Structural and Biological Basis of CTL Escape in Coronavirus-Infected Mice¹

Abstract

Cytotoxic T lymphocyte escape occurs in many human infections, as well as mice infected with the JHM strain of mouse hepatitis virus, which exhibit CTL escape variants with mutations in a single epitope from the spike glycoprotein (S510). In all CTL epitopes prone to escape, only a subset of all potential variants is generally detected, even though many of the changes that are not selected would result in evasion of the T cell response. It is postulated that these unselected mutations significantly impair virus fitness. To define more precisely the basis for this preferential selection, we combine x-ray crystallographic studies of the MHC class I (D^b)/S510 complexes with viral reverse genetics to identify a prominent TCR contact residue (tryptophan at position 4) prone to escape mutations. The data show that a mutation that is commonly detected in chronically infected mice (tryptophan to arginine) potently disrupts the topology of the complex, explaining its selection. However, other mutations at this residue, which also abrogate the CTL response, are never selected in vivo even though they do not compromise virus fitness in acutely infected animals or induce a significant de novo CTL response. Thus, while structural analyses of the S510/D^b complex provide a strong basis for why some CTL escape variants are selected, our results also show that factors other than effects on virus fitness limit the diversification of CD8 T cell epitopes.

Cytotoxic T lymphocytes recognize viral Ags displayed on the surface of infected cells in the context of MHC class I (reviewed in Ref. 1). CTL that recognize their cognate Ags are critically important for eliminating virus-infected cells. The resultant immune pressure can result in the selective outgrowth of viruses that have undergone mutation in targeted, immunodominant CD8 T cell epitopes. These CTL escape viruses contribute to enhanced pathogenesis and disease progression (2, 3). Such escape viruses have been identified in humans infected with HIV and hepatitis C virus (HCV)⁴, and in nonhuman primates infected with SIV and HCV (reviewed in Ref. 4). CTL escape in HIV-infected patients has been associated with progression to AIDS (5). Sequential mutation of CD8 T cell epitopes in nonhuman primates infected with HCV is also strongly associated with persistent infection and disease progression (6). Although these data provide strong evidence for the clinical relevance of CTL escape, only some CD8 T cell epitopes undergo escape and, within individual epitopes, specific residues are mutated preferentially. The limited variability of certain CD8 T cell epitopes is postulated to reflect the functional constraints of mutation on virus fitness, although this has not been directly demonstrated in vivo (7, 8, 9, 10, 11). Other factors may also contribute to the selection of a limited subset of CTL escape viruses, such as the development of de novo CD8 T cell responses (12, 13, 14) and cross-recognition of variant epitopes by existing CTL, which has been described in the setting of HIV infection (15). Furthermore, and paradoxically, CTL escape is sometimes detected in the presence of a robust immune response, without resulting in apparent enhancement of virus replication or disease progression. Perhaps the best example of this is shown in HIV-infected HLA-B57 and HLA-B∗5801-positive individuals, in which mutation in a dominant CTL epitope, TW10, commonly occurs but is not associated with progression to AIDS (16, 17, 18). Likewise, SIV-infected rhesus monkeys expressing Mamu-A∗01 progress to AIDS more slowly than do monkeys expressing other MHC class I haplotypes (19, 20), despite frequent escape in a well-defined immunodominant CD8 T cell epitope, CM9, restricted by Mamu-A∗01. The obvious explanation for these observations is that virus fitness has been compromised by mutation in TW10 and CM9 epitopes. Consistent with this, these mutations occur early in the infection, but variant virus is not predominant until disease progression occurs, and at least in the case of CM9, epitope variants are only selected in the presence of a second compensatory mutation that restores viral core assembly to wild-type (WT) levels (4, 20, 21).

Because CTL escape variant viruses are most commonly detected only in humans infected with HIV or HCV, and in nonhuman primates infected with SIV, mutational effects on virus fitness or prospective studies of de novo CD8 T cell responses that may suppress CTL escape variant virus replication are difficult or impossible to test in vivo. Thus, current data that define the basis for the selection of CTL escape variant viruses are, in large part, correlative. CTL escape also commonly occurs in C57BL/6 (B6) mice persistently infected with neurovirulent mouse hepatitis virus strain JHM (JHMV) and correlates with disease progression (22, 23). In these studies, suckling mice are inoculated with virus and nursed by JHMV-immune dams, resulting in maternal Ab-mediated protection from lethal acute encephalitis. However, between days 21 and 60 postinfection (p.i.), a proportion of mice develop hindlimb paresis/paralysis and demyelination on histological examination. Infectious virus is present in mice with clinical disease and is mutated in the immunodominant epitope S510 (spanning residues 510–518 of the spike (S) glycoprotein, CSLWNGPHL, H-2D^b-restricted). A subdominant CD8 T cell epitope from the same Ag (epitope S598, spanning residues 598–605 of the S glycoprotein, RCQIFANI, H-2K^b-restricted) is never mutated. In contrast to human and nonhuman primate infections, this murine model is easily manipulable, and reverse-genetic approaches allow the introduction of mutations into the JHMV genome (24), allowing in vivo testing of mutated viruses. Importantly, mutations are detected within epitope S510, facilitating direct in vitro and in vivo studies of the mechanisms of CTL escape.

As in humans or nonhuman primates infected with HCV, HIV, or SIV (10, 11, 25, 26), only a limited subset of possible CTL escape mutations in epitope S510 are detected in mice persistently infected with JHMV (22, 27, 28, 29). There is also selectivity within the set of mutations observed at a given residue. For example, almost 90% of CTL escape variants at position 4 of the S510 epitope (Wp4) are tryptophan to arginine substitutions (W513R) (Table I). These apparent constraints on antigenic plasticity are not predicted, as the region of the spike glycoprotein encoding epitope S510 is prone to mutation and deletion (reviewed in Ref. 30). Thus, the basis for the selection of a limited subset of CTL escape variant viruses in mice persistently infected with JHMV is unclear.

Herein, we have solved the crystal structure of the H-2D^b/S510 complex to gain insight into how specific mutations in the epitope abrogate recognition by CD8 T cells. CTL escape is observed in both MHC-anchoring residues and TCR-accessible residues. However, of the two most prominent solvent-accessible residues, only the tryptophan at position 4 is observed to generate escape variants. Molecular virological and immunological studies have been used to dissect out the basis of the near complete bias of tryptophan to arginine substitution at this position in escape variants isolated from afflicted mice.

Materials and Methods

X-ray crystallographic studies

Crystals of H-2D^bS510-Aba were grown at 21°C in 0.1 M of citrate (pH 7.5), 28% polyethylene glycol (PEG) 3350, and 0.2 M of LiSO₄ using a protein concentration of 3 mg/ml. Crystals were cryoprotected by equilibration against mother liquor containing 5% glycerol and flash frozen by placement in a nitrogen stream. The 2.1 Å dataset were collected at the Advanced Photon Source (Chicago, IL) synchrotron facility. Crystals of H-2D^bS510_W513S-Aba were grown at 21°C in 0.1 M of citrate (pH 6.4), 28% PEG 3350, and 0.15 M of LiSO₄ using a protein concentration of 6 mg/ml. Crystals were cryoprotected using perfluoropolyether before flash freezing. The 2.7-Å dataset was collected at the Australian Synchrotron Facility in Melbourne, Australia. In both cases the data were integrated in Mosflm (31) and scaled/merged using Scala (32). The structures were solved by molecular replacement in Phaser (33) against a previously solved H-2D^b complex (PDBid: 1BZ9). The resulting models were subjected to iterative cycles of refinement in Refmac5 (34), followed by model building/correction in Coot (35). The solvent structures were built using ARP/wARP (36) and then Coot. A summary of the processing and refinement statistics is presented in Table II.

Viruses

Recombinant, WT, and epitope S510 variants of JHMV were generated using overlapping extension PCR as previously described (24, 37). Primers that overlapped the tryptophan at residue 513 of the spike glycoprotein were (5′ to 3′) W513G: fwd, GTGAGTGTTCTCTTGGGAATGGGCCCCATTTGCGCTCGGC, rev, AGCGCAAATGGGGCCCATTCCCAAGAGAACACTCAC; W513L: fwd, GTGAGTGTTCTCTTTTGAATGGGCCCCATTTGCGCTCGGC, rev, AGCGCAAATGGGGCCCATTCAAAAGAGAACACTCAC; W513S: fwd, GTGAGTGTTCTCTTTCGAATGGGCCCCATTTGCGCTCGGC, rev, AGCGCAAATGGGGCCCATTCGAAAGAGAACACTCAC; W513R: fwd, GTGAGTGTTCTCTTCGGAATGGGCCCCATTTGCGCTCGGC, rev, AGCGCAAATGGGGCCCATTCCGAAGAGAACACTCAC. The outer primers were fwd, TGTTGATTGCGCCAGCAGCTACATTAG; and rev, ACCTACGGATTGAACGCTATCATTGAC. Recombinant viruses encoding variant S510 epitopes (rJ.S_W513G, rJ.S_W513L, rJ.S_W513S, rJ.S_W513R) were selected, propagated, and titered as previously described (38). At least two independent isolates of each recombinant virus were plaque purified and analyzed.

Mice

Specific pathogen-free B6 and BALB/c mice were obtained from National Cancer Institute (Bethesda, MD). Suckling mice were infected intranasally with 2–4 × 10⁴ PFU of recombinant JHMV at 10 days of age and nursed by JHMV-immune dams (39). After weaning, survivors of the acute infection were monitored for development of hindlimb paralysis out to 60 days p.i.. For experiments using epitope S510 variant viruses, half of each litter served as an internal control: one half of each litter was infected with variant virus, and the other half was infected with recombinant WT JHMV (rJ). All animal studies were approved by the University of Iowa Animal Care and Use Committee.

RNA sequence analysis

Total RNA was purified with TRIzol (Invitrogen) from the CNS of infected mice. The region encompassing epitope S510 was amplified by RT-PCR and PCR products sequenced directly as previously described (22).

One-step viral growth kinetics

Virus was inoculated onto confluent 17Cl-1 monolayers in a 12-well plate at a multiplicity of infection (MOI) of 1.0. Groups of cells were harvested at the indicated time points and total virus (cell-associated and cell-free) was titered as previously described (39).

In vitro and in vivo virus competition assays

For in vitro competition assays, equal PFU of rJ and rJ.S_W513G, rJ.S_W513L, rJ.S_W513S, or rJ.S_W513R were combined (each at an MOI of 1) and inoculated onto confluent 17Cl-1 monolayers. Cell-free supernatants from infected cultures were sequentially passaged every 24 h for 4 days. At passages 2 and 4, total RNA was isolated from infected cells, and the relative representation of WT vs variant virus template was determined by RT-PCR followed by direct sequencing of PCR products. This assay can specifically detect a given species of template when that species comprises at least 20% of a heterogeneous pool (40). Two isolates of each rJ.S_W513 variant were assayed in triplicate, and the results from all six samples were pooled. For in vivo competition assays, equal PFU (2–4 × 10⁴) of rJ and rJ.S_W513G, rJ.S_W513L, rJ.S_W513S, or rJ.S_W513R were combined and mice were inoculated intranasally. Total RNA was harvested from the brains of mice 7 days p.i. and the relative representation of WT vs variant template was determined as described above.

Dendritic cell-based vaccination

LPS-matured dendritic cells (5 × 10⁵) (DC) were prepared, coated with OVA or S510 peptides, and injected via tail vein into groups of 5-wk-old mice as previously described (41). Seven days following DC vaccination, mice were infected i.p. with 3 × 10⁵ PFU rJ.S_W513G or rJ.S_W513R. Seven days later, spleens were harvested from mice and the frequencies of epitope-specific CD8 T cells were determined by ex vivo peptide stimulation and intracellular cytokine staining as described below.

Intracellular cytokine staining and flow cytometry

Mononuclear cells were harvested from the brains of acutely ill mice 7 days p.i. and analyzed for expression of IFN-γ by an intracellular cytokine assay as previously described (42). Unless otherwise noted, peptides corresponding to the native S510 epitope or each S510 variant epitope were used at a final concentration of 1 μM. Cells were analyzed using a FACScan flow cytometer (BD Biosciences). Data sets were analyzed using FlowJo software (TreeStar). All Abs and reagents were purchased from BD Pharmingen.

Peptide binding and stability assays

In some experiments, peptide binding was assessed using RMA-S cells as previously described (22). In other cases, the thermostability of recombinant MHC-peptide complexes were used to assess peptide binding using circular dichroism (CD). CD spectra were measured on a Jasco 810 spectropolarimeter using a thermostatically controlled cuvette at temperatures between 20 and 90°C as described in detail elsewhere (43, 44). Far-UV spectra from 195 to 250 nm were collected and averaged over 10 individual scans; 218 measurements for the thermal melting experiments were made at intervals of 0.1°C at a rate of 1°C/min. The midpoint of thermal denaturation (Tm) for each protein was calculated by taking the first derivative of the elipticity data and identifying the inflection point.

T cell functional avidity determination

Mononuclear cells were harvested from the brains of individual rJ-, rJ.S_W513G-, rJ.S_W513L-, rJ.S_W513S-, or rJ.S_W513R-infected mice 7 days p.i. and stimulated ex vivo in the presence of EL-4 cells pulsed with 10-fold dilutions of the relevant peptide. After 5.5 h, cells were stained for intracellular IFN-γ as described above. For each epitope-specific population, data were normalized to the frequency of Ag-specific CTL recorded at the highest titration of peptide (1 μM).

TCR Vβ-chain usage

Cells were harvested from the CNS of mice 7 days p.i. and stimulated ex vivo with S510 or S510_W513R peptides. Cell aliquots were subsequently stained for CD8 (PE-Cy7-anti-CD8α) and each Vβ-chain (FITC-anti-Vβ2, 3, 4, 5.1/5.2, 6, 7, 8, 9, 10^b, 11, 12, 13, or 14) followed by intracellular staining for IFN-γ (PE-anti-IFN-γ). Data were collected using a BD LSR II instrument (BD Biosciences) at the University of Iowa Flow Cytometry Facility and are expressed as the frequency of Ag-specific CD8 T cells that express each Vβ-chain.

Results

Crystal structure of immunodominant JHMV CD8 T cell epitope H-2D^b/S_510–518

To evaluate the molecular nature of residues targeted for CTL escape during JHMV infection, we determined the crystal structure of the immunodominant JHMV CD8 T cell epitope H-2D^b/S_510–518 (S510; CSLWNGPHL). Initially, we observed that crystals of diffraction quality did not readily form due to oxidation of the N-terminal cysteine residue of S510 during in vitro refolding of the MHC-peptide complex. Substitution of this cysteine residue with l-α-aminobutyric acid (Aba), an isostereomer of cysteine, has previously been used to stabilize peptide epitopes (45). Aba-modified peptides are impervious to oxidative damage, cysteinylation, or dimerization, and in data not shown we demonstrated that Aba-modified S510 was equally stimulatory as native peptide when reacted with CNS-derived lymphocytes in IFN-γ intracellular staining assays. Furthermore, no additive effects were observed when mixtures of native and Aba-modified peptides were used to stimulate S510-specific CTL. Collectively, these results indicate that both peptides stimulate the same population of S510-specific CTL and validate the use of S510-Aba peptides for x-ray crystallographic analyses.

The structure of H-2D^b/S510-Aba consists of four heterodimers in the asymmetric unit, with each copy containing the S510-Aba peptide bound in the H chain’s Ag-binding cleft (Fig. 1). In all four heterodimers the S-510 peptide, residues are highly ordered (Table III) and occupy virtually identical conformations (with root mean square deviation (rmsd) values of only 0.14 Å for all peptide atoms). Therefore, unless otherwise stated the structural features described below were observed in four copies.

The peptide adopts an extended conformation with a backbone kink at Pro⁷. The side chains of Trp⁴ and His⁸ extend prominently out of the cleft and are predicted to dominate T cell recognition. MHC-anchoring interactions are made by Asn⁵ and Leu⁹ (Table IV). The peptide is anchored within the cleft primarily at positions 1, 2, 5, and 9. Aba¹ forms a hydrogen bond via its main chain with Tyr⁷, Tyr¹⁵⁹, and Tyr¹⁷¹, while its side chain is positioned within hydrogen bonding distance of Lys⁶⁶, suggesting a potential interaction for the original cysteine at that position. These interactions are consistent with a typical P1 amino acid complexed to H-2D^b, and they confirm the role of Aba as a potent peptidomimetic (Fig. 2). Ser² hydrogen bonds with Glu⁶³ and Lys⁶⁶, while Asn⁵ is hydrogen bonded to Gln⁷⁰ and Gln⁹⁷. The main chain of Leu⁹ hydrogen bonds with Ser⁷⁷, Asn⁸⁰, Tyr⁸⁴, and Thr¹⁴³, while its side chain is buried within a hydrophobic pocket formed by Trp⁷³, Leu⁸¹, Leu⁹⁵, Phe¹¹⁶, Tyr¹²³ and, Trp¹⁴⁷ (Fig. 2).

Unlike the anchoring residues, positions 3, 4, 6, 7, and 8 are involved in fewer interactions with H-2D^b (Fig. 1A). Trp⁴ is hydrogen bonded via its main chain carbonyl with His¹⁵⁵. Gly⁶ and Pro⁷ interact with Trp⁷³ and Tyr¹⁵⁶, while His⁸ interacts with Trp¹⁴⁷ (Fig. 2).

Crystal structure of mutant W513SH-2D^b/S_510–518

We also determined the crystal structure of H-2D^b in complex with the W513S mutant of S510 (Fig. 1B), which contains a tryptophan to serine substitution at position 4 of the peptide in addition to having the cysteine at position 1 replaced by Aba. The structure of W513S/H-2D^b/S_510–518 consists of two heterodimers in the asymmetric unit, with each copy containing the W513S S510-Aba peptide bound in the H chain’s Ag-binding cleft. In both heterodimers the W513S peptide residues appear highly ordered (Fig. 1, D and E, Tables II and III) and occupy virtually identical conformations (rmsd values of only 0.06 Å for all peptide atoms). The peptide adopted the same structure as observed in the WT complex (rmsd for peptide C-alpha atoms between the two structures is 0.35 Å) and forms equivalent interactions with H-2D^b (Fig. 1C). The side chain of Ser⁴ is oriented in the same way as the C-beta and C-gamma atoms of Trp⁴ in the index peptide. The only notable deviations in structure are observed at positions 3 and 4, where the peptide backbone of the W513S mutant is shifted down into the cleft by ∼0.7Å and is accompanied by a rearrangement in the side chain conformation of Leu³. These minor differences may be due to the loss of steric clashes at position 4 in the W513S mutant and may account for the enhanced thermostability of the mutant complex.

CTL escape targets both TCR-accessible and anchor residues within the S510 epitope

Mutations at anchor residues are the simplest way for CTL escape variants to be generated, because they result in failure of the variant sequence to bind to the MHC. Consistent with this assumption, >25% of all CTL escape variants bear mutations at positions 2 and 5 (Table I). Interestingly, no mutations at Leu⁹ were detected in mice, suggesting that mutation of anchor residues is not a global strategy for JHMV escape. Leu³, another buried amino acid residue, is also a common target of immune escape, while the partially buried Cys¹ is never mutated in infected animals. Mutations are also commonly observed at Gly⁶ and Pro⁷, and such changes in these residues would most likely change the peptide’s backbone conformation affecting TCR recognition. Mutations at the two most solvent accessible residues, Trp⁴ and His⁸, are expected to directly influence TCR recognition; however, only mutations at position 4 are detected. Moreover, the mutation at position 4 is restricted to Trp to Arg changes in >90% of persistently infected mice (Table I).

These data highlight the surprising diversity of CTL escape variants in this model and suggest multiple mechanisms of T cell escape. Even more surprising is the conservation of some variants, such as the propensity of Trp⁴ to Arg⁴ mutants. Single nucleotide changes in the Trp⁴ could also result in mutations to Gly, Leu, Ser, and Cys. To begin to understand the basis of selectivity of CTL escape at position 4, we compared the serological stability and thermostability of the D^b/S510 and four variant complexes (Arg⁴, Gly⁴, Leu⁴, and Ser⁴, also referred to as W513R/G/L/S) using the TAP1/2-deficient RMA-S cell line (46) and CD analyses. We found no difference in the ability of the variant peptides to bind H-2D^b in epitope stabilization assays (Fig. 3A). However, individual substitutions at Trp⁴ altered the thermostability of epitope S510/H-2D^b complexes. The Gly4 variant peptide was least able to stabilize the complex, relative to the other variant epitopes. This is consistent with a stabilizing role of van der Waals interactions between the indole side chain of the Trp residue and H-2D^b. In contrast, the Ser⁴ variant epitope enhanced the stability of H-2D^b complexes (Fig. 3B). Taken together, these results support the structural studies that indicate that position 4 of the S510 epitope is primarily involved in TCR recognition. However, some substitutions, such as Gly⁴ or Ser⁴, can also subtly influence the stability of the H-2D^b/epitope complex.

Next, we used the WT and W513S mutant H-2D^b/S510 complexes as templates for modeling how the mutations at position 4 could influence binding to S510-specific TCR. Of the four mutants, only the arginine substitution cannot be accommodated within the space occupied by the tryptophan’s indole ring, due to unfavorable contacts between its guanadinium group and H-2D^b, which results in an altered side chain orientation (Fig. 3, C–H). Consequently, the side chain in the Arg⁴ variant would be expected to extend out of the Ag binding cleft and interfere with WT-specific CTL recognition, while the Leu and Gly mutations could be accommodated with minimal structural perturbations, as was the case for the Ser⁴ variant peptide complex. The difference in polarity and size between these residues and the native Trp, however, is substantial and is predicted to significantly affect TCR recognition.

rJ and Trp⁴ variant viruses are equally fit in tissue culture cells and in mice with acute encephalitis

The structural data highlight why mutation at Trp⁴ results in evasion of the epitope S510-specific CD8 T response, but they do not explain the lack of outgrowth of Gly⁴, Leu⁴, and Ser⁴ variant viruses in persistently infected mice. To directly test whether functional constraints on virus fitness explain these results, we generated a series of isogenic recombinant viruses encoding Gly⁴, Leu⁴, and Ser⁴ substitutions (herein referred to as rJ.S_W513G, rJ.S_W513L, and rJ.S_W513S, respectively). Recombinant WT virus (rJ) and virus encoding the Arg⁴ substitution (rJ.S_W513R) were also generated. We found no differences among the viruses in analyses of one-step growth kinetics in 17Cl-1 cells (Fig. 4A). Furthermore, there were no marked differences following in vitro competition assays in which relative recovery of rJ vs each variant was assessed by RT-PCR and sequencing (Fig. 4B). In general, rJ outgrew variant viruses in 17Cl-1 cells, although mutated virus was still detected after four passages in 50% or more of cultures (Fig. 4C). In vitro analyses may not reflect the relative replicative capacity of viruses within the intact animal. However, all of the variant viruses grew as least as well as rJ in B6 mice and in BALB/c (H-2^d) mice, which do not recognize epitope S510 and therefore do not exert immune selection on this region of the virus (Fig. 4D).

Viruses encoding W513R/S but not W513G/L cause enhanced disease in Ab-protected infected mice

We previously showed that when Ab-protected, suckling mice are infected with naturally occurring CTL escape viruses, the mice exhibit increased morbidity and mortality, relative to infection with WT JHMV (23). Thus, as a measure of in vivo virus fitness, we infected maternal Ab-protected, suckling mice with WT or variant S510 recombinant viruses (Fig. 4E). As expected, ∼80% of mice infected with rJ survived the acute infection. By comparison, rJ.S_W513S and rJ.S_W513R behaved as expected for true CTL escape variants, with only 20% of infected mice surviving the acute infection (death before 14 days p.i.). Less expected were the results for mice infected with rJ.S_W513G and rJ.S_W513L viruses: we found substantially less morbidity and mortality during infection with either virus when compared with rJ-infected mice. These results are remarkable given that both rJ.S_W513G and rJ.S_W513L viruses cause lethal encephalitis in mice not protected by maternal Abs, and they compete with rJ in acutely infected adult animals. This inability to cause encephalitis suggests that the Leu⁴ and Gly⁴ mutations would not outgrow rJ even if the mutations arose in persistently infected mice. As described below, this may reflect differences in cellular tropism or inflammatory milieu in persistently as opposed to acutely infected mice. Subsequently, we focused on the Ser⁴ mutation, which should be (but is not) selected in persistently infected mice.

Recombinant viruses bearing Arg⁴, but not Ser⁴, substitutions prime CTL responses

Another possibility is that de novo CTL responses develop to the Ser⁴, but not Arg⁴, variants, thereby minimizing the likelihood that this variant would be selected in vivo. First, we demonstrated that CNS-derived WT S510-specific CD8 T cells do not recognize either variant epitope in direct ex vivo intracellular IFN-γ staining assays (Fig. 5A, Table V), confirming results obtained with splenic cells (29). Next, to determine whether either variant elicited a novel CTL response, we assayed CNS-derived lymphocytes from variant virus-infected mice for epitope-specific CD8 T cell responses. We found that mice mounted a low-level CD8 T cell response to the Arg⁴ epitope, but no response to the Ser⁴ epitope (Fig. 5B and Table V). Based on the modeled structures, the Arg⁴ variant is predicted to elicit a TCR profile different from that elicited by the WT epitope; consistent with this prediction, S510- and S510_W513R-specific CTL exhibit different TCR Vβ-chain profiles (Fig. 5C) with the emergence of Vβ4 and Vβ11⁺ T cell populations and retraction of the Vβ13⁺ population.

The lack of priming to Ser⁴ could reflect differences in Ag processing and presentation of this variant epitope or a hole in the TCR repertoire. To address the latter possibility, we evaluated the ability of the Ser⁴ epitope to prime CTL responses using peptide-pulsed DC-based vaccination. We found that the Ser⁴ epitope failed to prime a CTL response, while native S510-pulsed DCs primed a robust response (Fig. 5D). Therefore, the lack of priming to Ser⁴ likely reflects a hole in the TCR repertoire or tolerance to this epitope. These results further suggest that the Ser⁴ variant should be selected in persistently infected mice.

The response to epitope Arg⁴ was unexpected because virus expressing this variant epitope is recovered in vivo and is highly virulent. One possible explanation was that in mice persistently infected with rJ, the phenomenon of original antigenic sin occurred (expansion of native epitope-specific CTL, rather than priming a de novo variant-specific CTL response (47)). To explore this possibility, we used a combination of DC-S510 peptide priming followed by peripheral infection with rJ.S_W513R. However, as shown in Fig. 5, E and F, we found that these mice mount splenic CD8 T cell responses of similar magnitude to the Arg⁴ variant epitope, regardless of whether they were vaccinated with DC-S510 or DC-OVA (an irrelevant epitope), and they do not mount a secondary response to the WT epitope.

To further assess the biological significance of this modest S510_W513R-specific CD8 T cell response, we examined the functional avidity of these cells. As shown in Fig. 6, the relative functional avidity of CNS-derived cells responding to the native and the variant epitopes was not different. A low functional avidity may have explained why the response does not prevent the outgrowth of variant viruses expressing the Arg⁴ epitope; however, this does not appear to be the case for this escape variant.

Discussion

Herein, we combine structural studies of the H-2D^b/S510 epitope and a nonselected variant W513S S510 epitope with biological studies of viruses expressing variant epitopes to begin to understand selectivity in CTL escape development. While the structures of CTL variants of a single epitope (lymphocytic choriomeningitis virus (LCMV) gp33) have been solved (48), our study is the first to combine structural analyses with biological assays using recombinant viruses expressing variant epitopes. Our results show that selection is more complicated than previously postulated because a mutation, W513S, that results in CTL evasion without impairing virus replication is not selected in infected mice. Previous reports have concluded that some CTL escape variants are not selected because they compromise virus fitness. Moreover, many of those that do become selected are postulated to negatively affect virus fitness, as evidenced by rapid reversion to WT sequence when transmitted to hosts that lack the appropriate restriction element, or the requirement for co-selection of compensatory mutations that restore virus fitness (8, 49, 50). In a few cases, deleterious effects on virus fitness have been demonstrated directly in in vitro assays, using tissue culture cells or virus replicons (11, 50, 51). Consequently, a conclusion from these studies is that only a few mutations are selected because, despite their ability to evade CTL recognition, the vast majority negatively impact virus fitness. However, our results indicate that epitope diversification is also limited by features other than virus fitness, as discussed below.

The epitope S510 mutation at position 4 that is selected in vivo, Trp to Arg, alters the topology of the D^b/S510 complex and likely directly diminishes or abrogates binding by the TCR. Other studies showed that single amino acid changes that result in altered topology of the epitope/MHC complexes can significantly skew the TCR repertoire. For example, an Arg to Ala mutation at position 7 of the influenza A-specific PA224 epitope, which removes the most prominent feature of the complex (52), still induces a CD8 T cell response, but it is much less diverse compared with the response to the native epitope. Similarly, we observed that the Trp to Arg single amino acid change in epitope S510 significantly altered the TCR repertoire distribution of responding CTL based on Vβ usage (Fig. 5C). The importance of the Trp⁴ determinant is also illustrated by the complete abrogation of the response by other mutations, including Leu, Gly, and Ser. This focus on a single amino acid residue is especially notable because the S510-specific response is highly diverse, consisting of ∼1000–2000 different TCR clonotypes (53, 54); thus, although the response is highly diverse, it is functionally monospecific.

We expected that the lack of recovery of Gly⁴, Leu⁴, and Ser⁴ variants would be explained by significant deleterious effects on virus fitness, as has been described for certain mutations in CTL epitopes derived from HIV, SIV, or HCV (11, 50, 51). However, our results demonstrate that effects of mutations on fitness may be very subtle and appear only under certain conditions. For example, while there were no gross effects on fitness in either acutely infected adult mice or tissue culture cells (Fig. 4, A, C, and D), both the Gly⁴ and Leu⁴ substitutions result in a nearly complete loss of virulence in Ab-protected suckling mice (Fig. 4E). One difference between mice with acute lethal encephalitis and maternal Ab-protected mice is that different types of CNS cells are predominantly infected in each scenario. In mice with acute encephalitis, neurons are preferentially infected, whereas glial cells (oligodendrocytes, microglia, and astrocytes) are infected in persistently infected mice, including those protected by maternal Ab (39, 55, 56, 57, 58). Single amino acid changes in the JHMV S glycoprotein have been shown to change tropism from glia to neurons in previous studies (59, 60); one possibility is that rJ.S_W513G and r.J.S_W513L exhibit diminished growth in one or more type of glial cell in vivo when compared with rJ.S_W513S and r.J.S_W513R.

In addition to virus tropism, the inflammatory milieu of the acute vs persistently infected CNS may differ with perhaps important and direct effects on local T cell responses and the resultant selective pressure exerted by CTL. Inflammation, MHC class I expression, and CTL effector function wane dramatically during persistent JHMV infection (61, 62). Similar effects on CTL have been observed in mice persistently infected with LCMV and humans chronically infected with HIV (63, 64, 65, 66) and they correlate with PD-1 expression on dysfunctional CTL (67, 68). Thus, in maternal Ab-protected, persistently infected mice, the role of S510-specific CD8 T cells in driving diversification may be less potent than in acutely infected mice. If viruses, such as rJ.S_W513G and rJ.S_W513L, are subtly less fit in the persistently infected animal, a diminished CTL response would result in less immune pressure and perhaps in the lack of continued outgrowth of these viruses, even though they express mutations predicted to abrogate the CTL response.

Although these results provide a partial explanation for why some mutations (Leu⁴ and Gly⁴) that abrogate the S510-specific response are not selected in mice, the lack of selection of variant viruses encoding W513S mutations is puzzling. This mutation does not affect virus viability, and rJ.S_W513S behaves like a CTL escape virus, causing disease severity equivalent to that of rJ.S_W513R when inoculated into Ab-protected suckling mice (Fig. 4E). Furthermore, there is no recognition of this epitope in B6 mice (Fig. 5, B and D), suggesting a hole in the TCR repertoire. In support of this possibility, we have identified a nonamer in murine ADAM23 that is predicted to bind H-2D^b (CSLSNGAHC) and could potentially induce deletion of epitope S510_W513S-specific CD8 T cells during thymic selection. We also considered the possibility that the lack of outgrowth of the Ser⁴ variant might be explained by functional constraints related to codon usage or transition (vs transversion) mutation bias, as has been described for hepatitis G virus (HGV) (69). Transition mutations are favored by a factor of 10⁴ in HGV-infected cells, but both a transition (UGG to CGG) and a transversion (UGG to AGG) result in the W513R mutation and both occur with the same frequency.

Our results showing enhanced virulence of viruses encoding Arg⁴ or Ser⁴ substitutions demonstrate a critical protective role for S510-specific CTL in Ab-protected mice. Although worsened disease is likely related to enhanced virus replication, the effect may also involve immunopathologic mechanisms, with the anti-JHMV CD4 T cell response largely responsible for clinical disease (38, 42). Thus, it is possible that enhanced disease occurs in mice infected with CTL escape virus, because increased virus replication elicits a compensatory antiviral CD4 T cell response, with immunopathological consequences.

Collectively, our results suggest that, at least for some variant epitopes, the forces that influence the selection of CTL escape variant viruses are complex and extend beyond the postulates of virus fitness and de novo CTL recognition. Thus, additional understanding of the immune response during persistent viral infection may be key to understanding the mechanisms of CTL escape. For example, subtle changes in cell tropism and/or replicative capacity during persistence may shape the unique subset of CTL escape mutations that are selected. The monospecific nature of antiviral T cells responding to the immunodominant S510 epitope results in JHMV viral escape, suggesting that the avoidance of such focused responses by targeted vaccination programs is warranted.

Acknowledgments

We thank the staff at the Advanced Photon Source and the Australian Synchrotron for assistance with data collection and John T. Harty for critical reading of the manuscript.

Disclosures

The authors have no financial conflicts of interest.

References

Footnotes

Author notes