Identification and retrospective validation of T-cell epitopes in the hepatitis C virus genotype 4 proteome

Abstract

With over 150 million people chronically infected worldwide and millions more infected annually, hepatitis C continues to pose a burden on the global healthcare system. The standard therapy of hepatitis C remains expensive, with severe associated side effects and inconsistent cure rates. Vaccine development against the hepatitis C virus has been hampered by practical and biological challenges posed by viral evasion mechanisms. Despite these challenges, HCV vaccine research has presented a number of candidate vaccines that progressed to phase II trials. However, those efforts focused mainly on HCV genotypes 1 and 2 as vaccine targets and barely enough attention was given to genotype 4, the variant most prevalent in the Middle East and central Africa. We describe herein the in silico identification of highly conserved and immunogenic T-cell epitopes from the HCV genotype 4 proteome, using the iVAX immunoinformatics toolkit, as targets for an epitope-driven vaccine. We also describe a fast and inexpensive approach for results validation using the empirical data on the Immune Epitope Database (IEDB) as a reference. Our analysis identified 90 HLA class I epitopes of which 20 were found to be novel and 19 more had their binding predictions retrospectively validated; empirical data for the remaining 51 epitopes was insufficient to validate their binding predictions. Our analysis also identified 14 HLA class II epitopes, of which 8 had most of their binding predictions validated. Further investigation is required regarding the efficacy of the identified epitopes as vaccine targets in populations where HCV genotype 4 is most prevalent.

Keywords: :

• hepatitis C virus
• genotype 4
• T-cell epitopes
• vaccine
• immunoinformatics
• validation
• Immune Epitope Database

Introduction

Hepatitis C is an infectious inflammatory liver disease caused by the hepatitis C virus (HCV). According to the World Health Organization, around 150 million people worldwide suffer from chronic HCV infection, with an estimated 3–4 million people newly infected each year.¹ There are 6 major genotypes of the virus distributed worldwide; type 1, and to a lesser extent type 2, predominate in the US and Europe.^2,3 HCV genotype 4 (HCV-4) is the predominant type in the Middle East and Central Africa, accounting for >80% of infections in that region.⁴ In Egypt, where the prevalence of HCV infection is the highest worldwide (>15% of the population is chronically infected), more than 90% of the infections are attributed to HCV-4. Annually, 350 000 people are estimated to die as a result of HCV-related liver diseases worldwide.¹

HCV is mainly transmitted via blood contact between a healthy and an infected individual.^1,5 Following the onset of infection, the disease starts with an acute inflammatory phase that lasts for approximately 6 mo, within which 15–30% of patients are able to spontaneously clear the virus. In 70–85% of patients, however, the virus manages to evade the host’s immune response and the disease progresses to a chronic phase, which can last for decades and may eventually culminate in liver cirrhosis, hepatocellular carcinoma, and death.^1,5 The virus itself is an enveloped single sense-stranded RNA virus; its approximately 9600 bases-long genome harbors a single open reading frame that encodes a single precursor polyprotein molecule around 3000 amino acids in length. This molecule is processed during and after its synthesis via viral and host cellular machinery into 3 structural and 7 non-structural (NS) proteins.⁶ HCV replicates via the RNA-dependent RNA polymerase NS5B, which lacks proofreading capability, causing the virus to exhibit a high mutation rate. This allows HCV to evade the host’s adaptive immune response and genetically diversify into a wide array of variants that are classified based on the extent of their genetic variation into genotypes (30–35% variation), and further into subtypes (20–25% variation) and quasi-species.^3,6

Following HCV infection of the liver, the host’s innate immune system responds via the activation of natural killer lymphocytes and the production of type I and III interferons.^7-9 The adaptive humoral response involves the production of neutralizing antibodies that may block viral entry into healthy hepatocytes by targeting the viral envelope’s structural proteins E1 and E2.^6,10 The cellular immune response occurs through the cytotoxic activity of CD8⁺ T cells on infected hepatocytes and the immuno-modulatory and antiviral cytokine-secreting actions of the CD4⁺ T cells.⁷ Although spontaneous viral clearance is possible, in the majority of cases the virus succeeds in evading the immune response by several mechanisms, leading to persistent infection. Chief among the evasion mechanisms is the highly mutable nature of the virus, which allows it to produce escape variants and evade recognition by cells of the adaptive immune system. Other mechanisms include inhibition of type I IFN production, induction of regulatory T cells, and induction of T cell anergy (for a review see refs. 7 and 8). The outcome of HCV infection is believed to be determined by the strength and nature of the host’s immune response during the acute phase of the infection. For example, a strong base of evidence demonstrates that a vigorous, sustained, and multi-specific T cell response (both CD4⁺ and CD8⁺) to infection is associated with viral clearance. On the other hand, a weak, narrow, and short-lasting T cell response is associated with viral persistence.^8,10

Despite the need for one, no HCV vaccine has been developed to date. HCV vaccine development has been hampered by the obstacle of overcoming the virus’ immune escape mechanisms and by other technical obstacles. Due to the fastidious nature of the virus, HCV could not initially be produced in tissue culture systems.^11,12 Furthermore, the sole immuno-competent animal model for investigating the pathogenesis of the disease and the interaction of the virus with the immune system is the chimpanzee. As a result of the empirical, financial, and sometimes ethical constraints associated with the use of this animal in research, only a limited number of chimpanzees could be used in HCV vaccine testing, thus compromising the power of the conclusions reached from those studies.¹¹ A statistical meta-analysis was performed by Dahari et al. in 2010 on previous HCV vaccine studies performed on chimpanzees in order to reach meaningful conclusions from them. The conclusion of the analysis suggested that the inclusion of HCV structural proteins in an HCV vaccine may elicit T-cell responses that mitigate viral clearance.¹³

Despite the challenges associated with HCV vaccine design, several vaccine studies have been conducted on animal models in the past decade, a small number of which have actually progressed to human phase II clinical trials (for a review see refs. 11 and 12). However, the focus of most of these studies has been directed toward HCV genotypes 1 and 2 as targets, and very little attention has been given to HCV-4. A meta-analysis performed on all HCV-related epitopes on the Immune Epitope Database (IEDB) revealed that out of 3444 unique reported epitopes that tested positive in T cell and B cell assays, only 13 were reported for HCV genotypes 4, 5, and 6 combined.¹⁴ HCV-4 is highly prevalent in the Middle East and Central Africa, and is increasingly spreading to Europe and North America with travel and immigration.¹⁵

To address the unmet need for an HCV-4 vaccine, and to fill the wide gap in HCV-4 epitope analysis, we describe herein the identification of a set of immunogenic and conserved T-cell epitopes from 46 HCV-4 full genomic sequences using the validated immunoinformatics toolkit iVAX (EpiVax Inc.). The toolkit is a suite of immunoinformatics tools that can be used for the in silico design of epitope-driven vaccines derived from protein sequences of interest.¹⁶ To carry out this analysis, we collected HCV-4 sequences and uploaded them to the iVAX website.¹⁷ Following epitope prediction and analysis, we describe the comparison of our results to the curated HCV epitope data on the Immune Epitope Database (IEDB) as a method of retrospective validation of our results.¹⁸ Whereas this approach is not possible for novel pathogens, this method provides a cheap and fast alternative to the empirical validation of in silico predictions, thus accelerating the development of epitope-based immunotherapeutics for HCV-4. Furthermore, we performed an in vitro binding assay on 4 of the in silico-predicted HLA class II epitopes as a pilot evaluation of the reliability of our retrospective validation approach.

Results

Identification and validation of HLA class I epitopes

By using the epitope prediction and conservation analysis tools of the iVAX toolkit (EpiMatrix and Conservatrix, respectively), the amino acid sequences of the precursor polyprotein genes of 46 HCV-4 genomic sequences were parsed into 32 286 nine-amino-acid-long peptide frames—from here on referred to as 9-mers—overlapping by 8 amino acid residues, and each 9-mer frame was assessed for its percent conservation across the 46 analyzed sequences and scored for binding to 6 HLA class I supertype alleles covering over 90% of 5 major human populations, namely A*01:01, A*02:01, A*03:01, A*24:02, B*07:02, and B*44:03.¹⁹ Following this analysis, 151 9-mer peptide frames conserved in at least 90% of the HCV-4 sequences (42 out of 46) that had EpiMatrix binding prediction scores (Z-scores) ≥ 1.64 for at least one of the mentioned HLA class I alleles were selected for further analysis. Z-scores ≥ 1.64 for a given HLA allele indicate that the analyzed peptide is most likely a binder of that allele, falling among the top 5% of binders in a random set of peptides.¹⁶ The high conservation of the selected epitopes across the analyzed sequences indicates that they play an important structural or functional role—as part of the structural and non-structural proteins of HCV—in the HCV life-cycle, and hence are difficult to mutate without compromising viral fitness.²⁰

The 151 peptides were then assessed for homology to the human genome using JanusMatrix, a newly developed tool that can predict possible T cell receptor (TCR) cross-reactivity between selected epitopes of interest and epitopes derived from the human genome, human microbiome, or human pathogens; where a cross-reactive (homologous) epitope is defined as one that possesses identical TCR-facing amino acid residues when bound to HLA molecules and may trigger tolerogenic or auto-immune responses when administered in the final vaccine product.²¹ Such epitopes therefore should be excluded from the final vaccine. Based on the analysis, the 151 selected epitopes were filtered down to 90 that have no homologous counterparts—or ‘hits’—in the human genome. These epitopes were considered as good putative candidates for HCV-4 vaccine design. Table 1 lists, as an example, 5 of those epitopes selected at random along with their percent conservation across the analyzed HCV-4 sequences and EpiMatrix Z-scores for binding to each of the analyzed HLA alleles. An analysis of the location of the 90 epitopes on the HCV polyprotein molecule showed that 23 of the epitopes were located in the Core protein region, 21 in the NS3 region, 15 in the NS5B region, 14 in the NS4B region, 5 in the NS5A region, 4 in the E2 region, 2 in the NS2 region, and 1 in the E1 region. No epitopes were found to be located in the p7 and NS4A regions. The 5 remaining epitopes were located at the junctions between the sequences of non-structural proteins.

Table 1. Five HLA class I candidate epitopes for HCV-4 vaccine design

Epitope sequence	Percent conservation	EpiMatrix Z-scores						Hits	Location
		A*01:01	A*02:01	A*03:01	A*24:02	B*07:02	B*44:03
DVVCCSMSY	98%	2.52	−0.14	2.42	0.19	−0.21	0.6	2	NS5A/NS5B
RLLAPITAY	98%	1.86	1.07	2.64	0.09	0.93	2.02	3	NS2/NS3
FWAKHMWNF	96%	−0.21	−0.01	−0.54	3.11	1.16	0.97	1	NS4B
DPRRRSRNL	98%	−1.56	−1.12	−1.46	0.45	2.81	−0.49	1	Core
YLVAYQATV	93%	0.71	3.2	0.6	0.27	0.56	0.76	1	NS3

EpiMatrix Z-scores ≥2.32 indicate that the epitope is from among the top 1% binders of that allele in a random set of peptides, those ≥1.64 indicate that the epitope is from among the top 5% binders. Binding scores ≥ 1.64 are emphasized in bold. The ‘Hits’ column indicates the number of HLA alleles each epitope is predicted to bind to with a Z-score ≥ 1.64, representing a measure of the epitope’s binding promiscuity.

Due to the similarity of polyprotein AA sequences across HCV genotypes, we performed a search on the Immune Epitope Database, a repository of manually curated immune epitope data from peer reviewed literature, patents, and direct submissions from companies and institutions, to find out whether our 90 predicted epitopes have been empirically tested in previous literature for HLA binding and/or the ability to trigger a CD8⁺ cytotoxic T cell response in the context of other HCV genotypes. This approach would allow us to retrospectively validate our in silico predictions without having to empirically validate them from scratch, saving experimentation costs and accelerating the process of HCV-4 vaccine testing. To perform the search, the IEDB was queried for previous records of in vitro HLA binding assays and CD8⁺ T cell assays performed on the exact sequences of the 90 epitopes or on sequences harboring those epitopes as substrings.

The IEDB search revealed that 20 out of the 90 epitopes have no previous records of empirical testing, which suggests that these epitopes are novel and have not been characterized earlier. For the remaining 70 epitopes, entries were found for HLA binding assays and CD8⁺ T cell assays performed on the epitopes. Based on the results of these assays and the nature of the peptides tested in previous literature, the 70 epitopes were divided into three groups: (1) a group of 20 epitopes that tested positive in HLA binding and/or T cell assays as 9-mer peptides (exact match), (2) a group of 22 epitopes that tested positive in HLA binding and/or T cell assays as substrings of longer peptides, and (3) a group of 28 epitopes that have no positive tests described.

A closer look at the results of the first group showed that, for 19 out of the 20 epitopes, the empirically determined restricting HLA alleles to which the epitopes were shown to bind in HLA binding assays or which mediated a positive T cell response in T cell assays were the same HLA alleles that EpiMatrix predicted the HCV-4 epitopes to bind to, or belonged to the same supertype group.¹⁹ For only one epitope, the empirically determined HLA restriction was given a low Z-score by EpiMatrix (0.95); meaning that the EpiMatrix binding prediction for this epitope was a false negative. Overall, the results of the IEDB search retrospectively validate the results of the EpiMatrix predictions. Although entries with negative HLA binding and T cell assay results that contradicted with the EpiMatrix predictions were found for 9 out the 20 epitopes, these results were not used to draw conclusions regarding the validity of the EpiMatrix predictions, owing to the lack of standardization of experimental procedures and the variations between patient cohorts included in T cell assay studies that reported those negative results, and the fact that in chronic HCV patients the immune system frequently fails to develop a response to an epitope due to the capability of the virus to evade the immune system by other mechanisms rather than the epitope itself not being immunogenic. As an example of the validation results, Table 2 lists the EpiMatrix Z-scores for the same 5 epitopes listed in Table 1 and their empirically determined HLA restrictions, in addition to the Z-scores and empirically determined HLA restriction of the epitope that was given a false negative Z-score.

Table 2. Retrospective validation of 6 predicted HLA class I epitopes

Epitope sequence	EpiMatrix Z-scores						Empirically tested HLA restrictions (Supertype)	Ref.	Type of assay	Assay result
	A*01:01	A*02:01	A*03:01	A*24:02	B*07:02	B*44:03
DVVCCSMSY	2.52	−0.14	2.42	0.19	-0.21	0.6	A26 (A01)	22	HLA Binding Assay	Positive
								22	CD8^+ T cell Assay	Positive
								23	CD8^+ T cell Assay	Positive
RLLAPITAY	1.86	1.07	2.64	0.09	0.93	2.02	A*03:01 (A03)	24	HLA Binding Assay	Positive
FWAKHMWNF	−0.21	−0.01	−0.54	3.11	1.16	0.97	A*24:02 (A24)	25, 26, 27	HLA Binding Assay	Positive
DPRRRSRNL	−1.56	−1.12	−1.46	0.45	2.81	-0.49	B7 (B07)	28–32	CD8^+ T cell Assay	Positive
								33	CD8^+ T cell Assay	Negative
YLVAYQATV	0.71	3.2	0.6	0.27	0.56	0.76	A*02:01 (A02)	24, 25, 34-43	HLA Binding Assay	Positive
								24	CD8^+ T cell Assay	Positive
								41, 44, 45	CD8^+ T cell Assay	Negative
							A*02:02 (A02)	38, 39, 43	HLA Binding Assay	Positive
							A*02:03 (A02)	38, 39, 43	HLA Binding Assay	Positive
							A*02:05 (A02)	39	HLA Binding Assay	Positive
							A*02:06 (A02)	38, 39, 43	HLA Binding Assay	Positive
							A*02:07 (A02)	39	HLA Binding Assay	Negative
							A*68:02 (A02)	38, 39, 43	HLA Binding Assay	Positive
							A2 (A02)	38, 46–49	CD8^+ T cell Assay	Positive
QYLAGLSTL	−0.5	0.95	−0.69	2.73	0.55	1.13	A*02:01 (A02)	24	HLA Binding Assay	Positive

The table lists the retrospective validation results for the same epitopes listed in Table 1, in addition to an epitope that was assigned a false negative score by EpiMatrix. The Z-scores of the supertype alleles whose EpiMatrix predictions have been empirically validated are underlined and emphasized in bold. The supertypes of the empirically determined HLA restrictions (between parentheses) follow the Sette and Sidney supertype classification.^19,50 The references listed in the “Ref.” column are the research articles listed on IEDB where the HLA restrictions have been empirically tested by either HLA binding or T cell assays, giving positive or negative results.

For the second epitope group, the presence of positive results from T cell assays performed on larger peptides harboring the epitopes as substrings is an indicator that the positive response can be attributed to one of the component peptide frames of the larger peptides that may have undergone processing before epitope presentation, or due to the longer peptide itself if it is small enough to bind in the HLA class I binding groove. To confirm whether our predicted HCV-4 9-mers are responsible for the maximal positive response or not, truncation analysis must be performed; accordingly, these results cannot be used to validate the EpiMatrix predictions. Also, the results of HLA binding assays performed on longer peptides cannot be used for validation since they represent the binding affinity of the long peptide in itself. For the last group of peptides where only negative entries were found, no conclusions were drawn from the negative binding or T cell data for the same aforementioned reasons.

Identification and validation of HLA class II epitopes

For HLA class II epitope prediction, the same 32 286 9-mer frames parsed and assessed for conservation by Conservatrix across the 46 analyzed sequences were scored by EpiMatrix for binding to 8 HLA-DRB1 alleles covering over 90% of the human population; namely DRB1*01:01, DRB1*03:01, DRB1*04:01, DRB1*07:01, DRB1*08:01, DRB1*11:01, DRB1*13:01, and DRB1*15:01.⁵¹ The results of the Conservatrix and EpiMatrix analyses were then used by another tool, EpiAssembler, to construct extended immunogenic consensus sequence (ICS) peptides which represent the HLA class II epitopes to be included in the final vaccine product owing to the extended nature of HLA class II epitopes in comparison to those of class I. The EpiAssembler tool works by selecting the most conserved and promiscuously immunogenic 9-mer peptide frames analyzed and scored by Conservatrix and EpiMatrix, and extending those peptide frames at their N- and C-termini using the most highly immunogenic and conserved forms of their overlapping 9-mer peptide frames until a maximum user-defined length is reached, or until no more overlapping sequences matching the user-defined criteria exist in the analyzed set of peptides.¹⁶ From the 32,286 analyzed 9-mer frames, EpiAssembler used 156 frames to construct 14 ICS peptides, harboring within their ‘core’ sequence (excluding the 3 AA residues at their N- and C-termini) overlapping 9-mer frames that are collectively conserved in > 90% of the analyzed sequences, and which have a minimum EpiMatrix cluster immunogenicity score (an aggregate score of the EpiMatrix Z-scores of the 9-mer peptides contained within the core of the ICS) of 10. Each of the 14 ICS peptides harbor at least one promiscuous 9-mer frame in its core sequence that is predicted to bind at least 4 of the 8 analyzed HLA-DRB1 alleles, referred to as an EpiBar.¹⁶ As an example of the EpiAssembler results, 4 randomly chosen ICS peptides are listed in Table 3.

Table 3. Four EpiAssembler-constructed Immunogenic Consensus Sequence (ICS) peptides

ICS sequence	% ORF coverage	EpiMatrix cluster score	JanusMatrix cluster score	Location	Region
SQGYKVLVLNPSVAATLGFG	100	35.24	14.23	1246–1265	NS3
VSGIQYLAGLSTLPGNPA	100	18.78	4.58	1769–1786	NS4B
STQQTLLFNILGGWVAAQI	100	11.32	2.23	1802–1820	NS4B
EGAVQWMNRLIAFASRGNHVA	100	23.46	19.42	1910–1930	NS4B

The 9-mer frames of the core sequences of the ICS peptides (excluding the 3 AA of the N- and C-termini) are emphasized in bold and underlined in the ICS sequence column; the 3 terminal AA residues at the N- and C-termini of the peptides are involved in assisting peptide binding to the HLA class II binding cleft. The Open Reading Frame (ORF) coverage column displays the percentage of the analyzed HCV-4 sequences covered by each ICS sequence in terms of the conservation of its core 9-mer frames collectively across the 46 sequences.

JanusMatrix analysis of the ICS peptides revealed that all 14 have some degree of homology with the human genome as evident by their JanusMatrix cluster scores, which are equal to their EpiMatrix cluster immunogenicity scores minus penalty deductions for each cross-reactive hit with the human genome for each of the component 9-mer peptides of the core ICS sequence. The bigger the difference between the EpiMatrix and JanusMatrix cluster scores, the higher the homology with the human genome. Accordingly, ICS peptides with a smaller difference between both scores are more likely to be successfully included into the final vaccine; however, further empirical testing is required to assess the safety and efficacy of each peptide before deciding whether or not to include it in the final product. Table 3 lists the sequences of 4 ICS peptides, their conservation, EpiMatrix and JanusMatrix cluster scores, and their locations along the AA sequence of the HCV polyprotein gene. Table 4 lists, as an example, the EpiMatrix Z-scores and percent conservation of the 9-mer overlapping peptides that were used to construct one of the listed ICS peptides. Analysis of the locations of the ICS peptides revealed that 8 are located in the NS4B region, 3 in the NS5B region, 1 in the Core region, 1 in the NS3 region, and 1 at the junction between the NS2 and NS3 protein sequences.

Table 4. Percent Conservation and EpiMatrix Z-scores of the component 9-mer frames of ICS peptide NS3 (1246–1265)

Frame Sequence	% Conservation	EpiMatrix Z-Scores								Hits
		DRB1*01:01	DRB1*03:01	DRB1*04:01	DRB1*07:01	DRB1*08:01	DRB1*11:01	DRB1*13:01	DRB1*15:01
SQGYKVLVL	02%	−0.08	−0.37	0.14	0.89	0.26	−0.08	1.1	0.45	0
QGYKVLVLN	98%	−1.14	−0.83	−1.37	−0.81	0.73	−0.66	−0.37	−1.09	0
GYKVLVLNP	98%	−0.45	−0.47	0.36	0.15	−0.6	0.24	−0.11	−0.69	0
YKVLVLNPS	98%	1.81	1.93	2.02	1.43	2.6	3.08	1.35	2.1	6
KVLVLNPSV	100%	1.59	0.97	1.31	1.33	0.87	1.61	1.26	1.9	1
VLVLNPSVA	100%	1.9	1.39	1.73	1.48	1.31	1.74	1.49	1.58	3
LVLNPSVAA	100%	2.15	2.04	2.28	1.76	1.6	1.59	2	1.77	6
VLNPSVAAT	100%	0.62	0.02	1.15	0.41	0.35	0.64	0.9	0.19	0
LNPSVAATL	100%	1.97	1.83	1.39	1.9	0.82	1.45	1.68	1.38	4
NPSVAATLG	100%	0.22	−0.39	0.08	−0.69	−0.18	−0.65	−0.18	−0.06	0
PSVAATLGF	100%	−0.51	0.19	0.67	0.24	−0.34	−0.29	0.32	−0.25	0
SVAATLGFG	100%	−0.44	0.95	−0.35	−0.68	0.85	0.22	−0.17	1.05	0

Z-scores ≥2.32 indicate that the 9-mer frame is from among the top 1% binders of that allele in a random set of peptides, those ≥1.64 indicate that the frame is from among the top 5% binders Binding scores ≥ 1.64 are emphasized in bold.The 3 N- and C-terminal AA residues are underlined within the 9-mer frames that harbor them.

In order to retrospectively validate our EpiMatrix HLA class II binding predictions, a search was also performed on IEDB to find out whether the 14 constructed ICS peptides have been empirically tested in previous literature for HLA binding and/or the ability to trigger a CD4⁺ T-helper cell response in the context of other HCV genotypes. The search revealed that none of the 14 peptides has been tested earlier in the form of the exact same sequence, however, all 14 peptides have been tested as substrings of larger peptides, as 70–99% similar peptides (determined by performing a blast 70–90% search; see Materials and Methods section), and/or as smaller fragments of the peptides themselves. To validate our predictions, we only considered the IEDB entries where the sequence of the peptide tested in previous literature contained within its sequence at least one complete 9-mer frame from within the core sequence of our ICS peptides.

The IEDB search revealed that for 8 out of the 14 ICS peptides, entries with positive results were found for HLA binding and T cell assays that validated most of the binding predictions. More precisely, out of 51 EpiMatrix binding predictions performed for the 8 peptides collectively and for which corresponding empirical data was found on IEDB, 39 predictions have been validated as true positive (TP) or true negative (TN). Twelve of those predictions, however, were contradicted by the IEDB results. Before labeling these predictions as false positives (FP), it is suggested that further empirical testing be performed to confirm or refute the EpiMatrix predictions; especially considering that for most of the 12 predictions the contradicting results come from a single corresponding source. As an example of how the validations were performed, Table 5 lists the validation results for the ICS peptide NS3_1246–1265 (SQGYKVLVLNPSVAATLGFG). For each HLA binding or T cell assay entry found on IEDB for the ICS peptide, Table 5 lists the sequence of the peptide that was actually tested or to which the response was attributed, the references that mentioned testing that peptide, and whether the HLA restrictions that tested positive or negative for peptide binding confirm or refute the EpiMatrix predictions for each of the 8 analyzed HLA class II alleles.

Table 5. Retrospective validation of the ICS peptide NS3 (1246–1265)

Type of Assay	Test peptide sequence	Validation results and references
		DRB1*01:01	DRB1*03:01	DRB1*04:01	DRB1*07:01	DRB1*08:01	DRB1*11:01	DRB1*13:01	DRB1*15:01
HLA Binding Assay	GYKVLVLNPSVAAT	TP ^52 ^, ^53	—	TP ^53	TP ^53	TP ^53	TP ^53	TP ^53	TP ^53
	GYKVLVLNPSV	—	—	—	—	—	TP ^54	—	—
	GYKVLVLNPSVAATL	—	—	TP ^55	—	—	—	—	—
	KVLVLNPSVAATLGF	TP ^56	FP ^56	TP ^56	TP ^56	—	TP ^56	—	TP ^56
	GYKVLVLNPSVAATLGFGAY	TP ^57	—	TP ^57	TP ^57	TP ^57	TP ^57	TP ^57	TP ^57
CD4+ T cell Assay	QGYKVLVLNPSVAATLGFGA	—	—	—	—	—	TP ^58	—	—
	GYKVLVLNPSVAATL	—	—	TP ^55 ^, ^59	—	—	—	—	—
	GYKVLVLNPSVAAT	—	—	TP ^52	—	—	TP ^52	TP ^52	—
	VLVLNPSVAATLGFGAYM	—	—	TP ^60	—	—	—	—	—
Overall validation results		TP	FP	TP	TP	TP	TP	TP	TP

The parts of the test peptide sequences that are identical to the core sequence of the ICS peptide are underlined. True positive (TP) and false positive (FP) assignments listed in the table validate/refute the EpiMatrix binding predictions outlined in Table 4. Cells without assignments have no corresponding empirical data for comparison.

It is important to note that in all the IEDB results for the 8 ICS peptides whose EpiMatrix results were mostly validated, the sequences of the peptides that were actually tested or to which the response was attributed were composed mainly of the overlapping 9-mer frames of the core sequences of the ICS peptides; which makes our retrospective validations more reliable. It is also important to note that, where the IEDB search returned results in which the test peptides were 70–99% similar to the ICS peptides, the point mutations to which the dissimilarity was attributed mainly resided in the 3 N- and C-terminal AA residues of the ICS peptides, which are not directly involved in HLA class II binding. This is true for 7 out of the 8 ICS peptides; for the remaining peptide, a point mutation was present in the empirically tested sequences that resided in the core sequence of the ICS (data not shown). For this ICS in particular, 5 out of 7 EpiMatrix binding predictions for which corresponding empirical data were found were contradicted by the IEDB results. This suggests that the mutation affected the binding specificity of the test peptides, and further demonstrates the necessity of performing further empirical analysis on the ICS sequence itself before determining the validity of the EpiMatrix predictions.

Apart from the aforementioned results of the 8 ICS peptides, the results of the IEDB search for the remaining 6 peptides returned only results for CD4⁺ T cell assays where the HLA restriction was not determined. However, for 5 of the 6 peptides, the T cell assay results where positive, which encourages performing assays in the future to determine the HLA restrictions responsible for the positive response.

HLA class II in vitro binding assay

As a pilot assessment of the reliability of the performed retrospective validations for selecting targets for vaccine development, we performed a pilot in vitro binding assay on the 4 HLA class II ICS peptides listed in Table 3 by testing their affinity for binding to 4 purified HLA class II DR molecules, namely DRB1*01:01, DRB1*04:01, DRB1*07:01, and DRB1*15:01 in competition against biotinylated standards. The results of the assay are presented in Table 6. As expected, all ICS peptides were generally found to bind with high affinity to all 4 tested HLA class II molecules, which validates the EpiMatrix binding predictions, and confirms the results of the retrospective validation. The only exception was exhibited by the ICS peptide NS4B_1769–1786 (VSGIQYLAGLSTLPGNPA), which failed to bind to the allele DRB1*07:01 in contradiction to the EpiMatrix binding prediction. Interestingly, this negative result was repeated in HLA binding assays performed in three different references as demonstrated by the results of the retrospective validation.^57,61,62 Given the negative results of the previous assays and the result of this one, it is likely that the EpiMatrix binding prediction for this peptide for the DRB1*07:01 is a false positive. This demonstrates the reliability of the performed retrospective analysis for the validation of the in silico results.

Table 6. HLA class II binding assay results compared with the highest calculated Z-score per allele of the 9-mer frames of the core sequences of each ICS peptide

ICS sequence	DRB1*01:01		DRB1*04:01		DRB1*07:01		DRB1*15:01
	Highest Z-Score	IC50 (nM)	Highest Z-Score	IC50 (nM)	Highest Z-Score	IC50 (nM)	Highest Z-Score	IC50 (nM)
SQGYKVLVLNPSVAATLGFG	2.15	44	2.28	123	1.9	3,010	2.1	711
VSGIQYLAGLSTLPGNPA	2.54	148	2.25	694	1.93	Non-binder	2.25	1239
STQQTLLFNILGGWVAAQI	2.33	215	1.94	721	1.77	122 087	1.6	782
EGAVQWMNRLIAFASRGNHVA	2.54	2726	1.52	20 416	1.55	44 713	2.56	103

The binding affinity scale is divided as follows: (1) Strong Binder: IC₅₀ < 10 000 nM, (2) Moderate Binder: 10 000 < IC₅₀ < 100 000 nM, (3) Weak binder: IC₅₀ > 100 000 nM.

Discussion

In this article, we described the in silico identification of several highly immunogenic and conserved T-cell epitopes in the HCV-4 genome, using the validated iVAX immunoinformatics toolkit, that can be used for the development of an epitope-based vaccine against HCV-4 or for inclusion in an HCV vaccine that covers HCV-4 as one of its targets. We performed a retrospective validation of our results by searching the IEDB for previous records of HLA binding and T-cell assays performed on the identified epitopes, and we further performed a pilot in vitro HLA class II binding assay on 4 selected HLA class II ICS peptides to validate our in silico results and assess the reliability of our retrospective validation. Our in silico predictions lead to the identification of 90 highly conserved and immunogenic HLA class I epitopes that have no homologous counterparts in the human genome bearing identical TCR-facing AA residues, and to the construction of 14 ICS peptides (HLA class II) containing epitopes that have varying degrees of homology to epitopes derived from the human genome. Retrospective validation of the selected 90 HLA class I epitopes revealed that 20 epitopes have no record of previous empirical testing on IEDB, thus representing good targets for future investigation and testing of their role in the interaction between the immune system and HCV-4 and their efficacy as vaccine targets, and validated the results of the in silico predictions for 19 of the remaining 70 epitopes. As for the HLA class II ICS peptides, the IEDB search validated, for 8 out the 14 peptides, 39 EpiMatrix binding predictions out of 51 for which corresponding empirical data was found. The pilot in vitro HLA binding assay performed on 4 ICS peptides validated the EpiMatrix binding predictions, except for 1 peptide whose prediction for binding to the DRB1*07:01 allele was shown to be a false positive, in concurrence with the results of the retrospective validation. A flowchart summarizing the in silico epitope identification process and retrospective validation is presented in Figure 1; Table 7 summarizes the results of each step of this process in selecting the HCV-4 vaccine candidates.

Figure 1. Flowchart summarizing the identification and analysis of HLA class I and II T cell epitopes from HCV-4 and the retrospective validation of the results using the IEDB.

Table 7. Summary of the results of the in silico epitope identification and retrospective validation process

In silico analysis results	HLA class I	HLA class II
Total 9-mer frames scored by EpiMatrix and Conservatrix	32 286	32 286
Epitopes selected for further analysis	151	14*
Epitopes with no homologous hits in the human genome	90	0
Epitopes with matches on IEDB	70	14
Novel Epitopes (0 IEDB Matches)	20	0
Epitopes with retrospectively validated binding predictions	19	8

* HLA class II epitopes are the ICS peptides.

Analysis of the locations of the identified highly conserved HLA class I and II epitopes revealed that most of them lie in the Core, NS3, NS4B, and NS5B regions of the HCV polyprotein sequence. Twenty-three HLA class I epitopes and one ICS peptide were located in the Core protein region, the building block of the viral nucelocapsid. Twenty-one HLA class I epitopes and one ICS peptide were located in the NS3 serine protease-helicase NTPase protein region. Fourteen HLA class I epitopes and 8 HLA class II ICS peptides were found to be located in the NS4B region; which is involved in membranous web formation, an altered structure of the ER membrane that serves as a scaffold for the formation of the viral replication complex.⁶ Finally, 15 HLA class I epitopes and 3 class II ICS peptides were found to be located in the NS5B region. The key structural and functional roles that the sequences of those proteins play in the HCV life cycle make it less likely for these sequences to mutate in the course of viral evolution.²⁰ This explains why the highest proportion of the HLA class I and II epitopes that we identified are located within those sequences. On the other hand, very few epitopes were found in the E1 and E2 envelope glycoproteins in comparison to their size; being on the surface of HCV viral particles, these proteins exhibit frequent mutations in order to evade the host’s humoral immune response.⁶³

The presence of epitopes in the final vaccine product that are homologous to epitopes derived from the human genome can potentially trigger a suppressive immune response to the vaccine, mediated by T regulatory cells, or even a detrimental autoimmune response. Accordingly, such epitopes should not be included in the final vaccine. To screen for such epitopes in our predicted data set, we used the novel JanusMatrix tool, which searches for epitopes derived from the human genome that bear identical TCR-facing AA residues while allowing the MHC-facing residues (referred to as the ‘agretope’) to vary, with the variant agretope still predicted by EpiMatrix to bind to the same HLA alleles to which the screened epitopes were predicted to bind. Homology to the human genome was recently discovered in some commensal viruses using the same tool, and was found to differ quite significantly from the level of homology to self for ‘hit-and-run’ viruses like Variola, Ebola, and Marburg.⁶⁴ In reality, TCRs are able to recognize their target epitopes by AA residue side-chain availability; this means that JanusMatrix cannot predict all putative cross-reactive epitopes derived from the human genome based on sequence similarity alone, however it can still help narrow down the set of epitopes that would be later tested experimentally to prove whether or not they trigger suppressive or autoimmune responses.²¹ For our predicted HLA class I set, we stringently selected only the epitopes with zero cross-reactive hits in the human genome. For class II ICS peptides, we could not filter the epitopes based on the same stringent criteria due to their extended nature, which is required for proper HLA class II binding and the promiscuous coverage of the analyzed HLA-DR alleles using the most highly conserved forms of the 9-mer frames utilized for ICS construction. Filtering out 9-mer frames from the ICS peptides that exhibit similarity to the human genome may compromise allelic coverage, HCV-4 subtype coverage, and binding to HLA class II. A JanusMatrix analysis of a large set of empirically characterized T-effector cell epitopes has previously demonstrated that, within a certain ratio of cross-reactive hits per number of AA residues in the human genome database, a T cell epitope exhibiting cross-reactivity to the human genome may still successfully trigger a T-effector response.²¹ Accordingly, the decision of which peptides to use for the vaccine should be taken following future empirical analysis.

In conclusion, we have demonstrated the in silico identification and selection of a set of highly conserved and immunogenic T cell epitopes from within the HCV-4 proteome as a first step toward the development of an epitope-driven vaccine against this viral genotype using one of the most accurate T cell epitope prediction tools available online.⁶⁵ Compared with the traditional method of empirically testing all possible overlapping peptide frames in a target antigenic sequence for immunogenicity prior to the validation of their efficacy as vaccine targets in vivo, the initial in silico prediction of immunogenic epitopes narrows down the in vivo validation process to only a selected set of epitopes, thus saving the time and costs associated with the analysis. Furthermore, we performed a retrospective validation of our results using the repository of empirical data on the IEDB; this approach provides a faster and cheaper alternative to performing empirical validations from scratch, which can accelerate vaccine development when outbreaks of a different strain of the same pathogen occurs. Using this approach, the production of population-specific or individualized vaccines against pathogens such as HCV, HIV, and Coronaviruses may be facilitated. A recent example of utilizing this approach was performed by Duvvuri et al., where an IEDB search was performed to validate the immunogenic potential of in silico identified HLA class II epitopes from the novel avian influenza H7N9 virus relative to other human influenza A viruses.⁶⁶ In the future, we intend to empirically characterize the 20 novel HLA class I epitopes identified by the iVAX analysis, and to further elucidate the nature of the immune response toward the identified HLA class I and II epitopes in patient populations where HCV-4 is highly prevalent, as an assessment of the efficacy of those epitopes as vaccine targets. Although the future holds promise for increasing the chances of success of HCV therapy with the continuing approval of novel direct-acting antiviral drugs to enter the market (e.g., Sofosbuvir),⁶⁷ these new drugs remain very expensive, making it hard for a large sector of the infected population, especially in some countries of the Middle East and Africa, to afford treatment. The development of an effective preventative or therapeutic HCV vaccine can have a strong impact on HCV prevention and therapy in such populations.

Materials and Methods

HCV-4 Polyprotein sequence collection

The amino acid sequences of 46 non-redundant HCV-4 precursor polyprotein genes were retrieved from the Los Alamos HCV sequence database (LANL)⁶⁸ and uploaded to the iVAX toolkit server. The accession numbers of the full genomic HCV-4 sequences harboring those genes are: DQ418785, DQ418782, DQ418783, DQ418784, DQ418787; DQ418788; DQ418789, DQ516084, DQ988073, DQ988074, DQ988075, DQ988076, DQ988077, DQ988078, DQ988079, GU814265, NC_009825, FJ025854, FJ025855, FJ025856, FJ462435, FJ462436, DQ418786, DQ516083, EU392172, FJ462437, EF589160, EF589161, EU392169, EU392170, EU392174, EU392175, FJ462432, EU392171, EU392173, FJ462438, FJ839870, FJ462433, FJ462441, FJ462440, FJ462431, FJ462434, FJ462439, FJ839869, HQ537008, and HQ537009.

Epitope prediction and analysis using the iVAX toolkit

Following sequence upload, the tools of the iVAX toolkit were consecutively used in order to identify HLA class I and II epitopes from the HCV-4 sequences and filter the results to a selected set of epitopes that can potentially be used to develop the vaccine. The Conservatrix tool was used first; it was set to parse the sequences into 9-mer peptide frames overlapping by 8 amino acids and determine the percent conservation of each frame across the 46 sequences. The matrix-based T cell epitope prediction tool EpiMatrix was then used to score the parsed 9-mer frames for binding to the 6 HLA class I supertype alleles A*01:01, A*02:01, A*03:01, A*24:02, B*07:02, and B*44:03, and the 8 HLA-DRB1 alleles DRB1*01:01, DRB1*03:01, DRB1*04:01, DRB1*07:01, DRB1*08:01, DRB1*11:01, DRB1*13:01, and DRB1*15:01. To construct the HLA class II ICS peptides, the EpiAssembler tool was set to construct peptides of a maximum length of 21 AA residues that harbor at least one 9-mer frame that is immunogenic and conserved in ≥ 90% of the analyzed HCV-4 sequences, and that possess a minimum EpiMatrix Cluster Immunogenicity score of 10; an EpiMatrix cluster score of 10 or higher indicates that the constructed ICS has a significant immunogenic potential.¹⁶ The JanusMatrix tool was then used to identify possible homology between selected epitopes and ICS peptides, and epitopes derived from the human genome by searching for 9-mer epitope frames in a human protein database pre-uploaded into JanusMatrix that bear identical TCR-facing residues and were predicted by EpiMatrix to bind to the same HLA alleles as the analyzed HCV epitopes.²¹ Identifying the locations of the identified epitopes on the HCV genome was performed using the ‘Sequence Locator’ tool of the HCV Los Alamos database.

IEDB epitope search

The Immune Epitope Database (IEDB) was queried for the selected epitopes and ICS peptides via the home page search tool (http://www.iedb.org/). The database was queried for available data on MHC binding, MHC elution, and T cell response assays performed on epitopes/peptides that exactly match the query epitopes or harbor those epitopes as substrings. For HLA class II ICS peptides, the search was expanded to include peptide sequences that are 70–90% identical to the constructed ICS peptides using the BLAST 70–90% search option. The immune recognition context parameters of the search were confined to MHC class I- or II-restricted matches (according to the queried epitope), and T cell assays were confined to those performed on blood samples from human subjects by selecting Homo sapiens (id:9606) as the host organism.

HLA class II in vitro binding assay

In four 96-well reaction plates (Fisherbrand), each non-biotinylated ICS test peptide (21st Century Biochemicals) diluted in DMSO (Sigma-Aldrich) was incubated, in triplicates and over a wide range of concentrations, with the purified HLA-DR molecules DRB1*01:01, DRB1*04:01, DRB1*07:01, and DRB1*15:01 (Benaroya Research Institute)—one molecule per plate—and standard biotinylated peptides at 25 nM (21st Century Biochemicals) in 50 μL per well of pH 5.4 citrate-phosphate buffer solution at 37 °C overnight. After the binding mixtures reached a steady equilibrium, they were transferred to 96-well ELISA coated with anti-HLA-DRA antibodies (L243, BioXCell) and were incubated for 2.5 h at 37 °C. After the incubation, the plates were thoroughly washed with PBS+0.05% Tween-20, developed by adding Europium-labeled Streptavidin (Perkin-Elmer), and analyzed using a Victor³V Microtiter Plate Reader (Perkin-Elmer, MA, USA). IC₅₀ values were then calculated using SigmaPlot 11.1 software.

Abbreviations:
HCV	=	hepatitis C virus
HCV-4	=	hepatitis C virus genotype 4
UTR	=	untranslated region
NS	=	non-structural
TCR	=	T-cell Receptor
IEDB	=	Immune Epitope Database
AA	=	amino acid
HIV	=	human immunodeficiency virus

Disclosure of Potential Conflicts of Interest

Coauthors A.D.G. and F.T. are employees of EpiVax, a vaccine and therapeutic design company, and A.D.G. is a majority stockholder. These authors recognize the presence of potential conflicts of interest and affirm that the information represented in this paper is original and unbiased observations. K.A.H., H.M.E.A., A.H.G., and J.D. declare no competing interests.

Acknowledgments

This work has been funded by a grant from the American University in Cairo to Dr Hassan Azzazy and an ARRA supplement to the U19 grant to Prof. Anne De Groot. The iVAX toolkit has been made available for academic use through a unique arrangement with the University of Rhode Island. Thanks to Sheila Chandran for assisting with the in silico analysis and Lana Abdel-Hady for assisting with the artwork.

10.4161/hv.29177