Advancing therapeutic vaccines for chronic hepatitis B: Integrating reverse vaccinology and immunoinformatics

Abstract

Current treatments for chronic hepatitis B (CHB) are lifelong, often accompanied by side effects and the risk of drug resistance, highlighting the urgent need for alternative therapies such as therapeutic vaccines. However, challenges such as selecting appropriate antigens and addressing multiple hepatitis B virus (HBV) genotypes hinder the development of these vaccines. One approach to overcoming these challenges is reverse vaccinology (RV) combined with immunoinformatics. RV uses computational methods to identify antigens from pathogen genetic information, including genomic and proteomic data. These methods have helped researchers identify conserved epitopes across bacterial strains or viral species, including multiple HBV genotypes. Computational tools, such as epitope mapping algorithms, molecular docking analysis, molecular dynamics simulations, and immune response simulations, enable key epitope identification, predict vaccine candidates' binding potential to immune cell receptors, and forecast the immune response. Together, these approaches streamline therapeutic vaccine design for CHB, making it faster, more cost-effective, and accurate. This review aims to explore the potential role of RV and immunoinformatics in advancing therapeutic vaccine design for CHB.

Key Words: Chronic hepatitis B; Therapeutic vaccine; Reverse vaccinology; Immunoinformatics; Vaccine design

Core Tip: Chronic hepatitis B (CHB) remains a significant public health concern worldwide where effective alternative therapies are still urgently needed. This review aims to explore the potential role of reverse vaccinology combined with immunoinformatics in advancing therapeutic vaccine design for CHB.

INTRODUCTION

Chronic hepatitis B (CHB), the outcome of unresolved and persisted hepatitis B virus (HBV) infection, remains a significant public health concern due to its potential to cause cirrhosis, hepatocellular carcinoma (HCC), and hepatitis-related deaths[1,2]. As of 2024, an estimated 254 million people worldwide suffer from CHB, with a mortality rate higher than the previous year[1]. Hepatitis B vaccination for newborns prevents viral transmission from infected mothers to their children, however, the coverage of the recommended birth dose vaccine remains suboptimal and horizontal transmission still occurs in the community[3].

One contributing factor in the real time management of CHB is the limited effectiveness of current treatments. Current HBV treatment only resulted in partial cure, and not functional HBV cure which is defined as HBsAg loss and undetectable HBV DNA up to 6 months after stopping therapy[4]. Pegylated Interferon alpha 2a and nucleos(t)ide analogues (NA) are the most commonly used therapies for CHB[5-7]. These treatments can suppress viral replication efficiently shown by normalized serum liver enzymes and reduced HBV DNA levels, however, treatment must be given lifelong to patients[2]. These limitations reflect the inability of these therapies to fully address the complexities of CHB.

In CHB, there are complex issues involving the HBV covalently closed circular DNA (cccDNA) minichromosome and the host’s immune system (Figure 1)[8-11]. Upon infection, HBV genome is translocated to nucleus and converted from relaxed circular DNA (rcDNA) to cccDNA which acts as the transcription template for viral mRNAs and pregenomic RNAs[4,12,13]. This cccDNA interacts with both host histone proteins and HBV non-histone proteins, forming a minichromosome[8,14], which plays a key role in the persistent replication of HBV.

Figure 1 Overview of challenges in chronic hepatitis B and potential solutions through integrating reverse vaccinology and immunoinformatics in therapeutic vaccines design. A: T cell and B cell exhaustion in chronic hepatitis B (CHB) patients contribute to the inability to eliminate hepatitis B virus (HBV) covalently closed circular DNA minichromosomes, allowing for continuous viral replication; B: Therapeutic vaccines present a promising alternative treatment for CHB, but their design encounters two major challenges: The selection of appropriate antigens and the genetic variability of HBV genotypes; C: Reverse vaccinology, combined with immunoinformatics, addresses these challenges by facilitating the identification of conserved epitopes across all HBV genotypes and predicting stable, immunogenic vaccine candidates. HBV: Hepatitis B virus; HBc: Hepatitis B virus core protein; HBx: Hepatitis B virus X protein; HBs: Hepatitis B virus surface protein. Created in BioRender.com (Supplementary material).

Ideally, the host’s immune response would be capable of eliminating this cccDNA minichromosome. However, HBV has developed several mechanisms to evade immune defenses[15,16]. Furthermore, excessive exposure to HBV antigens can result in exhaustion and dysfunction of the patient’s T and B cells[17-19]. NA drugs target the polymerase region to inhibit viral replication, resulting in reduce formation of both rcDNA and double-stranded linear DNA and subsequent viral DNA integration[20]. However, it has no effect on the rcDNA to cccDNA conversion, thus cannot completely suppress viral DNA synthesis and cccDNA replenishment[21]. On the other hand, interferon treatment may affect cccDNA in infected cells by decreasing its transcriptional activity by decreasing the acetylation of cccDNA-bound histones[4]. Combining existing NA therapy with new antiviral agents may resulted in suppression of HBV DNAs and RNAs and depletion of cccDNA pools, but if there is no HBsAg loss, stopping the therapy will lead to viral relapse[4,15].

In the early HBV infection phase (HBeAg positive), almost all (> 95%) HBsAg production is derived from cccDNA transcriptions, while in the late infection phase (HBeAg negative), the integrated HBV DNA produced more than half of the circulating HBsAg proteins, including the small HBsAg proteins and the subviral particles[2]. The circulating presence of large amounts of HBsAg affect the host immune response through (1) interference with neutralizing antibody as a decoy; (2) downregulation of antigen specific T-cell immunity; and (3) suppression of innate immunity by dysregulation of natural killer cell and dendritic cell functions, possibly resulting in exhaustion of the immune response[2,4].

To overcome the challenges of CHB, a treatment that is safe, can be administered over a short period, and enhances the immune system's ability to target HBV—particularly the cccDNA minichromosome—is needed. One potential solution is a therapeutic vaccine. Despite therapeutic vaccines for CHB being in development for roughly the past two decades, none have yet achieved a sterilizing cure, which is characterized by the elimination of the HBV cccDNA minichromosome. Several challenges exist in developing therapeutic vaccines for CHB, particularly at the stage of designing vaccine candidates (Figure 1). One major issue is antigen selection. Most therapeutic vaccines utilize the HBV surface antigen (HBs), following the success of prophylactic vaccines[22-25]. However, the use of HBs in therapeutic vaccines has proven less effective, as it can lead to T cell exhaustion[23,25-27]. Some studies have combined HBs with other HBV proteins, such as HBV core protein (HBc), HBV polymerase, and HBV X protein (HBx), but the results remain suboptimal[28-32].

The second challenge is the genetic diversity of HBV. HBV infection persistence is related to its complex life cycle, with high replicative capacity, establishment of viral reservoirs for HBV replication and antigen production, high viral burden, and impaired host innate and adaptive immune responses against HBV[2,33]. HBV has high replicative capacity, producing more than 100 billion virions per day. However, due to the absence of proofreading activity of the viral reverse transcriptase, HBV is prone to mutations. In addition, recombination events may also occur during viral replication, resulting in high viral genetic diversity. As such, HBV exists as viral quasispesies, a population of viral variants that are closely linked genetically, capable to adapt quickly to a variety of selective pressures[33]. Different HBV variants have been known to be associated with different clinical manifestations and responsiveness to CHB treatment[34].

Based on the differences on HBV genome, up to now, there are 10 different HBV genotypes, each with distinct geographical distributions[35]. Genotypes A to D exhibit more than 8% nucleotide variation across the entire genome, while genotypes E to H show around 4% variation in the HBs coding gene[36]. Genotypes C and D are particularly known for having a higher frequency of core promoter and pre-S mutations compared to genotypes A and B[37]. Additionally, some genotypes contain mutations in the cytotoxic T lymphocyte (CTL) epitopes of the HBc gene, as well as point mutations in the HBx gene[38]. These genotype differences also influence the risk of developing cirrhosis and HCC[37]. This indicates that mismatches between the infecting virus genotype and the one used in the vaccine could lead to treatment failure. Immunogenic therapeutic vaccines designed to target all major HBV antigens might restore inadequate immune responses by boosting the existing or de novo viral specific T cell and B cell responses[15].

Therefore, new approaches are needed to achieve functional HBV cure. Restoring the host immune response, in combination with antiviral agents that targeted the cccDNA and integrated DNA, is now considered as a better approach to achieve sustained HBV control in CHB by eliminating the infected cells and/or block infection of new cells[2,15].

One potential solution to both issues is reverse vaccinology (RV). RV leverages computational techniques to identify antigens from the genetic data of pathogens[39]. Through RV, researchers can discover conserved antigenic regions or epitopes that are present across all HBV genotypes. Additionally, RV can be combined with immunoinformatics to predict the safety, immunogenicity, and immune responses of vaccine candidates, thereby saving time and reducing costs associated with laboratory testing (Figure 1)[40,41]. By integrating RV and immunoinformatics, only those vaccine candidates predicted to be safe, immunogenic, capable of inducing specific immune responses, and offering broad global population coverage will proceed to laboratory testing. Therefore, this article seeks to provide a deeper review of the potential role of RV and immunoinformatics in advancing the design of therapeutic vaccines for CHB.

RV IN THERAPEUTIC VACCINE DESIGN

The design of both prophylactic and therapeutic vaccines is traditionally initiated by cultivating the pathogen in a laboratory to identify its components[42,43]. Following this, potential vaccine candidates are selected by isolating each component of the pathogen. Once identified, these antigens must be mass-produced, either by growing the pathogen itself or through gene cloning[42]. The entire process is time-consuming, costly, and requires advanced laboratory equipment.

In contrast to the conventional method, RV offers several advantages. RV, also known as antigen- or epitope-based vaccinology, uses genomic and protein sequences from pathogens to identify antigens that can stimulate the host's immune response[44,45]. With RV and computational tools, researchers can predict epitopes recognized by T cells and B cells, which may serve as potential vaccine candidates without the need to handle the pathogen directly[39,41,46,47]. RV helps researchers identify key epitopes likely to trigger a strong immune response, reducing the need for extensive lab testing and lowering time and costs by focusing on the most promising candidates[40,41,48,49]. Additionally, RV helps identify conserved regions in the pathogen's genome or proteins, allowing for the design of vaccine candidates with broader protective capabilities[40,48-50]. The RV approach also facilitates the creation of multi-epitope vaccines, which combine epitopes from different parts of the pathogen, or even from different pathogens[41]. This is important because such vaccines can target multiple viral strains or different immune pathways simultaneously.

RV was first utilized in the 1990s to design a prophylactic vaccine candidate for Group B meningococcus[42]. At that time, the RV approach had a significant impact on Group B meningococcus vaccine development. In just a year and a half, 25 surface-exposed proteins were discovered, all capable of inducing antibodies[42,48]. This success led to the increasing use of RV in vaccine design, including therapeutic vaccines. The design of therapeutic vaccines using the RV approach begins with data collection, which can be either genomic or proteomic. When genomic data is used, gene prediction is required to identify coding sequences, which are then translated into protein sequences[51]. Data for RV can be obtained from sources like the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/) or the UniProt Knowledgebase (https://www.uniprot.org/)[51-53].

After data collection, researchers select proteins that naturally interact with the immune system[40,51]. These proteins must be non-redundant, non-homologous, non-allergenic, and non-toxic[51]. To identify conserved protein sequences across different pathogen species, multiple sequence alignment can be performed using various software tools such as ClustalW[54], ClustalX[55], Clustal Omega[56], MUSCLE[57], FAMSA[58], MAGUS[59], or web servers like MAFFT (https://mafft.cbrc.jp/alignment/server/index.html)[60] and T-Coffee (https://tcoffee.crg.eu/)[61]. The alignment results can be visualized and edited using JalView[62] or BioEdit[63]. Sequence conservation and variability levels can be measured by performing entropy analysis with AVANA[64], which is based on Shannon's entropy concept.

The next step involves predicting epitopes from the identified conserved protein sequences[40,43,51]. Three types of epitopes that can be utilized as candidates for therapeutic vaccines: CTL epitopes, helper T lymphocyte (HTL) epitopes, and linear-B cell (LBC) epitopes. CTL epitopes promote long-term immunity by targeting pathogens and infected cells, HTL epitopes play a role in both humoral and cell-mediated immunity, and LBC epitopes stimulate antibody production in response to antigens[50,65]. For therapeutic vaccine candidate, epitopes must meet several key criteria, including being antigenic, immunogenic, non-allergenic, non-toxic, and providing broad population coverage[48]. In some cases, researchers also consider anti-inflammatory epitopes, which help regulate the immune response to prevent harmful inflammation[66]. This consideration is particularly important for therapeutic vaccines, where the goal is not only to elicit an immune response but also to maintain a balance reaction that avoids tissue damage. The final step involves computational analysis and verification of the vaccine candidates[40,43], ensuring that they are safe, antigenic, structurally stable, and capable of inducing a targeted immune response.

IMMUNOINFORMATICS FOR THERAPEUTIC VACCINE DESIGN

In the initial stages of therapeutic vaccine design, immunoinformatics plays a key role in epitope prediction. Epitopes chosen for therapeutic vaccine design should be capable of inducing an immune response, specifically a cellular immune response mediated by CTLs and HTLs, as therapeutic vaccines are administered to patients with previously infected cells[46,67]. Therefore, the selected epitopes need to exhibit binding affinity with multiple major histocompatibility complex (MHC) class I and II alleles[68]. Several tools are available for CTL epitope prediction, such as NetCTL1.2[69], NetMHCpan4.0[70], and CTLPred (Table 1)[71]. Among the many web servers capable of predicting CTL epitopes, NetCTL1.2 is one of the most widely used tools due to its advantages over other prediction tools. Several tools can be used for HTL epitope prediction, including NetMHCIIpan 4.0[70], ProPred[72], and MARIA[73] (Table 1). Unlike NetMHCIIpan 4.0 and ProPred, MARIA does not predict epitopes recognized by CTLs but rather those presented by MHC class II[73].

Table 1 Web servers available for predicting Cytotoxic T Lymphocyte, Helper T lymphocyte, and B-cell epitopes.

Web server	Prediction methods	Website link	Ref.
Cytotoxic T lymphocyte epitope
NetCTL 1.2	Uses an ANN approach to combine MHC class I peptide binding, proteasomal C-terminal cleavage, and TAP transport efficiency	https://services.healthtech.dtu.dk/services/NetCTL-1.2/	Larsen et al[69], 2007
NetMHCpan 4.1	Predicts peptide binding to MHC molecules based on quantitative binding affinity and eluted ligands identified by mass spectrometry, using an ANN approach	https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/	Reynisson et al[70], 2020
CTLPred	Predicts CTL epitopes based on T cell epitope patterns, utilizing both ANN and SVM approaches	http://crdd.osdd.net/raghava/ctlpred/	Bhasin and Raghava[71], 2004
Helper T lymphocyte epitope
NetMHCIIpan 4.0	Predicts peptide binding to MHC II molecules (HLA-DR, HLA-DQ, HLA-DP) based on binding affinity and eluted ligands identified by mass spectrometry, using an ANN approach	https://services.healthtech.dtu.dk/services/NetMHCIIpan-4.0/	Reynisson et al[70], 2020
ProPred	Predicts MHC Class II (HLA-DR) binding regions within antigen sequences using QM	http://crdd.osdd.net/raghava/propred/	Singh and Raghava[72], 2001
MARIA	Predicts the likelihood of antigen presentation from a specific gene related to HLA class II alleles, using peptide sequences from mass spectrometry, antigen gene expression levels, and protease cleavage patterns with an ANN approach	https://maria.stanford.edu/	Chen et al[73], 2019
Linear B Cell epitope
ABCpred	Uses an RNN approach that considers peptide length to predict B cell epitopes within antigen sequences	http://crdd.osdd.net/raghava/abcpred/	Saha and Raghava[114], 2006
Bepipred Linear Epitope Prediction 2.0	Uses a random forest algorithm trained on annotated epitopes from antibody-antigen protein structures	https://services.healthtech.dtu.dk/services/BepiPred-2.0/	Jespersen et al[115], 2017
BCEPS	Predicts linear B cell epitopes using an SVM approach based on the tertiary structure of antibody-antigen complexes	http://imbio.med.ucm.es/bceps/	Ras-Carmona et al[116], 2021
SEMA	Applies a transfer learning approach using a pre-trained deep learning model to predict conformational B cell epitopes based on primary antigen sequences and tertiary structures	https://sema.airi.net/	Shashkova et al[117], 2022
LBtope	Uses SVM and Ibk approaches on a large dataset of experimentally validated B cell epitopes and non-epitopes to predict linear B cell epitopes	https://webs.iiitd.edu.in/raghava/lbtope/index.php	Singh et al[118], 2013
Bcepred	Predicts B cell epitopes using physicochemical properties, such as hydrophilicity, flexibility, accessibility, polarity, exposed surface, and turns	http://crdd.osdd.net/raghava/bcepred/	Saha and Raghava[77], 2004
COBEpro	Uses an SVM to predict short peptide fragments within query antigen sequences, calculating an epitope propensity score for each residue	https://scratch.proteomics.ics.uci.edu/	Sweredoski and Baldi[119], 2009
CLBTope	Combines alignment-based and alignment-free machine learning methods to predict B cell epitopes, using epitope and non-epitope sequence composition	https://webs.iiitd.edu.in/raghava/clbtope/	Kumar et al[120], 2024

ANN: Artificial neural network; CTL: Cytotoxic T lymphocyte; HLA: Human leukocyte antigen; Ibk: Instance-based k-nearest neighbor; MHC: Major histocompatibility complex; QM: Quantitative matrix; RNN: Recurrent neural network; SVM: Support vector machine; TAP: Transporter associated with antigen processing.

In addition to cellular immune responses, several studies report that epitopes used in therapeutic vaccine design should also be capable of inducing humoral immune responses[40,50,74,75]. Humoral response activation relies on the recognition of both linear (LBC) and conformational B-cell epitopes[76]. Initially, tools for predicting LBC epitopes, such as Bcepred[77], analyzed the sequential physicochemical properties of amino acids to make predictions. However, with the integration of machine learning algorithms data comparison across methods has improved (Table 1). This integration enhances prediction accuracy, producing more optimistic results and reducing misleading predictions[76]. Similarly, tools for predicting conformational B-cell epitopes have been developed to analyze the potential of epitopes to form different conformations, such as DiscoTope-3.0 (https://services.healthtech.dtu.dk/services/DiscoTope-3.0/)[78].

Predicted epitopes still require further selection before they can be used in therapeutic vaccine design. This selection process has become faster and easier with the use of immunoinformatics approaches. Numerous tools are now available to predict antigenicity, allergenicity, immunogenicity, toxicity, autoimmunity, and population coverage, each with varying levels of accuracy (Table 2). Additionally, tools have been developed to predict anti-inflammatory peptides and the ability to induce IFN-γ (Table 2), both of which are essential parameters for certain therapeutic vaccines[67,79]. In addition to the web servers listed in Tables 1 and 2, the Immune Epitope Database (IEDB) web server also provides tools to predict CTL epitopes (http://tools.iedb.org/mhci/), HTL epitopes (http://tools.iedb.org/mhcii/), and LBC epitopes (http://tools.iedb.org/bcell/) with a range of method options. The IEDB facilitates epitope prediction, particularly for CTL and HTL, as it includes human leukocyte antigen (HLA) MHC class I and II allele data, covering approximately 97% of the global population for CTL[80] and 99% for HTL[81]. Additionally, the server enables immunogenicity analysis and population coverage assessment for predicted CTL and HTL epitopes.

Table 2 Web servers available for epitope selection.

Parameter	Web server	Accuracy (%)	Link	Ref.
Antigenicity	Vaxijen v.2.0	70-89	https://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html	Doytchinova and Flower[121], 2007
	Antigenic	78.04-80.03	http://77.68.43.135:8080/Antigenic/	Rahman et al[122], 2019
	ANTIGENpro	76	http://scratch.proteomics.ics.uci.edu/	Magnan et al[123], 2010
Toxicity	ToxinPred3.0	93	https://webs.iiitd.edu.in/raghava/toxinpred3/prediction.php	Rathore et al[124], 2024
	ToxDL	83-90	http://www.csbio.sjtu.edu.cn/bioinf/ToxDL/	Pan et al[125], 2021
	ToxIBTL	96	http://server.wei-group.net/ToxIBTL	Wei et al[126], 2022
Immunogenicity	IEDB Immunogenicity	66	http://tools.iedb.org/immunogenicity/	Calis et al[127], 2013
	AbImmPred	72.73	https://www.genscript.com/tools/antibody-immunogenicity	Wang et al[128], 2024
	DeepImmuno	80-90	https://deepimmuno.herokuapp.com/	Li et al[129], 2021
Allergenicity	AllergenFP v.1.0	88	https://www.ddg-pharmfac.net/AllergenFP/index.html	Dimitrov et al[130], 2014
	AllerTOP v.2	85.3	https://www.ddg-pharmfac.net/AllerTOP/	Dimitrov et al[131], 2014
	AllerCatPro 2.0	84.7	https://allercatpro.bii.a-star.edu.sg/	Nguyen et al[132], 2022
Population coverage	IEDB Population Coverage	N/A	http://tools.iedb.org/population/	Bui et al[133], 2006
Autoimunity	miPepBase	N/A	http://proteininformatics.org/mkumar/mipepbase/index.html	Garg et al[134], 2017
Anti-inflammatory	AIPpred	73.4	http://211.239.150.230/AIPpred/AIPpredMethod.html	Manavalan et al[135], 2018
	PreAIP	76.7	http://kurata14.bio.kyutech.ac.jp/PreAIP/	Khatun et al[136], 2019
	PepNet	95	http://liulab.top/PepNet/server	Han et al[137], 2024
IFN-γ induction	IFNepitope	82.1	https://webs.iiitd.edu.in/raghava/ifnepitope/run_submit-old.php	Dhanda et al[138], 2013
	PIP-EL	74.8	http://www.thegleelab.org/PIP-EL/	Manavalan et al[139], 2018

N/A: Not applicable.

Immunoinformatics can also be used to predict the secondary and tertiary structures of vaccine candidates composed of selected epitopes. Several widely used tools are available for secondary structure prediction. These include PredictProtein (https://predictprotein.org/)[82]; Self Optimized Prediction Method with Alignment (https://npsa.lyon.inserm.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html)[83]; PSIPRED 4.0 (http://bioinf.cs.ucl.ac.uk/psipred/)[84]; and Jpred4 (https://www.compbio.dundee.ac.uk/jpred/)[85]. For tertiary structure prediction, tools such as Iterative Threading ASSEmbly Refinement (https://zhanggroup.org/I-TASSER/)[86], trRosetta (https://yanglab.qd.sdu.edu.cn/trRosetta/)[87], and AlphaFold2 (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb)[88] have been effectively used in research.

After successfully modeling the therapeutic vaccine candidate in 3D, interactions and binding strength between the designed vaccine candidate and various immune-response receptors can be tested through molecular docking analysis. Selected epitopes can also undergo docking to identify their binding pockets on the target receptors. Molecular docking analysis can be performed using tools such as HDOCK (http://hdock.phys.hust.edu.cn/)[89], ClusPro 2.0 (https://cluspro.org/help.php)[90], SwissDock (https://www.swissdock.ch/)[91], and MolModa (https://durrantlab.com/molmoda)[92]. To further predict interactions and binding potential between the vaccine candidate and target receptors, molecular dynamics (MD) simulations can be applied. MD simulations add water molecules and simulate temperature increases, replicating in vivo conditions and providing detailed insights into vaccine-receptor interactions. Commonly used software for MD simulations includes AMBER[93] and GROMACS[94]. In the final stage of vaccine design, ensuring the candidate’s ability to induce an immune response is essential. Immunoinformatics can assist in this phase by enabling immune response analysis through the C-ImmSim (http: //150.146.2.1/C-IMMSIM/index.php) server, which simulates antigen recognition at both single-peptide and whole-proteome levels[95].

APPLICATION OF RV AND IMMUNOINFORMATICS IN CHB THERAPEUTIC VACCINE DESIGN

Since 2017, several studies have explored epitopes and designed therapeutic vaccines for CHB using RV and immunoinformatics approaches (Table 3). Zheng et al[96] were interested in exploring epitopes on HBV polymerase due to its lower expression level. Zheng et al[96] utilized a consensus sequence created from 13 amino acid sequences of polymerase from NCBI, representing 13 HBV (sub)genotypes (A, A1, A2, B, C, D, D3, D4, E, F2, F4, G, H). After predicting CTL and LBC epitopes, they selected these based on immunogenicity, conservation, toxicity, and population coverage, using immunoinformatics approaches. They identified four conserved CTL epitopes and two LBC epitopes that showed strong binding to HLA-A*0201 in molecular docking analysis. However, they only predicted CTL epitopes recognizing HLA-A*0201, which is prevalent among Caucasians but expressed in about 40% of the global population and is not dominant in regions with high HBV prevalence, such as Asia and Africa[97].

Table 3 Applications of immunoinformatics in chronic hepatitis B therapeutic vaccine design.

Targeted antigens	Data type	Data source	Immunoinformatics applications	Ref.
HBV Polymerase	Proteomics	NCBI	Consensus sequence formation, epitope prediction, epitope selection (based on immunogenicity, toxicity, and conservation), population coverage calculation, and molecular docking analysis	Zheng et al[96], 2017
HBV polymerase; HBx	Proteomics	HBVdb	Epitope prediction and epitope selection (based on conservation and immunogenicity)	de Beijer et al[98], 2020
HBV polymerase	Proteomics	NCBI	Consensus sequence formation, epitope prediction, epitope selection (antigenicity, allergenicity, toxicity), population coverage calculation, physicochemical property prediction of vaccine candidates, secondary and tertiary structure prediction, molecular docking analysis, and molecular dynamics simulation	Ahmed et al[100], 2021
HBc; HBx	Proteomics	NCBI	Epitope prediction and epitope selection (conservation and autoimmunity)	Saeed et al[101], 2023
LHBs	Proteomics	NCBI	Epitope prediction, physicochemical property prediction, vaccine candidate antigenicity and allergenicity analysis, molecular docking analysis, molecular dynamics simulation, and immune response simulation	Zhu et al[104], 2024

HBc: Hepatitis B virus core protein; HBV: Hepatitis B virus; HBVdb: Hepatitis B virus database; HBx: Hepatitis B virus X protein; LHBs: Large hepatitis B virus surface protein; NCBI: National Center for Biotechnology Information.

In a different approach, de Beijer et al[98] predicted CTL epitopes from HBV polymerase and HBx using a larger dataset and more diverse HLA types. de Beijer et al[98] used 7489 amino acid sequences for polymerase and 8127 sequences for HBx from the HBVdb (https://hbvdb.lyon.inserm.fr/HBVdb/HBVdbIndex) database, which provides comprehensive genomic and proteomic data for HBV, covering genotypes A to H[99]. This study identified five novel HBx epitopes and 17 previously unreported polymerase epitopes[98]. Based on immunoinformatics analysis, all predicted epitopes were found to be conserved across eight HBV genotypes and demonstrated strong immunogenic potential. This was further verified through in vitro experiments, which showed that the predicted epitopes could bind to target HLA molecules and successfully induce IFN-γ production in peripheral blood mononuclear cells from individuals who had recovered from HBV infection[98].

Continuing the exploration of HBV polymerase, Ahmed et al[100] used a similar approach to design a vaccine candidate by combining several CTL, HTL, and LBC epitopes from polymerase. Unlike de Beijer et al[98], Ahmed et al[100] used a representative dataset of 148 HBV polymerase sequences from NCBI. They predicted CTL, HTL, and LBC epitopes using tools available on the IEDB web server. Epitopes in this study were selected based on antigenicity, allergenicity, and toxicity. Going a step further than previous studies, Ahmed et al[100] developed a multi-epitope polymerase vaccine candidate with an adjuvant, Mycobacterium tuberculosis 50S ribosomal protein L7/L12.

In addition to exploring HBV polymerase epitopes and vaccine design, other studies have targeted HBc and HBx epitopes for CHB therapy using immunoinformatics approaches. Saeed et al[101] gathered 237 HBx and 207 HBc amino acid sequences from NCBI, representing the 10 HBV genotypes. After epitope selection, they identified two LBC epitopes and 10 CTL epitopes from HBx, and 22 CTL epitopes from HBc, along with three HTL epitopes from HBx and 10 HTL epitopes from HBc, all with 100% conservation. The study also discovered several partially conserved epitopes (70%-90% conservation) in the C-terminal domain (CTD) of HBc, which contains four arginine-rich domains that can enhance pre-core gene transcription and either promote or inhibit host gene transcriptional activators[102]. Similarly, conserved epitopes in the CTD of HBx were found, where this domain includes two transactivation regions composed of several motifs, including a BH3-like motif that enables HBx to bind to anti-apoptotic proteins[103]. This information is highly valuable for designing CHB therapeutic vaccines, as it allows researchers to utilize specific epitopes from HBc and HBx rather than using the full-length proteins as primary vaccine components.

Although limited, there is a therapeutic CHB vaccine design using RV and immunoinformatics that has undergone in vitro and in vivo validation. In 2024, Zhu et al[104] predicted HTL and LBC epitopes from the LHBs protein and combined them with the immunoglobulin variable region of CTLA-4. Zhu et al[104] utilized amino acid sequences of LHBs and CTLA-4, obtained from NCBI, in their study. Selected LHBs epitopes were fused with the immunoglobulin variable region of CTLA-4 using a linker. Molecular docking analysis and MD simulation demonstrated that the vaccine candidate could strongly bind B7 molecules. These results were further corroborated by in vitro experiments, which showed that macrophages bound and engulfed the vaccine candidate more effectively than LHBs alone. ELISA and ELISPOT assays further indicated that the vaccine candidate could significantly stimulate HBV-specific T helper cell type 1 and type 2 responses. Western blot analysis also confirmed the good antigenicity of the vaccine candidate, as it effectively bound to specific antibodies in plasma from CHB patients. Immune simulation using the C-ImmSim server was further validated by in vivo experiments, where the levels of IFN-γ, IL-4, and total IgG in splenocyte culture supernatants were significantly higher in mice immunized with the vaccine candidate compared to those immunized with LHBs alone.

Although a perfect therapeutic vaccine design for CHB has yet to be achieved, various studies have demonstrated the crucial role of RV and immunoinformatics in addressing challenges in CHB vaccine design (Table 3). Through RV approaches, multiple studies have successfully identified numerous conserved epitopes in key antigens such as HBs[104], HBc[101], HBx[98,101], and polymerase[96,98,100] across all HBV genotypes, offering hope for developing a globally effective CHB therapeutic vaccine. The integration of immunoinformatics with RV has also proven to streamline the process of selecting optimal epitopes from different HBV antigens and combining them, for example, HBx with polymerase[98] or HBc with HBx[101], in vaccine candidate designs. Additionally, immunoinformatics enables researchers to design CHB therapeutic vaccine candidates without relying on certain antigen parts which could potentially support HBV replication, suppress host immune gene expression, or promote cancer development[101]. This approach has made the therapeutic vaccine design process faster, more cost-effective, and more precise.

OPPORTUNITIES AND CHALLENGES OF RV AND IMMUNOINFORMATICS IN CHB THERAPEUTIC VACCINE

After exploring several studies that successfully identified epitopes and designed therapeutic vaccines for CHB, it is clear that there are still many opportunities for advancing CHB therapeutic vaccine design through the integration of RV and immunoinformatics approaches. While epitope prediction from HBc and HBx has been conducted, it has not yet incorporated large-scale genomic data, such as the comprehensive datasets available on NCBI. Saeed et al’s study[101] utilized only hundreds of protein sequences representing 10 genotypes. Increasing the amount of data in these studies would better reflect the amino acid sequence variations across genotypes. Sequence variations can lead to differing consensus amino acid sequences, which would influence predicted epitopes. Additionally, although HBc and HBx have the potential to influence the stability and transcription of the cccDNA minichromosome[8], their epitopes have yet to be utilized in therapeutic vaccine design. Targeting HBc and HBx could be a strategic approach to disrupt cccDNA minichromosome.

On the other hand, epitope prediction of HBV polymerase has been conducted using large datasets from HBVdb[98], although it covers only eight genotypes. Additionally, studies predicting CTL epitopes have used only a limited selection of HLA class I alleles[96,98]. While some research has employed reference HLA class II alleles from the IEDB for HTL epitope prediction[100], CHB-associated HLA class II alleles, such as HLA-DPA1*01: 03-DPB1*04: 02, HLA-DPA1*01: 03-DPB1*04: 01[105], have not yet been considered. The selection of HLA alleles in epitope prediction can significantly influence the resulting predicted epitopes. There is an opportunity to expand this research by utilizing resources like the AlleleFrequencies[106] web server (https://www.allelefrequencies.net/hla6006a.asp), which provides access to frequently observed HLA class I and II alleles across global populations.

In prior studies, epitope selection has typically been based on factors such as antigenicity, allergenicity, immunogenicity, toxicity, autoimmunity, and conservancy[96,98,100,101,104]. However, additional parameters could enhance epitope selection to better address CHB, such as the ability to induce IFN-γ and being classified as anti-inflammatory peptides[107]. Additionally, epitopes can be filtered based on physicochemical properties, which are predictable using the ProtParam[108] web server (https://web.expasy.org/protparam/). Although previous studies have used this tool to assess physicochemical properties for vaccine candidates[104], it can also assist in epitope selection. For instance, HBx is a highly hydrophobic protein, indicating that its predicted epitopes may also be hydrophobic. Incorporating these epitopes fully into a therapeutic vaccine could increase the vaccine candidate’s overall hydrophobicity, though ideally, it should be hydrophilic. By selecting epitopes based on physicochemical properties, researchers could enhance the hydrophilicity of the vaccine candidate.

Another opportunity is to advance CHB therapeutic vaccine design through the optimal selection of adjuvants. To date, only a few studies have included adjuvants in CHB therapeutic vaccine design[100]. Immunoinformatics can facilitate this selection process, with tools such as the VaxinPad[109] web server (http://crdd.osdd.net/raghava/vaxinpad/) for peptide vaccine adjuvant design. In addition to enhancing immune responses, studies have shown that adjuvants in CHB therapeutic vaccines may help mitigate T and B cell exhaustion[17]. A recent study by Sacher et al[110], however, showed that amino-acid formulation efficiently stabilized hepatitis B therapeutic vaccine to break immune tolerance in AAV-HBV mice. It can be one of strategical approaches for therapeutic vaccines designed with RV and immunoinformatics.

Although previous studies have shown that RV and immunoinformatics can help address challenges related to antigen or epitope selection and HBV genotype diversity, these approaches have limitations, posing new challenges in CHB therapeutic vaccine design. RV can only identify epitopes encoded by the genome, while other macromolecules, such as carbohydrates and lipids, may also serve as antigens[111]. Vaccine candidates designed using RV and immunoinformatics still require laboratory validation and verification through comprehensive in vitro and in vivo studies. Predictions made with immunoinformatics depend on data quality and algorithm sophistication, so they cannot fully replace experimental research methods[112]. In addition, missing information between the relevance between HBV genotypes, DNA mutations, and clinical manifestations[113] might complicate the selection of the genomic and proteomic datasets for the vaccine design that might be useful for different populations of the patients.

As discussed in the previous section, available tools employ different predictive approaches, leading to variations in data accuracy. Additionally, these tools have other limitations. For example, the C-ImmSim web server can only predict immune responses in individuals with normal immune function[95]. It cannot make predictions for patients with altered immune statuses, such as those with immunodeficiency from HIV or diabetes. To date, only a few studies have demonstrated the effectiveness of CHB therapeutic vaccines designed with RV and immunoinformatics in preclinical and clinical trials. Even with careful design, these therapeutic vaccines have yet to show efficacy in addressing T and B cell exhaustion.

Future research should focus on improving the accuracy and reliability of computational tools by incorporating more comprehensive, high-quality genomic and proteomic datasets from diverse HBV genotypes. Advances in machine learning and artificial intelligence could further enhance epitope prediction and immune response modeling, leading to more robust vaccine candidates. Continued development of these technologies, combined with rigorous in vitro and in vivo validation followed by robust clinical trials, the maintenance of the stability and immunogenicity of the protein, will be essential to overcome current limitations and bring effective therapeutic vaccines for CHB closer to clinical application.

CONCLUSION

This article highlights the significant potential of RV and immunoinformatics in advancing the development of therapeutic vaccines for CHB. By enabling the identification of conserved epitopes across diverse HBV genotypes and streamlining the vaccine design process, these approaches offer promising strategies to address longstanding challenges in CHB treatment. Although their application has yielded encouraging in silico results, challenges related to data quality and experimental validation remain. Future efforts should prioritize enhancing computational accuracy through improved datasets and advanced algorithms, alongside comprehensive in vitro and in vivo validation, to successfully translate promising in silico findings into effective clinical interventions.

ACKNOWLEDGEMENTS

We would like to thank Husna Nugrahapraja and Reza Aditama for the discussion about reverse vaccinology and immunoinformatics.