Long-Term Immunity against Actual Poxviral HLA Ligands as Identified by Differential Stable Isotope Labeling¹

Viruses are one of the main factors that modify the repertoire of HLA ligands, the peptides associated with human MHC molecules. Viral HLA ligands are presented to T cells carrying the appropriate TCR to elicit a specific cellular immune response, thus making infected cells visible to the immune system. T cell responses were shown to play an essential role in clearance of poxvirus infections (1, 2, 3). Smallpox disease, caused by variola virus (VARV),³ was eradicated in the 1970s by vaccination with cross-protective vaccinia virus (VACV; Dryvax) (4). Although, in general, vaccinations aim to induce a strong Ab response to achieve viral clearance, vaccinees with T cell defects failed to control the infection after immunization (2, 3).

Despite the eradication of smallpox, there are several reasons for studying the cellular immune response to VACV. First, there is a constant threat that VARV may be reintroduced by acts of bioterrorism or that forms of new pathogenic poxviruses may evolve from, for example, zoonotic human monkeypox virus (MPXV) (5, 6). Because smallpox vaccination was stopped in the late 1980s, a large part of the population is unprotected. Dryvax, the only currently licensed vaccine against smallpox, carries the highest rate of side effects of any approved vaccine. Therefore, modified vaccinia virus Ankara (MVA), an attenuated replication-deficient strain of VACV, is currently being tested as a safer third-generation vaccine (5, 7, 8, 9). A more detailed understanding of the CTL response to MVA allows both the development of epitope-based vaccines promising a safe, stable, and handy alternative to traditional vaccination strategies, as well as the monitoring of clinical trials by following MVA-specific T cell responses (10, 11). Furthermore, MVA has been successfully introduced as a highly immunogenic recombinant viral vector vaccine for immunotherapy of infectious diseases and cancer (12), which requires the assessment of T cell epitope-specific responses elicited against vector and recombinant Ags.

Several CD8⁺ T cell epitopes of MVA and other VACV strains have been identified by indirect approaches applying epitope prediction (e.g., using www.syfpeithi.de) and subsequent T cell screening of immune donors, both human (13, 14, 15, 16, 17, 18, 19, 20, 21, 22) and mouse (23, 24, 25, 26, 27). Despite the increasing number of epitopes published, there is an ongoing need for further investigation. First, although thousands of peptides have been screened, covering ∼35% of the large viral DNA genome that encodes more than 200 nonoverlapping open reading frames (ORF), the published determinants represent only a fraction of the total antiviral CTL response (10). Second, only a small number of the identified epitopes have been validated to date (10). Third, our recent work challenges the indirect strategy to identify epitopes, which is based on the monitoring of T cell responses alone: upon a first vaccination with MVA, T cells were primed against viral peptides that were not necessarily presented by infected cells, but cross-presented by noninfected cells (28). During a second infection, however, only T cells with the ability to recognize viral peptides that were efficiently presented on infected cells participated in the recall response and mediated survival (29). Therefore, the identification of the viral HLA ligands that are actually presented on infected cells appears crucial for the design and monitoring of protective prophylactic vaccines.

Thus, we set out to identify peptide ligands presented on HLA class I of virus-infected cells by mass spectrometry (MS). MS allows the identification of the exact chemical composition of T cell epitopes, including those generated by posttranslational modifications. Thus, directly sequenced HLA ligands have been considered to be more reliable than determinants identified solely by T cell analysis (30), which is mandatory to accurately determine the magnitude of specific CD8⁺ T cell responses or the functionality of the CD8⁺ T cells that respond to an epitope. To date, only one MVA-derived HLA ligand has been identified by MS, however, without demonstrating immunogenicity (31). Ideally, immunologically relevant viral T cell epitopes are confirmed both as HLA ligands and as T cell stimulators.

The challenge in MS-based discovery of viral HLA ligands is to pinpoint the signals derived from viral sequences among the hundreds of signals produced by human self-peptides. In the past, nanoHPLC-coupled tandem MS analyses (LC-MS/MS) performed with chemically synthesized predicted epitopes and endogenously processed peptides isolated from virus-infected cells were compared to reveal the presence of the predicted epitope (32). To systemically search for viral HLA ligands, two strategies relying on the comparison of HLA ligands from virus-infected and noninfected cells have been applied, as follows: first, in silico subtraction of the two respective nanoHPLC-coupled MS analyses (LC-MS) (33), and second, metabolic stable isotope labeling of HLA ligands before purification (34, 35).

In this study, we describe a novel approach to identify viral HLA ligands by differential stable isotope labeling of HLA ligands purified from MVA-infected and mock-infected cells. This strategy is based on a technique recently established in our laboratory (36) to compare the repertoires of HLA ligands of tumor and healthy tissue (37, 38). We found 15 viral peptides, of which 12 were presented by HLA-A*0201 and 3 by HLA-B*0702. Nine peptides have not been described as CTL epitopes to date. All HLA-A*0201 ligands were actual memory CTL epitopes in MVA vaccinees. Eight of these epitopes were novel. All ligands were expressed early during the viral life cycle, although late protein synthesis was not impaired. Importantly, early viral HLA ligands mediated protection against a lethal respiratory challenge in mice, whereas late viral peptides previously described as CTL epitopes were inefficient.

The human B lymphoblastoid cell line (B-LCL) JY expressing HLA-A*0201 and B*0702 (European Cell Culture Collection catalog 94022533) and the human HLA-A*0201-transfected CML cell line K562/A*0201 (39) were used, as described (40), and maintained in RPMI 1640 (C.C. Pro) containing 10% FCS (Pan Biotech).

IgG2a Abs B1.23.2 (anti-HLA-B, C) (41) and W6/32 (anti-HLA-A, B, C) (42) were purified from hybridoma supernatants using protein A-Sepharose beads (GE Healthcare).

MVA was routinely propagated and titrated following standard methodology, as described (43). For infection with MVA, 1.4 × 10¹⁰ JY cells were incubated at 2.8 × 10¹⁰ cells/L with MVA to obtain a multiplicity of infection of 7. After 2 h, infected cell suspensions were diluted to 5.2 × 10⁹ cells/L and maintained for 10.5 h at 37°C.

Donors 1 and 2 (HLA-A*0201 positive) were immunized twice with MVA in an interval of 30 days. Blood was taken, as indicated in Fig. 3. Dryvax vaccinees were HLA-A*0201 positive, and blood samples were taken 25, 29, and 44 years postvaccination for donors 3, 4, and 5, respectively. This study was approved by the Ethics Committees of the University of Tubingen and of the Technical University Munich.

HLA ligands were obtained by immunoprecipitation of HLA molecules from 1.4 × 10¹⁰ cells of MVA- and mock-infected JY using a slightly modified protocol (44) that involves the Abs B1.23.2 and W6/32 coupled to CNBr-activated Sepharose (Roche), followed by acidic elution and size exclusion ultrafiltration.

Modification of peptides was conducted, as described (37). Peptide analysis was conducted, as described (36), using an Ultimate HPLC system (Dionex) with a gradient ranging from 15 to 55% solvent B within 170 min. The mix of the peptide samples from MVA- and mock-infected JY cells was recorded in an LC-MS experiment without fragmentation using a Q-TOF I (Micromass), as described (36, 44). For peptide sequence analysis, the sample derived from MVA-infected cells was analyzed in a separate LC-MS/MS experiment. Fragment spectra were evaluated manually, and database searches (National Center for Biotechnology Information; Expressed Sequence Tag) were conducted using the MASCOT search engine (www.matrixscience.com) (45). All viral HLA ligand sequences are available at www.syfpeithi.de.

HLA-A*0201-restricted peptides derived from MVA proteins (Table I), human CMV pp65_495–503 (NLVPMVATV), EBV BMLF1_259–267 (GLCTLVAML), Influenza virus M1_58–66 (GILGFVFTL), and HIV-1 RT_476–484 (ILKEPVHGV) were synthesized by standard Fmoc chemistry using an Economy Peptide Synthesizer EPS 221 (Abimed Analysen-Technik). Purity of peptides was analyzed by HPLC, and identity was confirmed by MS. For T cell experiments, all peptides were dissolved in 10% DMSO at 1 mg/ml.

Biotinylated rHLA class I molecules and fluorescent HLA tetramers for CD8⁺ T cell analysis were produced, as described (46).

PBMC were cultured in IMDM (BioWhittaker) containing 10% heat-inactivated human serum (PAA) and 50 μM 2-ME (Merck). IL-4 and IL-7 (5 ng/ml; R&D Systems) were added after thawing. On day 1, peptides (1 μg/ml) prepared as mixtures of five peptides each were added to PBMC, as follows: mix 1, F12L_404–412, G5.5R_27–35, B19R_207–215, A23R_273–281, and B8R_18–27; mix 2, C7L_74–82, A48R_187–195, B22R_79–87, H3L_184–192, and A47L_155–163; mix 3, O1L_247–255, E5R_93–101, D12L_62–70, D12L_251–259, and B22R_178–186; mix 4, O1L_335–344, J6R_303–311, and B15R_91–101; and IL-4 and IL-7 (5 ng/ml) were added. On days 3, 5, and 7, IL-2 (2 ng/ml; R&D Systems) was added. On day 12, part of the cells were used for IFN-γ ELISPOT assay. The remaining cells were restimulated by addition of peptide (1 μg/ml), followed by IL-2 (2 ng/ml) 24 h later and analysis by combined tetramer/intracellular IFN-γ staining on day 20.

IFN-γ ELISPOT was performed essentially as described (40), except 5 × 10⁵ PBMC/well were seeded in coated 96-well nitrocellulose plates (MSHAN4B50; Millipore) and 5 × 10⁴ K562/A*0201 cells/well and 1 μg/ml peptide were added. An HLA-A*0201-restricted HIV peptide was used as negative control; positive control wells contained PHA (Roche). After 26 h at 37°C, cells were transferred into a 96-well plate and maintained for further investigation. Anti-IFN-γ (7-B6-1; Mabtech) was used at 0.33 μg/ml. ExtrAvidin Alkalinphosphatase (1:100; Sigma-Aldrich) was used for 1 h at room temperature. Peroxidase staining was performed with 5-bromo-4-chloro-3-indolylphosphate toluidine/NBT (B5655, dissolved in water; Sigma-Aldrich) for 7 min. Spot numbers were automatically determined (Immunospot Image Analyzer, series 1; ImmunoSpot Software Version 3.2e; both Cellular Technology). To calculate the number of cells responding to a particular peptide, the mean spot numbers induced by the control peptide were subtracted from mean spot numbers induced by MVA peptides.

One day after ELISPOT analysis, the transferred cells were stained by tetramers (in each case using a tetramer containing a peptide other than the peptide used for stimulation in ELISPOT analysis) using PE tetramers for MVA peptides, allophycocyanin tetramers for control peptides, anti-CD8 PE Cy7, and anti-CD4 FITC (BD Biosciences). Cells were analyzed on a FACSCalibur cytometer (BD Biosciences).

After two rounds of peptide/IL-2 in vitro sensitization, PBMC were washed in IMDM, resuspended at 2 × 10⁷ cells/ml, and cultured for 7 h in IMDM containing either one of the MVA peptides or a control HLA-A*0201-restricted HIV peptide and Golgi-Stop solution (BD Biosciences). Stimulation with PMA/ionomycin was used as positive control. Cells were stained using the PE tetramers mentioned above, anti-CD8 PE Cy7, the Cytofix/Cytoperm Plus kit for permeabilization, and anti-IFN-γ FITC (BD Biosciences). Cells were analyzed on a FACSCalibur cytometer (BD Biosciences).

At 12.5 h postinfection, JY cells were used for analysis of intracellular proteins, as described (47), with slight modifications. Approximately 200 μg of intracellular proteins extracted from MVA- and mock-infected cells was separated by two-dimensional (2D)-PAGE (first dimension: pH 3–10NL, 24 cm (Bio-Rad), 70 kVh; second dimension: 12% SDS-PAGE). Gels were stained by Flamingo fluorescent staining (Bio-Rad) and scanned using a laser scanner (FLA 5100; Fujifilm), and by silver staining, as described (48), and scanned on a flatbed scanner (Powerlook 2100 XL; UMAX). Protein preparations from two infection experiments were subjected to 2D-PAGE in duplicates. All eight gels were comparatively evaluated using differential image analysis software (Progenesis SameSpots; Nonlinear Dynamics). All spots representing proteins that were unique or overexpressed upon MVA infection were excised manually from the gels and digested with trypsin, and peptide fragments were analyzed by LC-MS/MS, as described (47). Significant overexpression was defined for spots detected >1.6-fold higher upon MVA infection and an ANOVA value of <0.05 comparing the gels of both conditions. Even though the majority of analyzed proteins was overexpressed in both experiments, differential spots detectable on gels of only one of the two experiments were also analyzed. Peptide sequences were identified using the MOWSE algorithm as implemented in the MASCOT software (Matrix Science) (45), and using the National Center for Biotechnology Information database (as of 30/04/2007) containing human and MVA protein sequences.

For peptide vaccination, HLA-A*0201 transgenic β₂-microglobulin^−/− D^b−/− HHD II mice (13) were immunized s.c. with pools of synthetic peptides (0.03 mg/peptide; Biosynthan) and synthetic CpG oligodeoxynucleotide 1668 (10 nMol; TIB-Molbiol). Quantification of Ag-specific CD8⁺ T cell responses: PBMC isolated on day 7 from vaccinated mice were stimulated with indicated peptide pools for 5 h. HLA-A*0201-restricted control peptides were Tyr_369–377 (derived from human tyrosinase), FluM_58–66 (derived from the A/PR/8/34 Influenza virus matrix protein M1), and pp65_495–503 (derived from the human CMV internal matrix protein pp65). Brefeldin A (1 mg/ml; Sigma-Aldrich) was added for the last 3 h. Cells were live/dead stained with ethidium monoazide bromide (Molecular Probes) and blocked with anti-CD16/CD32 Fc Block (BD Biosciences). Surface markers were stained with allophycocyanin-conjugated anti-CD8 and anti-CD62L PE (Caltag Laboratories, now Invitrogen). Intracellular IFN-γ staining was performed with anti-IFN-γ FITC (clone XMG1.2) using the Cytofix/Cytoperm kit for permeabilization (BD Pharmingen). Data were acquired by FACS analysis on a FACSCanto (BD Biosciences) and were analyzed with FlowJo (Tree Star) software. Protection assays: Eight days after immunization with virus or after peptide immunization, mice were infected intranasally with VACV WR (originally provided by B. Moss (National Institutes of Health, Bethesda, MD)) diluted in 30 μl of PBS, and monitored for more than 3 wk with daily measurement of individual body weights, as described previously (13). Mice suffering from severe systemic infection and having lost >30% of body weight were sacrificed. The mean change in body weight was calculated as percentage of the mean weight for each group on the day of challenge.

To identify MVA-derived HLA ligands presented by HLA-A*0201 and B*0702, we differentially analyzed the ligands isolated from MVA- and mock-infected cells of the human B-LCL JY after 12.5 h of infection. Fig. 1 schematically illustrates the further experimental procedure. HLA-presented peptides were chemically modified by covalently linked stable isotope tags, as follows: peptides isolated from MVA-infected cells with heavy (deuterated (D₄)) nicotinic acid (NIC) and peptides isolated from mock-infected cells with light (hydrogenated) NIC. The two pools of tagged peptides were mixed, and the peptides were separated by nanoHPLC and analyzed online by MS. HLA ligands present on both MVA- and mock-infected cells were detected as doublets with a mass difference of 4 Da, due to the four deuterium atoms of D₄NIC (Fig. 1, lower central panel) replacing four hydrogen atoms present in hydrogenated NIC. In contrast, viral HLA ligands presented only by infected cells appeared as single peaks (Fig. 1, upper central panel; Fig. 2). The peptides corresponding to single peaks were then sequenced by fragmentation using LC-MS/MS analysis (Fig. 1, right panel; Fig. 2). All viral peptides identified were synthesized chemically and analyzed by the same procedure to verify their sequences (Fig. 2; data not shown).

We discovered 12 viral HLA ligands among the peptides isolated from HLA-A*0201 (Table I). Seven of these were novel (Fig. 2), whereas 4 had previously been published as CTL determinants (see Table VI for references), and 1 peptide (A48R_187–195: IVIEAIHTV) had been described before as an MVA-derived HLA-A*0201 ligand identified by LC-MS/MS (31), but had not been confirmed as CTL epitope. In addition, we detected 3 viral HLA ligands among the peptides isolated from HLA-B*0702 (Table I), 1 of which was also novel (Fig. 2).

In total, we confirmed 6 established CTL epitopes as well as 1 MVA-derived HLA ligand, and additionally found 8 novel ligands that represent potential CTL epitopes. All 12 viral HLA ligands were detected within an intensity range of one order of magnitude, suggesting an absolute quantity of ∼150–1500 specific peptide/MHC complexes per cell. This is in line with the absolute quantity of viral HLA ligands published for other viruses (49) (reviewed in Ref. 50). Peptides presented at lower levels may be missed by this approach.

The lower number of viral peptides identified for HLA-B*0702 in comparison with HLA-A*0201 reflects the more stringent HLA-B*0702 peptide-binding motif, which requires the relatively infrequent proline residue in position 2 within its ligands (www. syfpeithi.de). The HLA-A*0201 motif, in contrast, is less restrictive, with frequent amino acids as anchors in positions 2 (leucine, methionine, valine, isoleucine) and 9 (valine, leucine, isoleucine, alanine).

An intriguing characteristic of the proteins processed to the identified HLA ligands was their exclusive temporal expression early or early/late during the viral life cycle in MVA-infected B-LCL (Table I) despite the unimpaired expression of late genes in these cells (data not shown). This was surprising, because it had been shown for cells other than DC that proteins with late temporal expression can already be detected 4 h after infection, 6.5 h earlier than the time point at which the infected B-LCL were harvested for HLA ligand purification (43, 51). Therefore, we analyzed the levels of viral protein expression and viral HLA ligand presentation in infected B-LCL simultaneously. Comparative differential image analysis of 2D gels containing proteins from MVA- and mock-infected cells revealed several abundant proteins solely or predominantly expressed in the infected cells (data not shown). Furthermore, we specifically checked for expression of the proteins that gave rise to HLA ligands by predicting their theoretical spot coordinates on the gels. Differentially detected protein spots were digested by trypsin, and fragments were sequenced by LC-MS/MS. Peptide digestion products derived from proteins of both viral (Table II) and human origin (data not shown) were detected. Six of 24 identified viral proteins were late viral gene products (Table II), indicating that late proteins were available for proteasomal processing at the time cells were harvested for HLA ligand analysis. Under the conditions used for this experiment, several viral proteins would not be detected due to their extreme size or isoelectric point (pI): B22R and G5.5R would be missed due to their low molecular mass of 7.3 kDa each, whereas detection of E5R is unlikely due to its high pI of 10. Proteins detectable on the 2D gels had a molecular mass of more than 14 kDa and a pI between 4 and 9. Altogether, 69 MVA proteins do not match the above characteristics, and thus most likely should not be detected by this method.

Interestingly, there was little or no correlation between the relative abundance of intracellular viral proteins and directly processed viral peptides presented on HLA. Only 2 of the 24 most abundant MVA proteins were source proteins for identified HLA ligands, namely B8R (ligand KITSYKFESV_18–27) and C7L (ligand KVDDTFYYV_74–82), although CTL determinants for other proteins had been described previously (Table II). Interestingly, B8R and C7L provide immunodominant epitopes in mice and humans, respectively (14, 15, 17, 18, 25). In contrast, 11 of 13 source proteins for which we found HLA ligands were not detected using this approach, suggesting that most viral HLA ligands were derived from proteins of low abundance at 12.5 h postinfection.

We studied the immunogenicity of the identified HLA-A*0201 ligands by IFN-γ ELISPOT assay of PBMC derived from two HLA-A*0201-positive donors immunized twice with MVA. An initial screen of PBMC taken before vaccination and up to 30 days postboost (p.b.) indicated that immunization induced specific IFN-γ production in response to four of five HLA ligands tested (Fig. 3). Because specific responses against two peptides, A48R_187–195 and C7L_74–82, were seen in both donors and no responses were detected in PBMC isolated from preimmune samples of either donor, we concluded that these responses were induced by immunization and were not generated by in vitro peptide/IL-2 stimulation before analysis.

We extended our study to the complete panel of HLA ligands and analyzed in vitro expanded PBMC taken 2.5 years p.b. by flow cytometry using combined MHC/peptide tetramer and intracellular IFN-γ staining (Fig. 4 and Table III). For all peptides except B8R_18–27, tetramer-positive CD8⁺ T cell populations were found in donor 1 (Fig. 4, left panels), indicating that these HLA ligands are A*0201-restricted CTL epitopes and are simultaneously recognized as part of the long-term memory response to MVA. CD8⁺ T cell populations identified by tetramers specific for 8 of 11 epitopes were functional and responded to in vitro stimulation with the specific MVA peptide by down-regulating the TCR and producing IFN-γ (Fig. 4, right panels). Because the populations recognized by tetramers specific for the remaining three peptides, B19R_207–215, A23R_273–281, and D12L_251–259, were small (less than 0.06%), our assay may not have been sensitive enough to detect production of IFN-γ by a fraction of these cells reliably. However, TCR down-regulation as an indication of activation was clearly observed in response to two of these HLA ligands: B19R_207–215 and A23R_273–281. In donor 2, CD8⁺ T cell populations specific for 8 of 12 HLA ligands were detected by tetramer staining (Table III). Six of these populations specifically responded to in vitro stimulation with TCR down-regulation and IFN-γ production. Stimulation with E5R_93–101 induced IFN-γ-producing CD8⁺ T cells, which, however, were not stained by the matching tetramer, possibly due to a low-affinity TCR (Table III).

Analysis of the same samples by IFN-γ ELISPOT assay after a single round of in vitro expansion gave similar results: eight epitopes were recognized in at least one donor, and four were recognized in both (Fig. 5). In addition, IFN-γ production in response to A23R_273–281 by PBMC from donor 1 was observed, confirming this HLA ligand as a T cell epitope.

In summary, all 12 MVA-derived HLA-A*0201 ligands that we identified were immunogenic; 11 of these provided long-term T cell memory. We demonstrate that the cellular immune response to MVA infection is based on simultaneous recognition of many different CTL epitopes with donor-specific variations in the epitope-specific CD8⁺ T cell frequencies and in the epitope hierarchy. This finding is consistent with earlier analyses of human T cell responses to MVA (18) and other viruses such as CMV (52), EBV (53), and Influenza virus (54). Four HLA ligands proved to be common epitopes in the long-term response to MVA vaccination: C7L_74–82, the immunodominant epitope, as well as A48R_187–195, F12L_404–412, and O1L_247–255.

Because we observed T cells specific for the MS-identified HLA ligands more than 2 years after immunization in MVA vaccinees, we investigated long-term memory T cell responses specific for these peptides in PBMC from donors vaccinated with the VACV Dryvax vaccine more than 25 years ago. Because the sequences of the MVA-derived HLA-A*0201 ligands are identical with those in VACV Dryvax (Table IV), we anticipated that these ligands could also be immunogenic in the course of classical smallpox vaccination. Analysis of tetramer-specific CD8⁺ T cells derived from in vitro expanded PBMC of three HLA-A*0201-positive Dryvax vaccinees at 25–44 years p.b. revealed specificity for seven of the identified HLA-A*0201 ligands (Table V). All three donors contained CD8⁺ T cells specific for two of the common epitopes identified in the previous experiments with MVA vaccinees, C7L_74–82 and F12L_404–412, whereas two donors contained T cells specific for G5.5R_27–35 and O1L_247–255. These results suggest that one round of prime-boost vaccination with MVA or Dryvax was sufficient to induce a long-lived cellular immune response to several identical epitopes.

To determine whether the MVA epitopes described in this study can potentially cross-protect against infection by other VACV strains or orthopoxviruses, we compared the sequences of the MVA-derived HLA ligands with those derived from VACV Dryvax, VACV WR, and VARV, as well as monkey pathogenic MPXV, which has been described recently to cause human disease (55). Seven sequences were conserved between all strains (Table IV), including three of the common epitopes shown to be cross-reactive between MVA and VACV Dryvax (Table V). Although the immunodominant epitope C7L_74–82 is identical in the VACV strains, it differs by 2 aa in MPXV, making it unlikely that this epitope can provide protection against MPXV.

All of the HLA ligands identified in this study were peptides derived from early viral proteins. Recently, we found that T cells recognizing such peptides were capable of dominating the response to a secondary VACV infection. Therefore, we tested whether vaccination with the identified peptides would be able to clear an orthopoxviral infection. Importantly, HLA-A*0201 transgenic HHD mice were fully protected against a lethal respiratory challenge with the virulent VACV strain WR after a single immunization with a pool of three peptides derived from early gene products identified in this study (B8R_18–27, G5.5R_27–35, and C7L_74–82) (Fig. 6,B). In contrast, peptides from late viral proteins (A6L_6–14, H3L_184–192, and I1L_211–219) that dominate the primary response in this mouse model after MVA immunization (29) (Fig. 6,A) or induced much higher CD8⁺ T cell frequencies when applied in this study as pooled peptide vaccine (Fig. 6 A) were less protective. These animals showed a dramatic loss of weight (similar to the control group) and suffered prolonged disease progression (>25 days), whereas all mice in the early peptide group were fully recovered by day 14. Of note, the addition of CpG oligodeoxynucleotide as an adjuvant in all peptide vaccine preparations showed some unspecific protective capacity, as demonstrated by survival of control peptide-vaccinated mice, which was most likely mediated by the innate immune response.

Identification of viral HLA ligands by LC-MS/MS analysis has resulted in a better understanding of the cellular antiviral immune response. The major challenge has been to find a limited number of signals derived from the virus in the multitude of self-peptides. Planz et al. (32) used a tedious “predict-calibrate-detect” strategy to identify an HLA ligand from borna disease virus, whereas de Jong and van Els and colleagues (33, 35) have developed elegant approaches based on in silico subtraction and metabolic labeling to study measles and respiratory syncytial viruses. Unfortunately, these strategies have not become routine, and the number of viral HLA ligands known is still very limited. To date, only one MVA-derived HLA ligand has been found by LC-MS/MS analysis of peptides isolated from infected cells (31).

The strategy described in this study, differential analysis of HLA ligands by chemical stable isotope labeling after purification of peptides, combines several advantages. First, comparative measurements of HLA ligands from infected and mock-infected cells eliminate the need for ligand prediction. Second, identification of single peptide peaks in a survey LC-MS scan is time effective using manual evaluation, and algorithms providing automatic evaluation are expected to become available shortly. Third, the presence of a constant normalizing signal based on the self-peptides limits the requirement for reproducibility in chromatographic retention, peptide ionization, and selection for fragmentation. Finally, this approach can be applied to tissue taken from any organism, and may allow comparative analysis of different sites of infection.

Using differential stable isotope labeling of HLA ligands purified from infected and mock-infected cells, we discovered 15 MVA-derived ligands, 12 restricted to HLA-A*0201 and 3 to HLA-B*0702 (Table I). Eight ligands represent novel sequences. One peptide, A48R_187–195, had been described by Johnson et al. (31), and 6 ligands matched known CTL epitopes (references see Table VI). Nine proteins from which HLA ligands were derived are among the 29 previously described immunogenic early proteins (10), and 4 proteins, B19R, E5R, G5.5R, and B15R, were newly identified in this study to contain relevant human CTL epitopes (Table VI). The proteins bearing HLA ligands functionally belong to two groups: proteins with immunomodulatory or host range and virulence function (B8R, B15R, B19R, C7L, and F12L (51, 56, 57)), and proteins functionally connected to DNA replication or transcription (E5R, A48R, G5.5R, A23R, D12L, and J6R (51)). The function of O1L is unknown. The novel immunogenic proteins B15R and B19R are known cytokine receptors of VACV and play a pivotal role in the VACV-mediated interference with the immune response (56, 58). Less is known about the function of the two other proteins newly identified as T cell epitope sources: E5R is located in cytoplasmic sites of viral DNA replication, where it associates with the proteins H5R and E3L, which were both detected as abundant proteins in this study (59); G5.5R has been described as a subunit of the DNA-dependent RNA polymerase (60). Two proteins, O1L and D12L, bear two HLA ligands each, suggesting high immunogenicity.

Notably, we exclusively detected HLA ligands from viral gene products expressed early or early/late during the viral life cycle (51) (Table I). This is consistent with our previous observation that presentation of peptides derived from late viral Ags to specific T cells by infected mouse target cells was very inefficient, but could be restored by the expression of the same viral Ags under the control of early promoters (29). This finding indicates a bias for early viral Ags to be processed and presented on MHC class I molecules of infected APC. However, because T cell responses against late viral proteins are found in humans and mice, the data support the concept that T cells are efficiently cross-primed upon MVA vaccination (28), particularly, when considering that late protein synthesis is blocked in MVA-infected DC (43).

Several mechanisms can be invoked to explain the inability to detect HLA ligands from late proteins on infected cells. Early and late viral gene products are transcribed and translated in distinct cellular compartments (61, 62), possibly resulting in variable availability for Ag processing. Alternatively, the initiation of cell death during the course of infection may reduce the loading capacity of the cellular Ag-presenting machinery, thus reducing the abundance of HLA ligands from late viral proteins (63).

Several studies have described a number of early as well as late epitopes of MVA and replication-competent VACV strains based largely on T cell analysis of immunized donors (13, 14, 15, 16, 17, 18, 21). Our data concur with the finding that VACV-specific CTL epitopes are predominantly derived from early proteins (17). We also confirm C7L_74–82 as the immunodominant epitope, and F12L_404–412 and O1L_247–255 as subdominant epitopes. However, using the MS-based technique, we were able to detect only 4 of the 24 published HLA-A*0201-restricted epitopes as HLA ligands, but then only 2 of the remaining 20 were derived from source proteins, from which we found HLA ligands (Table VI). One reason might be that different virus strains have been used for these studies. Particularly, replication-competent VACV strains might differ in pattern of Ag presentation or immunogenicity compared with MVA. In addition, several other factors may affect this limited overlap between the actual repertoire of MVA HLA ligands described in this study and previously identified HLA-A*0201-restricted CTL determinants described for MVA or replication-competent VACV. First, technical restrictions within the LC-MS/MS analysis are likely to prevent the detection of all MVA-derived HLA ligands. Even if a peptide is presented by a sufficient number of HLA molecules to produce a signal with sufficient intensity, a peptide peak may be missed due to coelution of peptides of similar m.w. or suppression of peptide ionization by coeluting peptides (64). Furthermore, some peptide sequences are difficult to detect due to their chemical characteristics, as may be the case for the epitope B22R_79–87 (14, 18) containing a cysteine residue that can react by oxidation. Another possible reason for the limited overlap might be inherent to the in vitro infection model that we chose to generate the material for our HLA ligand analysis. We used one defined cell type and analyzed one time point postinfection. In addition, cross-presentation of epitopes might add to the repertoire of CTL determinants in vivo. A third explanation may be the individual heterogeneity of subdominant epitopes, e.g., many VACV Dryvax epitopes were characterized solely by IFN-γ production in a single donor (17). In contrast, 10 of 12 HLA-A*0201 ligands identified in this study were recognized by more than 1 of the 5 donors tested, suggesting that they are immunologically highly relevant. Finally, the limited overlap might also be a result of differing T cell assay protocols. In contrast to some other groups, we restimulated the PBMC of vaccinees twice to clearly detect the T cells with specificity for the HLA ligands presented by MVA-infected cells.

In summary, the MS-based technique used in this study seems to be a reliable method to identify clinically relevant viral CTL epitopes and could be applied to other large-genome pathogens or recombinant Ags expressed by MVA. We identified 12 HLA-A*0201 and 3 HLA-B*0702 ligands derived from MVA. Nine of these 15 peptides were novel. All HLA-A*0201 ligands were shown to be actual CTL epitopes in MVA-immune donors. These peptides, preferably common and more dominant epitopes such as C7L_74–82, F12L_404–412, G5.5R_27–35, O1L_247–255, and A48R_187–195, are essential to monitor CD8⁺ T cell responses to MVA-based vaccines in clinical trials and may be used as correlates of protection. In addition, they seem suitable to be included, for example, as an epitope-based component in a smallpox vaccine that might be considered as a low-cost, safe, and stable alternative to traditional vaccines against bioterrorist smallpox threats.

We thank K. Ehrhardt for assisting evaluation, L. Yakes for expert proofreading, A. Krefft for excellent work in animal experiments, and all voluntary blood donors. We owe many thanks to C. Gouttefangeas for helpful discussion and for providing T cell sensitization protocols. The technical assistance of P. Hrstic in peptide synthesis, and of I. Buchen, J. Madlung, and C. Fladerer in proteomic analysis is gratefully acknowledged.

The authors have no financial conflict of interest.