In Vivo Selection of a Lymphocytic Choriomeningitis Virus Variant That Affects Recognition of the GP33-43 Epitope by H-2D^b but Not H-2K^b

Abstract

CD8 T cells drive the protective immune response to lymphocytic choriomeningitis virus (LCMV) infection and are thus a determining force in the selection of viral variants. To examine how escape mutations affect the presentation and recognition of overlapping T-cell epitopes, we isolated an LCMV variant that is not recognized by T-cell receptor (TCR)-transgenic H-2D^b-restricted LCMV GP33-41-specific cytotoxic T lymphocytes (CTL). The variant virus carried a single-amino-acid substitution (valine to alanine) at position 35 of the viral glycoprotein. This region of the LCMV glycoprotein encodes both the D^b-restricted GP33-43 epitope and a second epitope (GP34-42) presented by the K^b molecule. We determined that the V-to-A CTL escape mutant failed to induce a D^b GP33-43-specific CTL response and that D^b-restricted GP33-43-specific CTL induced by the wild-type LCMV strain were unable to kill target cells infected with the variant LCMV strain. In contrast, the K^b-restricted response was much less affected. We found that the V-to-A substitution severely impaired peptide binding to D^b but not to K^b molecules. Strikingly, the V-to-A mutation did not change any of the anchor residues, and the dramatic effect on binding was therefore unexpected. The strong decrease in D^b binding explains why the variant virus escapes the D^b GP33-43-specific response but still elicits the K^b-restricted response. These findings also illustrate that mutations within regions encoding overlapping T-cell epitopes can differentially affect the presentation and recognition of individual epitopes.

CD8 T cells play a crucial role in the clearing of viral infections. Consequently, viruses have evolved multiple strategies to evade these CD8 T-cell responses, including modulation of surface major histocompatibility complex (MHC) expression and interference with peptide transport or cytokine production (2, 21). In addition, many viruses display considerable sequence heterogeneity, which favors the selection of virus variants that carry mutations in T- and/or B-cell epitopes. In the context of T-cell responses, even subtle changes can result in prevention of peptide processing, failure to bind MHC molecules, or a lack of recognition of the MHC-peptide complex by the T-cell antigen receptor (TCR). The emergence of such cytotoxic T lymphocyte (CTL) escape variants has been documented during several chronic viral infections in humans, such as human immunodeficiency virus type 1, hepatitis B and C viruses, and Epstein-Barr virus (3, 6, 8, 18).

CD8 T-cell responses are highly effective at controlling lymphocytic choriomeningitis virus (LCMV) infection (7). Depletion of CD8 T cells, either genetically or by antibody treatment, results in a failure to control acute LCMV infection (14). Given the importance of the CD8 T-cell response, there is strong selective pressure for the emergence of variants that carry mutated CTL epitopes (13, 15, 20, 29). In C57BL/6 (B6) mice, several D^b- and K^b-restricted CTL epitopes have been mapped in the LCMV nucleoprotein (NP) and glycoprotein (GP), including the D^b-restricted NP396-404, GP276-286, and GP92-101 epitopes and the K^b-restricted NP205-212 epitope (9, 22, 27). In addition, the viral glycoprotein also encodes two overlapping epitopes (11). The GP33-43 (KAVYNFATCGI) epitope (9) is presented by D^b, and the GP34-43 epitope (AVYNFATCGI) is K^b restricted (11). Both are immunodominant epitopes and, along with the NP396-404 epitope, comprise the three dominant LCMV CTL responses elicited during acute infection in H-2b mice.

Transgenic mice expressing a TCR specific for the D^b-restricted GP33-43 epitope have been developed (19). A large percentage (75 to 90%) of T cells in these mice express the transgenic TCR, allowing rapid control of low doses of LCMV (20). However, when these mice are infected with high doses of virus, there is an initial decline in virus titers followed by an increase, at which point CTL escape variants appear (20). We have used this system to generate LCMV variants with mutations in the GP33-43 epitope. This region (amino acids 33 to 43) of the LCMV glycoprotein contains the two overlapping D^b- and K^b-restricted epitopes (10). The development of MHC class I tetramers (4, 16) and the availability of D^b and K^b knockout mice (17) has allowed us to comprehensively dissect how mutations within these overlapping epitopes differentially affect the presentation and induction of responses to each individual T-cell epitope.

Here we present a detailed study of how substitutions within this overlapping epitope affect both the D^b-restricted GP33-43-specific response and the K^b-restricted GP34-43-specific response.

MATERIALS AND METHODS

Mice and infections.

B6 mice and B6 D2-TgN^TCRLCMV 327sdz transgenic (P14) mice were obtained from Jackson Laboratories (Bar Harbor, Maine). The P14 mice express an LCMV GP33-41-specific H-2D^b-restricted TCR (19, 20) and were back-crossed onto a B6 background. H-2D^b (D^b) or H-2K^b (K^b) knockout mice (17) were also constructed on a B6 background. For analysis of D^b and K^b responses, mice were inoculated intraperitoneally (i.p.) with 10³ PFU (unless otherwise indicated) of either wild-type LCMV (strain A22.2b) or variant LCMV and sacrificed 8 days after infection.

Isolation of LCMV variant viruses.

TCR-transgenic P14 mice were infected i.p. with 2 × 10⁶ PFU of LCMV strain A22.2b. Mice were bled retroorbitally at 29 days after infection. Virus in serum samples was titrated using Vero cell monolayers, and isolated plaques were picked from the plates. Virus stocks were propagated by growth in BHK-21 cells. Throughout the text, the parental A22.2b strain will be referred to as the wild type and the CTL escape variant (the L1 virus) will be referred to as the variant virus.

Preparation of viral cDNA and sequence analysis.

BHK-21 cells were infected with plaque-purified virus at a multiplicity of infection (MOI) of 0.2 and allowed to adsorb for 1 h at 37°C. Cultures were incubated for 2 days before cells were lysed in solution containing 4.0 M guanidine isothiocyanate, 0.1 M Tris-Cl [pH 7.5], 1% β-mercaptoethanol, and sodium dodecyl sulfate 0.5% and layered over a CsCl-EDTA density cushion. Total RNA was harvested from each cushion after centrifugation overnight at 30,000 rpm in an SW41 rotor (4°C). Between 5 and 6 μg of total RNA was used to prepare cDNA using reverse transcriptase (Boehringer Mannheim), according to the manufacturer's instructions. Subsequently, 2 μl of this reverse transcription reaction was used as the template for PCR amplification. PCR products (0.4 kb) were purified from an agarose gel before being sequenced (ABI systems). Approximately 300 nucleotides from the PCR DNA were analyzed. Oligonucleotide primer sequences were as follows: RT, 5′-GCTCGAAACTATACTCATGA-3′; PCR#1, 5′-TTCCTCTAGATCAACTGGGTGTCA-3′; PCR#2, 5′-GCAGAGGTCAGATTGCAAAAGTTG-3′; and SEQ, 5′-AATGTTTGAGGCTCTGCCTC-3′.

Ex vivo CTL assay.

Total splenocytes were harvested from infected mice and used in chromium release CTL assays as previously described (1).

Intracellular cytokine staining.

Detection of CD8 T cells producing gamma interferon (IFN-γ) in response to stimulation by virus-specific peptides in vitro was done as previously described (16, 25). Briefly, 10⁶ splenocytes were cultured with interleukin-2, the LCMV-specific epitope (0.1 μg/ml unless otherwise indicated), and brefeldin A for 5 h at 37°C. After this period, cell surface staining for CD8 was performed, followed by intracellular staining for IFN-γ with the Cytoperm/Cytofix kit (PharMingen, San Diego, Calif.) as described by the manufacturer.

Tetramer staining.

MHC class I tetramers specific for the D^b NP396-404, D^b GP33-43, and K^b GP34-43 epitopes were made as described (4, 16, 25). Tetramer staining was performed at 4°C.

Plaque assays.

Infectious virus was quantified by plaque assay as previously described. Briefly, 7.5 × 10⁵ Vero cells were grown overnight to confluency. Samples to be titrated were added in a 200-μl volume. After adsorption at 37°C for 1 h, cells were overlaid with agarose, and plates were incubated for 4 days at 37°C. Plaques were counted after overnight neutral red staining.

Peptide-MHC class I binding assay.

H-2K^b and H-2D^b peptide-binding assays were done as previously described (23, 27, 28). The radiolabeled probes used were a Pro-to-Tyr (position 7) analog of the adenovirus E1A epitope (SGPSNTYPEI) for D^b and the vesicular stomatitis virus (VSV) NP52-59 epitope (RGYVFQGL) for K^b. The 50% inhibitory concentrations (IC₅₀s) were 4.4 nM for the E1A epitope and 3.1 nM for the VSV NP52-59 epitope.

RESULTS

Selection of LCMV CTL escape variants in vivo.

Infection of B6 mice with a high dose of LCMV strain A22.2b (which will be referred to as the wild-type virus) resulted in a high-grade viremia during the first 2 weeks, followed by a gradual resolution of the infection (Fig. 1A). In contrast, TCR-transgenic mice initially controlled the infection, and virus titers in the serum were approximately 100-fold lower at day 8 postinfection than in nontransgenic B6 mice. Subsequently, virus titers increased, suggesting a loss of immune-mediated virus control. To determine whether this reemergence of virus was due to the selection of CTL escape variants, virus was isolated from the serum of infected mice at 29 days after infection. The nucleotide sequence spanning the D^b-restricted GP33-43 epitope and its flanking sequences was determined for seven plaque-purified viral isolates and also for the parental A22.2b strain. The nucleotide sequence analysis revealed that all seven isolates harbored the same nucleotide substitution (U to C at position 181). This mutation resulted in a change from valine to alanine at amino acid position 35 of the glycoprotein, yielding the variant epitope KAAYNFATCGI instead of the wild-type sequence KAVYNFATCGI. We did not find any other mutations in the region of the glycoprotein sequenced. For the purpose of these studies, one isolate (which will be referred to as the variant virus) was chosen for all subsequent experiments.

FIG. 1.

(A) P14 transgenic mice fail to control LCMV infection. Transgenic mice were infected with 2 × 10⁶ PFU of A22.2b virus i.p. (n = 2, open squares). B6 mice were infected with 10⁶ PFU of A22.2b virus i.p. (n = 3, solid circles). Serum was taken for titration at several time points. (B) TCR-transgenic GP33-specific CTL do not recognize LCMV L1-infected cells. Splenocytes were harvested from transgenic mice infected with A22.2b (LCMV wild type, left panel) or LCMV-L1 (LCMV variant, right panel), pooled, and measured for their ability to kill virus-infected target cells in direct ex vivo CTL assays. E:T, effector-to-target cell ratio.

We first wanted to determine whether the V-to-A mutation in the variant virus caused CTL escape. If so, it would be expected that infection of P14 transgenic mice with the variant virus would not induce any CD8 T-cell response and that GP33-specific effector cells would not recognize variant virus-infected target cells. We found that CTL isolated 8 days after infection of transgenic mice with the wild-type strain (2 × 10⁶ PFU) lysed wild-type-virus-infected targets, but were unable to kill target cells infected with the variant virus (Fig. 1B, left panel). When transgenic mice were infected with the variant virus, no detectable CTL response against target cells infected with either virus was observed (Fig. 1B, right panel). These results are consistent with the hypothesis that this mutation affects MHC class I binding, processing, and/or T-cell recognition of this epitope.

Immune response to LCMV L1 variant virus in B6 mice.

We next wanted to determine the effect of this mutation on a polyclonal response, which would include the K^b-restricted GP34-42-specific response. To test this, B6 mice were infected with the variant virus. As a control, we infected transgenic mice with the same variant virus. Splenocytes were harvested on day 8 postinfection and tested for their capacity to generate IFN-γ in response to stimulation with variant GP33-43 peptides (KAAYNFATCGI). CD8 T cells from variant-virus-infected transgenic mice failed to respond to the variant GP33 peptide (Fig. 2A). However, CD8 cells from nontransgenic B6 mice infected with the same virus did respond to the variant GP33 peptide (Fig. 2B). The observation that mice infected with the escape variant can still mount a GP33-specific response is in apparent contradiction to the results shown in Fig. 1. However, the GP33-43 peptide can be presented by both H-2D^b and H-2K^b MHC class I molecules (11). Whereas the transgenic mice used in the generation of this escape variant contain only T cells carrying TCRs that are H-2D^b restricted, infection of normal mice should yield D^b- and K^b-restricted responses. As such, it is possible that the variant virus has escaped only the H-2D^b-restricted response, leaving the H-2K^b response intact. Alternatively, it is possible that the polyclonal response in nontransgenic mice includes additional fine specificities of D^b-restricted CD8 T cells that recognize the variant epitope.

FIG. 2.

(A) Splenocytes were harvested from P14 transgenic mice 8 days after infection with either the parental A22.2b strain (wild type) or the L1 variant strain (variant). Cells were stimulated with the variant GP33-43 peptide (KAAYNFATCGI) and stained for intracellular IFN-γ production (n = 2 for each group). The percentage of CD8 T cells producing IFN-γ is shown in the top right corner of each plot. (B) Splenocytes from B6 mice infected with A22.2b or LCMV L1 were harvested 8 days postinfection and stimulated with the variant GP33-43 peptide before staining for cytokine production (n = 6 for each group). The percentage of CD8 T cells producing cytokine is listed in the top right corner of each plot.

To test these possibilities, splenocytes obtained at day 8 after infection of nontransgenic B6 mice with either the variant virus or the wild-type virus were stained with tetramers specific for D^b GP33, K^b GP34, and, as a positive control, D^b NP396. We found that the number of D^b-restricted GP33-specific cells was more than 20-fold lower in variant-virus-infected mice, whereas the numbers of K^b-restricted GP34-specific CD8 T cells were comparable in wild-type- and variant-virus-infected mice (Fig. 3). Thus, a single-amino-acid change at position 35 of the viral glycoprotein is capable of selectively affecting the D^b- and not the K^b-restricted response.

FIG. 3.

LCMV L1-primed B6 mice lack D^b GP33-specific CD8 T cells but retain the ability to generate a K^b-restricted GP34 response. (A) Splenocytes from mice infected with A22-2b (wild type) or LCMV L1 (variant) were harvested 8 days postinfection and stained with tetramers specific for D^b NP396, D^b GP33, and K^b GP34 epitopes. All cells were gated on the CD8-positive population before being analyzed for tetramer versus CD44 expression. The percentage of CD8 T cells that are tetramer positive is listed in the top right corner of each plot. (B) Enumeration of LCMV-specific CD8 T cells from wild-type (WT)- and variant-infected mice (n = 4 to 7 mice).

Immune response to variant virus in H-2D^b and H-2K^b knockout mice.

To further investigate how the V-to-A change at residue GP35 affects CD8 T-cell responses, we infected D^b and K^b knockout mice as well as B6 mice with the wild-type and variant strains. T-cell responses were monitored by measuring IFN-γ production after peptide stimulation.

In B6 mice infected with wild-type virus, the wild-type epitope was more efficiently recognized than the variant epitope (Fig. 4), whereas infection with the variant virus resulted in the opposite pattern. The total GP33-specific response in B6 mice infected with the variant virus was less than in wild-type virus-infected mice (Fig. 4), most likely because only the K^b-restricted response was induced in B6 mice infected with the variant virus.

FIG. 4.

Escape mutation affects epitope recognition by D^b- and K^b-restricted CTL and the induction of the D^b-restricted CTL response. (a) Splenocytes from B6 (+/+), K^b knockout (K^b −/−), or D^b knockout (D^b −/−) mice infected with LCMV A22-2b (WT) or LCMV L1 (variant) were harvested 8 days postinfection and stimulated with wild-type (WT) GP33 or variant GP33-43 peptide before staining for IFN-γ production. The percentage of CD8 T cells producing cytokine is given in the top right corner of each plot. (B) Intracellular IFN-γ staining results from wild-type- and variant-infected mice (n = 3 to 5 mice).

The D^b-restricted response to the GP33-43 epitope in K^b knockout mice was severely affected by the V-to-A substitution at position 35 (Fig. 4). CD8 T cells from K^b knockout mice infected with the variant virus failed to respond to either wild-type or variant peptide, reflecting a dramatic defect in GP33-specific CD8 T-cell induction (Fig. 4). Knockout mice infected with the wild-type virus recognized wild-type epitope much more efficiently than the variant epitope.

Infection of D^b knockout mice with the wild-type virus showed that the K^b-restricted response recognized both the variant and wild-type epitopes, although stimulation with the wild-type peptide yielded higher numbers of cytokine-producing cells than stimulation with the variant peptide (Fig. 4). The opposite was true for variant-virus-infected mice. Thus, K^b-restricted variant-specific (AAYNFATCGI) cells are qualitatively different from K^b-restricted wild-type-specific (AVYNFATCGI) cells. The total GP33-specific responses in D^b knockout mice infected with either the wild-type or variant virus were similar. Thus, the magnitude of the K^b response was not affected by the mutation.

In conclusion, these experiments provide clear evidence that the V-to-A amino acid change in the GP33 epitope had very different effects on the D^b- and K^b-restricted responses. Clearly, the D^b-restricted response is severely reduced by the mutation, whereas the K^b-restricted response was hardly affected. Thus, it appears that the mutation selectively affects processing or presentation of the epitope.

Effect of mutation on MHC class I binding.

To determine whether the V-to-A substitution affects the MHC class I binding of the epitope, we performed two experiments. First, the ability of the peptides to bind and stabilize MHC class I complexes was measured directly in an in vitro peptide-MHC class I binding assay. This assay is based on competition for binding of the peptide of interest with a radiolabeled standard peptide that binds the class I molecules in question with high affinity (23, 28). The concentration necessary to inhibit binding of the standard peptide to solubilized class I molecules by 50%, the IC₅₀, is measured and approximates the affinity of a given peptide-MHC interaction.

The binding experiments revealed that the V-to-A substitution had a dramatic effect on the D^b binding affinity of the peptide: the mutation changed the IC₅₀ value from 395 nM for the wild-type peptide (KAVYNFATCGI) to 17,089 nM for the variant peptide (KAAYNFATCGI). As a result, the variant peptide would be predicted not to bind to D^b at all (24). Binding to K^b was much less affected: the substitution reduced the IC₅₀ value from 22 to 49 nM (Table 1). We have also tested a 9-mer GP33-41 peptide (KAVYNFATM) which carries a methionine residue at position 41 instead of the cysteine residue that is encoded by the virus at this position. This methionine-substituted peptide has been used extensively to study GP33-specific responses (5, 12, 26). We found that the presence of the methionine residue at the C-terminal anchor position stabilized the peptide class I interaction for D^b (IC₅₀ = 21 nM, Table 1). Consistent with this, the variant 9-mer peptide still had an IC₅₀ of 329 nM, which suggests that this peptide should still bind D^b (24). Again, the effect of the V-to-A substitution on K^b binding was much less dramatic (Table 1).

TABLE 1.

Binding of wild-type and variant peptides to D^b and K^b molecules

Peptide	IC50 (nM)
	Db	Kb
KAVYNFATM	21	7
KAAYNFATM	329	20
KAVYNFATCGI	395	22
KAAYNFATCGI	17,089	49
KAVYNFATC	5,429	NDa

^aND, not determined. 

In a second set of experiments, we estimated class I peptide binding by pulsing stimulator cells with different concentrations of peptides and then quantitating IFN-γ production by responder cells (30). To ensure the D^b and K^b specificity of the responses, we used effector cells from K^b and D^b knockout mice, respectively. K^b and D^b knockout mice were infected with LCMV Armstrong (2 × 10⁵ PFU i.p.), and splenocytes were harvested at 8 days postinfection. These cells were stimulated with the four relevant peptides (i.e., the wild-type and variant 9-mer and 11-mer peptides), and we measured the percentage of responding cells. The results obtained with the K^b knockout mice are summarized in Fig. 5 and are consistent with the binding data: dilution of the variant 11-mer peptide led to a rapid decline in the number of T cells responding with IFN-γ production. The wild-type 9-mer and wild-type 11-mer peptides behaved similarly, whereas the variant 9-mer peptide appeared to have somewhat less class I binding than the wild-type peptides (Fig. 5). Thus, these data show that the V-to-A substitution strongly impaired D^b binding of the 11-mer peptide, whereas class I binding of the 9-mer variant peptide was much less affected. It is notable that at very high peptide concentrations (1 μg/ml), responses to the variant 11-mer peptides are comparable to the responses specific for the other peptides.

FIG. 5.

Binding of wild-type and variant peptides to the MHC class I D^b molecule. K^b knockout mice were infected with LCMV strain Armstrong (2 × 10⁵ PFU i.p.), and splenocytes were harvested at 8 days postinfection. Splenocytes were stimulated with 1 to 10⁻⁵ μg of the wild-type and variant 9-mer and 11-mer peptides per ml. Responses were quantitated by intracellular IFN-γ staining. Peptide sequences are shown.

DISCUSSION

This study describes a CTL escape mutant that carries a single-amino-acid substitution in the viral glycoprotein (valine to alanine at position 35). This LCMV variant was selected in a transgenic mouse expressing the TCR specific for the D^b-restricted GP33-43 epitope after infection with the LCMV A22.2b strain (wild type). The mutation changes the sequence of the D^b-restricted epitope from KAVYNFATCGI to KAAYNFATCGI and has a very strong negative effect on the D^b-binding affinity of this peptide. The surprising result of our study is that this mutation has a much weaker effect on class I binding of the overlapping K^b-restricted epitope (GP34-43), which changed from AVYNFATCGI to AAYNFATCGI. In neither case does the substitution affect the primary or auxiliary anchor residues (22), and the dramatic effect on D^b binding was therefore unexpected. Clearly, the amino acid residue at position 3 plays a role in MHC binding. This conclusion is consistent with the recent determination of the H-2D^b-GP33 crystal structure, in which the position 3 side chain is buried in the H-2D^b binding cleft (26).

The phenotypic effects that we have observed correspond very well with the binding data. The very low binding affinity of the variant epitope explains why the variant virus escaped the monoclonal CD8 T-cell response in the transgenic mice and why target cells infected with the variant virus are not lysed by transgenic D^b-restricted effector cells. This very low binding affinity also accounts for the inability of K^b knockout mice to elicit a D^b-restricted, GP33-specific response following infection with the L1 variant. In apparent contrast to these data, GP33-specific, D^b-restricted effectors (for instance, from the K^b knockout mice) do respond to the variant peptide, as detected by IFN-γ production. This indicates that their TCRs do recognize the variant peptide and that exogenously added peptide is apparently presented to these T cells. The peptide dilution experiments provide an explanation for this by showing that IFN-γ production in response to the variant peptide is a function of peptide concentration. IFN-γ production is apparent if stimulator cells are pulsed with high concentrations of peptide but is no longer observed at lower peptide concentrations. This result confirms our conclusion that the major defect of the variant epitope is its impaired D^b binding. Pircher and coworkers have also generated a CTL escape mutant containing a substitution at position 35 (valine to leucine) (20), and it is possible that their mutation may have effects on D^b and K^b responses similar to those described here.

A complicating factor in the interpretation of our data is that many studies have been done with a 9-mer peptide in which the natural cysteine residue at position 41 was changed to a methionine (i.e., KAVYNFATM instead of KAVYNFATC) (5, 12, 26). This cysteine-to-methionine substitution prevents the formation of cysteine dimers by the synthetic peptides. Our data show that the cysteine-to-methionine substitution also increases the apparent MHC class I binding: the IC₅₀ for the KAVYNFATM peptide is 21 nM, whereas the 9-mer KAVYNFATC peptide has an IC₅₀ of >5,000 nM (27). The result of this binding stabilization is that the variant 9-mer peptide (KAAYNFATM) still bound to D^b molecules with intermediate affinity (IC₅₀ = 329 nM). Introduction of the mutation in the 11-mer peptide (which carries the C at position 41) decreases the binding affinity from 395 to 17,089 nM.

In our experiments, we observed the same recognition specificity for the 9-mer peptides as for the 11-mer peptides, although the differences in cytokine production by CD8 T cells stimulated with the wild-type peptide versus variant-peptide-stimulated CD8 T cells was less exaggerated. The strong phenotypic effect of the V-to-A mutation on in vivo T-cell responses shows that this conservative amino acid change has a remarkable effect on this overlapping epitope. Although a previous study (10) has eluted this epitope from virus-infected cells, it is clear that, given our findings, the direct sequencing of the naturally processed epitope is warranted. Moreover, we believe that data regarding binding affinities and the fine specificity of the GP33-specific response, obtained using methionine-substituted peptides (5, 12, 26), should be interpreted with caution.