely. Green marks that are tall are erected at some points along the line and indicate where sugar tags can be fixed. When one of those marks exceeds the cutoff it is counted as a probable attachment point. Beyond that point you can see several of the high points, here and there. Their positioning is not even that, there is more of them in certain areas and little in others. The peaks all denote the points at which the probability of tagging surges above 50%. Observing potential N-glycosylation sites, five correspond to the common N-X-S/T motif X, but none of them contains proline of this amino acid in 159–302. The motifs are not repeated using and instead NTTK and NTTL are paired together, their scores ranging between 0.5496 and 0.7004, suggesting the two motifs are probably altered.

Fig. 11.

Predicted N-glycosylation sites in the vaccine sequence.

Molecular docking results

Vaccine interaction with TLR7, TLR8, and TLR9

Table 11 and Fig. 12 present the binding affinity between the vaccine and various TLR proteins, along with a three-dimensional model of the complexes formed. The highest correlation is observed between the vaccine and TLR7 (docking score of −323.06 kcal/mol). The next highest correlations, between the vaccine and TLR8 and TLR9, are − 294.92 kcal/mol and − 221.82 kcal/mol, respectively. This trend suggests that the vaccine has a higher affinity for TLR7 and TLR8. This is confirmed by the confidence levels: 0.9696 for TLR7 and 0.9478 for TLR8, which are satisfactory values, while 0.8079 for TLR9 indicates lower certainty for the model.

Table 11.

Binding energies between vaccine and TLR proteins.

	Vaccine-TLR9	Vaccine-TLR8	Vaccine-TLR7
Docking score	− 221.82	− 294.92	− 323.06
Confidence score	0.8079	0.9478	0.9696
Ligand rmsd (Å)	59.70	283.90	309.74

Fig. 12.

3D docked complexes between the constructed vaccine and TLR proteins.

Mapping of H-bonding and hydrophobic interactions in vaccine–TLR complexes

Figure 13 illustrates the connectivity of amino acids from two parts of the protein, called chains A and B, with TLR7 where they bind. The segments of chain A appear in blue; those of chain B, in green. Clear connections are observed: between Gln162A and Asp161B, and between Lys175 and Asn176A, on the one hand, and Glu167B and Glu171B, on the other. Another bond forms when A attempts to establish a bond between Arg179 and Ser184 on the one hand, and Arg164 of B on the other. A bond is also observed between Lys136 of A and Val132 of B. In the interaction between the vaccine and TLR9, the important amino acids are linked by hydrogen bonds, illustrated by green dashed lines. This includes not only the Gln162(A)-Asp161(B) bond, but also contacts such as Lys175(A), Asn176(A) with Glu167(B) and Glu171(B). Arg179(A) and Ser184(A) are then observed to be bonded to Arg164(B), and His117(A) to Tyr116(B). Lys136(A) is bonded to Val132(B), and Glu99(A) to Tyr127(B). Each bond reinforces the structure and holds it firmly in place. Simultaneously, hydrophobic contacts are highlighted by the red curves. The side chains of Leu169(A), Phe118(A), Leu124(A), Tyr121(A), and Leu172(A) are closely associated with those of Val166(B). The interaction between TLR8 and the vaccine follows two angular directions. Blue represents the vaccine’s A chain; green, that of TLR8 (chain B), linked by hydrogen bonds that form the structure. The A chain is bonded to Arg569, Ala263, and Cys260, while Pro264 is oriented toward Ala571 and Ile570. Asp628 from A binds to Lys185 from B, creating additional anchoring points. A twist is added: the bottom figure is symmetrical. Cys260 and Ala263 from A bind to Arg569, Ile570, and Ala571 from B, demonstrating a balance of binding strength.

Fig. 13.

Mapping of H-bonding and hydrophobic interactions in vaccine–TLR complexes.

Normal mode anlysis findings of vaccine–TLR complexes

Figure 14 presents the normal mode analysis (NMA) of the vaccine-TLR complexes. This analysis revealed distinct variations in deformability, residue correlation, and factor B profiles for the vaccine-TLR9 and vaccine-TLR8 systems. In the vaccine-TLRα complex, deformability values remained generally low for most residues, fluctuating mainly between 0.0 and 0.4, with a major peak near 1.0 around residue indices 250–300 and smaller fluctuations near the terminal regions. The covariance matrix revealed alternating correlated and anticorrelated movements of moderate intensity, distributed across residual indices up to approximately 1100. The B factor profile showed NMA values primarily between 0.1 and 0.5, whereas experimental B factors from PDB fluctuated more broadly between 0.2 and 1.0, with several peaks observed between residual indices 400 and 900. In the Vaccine-TLR8 complex, deformability values were relatively low in the first half of the structure, remaining mostly below 0.3 up to residual index 1200, then gradually increasing with peaks reaching approximately 0.8–0.9 between residual indices 1500 and 1850. The covariance matrix extended to approximately 1900 residuals and showed more extensive regions of correlated movements than in the complex TLR9. The B factor graph indicated that the NMA-derived values remained mainly between 0.1 and 0.4 on most of the residuals. The relatively low deformability values, moderate residual fluctuations, and stable distributions of factor B observed for the Vaccine-TLR9 and Vaccine-TLR8 complexes suggest that the complexes possess acceptable structural stability and maintain a stable interaction profile throughout the simulation analysis. Regarding the NMA for the complex between the vaccine and TLR7, there is observed global stability of the complex structure. There is no large flexibility, since the deformability of the vast majority of residues lies within the range of 0.0 to 0.4, with only a few peaks being found in the range of 0.8–1.0. These represent the hinges responsible for the formation of conformational changes. The positive correlation of covariance motion is high; thus, motions of certain residues are related and correlated in the protein complex. Negative correlations have less frequency than positive correlations, being much weaker. PDB B-factors vary between 0.2 and 1.1, while NMA predicted B-factors lie in the range of 0.2 to 0.5.

Fig. 14.

Normal mode analysis of vaccine–TLR complexes by NMA.

Dynamic simulation findings

Figure 15 presents the dynamic simulation findings of vaccine and TLR proteins. According to this figure, the findings of the molecular dynamics analyses demonstrated differences between the stability and flexibility properties of the three vaccine-TLR complexes. Specifically, the RMSD of the Vaccine-TLR9 complex showed a rapid rise initially followed by stabilization at about 10–11 Å after around 10 ns, suggesting that the complex reached a relative stability structure following some adjustments within its initial structure. The RMSF plot for this complex indicated that there were only slight fluctuation changes among all of the residues, with major peaks appearing around residues 40–60 and 150–180. This suggests that there were flexible loops in the structure while the rest was stable. On the other hand, the analysis of the Vaccine-TLR7 complex showed a steady rise in the RMSD value of around 3.5–12 Å over the whole 100 ns run. This indicates a progressive conformation change process that led to less stability than in the case of TLR9. Also, the high values in the RMSF plot of the Vaccine-TLR7 complex suggested flexible residues, especially those located in the region of 150–500 and 1700–2000. For the Vaccine-TLR8 complex, the lowest RMSD values were obtained, gradually ranging from about 2.0 Å to 4.5 Å. In the case of RMSF analysis for this system, comparatively low fluctuations in residue movement were found throughout the structure, with some increased fluctuations restricted only to the N-terminal end (150–400 residues), which means that there is no large degree of movement in this particular system.

Fig. 15.

Dynamic simulation findings between TLR proteins and constructed vaccine.

Antigenic propensity of vaccine constructed sequence

The Fig. 16 presents the antigenicity of the vaccine along its complete sequence. The figure illustrates the variations in the average antigenicity of the protein sequence. Peaks indicate high antigenicity of some segments, while troughs indicate low antigenicity of other segments. According to this Table, the average value of antigenic potential is 1.0281, with variation across different lengths of the sequence, ranging from 0 to 376. In total, 18 regions of antigenicity have been identified, starting from position 5 with the sequence of amino acids “STDELLD” and ending with position 372 of “RTLEVFKELSKIHH.” Some regions are shorter, like “SGGLVVFW” from position 226 to 233 and from 253 to 260, while others are longer, such as “AKLEAAGATVTVKEAAAKYVWLKSILVAILVFG” (Table 12).

Fig. 16.

Antigenic plot for vaccine sequence.

Table 12.

Antigenic determinants in vaccine sequence.

n	Start position	Sequence	End position
1	5	STDELLD	11
2	17	TLLELSDFVK	26
3	32	FEVTAAAPVAVAAAGAAPAGAAVEA	56
4	61	SEFDVILE	68
5	73	KKIGVIKVVREIVSGLGL	90
6	92	EAKDLVGAPKPLLEKVAK	109
7	117	AKLEAAGATVTVKEAAAKYVWLKSILVAILVFG	149
8	189	TSLQAPPSYEESVCPG	204
9	211	DSYHFYVAP	219
10	226	SGGLVVFW	233
11	237	YNTTLKYVAP	246
12	253	SGGLVVFW	260
13	265	DSYHFFTY	272
14	276	RGYDVVGV	283
15	285	TTKDVVGV	292
16	305	LKYAAYVDQ	313
17	315	RTLEVFK	321
18	359	RTLEVFKELSKIHH	372

Cell kinetics and immune mediators’ post-vaccination

Figure 17A presents the humoral immune response to antigen exposure, showing antigen concentration, immunoglobulin production, and total antibody levels over time: Antigen concentration dropped steeply in the first few days after exposure, IgM concentrations rose early, reaching a peak around day 10 and declining, total antibody levels followed IgM initially, IgG1 and IgG2 responses were delayed and continued to be detected throughout the simulation, and IgG1 levels exceeded IgG2 levels. Figure 17B shows the dynamics of the epithelial cells and APCs, with the epithelial cell population remaining relatively stable over time, infected epithelial cells showing transient increases following antigen exposure, and APC populations increasing progressively during the simulation. The cytokine and interleukin profiles expressed in ng/mL (Fig. 17C) indicate that IFN-α and IFN-γ increased early in the simulation, whereas IL-1, IL-6, and TNF-α were present at high concentrations during the inflammatory phase, and IL-10 and IL-12 increased more gradually over time. The dynamics of total B cells, memory B cells, and immunoglobulin-associated B cell populations are shown in Fig. 17D. The total number of B cells increased gradually over time, memory B cells accumulated gradually, and IgG1-associated B cells were at the highest levels, whereas IgA responses were transient and IgG2-associated cells remained relatively low. The functional states of B cells (active, antigen-presenting, proliferative, quiescent, and class-switching populations) are described in Fig. 17E, which shows that active and antigen-processing B cells increased rapidly at the start of the simulation, followed by an increase in proliferative and antigen-presenting populations, whereas quiescent and class-switching populations remained low throughout.

Fig. 17.

Kinetics of the humoral immune response (A), dynamics of epithelial cells (EP) (B), and evolution of cytokine and interleukin concentrations (C). Dynamics of total B lymphocytes, memory cells and those expressing IgA, IgG1 and IgG2 (D), distribution of B lymphocytes (E), B plasma cells subdivided according to IgA, IgG1 and IgG2 isotypes (F), Dynamics of total and memory CD4 T lymphocytes (G), subdivision of CD4 helper T cells (H), dynamics of regulatory CD4 T lymphocytes (I). Dynamics of total and memory cytotoxic CD8 T cells over time (J), distribution of cytotoxic CD4 T lymphocytes according to their functional state (K), evolution of the total number of NK (Natural Killer) cells over time (L), dynamics of dendritic cells (DCs) (M) macrophage (MA) population subdivided (N).

Figure 17F shows the kinetics of plasma B cells that produce IgA, IgG1, and IgG2, which shows a rapid increase and subsequent decline in plasma cell populations following antigen exposure, with the IgG1-producing plasma cells predominating throughout the simulation. Figure 17G shows the kinetics of total and memory CD4 T cells, which shows an initial increase followed by a decline in total CD4 T cells and a gradual accumulation of memory CD4 T cells. Figure 17H shows the kinetics of CD4 helper T cells, divided into active, proliferative, quiescent, and memory populations, which shows an increase in activated and proliferative populations early in the response and the appearance of quiescent populations in later phases. Figure 17I shows the kinetics of CD4 regulatory T cells, which shows an increase in total and memory regulatory T cells and a peak in the active regulatory T cells. Figure 17J describes the kinetics of total and memory CD8 cytotoxic T cells. The total CD8 T cell population increases rapidly in the early phase and then gradually decreases, while memory CD8 T cells accumulate gradually. Figures 16L and 17K show the functional distribution of effector CD8 T cells and NK cells, respectively. Activated and proliferative CD8 T lymphocytes exhibit initial peaks followed by a decline, while quiescent and memory populations become predominant in later stages. NK cell populations show rapid early activation followed by stabilization. Figure 17M illustrates the dynamics of dendritic cells (DCs), including their antigen-presenting, infection, resting, and activation states. Antigen-presenting and activated DC populations increase rapidly in the early phase and remain detectable throughout the simulation. Figure 17N shows the kinetics of the macrophage population, including their infected, antigen-presenting, quiescent, and activated states.

Codon optimization findings

Figure 18 depicts the Codon Adaptation profile of the vaccine sequence. This Figure reveals that there have not been any significant variations in the relative adaptability profile during the course of optimization and the range varies from 10 to 15% to 100% with an average value of 50%. On the other hand, there has been a marked decrease in the GC content profile of the sequence and the.

Fig. 18.

GC content before/after codon optimization and MFE secondary structure.

Cloning vaccine

Figure 19 illustrates the PET-28a (+) vector-cloned vaccine. The vaccine insert is placed within the multiple cloning site (MCS) section and is colored in red. The insert exists in a particular segment of the plasmid ranging from the SexAI restriction site to the group of restriction sites comprising BspEI, BsaBI, and MflI. This positioning ensures that the construction of the artificial vaccine gene is successfully incorporated into the vector without interfering with any important parts of the plasmid. The vector incorporates an origin of replication (Ori) which is found at the lower left side of the plasmid, guaranteeing its ability to replicate autonomously in the host bacteria. Rop gene found at the lower right side regulates the number of copies of the plasmid. The vaccine insert is placed in accordance with the structure of the vector.

Fig. 19.

Representation of the pET-28a (+) vector, used for the expression of a vaccine antigen in E. coli.

Discussion

This study aimed to design a multiepitope vaccine against human cytomegalovirus (CMV) containing various immunogenic strains targeting B lymphocytes, major histocompatibility complex (MHC) class I and class II, and inducing an effective and durable immune response^69,70. Several candidate proteins of human cytomegalovirus were selected, and glycoprotein B was chosen as the most promising based on our multi-parameter classification system according to Zahraei et al.⁴¹. The two models utilize a multi-parameter evaluation system to rank antigens; however, their applications differ. Zahraei et al.⁴¹ suggest applying a cancer-based framework with parameters that consider protein location, tumor specificity, and the clinical importance of specific protein types. In comparison, our analysis addresses the topic of viral infection (human cytomegalovirus), ranking the parameters by criteria associated with vaccine immunogenicity, including antigenicity, toxicity, allergenicity, and protein availability. Similar in their classification principles, the two models use distinct parameters and have varied biological targets for their ranking systems. The choice of gB antigen can be also justified by its known biological importance within HCMV infections^71,72. As one of the main envelope proteins, gB acts as an agent of virus binding, membrane fusion, and cell entry, which makes it essential for its replication process^71,72. Moreover, the previous experimental evidence indicates the role of gB as a dominant target of neutralizing antibodies and cellular immunity, which makes it important as a component of vaccine^25,73.

The results of this study will also be applicable to the application of rational designs described in the literature, which demonstrate the potential of multi-epitope vaccines to induce diverse immune responses, generally associated with HCMV and other DNA viruses^74–76. Thus, combining different types of epitopes could enhance overall immunogenicity and reduce the risk of adverse vaccine effects⁷⁷. Moreover, the 50 S ribosomal protein L7/L12 was selected as an adjuvant vaccine because of its potent capability in stimulating Toll-like receptor activity⁷⁸. It facilitates the generation of pro-inflammatory cytokines and increases dendritic cell maturation⁷⁷. The L7/L12 induces the generation of an appropriate Th1/Th2 balance, hence, enabling the stimulation of both cellular and humoral immune responses⁷⁹. Unlike conventional adjuvants, this compound has the capacity to be genetically attached to vaccine antigens. Furthermore, this compound demonstrates high immunogenicity, excellent solubility, and no toxicological or allergenic properties⁸⁰. Single dose immunological simulation approaches are consistent with previous studies on vaccine design using immunoinformatics that always start by estimating the immunogenicity of the vaccine candidate through a simple dosing approach.

The physicochemical properties of the vaccine reveal that its constituents and molecular attributes are compatible to be soluble in a human biosystem and properly fold. Its acidic population, as well as the abundance of polar amino acids, including E; threonine and serine, should give the best access to epitopes by immune cells and T and B lymphocytes. The cysteine content is also low and restricts the spontaneous occurrence of disulfide bonds formation. Similarly, the proportion between mainly hydrophobic and hydrophilic amino acids may provide enough stability to sustain the three-dimensional structure of the vaccine^28,29,39. These properties align with data regarding the development of multi-epitope vaccines in silico, where soluble and slightly acidic proteins are linked to excellent recombinant expression and reaching immune^28,29,39. The vaccine design is predicted to confer geometric rigidity and local plasticity which is thought to be a key element in producing successful proteins that show multi-epitope vaccines. This hypothesis of the geometric structure can also be applied to predicting N-glycosylation sites since such sites are of importance in determining the stability and presentation of an epitope. This information on the prediction of N-glycosylation sites means that the vaccine is likely to have glycosylation in the centre of the protein in such a way that glycosylation will serve to stabilise the protein further and also guard against the degradation of the protein by proteolytic enzymes^81,82. Glycosylated motifs can be recognized by dendritic cells with greater ease and assist T and B lymphocytes to develop and display antigens, which results in better immune responses^83,84. Native epitope structures may also be provided by glycosylated motives, which are required to produce a robust and specific HCMV immune response^81,85,86. The above pieces of evidence are consistent with the findings of prior in silico vaccine modelling studies that have associated glycosylation with increased immunogenicity and closer similarity to natural viral proteins, hence making the argument that glycosylation is an important component of vaccine success^87,88. This evolution of the epitope into the in silico modelling will enable the assessment of the lymphocyte activation and the generation of long-term memory^89,90.

The conservancy analysis of the epitopes proved to be quite uneven in terms of the conservation levels observed in the examined predicted vaccine epitopes, which is indicative of the heterogeneity of the chosen human cytomegalovirus glycoprotein B sequences. Several epitopes had rather high levels of conservancy, with some having up to 100% identity and conservation in more than 60% of the analyzed sequences. Such high levels of conservancy make those epitopes good candidates for the vaccine development process, since conserved regions usually stimulate cross-protection in multiple virus strains. This approach has also been used by other authors who have demonstrated similar conservation of epitopes in the vaccine constructed against Sars-Cov2^91,92. However, the conservation of epitopes approach employed in the current study utilizes the method that involves a 100% identity approach applied in the IEDB conservation tool. It is extensively used in vaccine-related epitope studies owing to its straightforwardness, simplicity, and comparability between different strains. Nevertheless, it is highly stringent and fails to take into consideration conservative amino acid changes that still conserve immunologic recognition. In addition, its precision is restricted by a small number of sequences analyzed. There exist more complex computational models for the large-scale epitope conservation analysis such as Java-based model developed by Basmenj et al.⁷⁶. It is used for high-throughput analysis of thousands of virus sequences and involves application of evolutionary sequence processing approaches for systematic conservation analysis. The approach employed in this study is less extensive than those mentioned above but provides better interpretability and conservation precision.

Epitopes toxicity and allergenicity these have been demonstrated as the most important results of assessments the majority of epitopes are safe which reduces the risk of adverse reactions. Such theoretical data is upheld by multi epitope design strategies in HCMV vaccine designs by varying combinations of CTL Epitopes and HTL epitopes that have been linked to a strong, specific immune response with few adverse effects. The modular composition of the vaccine (N-Terminal adjuvant with strategic linkers) could also be useful in further proving this hypothesis concerning the vaccine immune efficacy. Hypothetically, the modular nature of the vaccine (adjuvant, rigid flexible linkers and His Tag) could hypothetically offer an optimal exposure to epitopes and an efficient response by the innate immune^93,94. Switching the vaccine to docking would ultimately give predictions to comparisons with a direct interaction with TLRs, thus giving an addition to the idea of early immune activation⁹⁵. However, the conservability of epitopes in the constructed vaccine remained lower than that obtained by Srinivasan et al.⁹⁶ in predicting a multi-epitope vaccine against SARS-CoV-2, and higher than certain epitopes for the vaccine constructed by Fatahi et al.⁹⁷ against the Epstein-Barr virus.

From the results obtained from the molecular docking study, it was evident that the designed multi-epitope vaccine had higher binding affinities towards TLR7 and TLR8 compared to TLR9. In fact, TLR7 and TLR8 are the pattern recognition receptors that play an important role in the induction of antiviral immune responses⁹⁸. Induction of these two receptors is important for the activation of innate immune system because it is known that activation of TLR7/8 leads to the maturation of dendritic cells, release of cytokines, and increased antigen presentation leading to the development of adaptive immune response^99,100. It is also evident that TLR7/8 pathway is important for the induction of Th1-based immune response because it is essential for the control of virus infections by activating cytotoxic T lymphocytes and antiviral cytokines¹⁰¹. The relatively lower affinity obtained in this study for TLR9 suggests that it may cause moderate activation of this signaling pathway. Moreover, the fact that several hydrogen bonds and hydrophobic interactions were identified in the vaccine-TLR complexes suggests the occurrence of stable receptor-ligand binding, which is essential for efficient downstream immune signaling. In order to further understand the structural properties of these complexes, normal mode analysis and molecular dynamics were performed. According to the results, the vaccine-TLR complexes exhibited structural stability and sufficient flexibility in order to accommodate necessary conformational changes for efficient receptor-ligand interaction and immune signaling. It is interesting to note that while the peripheral domains of the vaccine construct were more flexible, the central domains were relatively rigid¹⁰². Such an arrangement of the structural components may contribute to the proper presentation of epitopes on the surface of the vaccine construct, making it easier for immune receptors to recognize these epitopes and interact with them^103,104. The fact that the vaccine construct is characterized by the right balance between flexibility and rigidity makes it suitable for use as a multi-epitope vaccine. Overall, the results of docking and simulation studies suggest that the vaccine construct has the appropriate immunological and structural features for binding with innate immune receptors. Thus, it can stimulate antigen-presenting cells, activate T cells, and induce humoral immunity.

Limitations of the study

The main limitation of this study is that such studies are predictive in nature. Thus, stability, safety and immunogenicity of the multi-epitope HCMV vaccine candidate should be confirmed in vitro and in vivo to confirm the findings. In addition, the personal differences in the abilities of the human HLA system besides the complexity of the immune interaction might have an effect on the real response induced to the use of such multi-epitope vaccine construct. Irrespective of these setbacks, this study provides a platform on the direct development of HCMV vaccine and the foundation of subsequent research based on the above experiments in order to determine the possibility of the multi-epitope HCMV vaccine. The integration of MM-PBSA and trajectory analyses and additional parameters derived from the trajectory, such as hydrogen bond analysis, radius of gyration (Rg) and solvent accessible surface area (SASA), would further improve the characterization of vaccine-TLR interaction in future research.

Conclusion

According to the findings of the conducted research, it was feasible to produce a multi-epitope vaccine using in silico tools which facilitated the choice and incorporation of B, CD4 + T and CD8 + T epitopes to create the most effective humoral and cellular immune responses. Besides establishing the stability, hydrophilic nature and solubility of the proposed construct through physicochemical analysis, the assumptions about the structure also suggest that the construct is likely to have a correctly-folded 3D structure which will be able to allow exposing the various epitopes. Simulation of glycosylation, and binding experiments showed that the proposed construct would interact well, in terms of stimulating the innate immune response, with both TLR7 and TLR8. These results imply that the suggested multi-epitope vaccine candidate would be able to make the immune response stimulation more active. Simulations of the immune response in theory predicted the production of early IgM, followed by IgG antibody, activation of CD4 + and CD8 + T cells and interaction of dendritic cells, macrophages and NK cells facilitates a full and sustained cellular immune response.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

Ag

Antigen

AA

Amino acid

CTL

Cytotoxic T lymphocyte

CD4⁺ T cells

Cluster of differentiation 4 positive T lymphocytes

CD8⁺ T cells

Cluster of differentiation 8 positive T lymphocytes

BCR

B cell receptor

DC

Dendritic cell

MA

Macrophage

NK

Natural killer cell

MHC-I

Major histocompatibility complex class I

MHC-II

Major histocompatibility complex class II

HLA

Human leukocyte antigen

TLR

Toll-like receptor

TLR7/TLR8/TLR9

Toll-like receptors 7, 8, and 9

IgM

Immunoglobulin M

IgG

Immunoglobulin G

IgG1/IgG2

IgG subclasses 1 and 2

IgA

Immunoglobulin A

IL

Interleukin

IFN-α/IFN-γ

Interferon alpha/interferon gamma

TNF-α

Tumor necrosis factor alpha

NMA

Normal mode analysis

RMSD

Root mean square deviation

pI

Isoelectric point

MW

Molecular weight

N-glyc

N-glycosylation

GRAVY

Grand average of hydropathy

3D

Three-dimensional structure

Author contributions

Conceptualization, Mengue Ngadena Yolande Sandrine; methodology, Ownoa Pascal Emmanuel; software, Bilanda Danielle Claude; validation, Maha A Aljumaa and Hanan Abdulrahman Sagini; formal analysis, Akingbolabo Daniel Ogunlakin; investigation, Bidingha A Goudani Ronald; resources, Tariq Aziz; data curation, Dzeufiet Djomeni Paul Desire.; writing—original draft preparation, Ownoa Pascal Emmanuel.; writing—review and editing, Tariq Aziz; visualization, Shaza N. Alkhatib; supervision, Tariq Aziz and Mengue Ngadena Yolande Sandrine.; project administration, Mengue Ngadena Yolande Sandrine; funding acquisition, Tariq Aziz.

Funding

The authors are thankful to Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2026R456), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

Data availability

All the data generated in this research work has been included in the manuscript.

Declarations

Competing interests

The authors declare no competing interests.