Peptide synthesis.

Peptides were synthesized as crude material by Pepscan Systems (Lelystad, The Netherlands) as described previously (26). A total of 400 K^b and D^b algorithm-selected peptides, along with a set of 664 15-mer peptides, overlapping by 10 amino acids, spanning the entire LCMV proteome, were synthesized. The set of 664 15-mers were separated into 83 pools containing 8 peptides per pool. Candidate epitopes were identified in pool deconvolution studies. For mapping the minimal epitope within each positive 15-mer peptide, an additional set of 550 peptides of 8, 9, 10, and 11 amino acids in length were synthesized. Optimal LCMV-specific epitopes were resynthesized by A and A Labs and purified to 95% or greater homogeneity by reverse-phase high-performance liquid chromatography.

MHC peptide-binding assay.

Quantitative assays to measure the binding affinity of peptides to purified K^b and D^b molecules are based on the inhibition of binding of a radiolabeled standard peptide and were performed essentially as described elsewhere (26, 32). After a 2-day incubation, binding of the radiolabeled peptide to the corresponding MHC class I molecule was determined by capturing MHC-peptide complexes on Greiner Lumitrac 600 microplates (Greiner Bio-One, Monroe, NC) coated with either Y-3 antibody (Ab) (K^b) or 28-14-8S Ab (D^b) and measuring bound cpm using the Topcount microscintillation counter (Packard Instruments).

Bioinformatic analyses.

Experimental data of peptides binding to H-2^b molecules previously generated in our laboratory were used to develop binding predictions. This data set comprised 521 8-mer and 230 9-mer peptides binding to K^b and 319 9-mer peptides and 149 10-mer peptides binding to D^b. In addition, combinatorial peptide libraries described in reference 30 were available for peptides of 8 and 9 amino acids in length binding to K^b and 9 amino acids in length binding to D^b. These two sources of data were combined to calculate scoring matrices that quantify the contribution of each residue in a fixed-length peptide to binding to an MHC molecule, as described in detail in reference 23. The entire proteome of LCMV Armstrong (clone 53b) was then scanned for peptides of 8, 9, and 10 amino acids in length. The MHC binding affinity of these peptides was predicted using the scoring matrices described above, and peptides with affinities in the top 3% were synthesized as potential epitope candidates.

Mice.

C57BL/6J and B6.SJL/J female mice (The Jackson Laboratory, Bar Harbor, ME) were housed in the animal facilities at the La Jolla Institute for Allergy and Immunology (San Diego, CA) and were used between 6 to 12 weeks of age. All mouse studies followed guidelines set out by the National Institutes of Health and Institutional Animal Care and Use Committee-approved animal protocols.

Virus.

The Armstrong clone 53b strain of LCMV was obtained from Matthias G. von Herrath (La Jolla Institute for Allergy and Immunology). The LCMV stock was prepared in BHK-21 cells, and viral titer was determined through plaque assays on Vero cells in three independent experiments.

Infection and immunization.

B6 mice were infected intraperitoneally (i.p.) with 1 × 10⁴ to 1 × 10⁵ PFU of LCMV in phosphate-buffered saline (PBS). At 8 days postinfection, mice were sacrificed, and purified splenic CD8⁺ T cells or splenocytes were used for ex vivo mouse IFN-γ ELISPOT or for IFN-γ intracellular cytokine staining (ICCS) assays, respectively. For peptide immunization, mice were immunized subcutaneously with a mixture of peptide (50 μg/mouse) and the helper IA^b-restricted epitope, hepatitis B virus (HBV) core 128 to 140 (140 μg/mouse) (18) in PBS-9.5% dimethyl sulfoxide (DMSO) emulsified in incomplete Freund's adjuvant. After 12 days, mice were challenged with 1 × 10⁵ PFU LCMV i.p. On day 4 postchallenge, titers were determined from the spleen by plaque assays on Vero cells. Virus titers were calculated per gram of spleen.

Ex vivo IFN-γ ELISPOT assay.

The mouse IFN-γ ELISPOT assay was performed as previously described (28). Each assay was performed in triplicate wells, and the experimental values were expressed as the mean net spots/10⁶ CD8⁺ T cells ± standard error of the mean for each peptide tested. Responses against DMSO alone were measured to establish background values that were subtracted from experimental values. Statistical significance was determined by a Student's t test. For all tests, a P value of ≤0.05 using the mean of triplicate values of the response against the relevant peptide versus the response against DMSO alone was considered to indicate statistical significance. The net number of spots/10⁶ effector cells was calculated as follows: [(number of spots against relevant peptide) − (number of spots against DMSO control)] × [(1 × 10⁶)/(number of effector cells/well)]. The stimulation index (SI) was calculated as follows: (number of spots against relevant peptide)/(number of spots against DMSO control).

Intracellular IFN-γ staining.

Splenocytes from day 8 LCMV-infected B6 mice were cultured in the presence of 10 pg/ml to 1 μg/ml of MHC class I peptide, 10 μg/ml brefeldin A (BFA; Sigma-Aldrich), and 50 U/ml recombinant human interleukin-2 (rhIL-2; Roche, Palo Alto, CA) in complete medium. After 5 h, cells were harvested and stained with allophycocyanin-conjugated anti-mouse CD8α monoclonal Ab (MAb, clone 53-6.7; BD Biosciences, San Jose, CA). In some experiments, cells were also stained with fluorescein isothiocyanate-conjugated anti-mouse CD44 MAb (clone IM7; BD Biosciences). Cells were fixed and permeabilized and then stained for intracellular IFN-γ and tumor necrosis factor alpha (TNF-α) with a phycoerythrin-conjugated anti-mouse IFN-γ MAb (clone XMG1.2; BD Biosciences) and fluorescein isothiocyanate-conjugated anti-mouse TNF-α MAb (clone MP6-XT22; BD Biosciences), respectively. After washing, samples were resuspended in 1% paraformaldehyde-PBS and acquired on a FACSCalibur flow cytometer (BD Biosciences). The frequency of CD8⁺ T cells responding to each MHC class I peptide was quantified by determining the percent total of gated CD8⁺ and IFN-γ⁺ or TNF-α⁺ cells using CellQuest software (Becton Dickinson, San Jose, CA). The peptide-specific responses were calculated by subtracting the percentage of cells that scored positive for IFN-γ or TNF-α⁺ production in the absence of peptide (no peptide control).

In vivo cytotoxicity assay.

Splenocytes from naive B6.SJL/J mice were treated with red blood cell lysing buffer, washed, and divided into two equal cell populations at 10 × 10⁶ cells/ml. One group of cells was incubated with the relevant CTL peptide at 0.5 μg/ml (targets) and the other group with DMSO (marker population) for 90 min at 37°C. Cells were washed and resuspended at 10 × 10⁶ cells/ml with PBS-0.1% bovine serum albumin. Target and marker cell populations were incubated with 1.0 μM and 100 nM CFSE (Invitrogen Molecular Probes, Carlsbad, CA), respectively, for 10 min at 37°C. Excess CFSE was removed by washing, the two cell populations were combined in 1:1 ratio, and 10 × 10⁶ total cells were adoptively transferred intravenously into day 8 LCMV-infected recipient B6 mice. After 5 h, spleens were harvested from recipient mice. Splenocytes were stained with biotin-conjugated mouse anti-CD45.1 MAb (clone A20; BD Biosciences), followed by streptavidin-allophycocyanin conjugate to further differentiate donor cells from recipient. The ratio of target (CFSE^high) to marker (CFSE^low) donor cells was assessed by flow cytometry on a FACSCalibur. Specific killing was calculated as follows: 100 − [(target-to-marker ratio in LCMV-infected mice × 100)/(target-to-marker ratio in naive mice)].

CD8⁺ CD44^hi T-cell response in LCMV-infected H-2^b mice.

In preliminary experiments, a total of 91% of CD8⁺ CD44^hi T cells were detected in splenocytes from day 8 LCMV-infected H-2^b mice. In contrast, we observed repeatedly that a maximum of 45% of CD8⁺ T cells could produce IFN-γ in response to all of the previously characterized epitopes (GP33-41, GP34-41, GP276-286, GP118-125, GP92-101, NP205-212, NP166-175, and NP235-243), except GP70-77, which did not evoke a significant IFN-γ response over background.

This discrepancy might be due to the presence of unaccounted antigen-specific CD8⁺ T cells or cytokine-mediated bystander activation. Experiments using LCMV-infected MHC class I-matched cells as stimulators to enumerate antigen-specific CD8⁺ CD44^hi T cells yielded highly variable results (data not shown). Therefore, we chose to search for the additional antigen-specific CD8⁺ T cells by screening the antigenicity of LCMV-derived peptides predicted to bind to either K^b or D^b molecules.

Identification of LCMV-derived potential CD8⁺ epitopes using predicted peptides.

A total of 400 predicted determinants were synthesized, comprising the top 100 each of K^b 8-mers and 9-mers and D^b 9-mers and 10-mers. The antigenicity of these predicted peptides was assessed at the peak of the primary T-cell response by using purified splenic CD8⁺ T cells from B6 mice 8 days after i.p. infection with LCMV. Antigenicity was evaluated by measuring IFN-γ spot formation in at least two independent ELISPOT assays utilizing 0.5 and 10 μg/ml of each peptide. Representative data from the screen of the 100 K^b 9-mer peptides tested at 0.5 μg/ml is shown in Fig. 1. Peptides with an average of ≥20 spot-forming cells (SFCs)/10⁶ CD8⁺ T cells and an average SI of ≥2.0 in replicate assays were considered positive. Using this approach, we identified 18 K^b 8-mer, 14 K^b 9-mer, 11 D^b 9-mer, and 5 D^b 10-mer potential epitopes. Of these, two corresponded to the well-characterized NP epitopes, NP396-404 and NP205-212, and five were previously characterized GP epitopes (GP33-41, GP34-41, GP118-125, GP92-101, and GP70-77) (data not shown).

Identification of LCMV-derived CD8⁺ epitopes using overlapping 15-mer peptides.

Previous work has established that CD8⁺ T-cell epitopes can be identified by using overlapping peptides spanning complete viral proteomes (1, 27). Herein, we measured the antigenicity of a library of 15-mer peptides, overlapping by 10 amino acids that spanned the entire LCMV Armstrong proteome. This method was utilized to confirm putative epitopes identified by the predictive approach and also to identify potentially novel epitopes overlooked by the predictive approach.

A set of 664 15-mer peptides was divided into 83 peptide pools with 8 peptides per pool. The antigenicity of these pools was evaluated in two independent IFN-γ ELISPOT assays using purified splenic CD8⁺ T cells from day 8 LCMV-infected B6 mice. The two peptide pools spanning the Z protein were not positive in these assays. By contrast, 24 pools derived from the NP, GP, and L proteins generated positive responses following LCMV infection. A total of 42 15-mer peptides generated reproducible responses (data not shown). Each peptide pool contained one or more positive 15-mer peptide(s), except one pool, for which no definitive individual peptide stimulating a response could be identified.

Optimal epitopes map within antigenic 15-mer peptides.

Of the 42 antigenic 15-mers, 25 were nested and produced a similar CD8⁺ T-cell response to an already identified predicted minimal peptide (data not shown). In several cases, two overlapping 15-mer peptides matched a single predicted determinant, thus decreasing the number of candidates from 25 to 16. Furthermore, 17 antigenic 15-mers did not contain a predicted peptide. To identify the optimal candidate epitope within each of these “unmatched” 15-mer peptides, sets of 8-mers, 9-mers, 10-mers, and 11-mers were synthesized and tested in the IFN-γ ELISPOT assay using purified splenic CD8⁺ T cells from day 8 LCMV-infected B6 mice as effectors.

Representative data for the truncated peptides derived from the NP236-250 15-mer are shown in Fig. 2A. At the 0.5-μg/ml dose, three peptides of 11 amino acids in length were able to activate CD8⁺ T cells to a level similar to that of NP236-250, whereas the remaining peptides generated either weak or no responses. The optimal 11-mer peptide was mapped to NP238-248 by analyzing the dose-dependent CD8⁺ T-cell response to these three 11-mer peptides (Fig. 2B). Of these 11-mers, NP238-248 was the only peptide capable of activating CD8⁺ T cells at both 5 and 0.5 ng/ml.

Accordingly, we identified an additional seven optimal peptides (data not shown). No clear candidate epitope could be identified for 10 of the unmatched antigenic 15-mer peptides, because of weak and/or inconsistent results with the original 15-mer and/or truncated peptide sets. The well-characterized GP-specific D^b 11-mer epitope, GP276-286, was independently identified here. We also identified two NP-specific 11-mers, NP165-175 and NP238-248, that were similar to the previously known D^b 8-mer, NP166-173, and K^b 9-mer, NP235-243, respectively. In addition, we identified 4 novel L protein-derived determinants, including 2 K^b 8-mers, 1 K^b 11-mer, and 1 D^b 11-mer. These potential epitopes were missed by the predictive approach because they contained a noncanonical K^b motif (8-mers) or were of noncanonical size (11-mers).

Enumeration of antigen-specific CD8⁺ T cells during LCMV infection.

In summary, the prediction approach identified 48 candidate epitopes. The overlapping peptide approach identified a total of 23 potential epitopes. A combined total of 55 unique putative LCMV-specific determinants were detected by both epitope identification methods, the majority of which were newly identified and derived from the L protein (data not shown). To further examine the antigenicity of the 55 candidate epitopes defined in the ELISPOT assays, we tested them in an IFN-γ ICCS assay utilizing 0.1 μg/ml of peptide. Under these stringent conditions, a total of 28 different epitopes yielded positive results (at least >1.5 standard deviation above the average background). This threshold was based on the lowest standard deviation measured for the known epitopes. Nine of these were previously known (GP70-77 was negative), and 19 represent novel responses (representative data are shown in Fig. 3).

A total of 15 different L-specific novel epitopes were detected with responses in the 9.2 to 0.45% range (Table 1, background values are subtracted from the percentage of CD8⁺ IFN-γ⁺ cells averaged from two or more experiments). Of the four known NP epitopes identified, the overall strongest CD8⁺ T-cell response was generated against the NP396-404 (12%) epitope and followed by NP205-213 (2.8%), NP165-175 (1.9%), and NP238-248 (1.8%). Within the seven GP-derived epitopes, five were previously known with responses in the 12 to 0.42% range and four (GP44-52, GP221-228, GP365-372, and GP166-173) with responses of 4.0%, 2.8%, 1.2%, and 0.59%, respectively, were novel.

Evaluation of relative efficacy and accuracy of different epitope identification methods.

To assess the effectiveness of the predictive approach, the number of epitopes from Table 1 was plotted versus their rank for each H-2^b allele-size combination (Fig. 4). A strong correlation was observed between peptide antigenicity and prediction rank. In fact, each of the K^b 8-mer, D^b 9-mer, and D^b 10-mer with a prediction rank of one were antigenic, and all epitopes had a prediction rank of ≤38 (Table 1). Therefore, testing the top 1.2% (equivalent to 40) highest-ranking predicted epitopes (from each allele-size combination) would have been sufficient to identify all 11 K^b 8-mers, 5 K^b 9-mer, 6 D^b 9-mers, and 1 D^b 10-mer epitopes. We also measured the binding capacity of the LCMV-specific epitopes identified through the predictive approach to their predicted H-2^b molecule (Table 1). Of these epitopes, 13 bound with high affinity (50% inhibitory concentration [IC₅₀] < 50 nM), 8 with intermediate affinity (IC₅₀ < 500 nM), and 2 with low affinity (IC₅₀ < 6,000 nM).

We then compared the effectiveness of the predictive approach to the overlapping peptide approach. A total of 28 unique epitopes were identified by predictive and overlapping peptide methods, which included 9 of the previously reported epitopes. The cumulative CD8⁺ T-cell responses generated by individual peptide stimulation in the IFN-γ ICCS assays were defined as 100% of the response. Of the 28 epitopes identified by our analyses, the 3 noncanonical 11-mers made up 8.2% of the response (Fig. 5). The remaining 91.8% of the response was composed of 25 epitopes of canonical size. Of the canonical epitopes, 11 were identified by both methods (54.1%), 2 by 15-mer peptides only (2.9%), and the other 12 by predicted peptides only (34.8%).

In total, 54.1% of the anti-LCMV response is composed of epitopes that were identified by both methods. The 15-mer approach including truncated peptide sets required synthesis and testing of 1,214 peptides and identified approximately 65.2% of the overall response. By contrast, the predictive approach required synthesis and testing of 400 peptides (or 160 if only the top 1.2% from each allele would have been synthesized) and identified approximately 88.9% of the total response.

Further characterization of the new LCMV-specific epitopes.

In the next series of experiments, the newly identified epitopes were further characterized in comparison to the previously known epitopes. First, we measured the frequency of TNF-α-producing cells. In agreement with a previous study, the percentage of TNF-α-producing CD8⁺ T cells was consistently lower than that of IFN-γ (2); however, all 28 LCMV peptides generated a reproducible TNF-α response at 0.1 μg/ml (data not shown). Subsequently, we extended the IFN-γ ICCS dose-response studies to determine 50% effective concentrations (EC₅₀s, defined as the peptide concentration inducing 50% of the maximal CD8⁺ T-cell response) for candidate known and new epitopes.

The CD8⁺ T-cell responses directed against the NP-specific 11-mer epitopes, NP165-175 and NP238-248, and the well-characterized NP396-404 and NP205-212 were in a similar EC₅₀ range (10 to 100 pg/ml) (Fig. 6A). The EC₅₀s for the newly identified GP221-228 and GP44-52 epitopes (100 ng/ml) were within the higher range of the 5 known GP-specific epitopes (50 pg/ml to 100 ng/ml) (Fig. 6B). The epitopes derived from the L protein had EC₅₀s (50 pg/ml to 100 ng/ml) (Fig. 6C) similar to the known GP- and NP-derived epitopes. Overall, these titration experiments demonstrated that peptide avidities of the newly identified LCMV-specific epitopes are within the same range as those of the known epitopes.

We also compared the frequency of memory CD8⁺ T cells against the known and newly identified epitopes at 28, 115, and 239 days after LCMV infection. In general, long-term memory CD8⁺ T-cell frequencies remained at approximately 20 to 40% of the day 8 response for all epitopes (data not shown). Taken together, these experiments show that the novel epitopes identified herein induce significant primary and memory CD8⁺ T-cell responses as well as possessing similar peptide avidities to the previously characterized LCMV-derived epitopes.

Total size of antigen-specific CD8⁺ T-cell response during LCMV infection.

Intracellular IFN-γ staining was used to assess the percentage of LCMV-specific CD8⁺ T cells responding to the known and new epitopes. Splenocytes from B6 mice 8 days after LCMV infection were cultured in vitro either without or with peptides at 0.1 μg/ml each for 5 h in the presence of BFA and rhIL-2. A peptide dose of 0.1 μg/ml was utilized, since all of the known epitopes retained their activity at this concentration (Fig. 6). By adding CD8⁺ T-cell responses generated by stimulation with individual peptides, we found that the 9 previously characterized epitopes (9 known) stimulated 45% of CD8⁺ T cells, whereas the 19 newly identified epitopes (19 new) stimulated 42% of CD8⁺ T cells (Fig. 7A). Notably, when responses to all 28 peptides were added together (9 known plus 19 new), 87% of CD8⁺ T cells were stimulated. The data demonstrates that the T cells recognizing the new epitopes make up a significant portion (about one half) of the LCMV-specific response.

Next, we examined the sum total of CD8⁺ T-cell responses directed against the 28 LCMV-specific epitopes, categorized by the viral antigen from which they were derived. Using peptides at 0.1 μg/ml, we found that NP epitopes stimulated 18% of CD8⁺ T cells, whereas GP epitopes stimulated 35% of CD8⁺ T cells (Fig. 7B). The L epitopes contributed to 34% of the CD8⁺ T-cell response, again emphasizing that epitopes derived from the L protein make up a significant portion of the overall CD8⁺ T-cell response to LCMV.

Finally, we compared the percentage of antigen-specific CD8⁺ T cells to the population of activated CD8⁺ CD44^hi T cells in the spleen. We found that approximately 91% ± 3.7% of CD8⁺ T cells stain for CD44, which closely correlates with the frequency of CD8⁺ T cells responding to the 28 epitopes identified in this study (Fig. 7C and D).

In vivo cytotoxic killing of L peptide-pulsed target cells during LCMV infection.

After establishing that L-specific peptides are recognized by a significant fraction of CD8⁺ T cells during LCMV infection of B6 mice, we examined the specific killing of L peptide-coated targets in vivo to further characterize the primary antiviral response against the L protein. Cytotoxicity was assessed by transferring peptide-labeled B6.SJL/J splenocytes into day 8 LCMV-infected B6 recipient mice.

After 5 h in vivo, we observed a 98% ± 2.2% and 90% ± 3.5% reduction in targets coated with the control GP33-41 and GP276-286 peptides, respectively, compared to the unlabeled marker population (representative data are shown in Fig. 8A). As expected, peptide-coated targets transferred into naive B6 mice were not killed. The specific killing induced by L455-463, L2062-2069, and L156-163 epitopes, which stimulated a high, medium, and low percentage of CD8⁺ T cells compared to the known epitopes, respectively, was measured next. A high level of cytotoxicity (98% ± 0.80%) was observed against targets labeled with L2062-2069 (Fig. 8A), similar to the degree of cytotoxicity observed for GP33-41- and GP276-286-coated targets. A lower level of cytotoxicity was seen for L156-163 (62% ± 16%) and L455-463 (25% ± 14%) peptide-coated target cells.

We also observed that the level of killing did not correlate with the percentage of CD8⁺ IFN-γ⁺ T cells specific for the L455-463 epitope. To further investigate the relationship between cytolytic activity and IFN-γ production, the in vivo cytotoxic potential for 11 of the newly defined epitopes was measured (the percentage of specific killing averaged from two experiments is shown in Table 1) and compared to GP33-41, GP276-286, NP165-175, and NP238-248. The averaged cytotoxicity values were then plotted versus the frequency of CD8⁺ IFN-γ⁺ T cells for each epitope (Fig. 8B). The curve represents the fit of a rate equation model, assuming a rate of specific killing proportional to the number of antigen-specific CD8⁺ IFN-γ⁺ T cells times the number of peptide-pulsed target cells. Overall, a positive correlation was observed between the level of cytotoxicity and IFN-γ production for the majority of epitopes. However, three outlying epitopes, L349-357, L455-463, and NP165-175, on the lower right of Fig. 8B were not included in the model fit. Interestingly, these outlying T-cell subsets produced significant levels of IFN-γ but only generated low levels of in vivo killing.

Protection from LCMV challenge is dependent on the level of in vivo cytotoxicity.

Since epitopes derived from the L protein were effective at generating IFN-γ and TNF-α responses as well as inducing significant levels of cytotoxicity, we examined whether priming of B6 mice with L peptides would confer protection against a subsequent virus challenge. B6 mice were immunized with GP33-41, L2062-2069, L156-163, L455-463, or a mock control with DMSO. Twelve days after peptide immunization, mice were challenged i.p. with a 1 × 10⁵-PFU dose of LCMV Armstrong. As shown in Fig. 8C, priming the mice with GP33-41 peptide conferred protection against LCMV challenge. Protection against infection was also seen for the L epitope, L2062-2069. For both GP33-41 and L2062-2069 peptide-primed mice, significant reductions in virus titers at day 4 postinfection were observed (Fig. 8C). L156-163 peptide-primed mice were only moderately protected following virus challenge. In contrast, the L455-463 peptide-primed and mock-primed mice were not able to control LCMV infection.

We then determined whether the protective capacity of the L peptides was dependent on the cytolytic activity and/or IFN-γ production of the L-specific CTLs. For L2062-2069, L156-163, and L455-463 epitopes, we observed that viral load in peptide-primed mice was inversely correlated with the cytolytic activity of the L-specific CD8⁺ T cells. Furthermore, the protective capacity of the L peptides seemed independent of IFN-γ production since L2062-2069- and L455-463-specific CD8⁺ T cells produced moderate and high amounts of IFN-γ, respectively, but vastly differed in their protective capacity. Thus, priming with a peptide derived from the L protein can confer protective immunity to virus challenge and is largely dependent on the cytotoxic potential of the L-specific CD8⁺ T cells.