Escape from HLA-B*08-Restricted CD8 T Cells by Hepatitis C Virus Is Associated with Fitness Costs

Abstract

The inherent sequence diversity of the hepatitis C virus (HCV) represents a major hurdle for the adaptive immune system to control viral replication. Mutational escape within targeted CD8 epitopes during acute HCV infection has been well documented and is one possible mechanism for T-cell failure. HLA-B*08 was recently identified as one HLA class I allele associated with spontaneous clearance of HCV replication. Selection of escape mutations in the immunodominant HLA-B*08-restricted epitope HSKKKCDEL_1395-1403 was observed during acute infection. However, little is known about the impact of escape mutations in this epitope on viral replication capacity. Their previously reported reversion back toward the consensus residue in patients who do not possess the B*08 allele suggests that the consensus sequence in this epitope is advantageous for viral replication in the absence of immune pressure. The aim of this study was to determine the impact of mutational escape from this immunodominant epitope on viral replication. We analyzed it with a patient cohort with chronic HCV genotype 1b infection and in a single-source outbreak (genotype 1b). Sequence changes in this highly conserved region are rare and selected almost exclusively in the presence of the HLA-B*08 allele. When tested in the subgenomic replicon (Con1), the observed mutations reduce viral replication compared with the prototype sequence. The results provide direct evidence that escape mutations in this epitope are associated with fitness costs and that the antiviral effect of HLA-B*08-restricted T cells is sufficiently strong to force the virus to adopt a relatively unfavorable sequence.

The inherent sequence diversity of hepatitis C virus (HCV) represents a major hurdle for the adaptive immune system to control viral replication. The virally encoded RNA-dependent RNA polymerase of HCV lacks a proofreading function and is therefore characterized by ongoing high error rates (7). With these error rates—and a high turnover rate of an estimated 10¹² virions per day (20)—theoretically, all possible mutations in every single position of the genome will be generated in one infected host every day. Many mutations are likely to be disadvantageous or even deleterious for replication, and these variants are eliminated in a negative selection process. However, some mutations may have no negative impact on replication, and a few may even be beneficial and confer a replication advantage. Variants harboring these beneficial mutations will outcompete others in a dynamic process of continuous positive selection.

Recent studies suggest that the adaptive immune system plays an important role in this selection process. Continuous selection of viral variants in the envelope (E2) protein conferring resistance to neutralizing antibodies has been shown (30). Moreover, mutational escape within targeted CD8 epitopes during acute HCV infection has been well documented and is an important mechanism for T-cell failure (6, 27, 28). Some of these CD8 escape mutations are reproducibly selected in subjects sharing the same HLA allele and are visible at a population level as HLA “footprints” (11, 21, 29). However, little is known about the impact of escape mutations on the replication capacity of HCV. In human immunodeficiency virus (HIV), the dramatic fitness costs of some CD8 escape mutations have been described (4, 18, 24). In HCV, reversion of CD8 escape mutations upon transmission to a new host who does not possess the relevant restricting HLA allele has been observed, suggesting the significant fitness costs of some mutations (21, 28).

The aim of this study was to determine the impact of sequence variation in an immunodominant CD8 epitope on replication capacity. We analyzed the HLA-B*08-restricted cytotoxic T lymphocyte epitope HSKKKCDEL_1395-1403 in HCV NS3 with a cohort with chronic HCV genotype 1b infection and with patients from a single-source genotype 1b outbreak. Sequence changes in this highly conserved epitope region are rare and selected only in the presence of the HLA-B*08 allele. The impact of sequence variation in this epitope on viral replication capacity was assessed and compared to the impact of an escape mutation in a second CD8 epitope in NS3 (HLA-B*35 restricted) and a drug resistance mutation selected in the presence of an HCV-specific protease inhibitor.

MATERIALS AND METHODS

Patients.

Patients were recruited from the hepatology outpatient clinics at the university hospitals of Freiburg and Essen and the clinic for addiction medicine at the University of Essen after informed consent and in agreement with federal guidelines and the local ethics committee. Additional samples collected between 1995 and 1998 were obtained from patients infected by an isolate from a single source (AD78) in contaminated immunoglobulin preparations between 1978 and 1979 (31). All patients were HLA typed using standard serological or molecular techniques.

Amplification and sequencing of HCV RNA.

Viral RNA extraction, reverse transcription, and amplification by nested PCR were done as previously described (29) with the following primers: HCV1b 4a-F (5′-ATGGAAACTACYATGCGG [nucleotides (nt) 3942 to 3959 as aligned to H77]) and HCV1b 4d-R (5′-CCAGGTGCTVGTGACGACC [nt 5303 to 5321]) for the first round of PCR and HCV1b-4b-F (5′-AAGGACCATCACCACGGG [nt 4187 to 4204]) and HCV1b-4c-R (5′-GGTGTATTTAGGTAAGCCCG [nt 4959 to 4978]) for the second round of PCR. PCR products were population sequenced on an ABI 3730 XL automated sequencer. Sequences were aligned and edited with CodonCode Aligner (Dedham, MA) and Se-Al (http://evolve.zoo.ox.ac.uk).

Construction of variant subgenomic replicons.

The bicistronic subgenomic HCV replicon Con1/ET (pFK PI-luc/NS3-3′/Con1/ET) was previously described (17) and served as a backbone for the generation of a series of variant replicons. Mutations were introduced by site-directed mutagenesis utilizing the QuickChange II XL site-directed mutagenesis kit (Stratagene), according to the manufacturer's instructions, in an intermediate plasmid containing a 3,520-bp fragment from the original replicon. The engineered mutations were confirmed by sequencing, and the resulting plasmid was digested with BssHII and HindIII to obtain a 2,857-bp fragment that contains the mutated epitope. This fragment was ligated into the original pFK PI-luc/NS3-3′/Con1/ET replicon, which was treated with the same restriction enzymes. The correct sequences of the insert and the ligation sites were again confirmed. The defective replicon pFK PI-luc/NS3-3′/GND was previously described (17) and served as a negative control.

In vitro transcription.

Plasmid DNA was restricted with PvuI and ScaI (New England Biolabs), and after purification by phenol chloroform extraction and ethanol precipitation, the DNA pellet was dissolved in RNase-free water. In vitro transcription was performed at 37°C for 3 h with 1 μg of linearized and purified plasmid DNA using the T7 Megascript high-yield transcription kit (Ambion). After termination of the transcription reaction, RNA was purified by Trizol (Invitrogen), precipitated with isopropanol, and dissolved in diethyl pyrocarbonate-treated water (Ambion).

Transfection and transient replication assay.

Huh7-Lunet cells were transfected as previously described (17). Briefly, 4 × 10⁶ cells were suspended in a cytomix buffer and mixed with 5 μg of in vitro-transcribed RNA and pulsed at 270 V and 960 μF with the Gene Pulser apparatus (Bio-Rad). Transfected cells were resuspended in complete Dulbecco's modified eagle medium (supplemented with 2 mM l-glutamine, nonessential amino acids, 100 U penicillin per ml, 100 μg streptomycin per ml, 10 mM HEPES buffer, and 10% fetal calf serum) seeded on a six-well plate. Luciferase activities after 4 h, 24 h, 48 h, and 72 h in transfected cells were determined using the Luclite luminescence reporter gene assay system (Perkin Elmer) with the TopCount microplate scintillation and luminescence apparatus (Perkin Elmer). Results were analyzed in relation to the luciferase activity after 4 h (set to 100%).

Culture of G418-resistant cell clones.

The previously described plasmid pFK PI-neo/NS3-3′/Con1/ET served as a backbone for a second series of variant vectors. Single point mutations coding for amino acid changes K1397R and K1398R in the HSKKKCDEL_1395-1403 epitope were introduced as described previously, and the resulting plasmid served as a template for RNA transcription in vitro followed by transfection of Huh7-Lunet cells. In a second series of experiments, the RNA was mixed with prototype RNA in a ratio of 9:1 before transfection. G418-resistant clones were selected in complete Dulbecco's modified Eagle medium in the presence of 0.5 mg/ml G418 and cultured for 1 year. An aliquot of the cells was treated with 2 μl of cell lysis buffer (Ambion) and directly used as a template for reverse transcription and amplification, by following the same protocol as described previously. PCR products were directly purified and cloned (Topo TA; Invitrogen). Individual clones were sequenced on an ABI 3730 XL automated sequencer.

Polyclonal antigen-specific expansion of T cells.

Peripheral blood mononuclear cells (PBMC) (4 × 10⁶) were resuspended in 1 ml of complete medium (RPMI 1640 containing 10% fetal calf serum, 1% streptomycin/penicillin, and 1.5% HEPES buffer 1 mol/liter) and stimulated with peptide HSKKKCDEL or variants (10 μg/ml) and anti-CD28 (0.5 μg/ml; BD PharMingen). On days 3 and 7, 1 ml of complete medium and recombinant interleukin-2 (20 U/ml; Hoffmann-La Roche) was added. On day 10, the cells were tested for gamma interferon (IFN-γ) secretion after 5 hours of stimulation with the prototype or variant peptides by intracellular cytokine staining as previously described (28).

Nucleotide sequence accession numbers.

All sequences were submitted to GenBank and are available under accession numbers EU078792 through EU078840 and FJ002530 through FJ002569.

RESULTS

Mutations in the HSKKKCDEL_1395-1403 epitope are rare and selected in the presence of the HLA-B*08 allele.

Previous studies report selection of escape mutations in the HSKKKCDEL_1395-1403 epitope in HLA-B*08-positive subjects (11, 28, 29). To analyze this with our cohort of subjects with chronic genotype 1b infection, a region spanning this epitope was amplified and sequenced. Figure 1 shows all sequences, derived from 51 subjects, aligned to the genotype 1b consensus sequence retrieved from the Los Alamos National Laboratory (LANL) HCV database (16). Eight out of 51 sequences (15.7%) differ from the consensus sequence in the B*08-epitope region. As expected, differences from the consensus sequence are significantly more frequent in subjects expressing the HLA-B*08 allele than in HLA-B*08-negative subjects. Six of 15 HLA-B*08-positive subjects (40.0%) harbor a virus with a variant sequence. In contrast, this region is highly conserved in HLA-B*08-negative subjects, where only 2 out of 36 subjects (5.6%) are infected with an isolate that differs in this epitope region (P = 0.004, Fisher's exact test). In a phylogenetic tree, all variants from HLA-B*08-positive subjects fall into different lineages, supporting the idea that these mutations occurred independently (data not shown).

FIG. 1.

Alignment of HCV sequences (amino acids 1390 to 1413 of the polyprotein) from subjects with chronic HCV genotype 1b infection. The HLA-B*08-restricted CD8 epitope HSKKKCDEL_1395-1403 is highlighted in gray. Viral loads of the samples are indicated (in IU/ml). ND, no data. Dots indicate that the residue corresponds to the consensus residue.

The frequencies of these polymorphisms in our cohort of subjects with chronic genotype 1b infection are similar to the frequencies observed in the LANL HCV database (Table 1). This region is overall highly conserved, with 93.3% of all genotype 1b sequences and 88.8% of all genotype 1a sequences carrying the prototype sequence. Interestingly, in the remaining sequences, mutations are heterogeneous and are located in different positions within the epitope. In genotype 1b, the most frequent mutation is located in position 3 of the epitope (K1397R), which is present in 4 out of 51 sequences in our cohort (7.8%) and 7 out of 238 sequences in the LANL database (2.9%). Other variants that were reproducibly found in the LANL database in genotype 1b are K1398R (0.9%) and L1403F (0.9%). In genotype 1a, the most frequent polymorphism is the K1398R substitution (8 out of 232; 3.4%). Other reproducible substitutions in the LANL database in genotype 1a are K1397R (1.7%), D1401A (1.3%), E1402D (1.3%), L1403F (1.3%), and L1403V (1.7%).

TABLE 1.

Frequency of sequence variants in the B8 HSKKKCDEL_1395-1403 epitope and B35 HPNIEEVAL_1359-1367 epitope

Epitope	Variants at amino acid sitesa									GT1b cohort		GT1b databaseb		GT1a databaseb
										No. of indicated sequences or variants	% of total	No. of indicated sequences or variants	% of total	No. of indicated sequences or variants	% of total
B8 1395-1403
Consensus sequence	H	S	K	K	K	C	D	E	L	43	84.3	222	93.3	206	88.8
Frequent variants	.	.	R	.	.	.	.	.	.	4	7.8	7	2.9	4	1.7
	.	.	.	R	.	.	.	.	.	2	3.9	2	0.9	8	3.4
	.	.	.	.	.	.	A	.	.	0		0		3	1.3
	.	.	.	.	.	.	.	D	.	0		1	0.4	3	1.3
	.	.	.	.	.	.	.	.	F	1	2.0	2	0.9	3	1.3
	.	.	.	.	.	.	.	.	I	1	2.0	1	0.4	0
	.	.	.	.	.	.	.	.	V	0		0		4	1.7
Rare variants	P	.	.	.	.	.	.	.	.	0		1	0.4	0
	.	.	.	S	.	.	.	.	.	0		1	0.4	0
	.	.	.	.	.	Y	.	.	.	0		1	0.4	0
.	.	.	.	R	.	.	E	D	.	0		0		1	0.4
B35 1359-1367
Consensus	H	P	N	I	E	E	V	A	L	36	76.6	212	83.5	259	87.8
Frequent variants	.	S	.	.	.	.	.	.	.	5	10.6	16	6.3	14	4.7
	.	H	.	.	.	.	.	.	.	1	2.1	1	0.4	5	1.7
	.	A	.	.	.	.	.	.	.	0		1	0.4	3	1.0
	.	.	S	.	.	.	.	.	.	0		4	1.6	3	1.0
	.	.	.	.	Q	.	.	.	.	0		0		2	0.7
	.	.	.	.	.	.	A	.	.	0		4	1.6	3	1.0
	.	.	.	.	.	.	I	.	.	0		6	2.4	0
	.	.	.	.	.	.	.	G	.	2	4.3	10	3.9	6	2.0
Rare variants	S	.	.	.	Q	.	.	.	.	0		1	0.4	1	0.3
	S	.	.	.	.	.	I	.	.	1	2.1	0		0
	L	.	.	.	.	.	.	.	.	0		1	0.4	0
	.	.	.	.	Q	.	.	G	.	1	2.1	2	0.8	1	0.3
	.	.	.	.	D	.	.	.	.	1	2.1	0		0
	.	.	.	.	.	.	S	.	.	0		2	0.8	0
	.	.	.	.	.	.	I	G	.	0		1	0.4	0
	.	.	.	.	.	.	.	P	.	0		1	0.4	0

^aDots indicate that the residue corresponds to the consensus residue.

^bData retrieved from the LANL HCV database (http://hcv.lanl.gov).

All subjects with HCV genotype 1b infection in our cohort were recruited during the chronic phase of the disease. Therefore, the exact sequence at the time of transmission is unknown, and we cannot exclude that a virus harboring a polymorphism was transmitted to the next host. A unique opportunity to overcome this problem is analysis of single-source outbreaks where the viral inoculum sequence is known. Ray et al. previously published sequence data from an Irish cohort that was infected by an isolate from a single source in contaminated anti-D immune globulins in 1977 and 1978 (21). Interestingly, the inoculum sequence harbored the K1397R mutation (HSRKKCDEL). This residue remained fixed in all HLA-B*08-positive subjects but has reverted back to the prototype in 50% of HLA-B*08-negative subjects 18 to 22 years after the transmission event (Table 2). Here we analyzed sequences from a very similar cohort also infected by contaminated anti-D immune globulins in a single-source outbreak in 1978 and 1979 (reported in reference 31). In this outbreak, the source virus (AD78) had the prototype sequence HSKKKCDEL (22). Between 16 and 20 years after transmission, 5 out of 11 (45.5%) HLA-B*08-positive subjects carried a mutation in the HSKKKCDEL_1395-1403 epitope (Table 2). In contrast, none of the HLA-B*08-negative subjects developed a mutation in this region. Interestingly, the most frequent polymorphism in genotype 1b in the LANL HCV database (K1397R) was present only once as a double mutant, together with the K1398R substitution.

TABLE 2.

Evolution in the B8 HSKKKCDEL_1395-1403 and B35 HPNIEEVAL_1359-1367 epitopes in two single-source outbreaks

Epitope	East German cohort										Irish cohortb
	Variants at amino acid sitesa									No. of subjects with sequence/total no. of subjects	Variants at amino acid sitesa									No. of subjects with sequence/total no. of subjects
B8 1395-1403
Inoculum sequence	H	S	K	K	K	C	D	E	L		H	S	R	K	K	C	D	E	L
B*08 positive	.	.	.	.	.	.	.	.	.	6/11	.	.	.	.	.	.	.	.	.	6/6
	.	.	R	R	.	.	.	.	.	1/11
	.	.	.	R	.	.	.	.	V	1/11
	.	.	.	R	.	.	.	.	.	1/11
	.	.	.	.	.	.	.	.	V	1/11
	.	.	.	.	.	.	.	.	F	1/11
B*08 negative	.	.	.	.	.	.	.	.	.	0/31	.	.	K	.	.	.	.	.	.	8/16
											.	.	.	.	.	.	.	.	.	8/16
B35 1359-1367
Inoculum sequence	H	P	N	I	E	E	V	A	L		H	P	N	I	E	E	V	A	L
B*35 positive	.	.	.	.	.	.	.	.	.	6/9	.	S	.	.	.	.	.	.	.	3/3
	.	S	.	.	.	.	.	.	.	2/9
	.	H	.	.	.	.	.	.	.	1/9
B*35 negative	.	.	.	.	.	.	.	.	.	0/32	.	.	.	.	.	.	.	.	.	0/19

^aDots indicate that the residue corresponds to the inoculum residue.

^bData was previously published in reference 21.

In summary, viral sequence polymorphisms in the HSKKKCDEL_1395-1403 epitope region are overall rare in subjects with chronic HCV genotype 1b infection and are associated with expression of the HLA-B*08 allele.

We similarly analyzed a second CD8 epitope restricted by HLA-B*35 (HPNIEEVAL_1359-1367) located 36 amino acids N terminal of the HSKKKCDEL_1395-1403 epitope. This epitope represents the immunodominant target in HLA-B*35-positive subjects during acute HCV infection (5), and mutational escape was observed with a P1360S mutation at the anchor residue (6). In our cohort of patients with chronic HCV infection, 11 of 47 sequences (23.4%) differ from the prototype sequence in the HPNIEEVAL_1359-1367 epitope, with the P1360S substitution being the most frequent (10.6%), which is also reproducible in other HCV genotype 1b and genotype 1a sequences from the public database (Table 1). Of note, even though this mutation was significantly more frequent in isolates from HLA-B*35-positive subjects in a cohort with chronic HCV genotype 1a infection from Boston (29), it is not enriched in HLA-B*35-positive subjects from our cohort with chronic HCV genotype 1b infection. Interestingly, in HCV genotype 1b isolates from the database, sequence variants in this region are more frequent than in the HSKKKCDEL_1395-1403 epitope (B35, 83.5% prototype; B8, 93.3% prototype; Table 1) and are stable in numerous HLA-B*35-negative patients in our cohort, indicating that substitutions are better tolerated in the HPNIEEVAL_1359-1367 epitope. We also analyzed this epitope with the German single-source outbreak where the source contained the prototype sequence in this epitope. Similar to observations in the Irish cohort (21), mutations in this epitope are selected only in the presence of the HLA-B*35 allele (Table 2). A P1360S mutation is selected in two patients, and a P1360H mutation is selected in one additional patient. Therefore, the selection of mutations in the HPNIEEVAL_1359-1367 epitope is reproducible in HLA-B*35-positive subjects from two single-source genotype 1b outbreaks.

Replication capacity of sequence variants in the HSKKKCDEL_1395-1403 epitope.

In order to address the impact of sequence variation in the HSKKKCDEL_1395-1403 epitope on replication capacity, we tested various mutations in a transient replication assay utilizing the genotype 1b-based subgenomic replicon Con1/ET as a backbone (17). In three independent experiments, we tested mutations that we observed in naturally circulating isolates (K1397R, K1398R, and L1403F) as well as additional mutations in positions 1397 and 1398 that are also the result of single nucleotide substitutions but are not observed in the database. Mutations that are not observed in circulating isolates (K1397M, K1397Q, K1398M, and K1398Q) do not support replication in this assay (Fig. 2A). Similarly, all other possible amino acid changes in position 1398 that are the result of single nucleotide mutations in the corresponding codon but not observed in vivo (K1398N, K1398T, and K1398E) completely abolished replication (data not shown). In contrast, all mutations in the HSKKKCDEL_1395-1403 epitope region that are observed in circulating isolates in vivo were replication competent in the transient replication assay (Fig. 2A). Interestingly, analysis of the replication efficiency of these competent constructs revealed a clear hierarchy. The original Con1/ET replicon bearing the prototype sequence showed reproducibly the highest replication level. The K1397R or K1398R substitution diminished replication to a level of 60.7% ± 11.9% standard deviation (SD) or 57.8% ± 16.3% SD, respectively, compared to the original Con1/ET construct 72 h after transfection (Fig. 2A). Less impairment of replication was observed for the L1403F variant that replicated at levels of 90.3% ± 0.8% SD than for the original Con1/ET after 72 h. We also tested an additional construct with a silent mutation in position K1399 as a control. As expected, the replication level was similar (100.4% ± 6.2% SD) to that of the original Con1/ET (Fig. 2A).

FIG. 2.

Relative replication efficiency of different variants compared to the Con1/ET replicon. (A) Luciferase activities after 48 h (top) and 72 h (bottom) are shown relative to the Con1/ET prototype (set to 100%). (B) A second series of experiments with independently constructed replicons was made. Relative luciferase activities after 48 h (top) and 72 h (bottom) are shown. Error bars represent SDs of three independent experiments.

In a second series of experiments that used for RNA transcription plasmid templates that were independently constructed, similar results have been obtained (Fig. 2B). Again, the original Con1/ET construct reproducibly showed the highest replication capacity, followed by the replicon harboring the L1403F substitution (84.9% ± 11.9% SD compared to Con1/ET) and the replicon harboring the K1398R substitution (57.3% ± 22.7% SD compared to Con1/ET). In this series of experiments, an additional E1402D substitution that has been reproducibly observed in published genotype 1a sequences and was transiently detected in a subject with acute HCV genotype 1a infection targeting this epitope (28) was also included. The E1402D substitution impaired replication to a level of 29.5% ± 7.7% SD compared to Con1/ET (Fig. 2B).

Because reversions of the K1397R mutation in genotype 1b and of the K1398R mutation in genotype 1a back to the prototypes have been observed in vivo (21, 28), we tested these mutations in a long-term replication assay, utilizing the selectable subgenomic replicon PI-neo/NS3-3′/Con1/ET. After 6 and 12 months of culture, total RNA was extracted and a fragment of the HCV RNA encoding the mutated HSKKKCDEL_1395-1403 epitope was amplified, cloned, and sequenced. Analysis of clonal sequences after 12 months revealed not a single case of reversion in a total of 23 sequenced clones (data not shown). In an attempt to simulate the quasispecies nature of the viral population, we repeated this experiment by transfecting a mixture of prototype (10%) and variant (90%) RNA. After 2 weeks, total RNA was extracted, amplified, and sequenced. Analysis of clonal sequences revealed the prototype in 5 out of 9 (55.6%) clones after 2 weeks.

To further evaluate if the observed differences in the replication rate potentially play a role in vivo, we decided to compare the in vitro replication results obtained for the B8 epitope with those of another mutation for which reversion back to the prototype residue in the absence of selection pressure is well documented. We therefore analyzed a drug resistance mutation in the HCV NS3 protease domain, A1182T (A156T in the protease), that confers high-level resistance to most protease inhibitors (13). Isolates harboring this substitution in patients receiving treatment with a protease inhibitor rapidly reverted back to the prototype within a few weeks when treatment was discontinued (23). When we tested replication of this substitution in the Con1/ET context, the replication rate after 72 h was decreased to 54.2 ± 21.6% SD compared to that of the prototype. Therefore, the fitness costs of the K1397R and K1398R mutations in the HSKKKCDEL_1395-1403 epitope in vitro are similar to those of a drug resistance mutation in the HCV NS3 protease domain (A1182T). Finally, we tested the replication rate of the P1360S variation in the HPNIEEVAL_1359-1367 epitope, which was selected in HLA-B*35-positive subjects in two single-source outbreaks infected with HCV genotype 1b. Here, a reduction of replication to levels of 71.2 ± 10.6% compared to the Con1/ET prototype was observed.

Impact of sequence variants in the HSKKKCDEL_1395-1403 epitope on T-cell recognition.

Next we analyzed the impact of sequence variation in the HSKKKCDEL_1395-1403 epitope region on recognition by specific T cells. We included analyses of PBMC from one subject with chronic HCV infection (genotype 1b), one subject with spontaneously resolved HCV infection (unknown genotype), and one subject with acute HCV infection (genotype 1a). The patient with chronic HCV infection is part of the German anti-D cohort (AD-54) and harbored a virus with two mutations (K1397R plus K1398R). When PBMC were stimulated with the prototype for 10 days, up to 12% of specific T cells (relative to all CD8⁺ T cells) were expanded (Fig. 3). These T cells were cross-reactive with the K1397R and L1403F variants but not with the K1398R variant (Fig. 3). Unfortunately, due to limited cell numbers, we were not able to test the autologous sequence containing the double mutant (K1397R plus K1398R). Similar patterns of cross-recognition of prototype-specific T cells were also observed for the resolver (R66) and the subject with acute HCV infection (06-75). Prototype-specific T cells from both subjects cross-reacted with the K1397R variant and to a lesser extent with the L1403F variant but not with the K1398R variant (Fig. 3).

FIG. 3.

Peptide-specific IFN-γ secretion of T cells stimulated for 5 h with different dilutions of synthetic peptide HSKKKCDEL (•), HSRKKCDEL (▾), HSKRKCDEL (▴), or HSKKKCDEF (⧫). Numbers of IFN-γ-positive cells as percentages of all CD8-positive T cells are shown as the result of intracellular cytokine stainings. Before restimulation, specific T cells were expanded in vitro for 10 days in the presence of the prototype peptide HSKKKCDEL.

Finally, we compared in the resolver and the subject with acute HCV infection the ability of specific T cells to proliferate upon stimulation for 10 days in the presence of the prototype or the variants in vitro (Fig. 4). Similar frequencies of fully cross-reactive T cells were expanded in the presence of the prototype and the K1397R variant in R66. In line with the degree of cross-recognition of prototype-specific T cells, the number of specific cells was substantially lower after expansion in the presence of the L1403F variant and no specific T cells were detectable after expansion in the presence of the K1398R variant (Fig. 4). Interestingly, this pattern was different in the cultures from the patient with acute HCV infection. When PBMC were stimulated in the presence of the K1398R variant, high frequencies of variant-specific T cells were expanded (46.8%) that were not cross-reactive with the prototype or any other of the variants tested (Fig. 4). This suggests the expansion of an alternative variant-specific T-cell population and the coexistence of both variant-specific and prototype-specific T cells in the same subject.

FIG. 4.

Cross-recognition of T cells expanded for 10 days from the PBMC of R66 (top) or 06-75 (bottom) in the presence of the prototype or variant peptides, as indicated in the row labels by bold type and underlining. After in vitro expansion, cells were restimulated with the prototype or variant peptides (10 μg/ml), as indicated in the column labels, before intracellular cytokine staining for IFN-γ. The number of IFN-γ-positive and CD8-positive T cells is indicated in the upper right quadrant of each dot plot (relative to all CD8-postive T cells [percentage]).

Taking the results together, we observed substantial cross-recognition between prototype-specific T cells and the K1397R and L1403F variants. In contrast, recognition of the K1398R variant by prototype-specific T cells is completely abolished in all subjects tested here. Interestingly, we identified also one subject in whom two T-cell populations targeting this region—one prototype specific and one variant specific—coexisted at the same time.

DISCUSSION

The impact of immunity-driven escape mutations in HCV on viral fitness is largely unknown. The observation that escape mutations in the HSKKKCDEL_1395-1403 epitope revert in the absence of immune pressure back to the prototype sequence (21, 28) suggests an impairment of viral replication capacity in vivo. Here we analyzed in vitro replication of various mutations in this HLA-B*08-restricted CD8 epitope in a transient replication assay utilizing a genotype 1b subgenomic replicon and provided direct evidence that the prototype sequence is advantageous in this epitope.

So far, it has been unclear if mutations selected by T-cell immune pressure do cause fitness costs in HCV. A previous study by Soderholm et al. tested different alanine substitutions in a frequently targeted HLA-A*02-restricted epitope in HCV NS3 and similarly demonstrated fitness costs associated with such mutations (25). However, the alanine substitutions are not observed in circulating isolates and are not selected in the presence of HLA-A*02-mediated immune pressure. In our study, CD8 escape mutations in the HSKKKCDEL_1395-1403 epitope that are observed in vivo are replication competent; however, compared with that of the prototype sequence, replication is reduced to levels of 60%. Of note, an escape mutation in an adjacent HLA-B*35-restricted epitope, HPNIEEVAL_1359-1367, has a weaker effect on replication, consistent with the observation that sequence polymorphisms are more frequent in this epitope than in the HSKKKCDEL_1395-1403 epitope in HCV genotype 1b isolates. The B8 epitope was selected for this study because it is so far the only one in which both selection of CD8 escape mutations and reversion back to the prototype in the absence of immune pressure were observed in vivo. However, in the Irish cohort, not all sequences from HLA-B*08-negative subjects reverted back to the prototype. Possible reasons for the absence of reversion are compensatory evolution at other sites in the polyprotein and the lack of sufficiently fit viral clones that harbor the prototype sequence in the quasispecies pool (e.g., due to a bottleneck upon transmission), as reversion in most cases likely reflects outgrowth of a prototype quasispecies that was previously suppressed to low frequencies in the presence of selection pressure. In line with the latter notion, we observed preferential outgrowth in vitro only when a mixture of replicon RNA containing both the prototype and variant was transfected. However, the replicon system is limited because it is characterized by replication of an RNA clone that is not able to leave the cells, and consequently no viral spread occurs. Infectious cell culture systems for HCV may help to address this in the future. Unfortunately, no systems that contain the nonstructural proteins of HCV genotype 1b that support production of high titers of viral particles are available yet (3).

To further address if the observed twofold difference in replication between the prototype and escape variant in this epitope is relevant in vivo, the results were evaluated and compared to the replication levels of a drug resistance mutation. For this analysis, the A156T substitution (A1182T in the polyprotein) conferring high-level resistance to treatment with HCV-specific protease inhibitors was chosen. This substitution was rapidly selected in a subgroup of patients treated with telaprevir who reverted back to the prototype within a few weeks after treatment was stopped (23). The replication level of 50% of a replicon harboring this drug resistance mutation is comparable with that of the escape variant in the B8 epitope but is slightly higher than the level in previously published results obtained with a different genotype 1b replicon system (13). Of note, as with the low frequency of substitutions in the HSKKKCDEL_1395-1403 epitope, the drug resistance mutation A156T is not observed as the predominant quasispecies in isolates obtained from treatment-naïve patients. Taken together, these results support the idea that even subtle differences in replication capacity in vitro are potentially important for selection in vivo in the presence of a high viral turnover rate.

The HSKKKCDEL_1395-1403 epitope analyzed here is located in the helicase domain of HCV NS3 and overlaps with motif IV of the encoded protein (9). Motif IV is located in domain 2 of the helicase and is believed to interact with RNA (15, 32), a concept which is supported by analysis of different mutants of the expressed protein (10). Structural analysis revealed that the lysine cluster at residues 1397 to 1399 enables water-mediated interaction with nucleotides of the substrate RNA (15, 32). Substitution by the chemically related amino acid arginine may impair this interaction, though apparently only to a small extent. Substitutions by nonbasic amino acids, such as methionine or glutamine, likely have a much more dramatic impact on this interaction. Assuming that intact helicase activity is crucial for efficient replication in vitro, it may not be surprising that the latter substitutions, which are also the result of single nucleotide changes in the corresponding codon, completely abolish replication in the in vitro assay and are not observed as the predominant quasispecies in circulating isolates.

Interestingly, substitutions in this epitope are not reproducibly selected at the same site. Differences in the relative replication capacities of various isolates may partly account for this. Of note, the E1402D mutation, which had the most marked effect on viral fitness in our assays, was only transiently selected during acute infection (28) and is not observed during chronic HCV genotype 1b infection. The most frequently selected sites in circulating isolates from subjects with chronic genotype 1 infection are in positions K1397 and K1398, both with a substitution to arginine. K1397 is located in position 3 of the epitope and represents one of the anchor residues in the B*08 motif. HLA-B*08-restricted epitopes are characterized by a lysine residue in positions 3 and 5 and leucine at the C-terminal position (26). Of note, arginine in position 3 is less frequent but is still accepted as a possible anchor for HLA-B*08 binding (26). In line with this, we observed substantial cross-recognition between the prototype and the K1397R variant in our immune assays. In contrast, prototype-specific T cells were not cross-reactive with the K1398R variant in any of the subjects. This last substitution likely directly impairs interaction of the peptide-HLA class I complex with the T-cell receptor. Interestingly, there is no difference between the replication efficiency of the K1397R and K1398R variants in vitro. Therefore, when both fitness and immunological results are considered, it is surprising that the K1397R variant is so frequently selected in circulating isolates.

Several explanations may account for this apparent discrepancy. Firstly, replication results obtained in vitro may not entirely represent the in vivo situation and relevant fitness differences between the two variants may have been missed in our assay. Secondly, this substitution was also observed in two HLA-B*08-negative subjects, and it is therefore possible that factors other than B*08-mediated immune pressure selected this mutation. Of note, a second HLA-A*03-restricted CD8 epitope (LIFCHSKKK_1391-1399 [33]) overlaps with the HSKKKCDEL_1395-1403 epitope. Three of four subjects in our cohort carrying the K1397R substitution with chronic HCV genotype 1b infection are HLA-A*03 positive (including both HLA-B*08-negative subjects). Therefore, HLA-A*03-mediated cytotoxic T lymphocyte pressure may have selected this mutation in these cases. However, T cells specific for this epitope were not detected in any of these three subjects (data not shown), suggesting either true absence or dysfunction of such T cells or insufficient frequencies in PBMC, e.g., due to their compartmentalization in the liver. Thirdly, we did not include T-cell assays that involve presentation of the endogenously processed antigen. In a previous study for this epitope, such assays revealed that the immunological effects of the L1403F substitution are missed only when synthetic peptides are used (28). And finally, the observation of specific T cells targeting the K1398R variant is interesting in this context. This suggests that a de novo response against the K1398R variant was mounted in 06-75. Such an escape-specific T-cell response was described for HIV (1) and may cause continuous evolution and deletion of the K1398R variant in some cases. The pattern of evolution is therefore potentially influenced by the T-cell repertoire recruited in each individual. We identified only one HLA-B*08-positive subject with chronic HCV infection and a robust response against the HSKKKCDEL_1395-1403 epitope for which we could directly correlate the autologous sequence with the T-cell response. Even though we unfortunately were not able to test the autologous sequence carrying the double mutation (K1397R plus K1398R), the lack of cross-recognition of the single K1398R variant strongly suggests that the autologous sequence is similarly not targeted.

The overall impact that such escape mutations associated with fitness costs may have on immune control is incompletely understood. In HIV and HCV, mutational escape has been associated with loss of control (12) leading to viral persistence (6, 8). More recently, it has been highlighted that HLA alleles associated with delayed disease progression in HIV select escape mutations in key epitopes and that the resulting fitness costs are potentially a correlate of delayed disease progression (2, 4, 24). In HCV, a recent large cohort study identified HLA-B*08 as one allele associated with spontaneous elimination of HCV (14). Interestingly, in the Irish anti-D cohort, where the source virus already harbored a substitution in the HSKKKCDEL_1395-1403 epitope, HLA-B*08 was associated with viral persistence (19). The results in this study demonstrate impairment of HCV replication associated with mutational escape in this immunodominant HLA-B*08-restricted epitope. This suggests that the antiviral effect of HLA-B*08-restricted T cells directed against HCV is sufficiently strong to force the virus to adopt a relatively unfavorable sequence. Similar to the concept of combination drug therapy, simultaneous selection pressure on multiple sites in the polyprotein may reduce the ability of HCV to select replication-competent escape mutations, which may have important implications for vaccine design.