Shifting immunodominance pattern of two cytotoxic T-lymphocyte epitopes in the F glycoprotein of the Long strain of respiratory syncytial virus

Respiratory syncytial virus (RSV) is a major cause of lower respiratory tract infection in infants and young children (Collins et al., 2001), affecting also immunocompromised patients and the elderly. Reinfection is common and currently no effective vaccine is available (Simoes, 1999).

Immune mechanisms involved in RSV disease and protection are not well understood. Protection is mediated mainly by neutralizing antibodies directed to the two virion surface glycoproteins, while clearance of virus-infected cells requires CD8⁺ T lymphocytes. Studies in the mouse model of human RSV have shown that CD8⁺ T lymphocytes play a role both in lung pathology and viral clearance (Cannon et al., 1988; Graham et al., 1991). CD8⁺ T lymphocytes are thought to regulate differentiation and activation of Th2 CD4⁺ T lymphocytes, which mediate lung pathology by recruiting eosinophils into the lungs during RSV infection (Srikiatkhachorn & Braciale, 1997). On the other hand, CD8⁺ T lymphocytes specific for the matrix 2 (M2) M2₈₂₉₂ epitope have been found to be the sole mediators of resistance to RSV infection in BALB/c mice infected with a recombinant vaccinia virus (rVV) expressing the M2 protein of RSV (Kulkarni et al., 1995). Further studies found that RSV infection altered CD8⁺ effector activity and memory T lymphocytes selectively in the lungs (Chang & Braciale, 2002; Chang et al., 2001). This RSV-induced immune dysregulation of virus-specific CD8⁺ T lymphocytes has been suggested as a possible mechanism for the absence of durable long-lived immunity to RSV infection (Chang & Braciale, 2002).

The fusion (F) protein of RSV, one of the two major surface glycoproteins in the virion, has long been recognized as a major vaccine candidate as it is an important target antigen for virus-specific cytotoxic T lymphocytes (CTL) (Pemberton et al., 1987) and neutralizing antibodies (Olmsted et al., 1986). Monoclonal antibodies to the F glycoprotein, which is highly conserved among the two antigenic groups of human RSV, passively protect against human RSV challenge in the mouse (Taylor et al., 1984) and reduce the severity of disease in premature and newborn babies (The IMpact-RSV Study Group, 1998). Furthermore, immunization of mice with rVV (Olmsted et al., 1986; Stott et al., 1987) or plasmid DNA encoding the F protein (Bembridge et al., 2000; Li et al., 1998), which induce F-specific antibodies and CTL, confers protection against challenge with RSV of either antigenic group.

Human RSV F glycoprotein is synthesized as a 574 aa F0 precursor that is processed at two cleavage residues (109 and 136) by furin-like proteases (González-Reyes et al., 2001; Zimmer et al., 2001). Cleavage yields two chains that remain linked by a disulfide bond. The larger carboxy-terminal F1 chain hosts the fusion peptide, the transmembrane region and several neutralizing antibody antigenic sites (López et al., 1998). The amino-terminal signal sequence is followed by the F2 chain, which has recently been found to be responsible for RSV host-cell specificity suggesting it is the binding partner for specific RSV entry receptors (Schlender et al., 2003). Furthermore, two murine H-2K^d-restricted epitopes recognized by CTL, F8593 and F92106, have been identified in the F2 chain of the A2 strain of human RSV (Chang et al., 2001; Jiang et al., 2002).

In this study, we used a novel approach to generate RSV-specific CTL lines by stimulating primed mouse splenocytes with the BCH4 fibroblast cell line that is persistently infected with the Long strain of RSV (Fernie et al., 1981). We found that the F92106 epitope is not present in RSV Long whereas the F8593 epitope is conserved between strains. We searched for CTL epitopes in the F1 chain by using CTL lines, epitope prediction programs and an rVV encoding a mutant F protein, in which the F2 chain was deleted. A novel H-2K^d-restricted epitope, F249258, was identified. No hierarchy in CD8⁺ T-lymphocyte responses to F8593 and F249258 epitopes was found in vivo during a primary response to an rVV expressing the full-length F glycoprotein. In contrast, F249258 was found subdominant with respect to F8593 in in vivo memory and secondary responses, and in CTL lines generated in vitro.

Viruses.
Human RSV Long was grown in HEp-2 monolayers and purified from culture supernatants as described previously (García-Barreno et al., 1988).

Construction of recombinant vvF containing the F gene of RSV Long (VRBF) has been described (Bembridge et al., 1998). An rVV expressing an F protein with aa 34128 (both included) deleted, vvF-ΔF2, was obtained as follows: plasmid LF1, containing a cDNA copy of the entire F protein gene of the Long strain (Cristina et al., 1990), was mutagenized to introduce a restriction site recognized by BbsI (nt 106, Long sequence; López et al., 1988) and AflII (nt 397), using the Quick-Change site-directed mutagenesis kit (Stratagene) and appropriate oligonucleotides. The resulting plasmid was digested with BbsI and AflII, treated with Klenow polymerase to make it blunt-ended, and ligated before being used to transform DH5 bacterial cells. Transformants were tested by PCR for the presence of plasmid with the deleted sequence that removed aa 34128 (both inclusive) from the F2 chain of the F protein. In addition, residue 31 was changed from glutamic to aspartic acid as a consequence of the cloning strategy. The entire F segment of one of the plasmids (LF1-ΔF2) was sequenced and ΔF2 insert was subcloned into plasmid pRB21 (Blasco & Moss, 1995) after digestion with BamHI and EcoRI. This plasmid was used to select an rVV of the vRB12 strain by the method of Blasco and Moss (1995). The rVV was plaque-purified three times and expression of F protein was tested by immunofluorescence with specific monoclonal antibody (mAb) 47F (García-Barreno et al., 1989). rVV stocks were grown in CV-1 monolayers and consisted of clarified sonicated cell extracts purified by ultracentrifugation through a 36 % sucrose cushion.

Cell lines.
P13.1 cell line, a derivative from P815 mastocytoma cells (H-2^d) by transfection with lacZ gene encoding β-galactosidase (Carbone & Bevan, 1990), was provided by H.-G. Rammensee (Tübingen University, Tübingen, Germany) and was cultured in Iscove's modified Dulbecco's medium supplemented with 10 % fetal bovine serum and 5x10⁵ M 2-mercaptoethanol. Human lymphoblastoid T2 cells deficient in transporters associated with antigen processing and transfected with K^d (Zhou et al., 1994) were provided by G. Hämmerling (German Cancer Research Centre, Heidelberg, Germany) and were cultured in RPMI 1640 medium supplemented with 10 % fetal bovine serum and 5x10⁵ M 2-mercaptoethanol. BALB/c fibroblast cells and human RSV Long persistently infected BCH4 fibroblast cells (Fernie et al., 1981) were obtained from B. Fernie (Georgetown University School of Medicine and Dentistry, Manassas, VA) through G. Taylor (Institute for Animal Health, Compton, UK), and were cultured in Dulbecco's modified Eagle medium supplemented with 10 % fetal bovine serum. All cell lines were cultured at 37 °C in a 5 % CO₂ atmosphere.

Synthetic peptides.
Peptides were synthesized in an Applied Biosystems peptide synthesizer (model 433A) and purified. Identity was confirmed by electrospray mass spectrometry by D. López (Centro Nacional de Microbiología, Madrid, Spain). All peptides that tested positive in cytotoxicity assays were found to be homogeneous by reversed-phase high-performance liquid chromatography. Peptides are named indicating the amino and carboxyl termini in the amino acid sequence of the F protein of RSV Long. Sequences are F3140, EFYQSTCSAV; F3240, FYQSTCSAV; F8593, KYKNAVTEL; F92106, ELQLLMQSTPAANNR; F249257, TYMLTNSEL; F249258, TYMLTNSELL; F365373, VFCDTMNSL and F467475, LYVKGEPII.

Polyclonal CTL lines and cytotoxicity assays.
Female BALB/c mice (H-2^d haplotype) bred in our colony were infected by intraperitoneal injection of 10⁷ p.f.u. of vvF or vvF-ΔF2. Three weeks post-infection an identical booster was given, and three or more weeks later mice were sacrificed and spleen cells obtained. CTL lines were named indicating the virus used for priming in vivo and the agent used for restimulation in vitro (Table 1). Thus CTL F/RSV derive from vvF-primed mice and were restimulated in vitro with RSV-infected splenocytes, mimicking a described method (Gaddum et al., 1996). Splenocytes from vvF-primed mice were stimulated with naïve splenocytes infected for 90 min with RSV Long at a m.o.i. of 0·4 p.f.u. per cell. Recombinant human interleukin 2 (IL2), generously provided by Hoffman-La Roche, was added 5 days later at 25 U ml¹. Long-term cultures were restimulated weekly with IL2 and RSV-infected splenocytes at an effector : stimulator ratio of 3 : 1. The CTL F/BCH4 and CTL ΔF2/BCH4 lines were generated by stimulating 5x10⁶ splenocytes ml¹ with 2x10⁵ BCH4 cells ml¹ treated with 90 µg mitomycin C ml¹ (Sigma). IL2 (25 U ml¹) was added after 5 days. Long-term cultures were restimulated weekly with IL2 and mitomycin C-treated splenocytes and BCH4 cells at an effector : stimulator ratio of 3 : 1 and 5 : 1, respectively. The CTL F/F8593 and CTL F/F249257 lines were generated by stimulating 10⁷ splenocytes ml¹ with 10¹⁰ M of the relevant peptide. Five days after selection, IL2 (25 U ml¹) was added. Long-term cultures were maintained by weekly restimulation with mitomycin C-treated 10⁶ M peptide-pulsed splenocytes (effector : stimulator ratio of 3 : 1) and cultured in medium with IL2 and 10¹⁰ M peptide. All CTL lines were cultured in minimal essential medium α supplemented with 10 % inactivated fetal bovine serum and 5x10⁵ M 2-mercaptoethanol at 37 °C in a 5 % CO₂ atmosphere. Standard 34 h ⁵¹Cr release cytotoxicity assays (Del Val et al., 1988) were performed incubating CTL and target cells that had been previously labelled for 6090 min with Na⁵¹CrO₄ at 37 °C in the presence or absence of 10⁶ M peptide and washed. Down to 100-fold lower peptide concentrations gave similar results. Alternatively, P13.1 cells were infected with rVV at a m.o.i. of 310 p.f.u. per cell as described (Eisenlohr et al., 1992), virus inoculum was thoroughly washed after 1 h adsorption, and infection was allowed to proceed for 3 h prior to cytotoxicity assay.

Table 1. CTL lines used in this study

Intracellular cytokine staining.
Intracellular cytokine staining assays were performed as described previously (Chen et al., 2000). CTL lines were stimulated in the presence of 10 µg brefeldin A ml¹ (Sigma) for 4 h with P13.1 target cells previously pulsed for 30 min with an excess 10⁵ M peptide. Pooled splenocytes were stimulated with an excess 10⁵ M peptide for up to 2 h, and stimulated for a further 3 h in the presence of brefeldin A. Following stimulation, cells were incubated with FITC-conjugated anti-CD8α (clone 53-6.7) mAb for 20 min at 4 °C, fixed with Intrastain kit (DakoCytomation) reagent A, and incubated with phycoerythrin-conjugated mAb to interferon-γ (IFN-γ) in the presence of Intrastain kit permeabilizing reagent B for 30 min at 4 °C. All antibodies were purchased from BD PharMingen. Events were acquired using a FACSCalibur flow cytometer (BD Biosciences) and data were analysed using CellQuest software (BD Biosciences). In vitro generated CTL lines specific for the F glycoprotein of the Long strain of RSV identify F8593 as a conserved epitope
Polyclonal F-specific CTL lines were generated in vitro from splenocytes obtained from BALB/c mice primed with vvF, an rVV encoding wild-type fusion glycoprotein F of RSV Long. Initially, polyclonal CTL lines were generated using procedures that kept them multispecific, that is, capable of recognizing as many F glycoprotein epitopes as possible. Thus, the CTL F/RSV line was generated by in vitro restimulation with RSV Long-infected splenocytes (Gaddum et al., 1996). Using a novel approach the CTL F/BCH4 line was generated by stimulation with BCH4 cells. RSV Long persistently infected BCH4 cells (Fernie et al., 1981) have been extensively used as targets of F-specific CTL despite their reduced synthesis of F and G viral glycoproteins (Martínez et al., 2001). However, they have not been previously used as stimulators to obtain F-specific CTL lines. Functional characterization of CTL lines was carried out in cytotoxicity assays with ⁵¹Cr-labelled BCH4 cells as RSV-infected targets (Fig. 1a). Both CTL F/RSV and CTL F/BCH4 lines were able to recognize RSV persistently infected BCH4 targets, whereas no significant lysis was observed with non-infected P13.1 targets. Poorer performance of CTL F/RSV might reveal a suboptimal restimulation protocol. BALB/c fibroblasts (Fernie et al., 1981) were not used as a negative control as, unexpectedly, no evidence was found of presentation of any synthetic peptide representing known viral epitopes to proven active CTL. Lack of recognition of wild-type vaccinia-infected cells confirmed the absence of vaccinia-specific CTL in CTL F/BCH4 that were otherwise able to efficiently detect presentation of F protein epitopes in vvF-infected cells (Fig. 1b).

(27K): Fig. 1. In vitro generated CTL lines specific for the F glycoprotein of the Long strain of RSV identify F8593 as a conserved epitope. Long-term cultures of multispecific CTL were generated from BALB/c mice infected with an rVV encoding the F protein of RSV Long (vvF). Splenocytes obtained from vvF-primed mice were stimulated with RSV-infected splenocytes (CTL F/RSV) or RSV persistently infected BCH4 cells (CTL F/BCH4). (a) CTL F/RSV and CTL F/BCH4 were used as effectors in a cytotoxicity assay, and RSV persistently infected BCH4 cells () or non-infected P13.1 cells () as targets. Percentage specific lysis calculated from 51Cr release at each E : T ratio is represented. (b) CTL F/BCH4 were used as effectors and P13.1 cells infected with vvF () or its parental strain WR () as targets. (c) Cytotoxicity assay performed with CTL F/RSV and CTL F/BCH4 as effectors and P13.1 targets pulsed with no peptide (), F92106 peptide (x) or F8593 peptide (), corresponding to the epitopes that have been defined within the F2 chain of the A2 strain of RSV. (d) Titration cytotoxicity assay performed with P13.1 targets pulsed with increasing concentrations of F8593 synthetic peptide at a 5 : 1 E : T ratio with CTL F/BCH4 () and CTL F/F8593 () as effectors. The latter CTL line was generated from mice infected with vvF by stimulation with F8593 peptide. Percentage specific lysis calculated from 51Cr release obtained at different concentrations of F8593 used to pulse targets is represented.

F glycoprotein epitopes in A2 strain of RSV are known to be presented only by MHC class I molecule K^d (Chang et al., 2001) in BALB/c mice, which express K^d, D^d and L^d. To assess if this is also true for F glycoprotein epitopes in RSV Long, cytotoxicity assays were performed with CTL F/BCH4 and vvF-infected targets transfected separately with each MHC class I molecule. Efficient presentation to active CTL was found with targets transfected with K^d but not with L^d or D^d (data not shown). Thus, as for RSV A2, in the H-2^d haplotype CTL epitopes of the fusion glycoprotein of RSV Long are only presented by K^d.

To identify epitopes presented by K^d to multispecific CTL F/RSV and CTL F/BCH4, we first assessed in cytotoxicity assays whether protein epitopes known in the A2 strain were recognized. Peptides F8593 and F92106 corresponding to these epitopes were synthesized and used to sensitize cells. We found that peptide F92106 is not recognized by either CTL line (Fig. 1c) and also fails to restimulate F-primed CTL (data not shown). Despite 98 % aa identity between the F glycoproteins of RSV strains A2 and Long (López et al., 1988), residues 11 and 12 of this 15-mer F92106 epitope are not conserved (PT in A2 and AA in Long) and may be critical for the unpredictable non-canonical binding of the epitope to K^d. It cannot be excluded either that the F92106 peptide is not a good reagent to mimic this potential epitope in the Long strain. Therefore, in RSV Long we have no evidence of the presence of this epitope.

On the other hand, both multispecific CTL F/RSV and CTL F/BCH4 lines were able to recognize F8593 sensitized cells (Fig. 1c). Furthermore, a long-term polyclonal monospecific CTL F/F8593 line was generated by stimulating splenocytes from vvF-primed mice with F8593 peptide. Titration assays were performed by pulsing targets with increasing concentrations of peptide, and both multispecific CTL F/BCH4 and monospecific CTL F/F8593 were found to recognize down to 10¹¹ M F8593 (Fig. 1d). Therefore, as expected from the conserved sequence within the epitope and flanking regions, F8593 is also a CTL epitope in the Long strain.

Identification of CTL epitope F249258 in the F glycoprotein of the Long strain of RSV
To identify other potential CTL epitopes in the F glycoprotein of the Long strain we constructed an rVV, vvF-ΔF2, encoding a mutant F protein in which aa 34128 have been deleted, and which therefore does not contain the F8593 epitope. Although the mutant was constructed in such a way as to allow maturation of the full F1 chain, reduced levels of expression compared with the wild-type protein were detected (data not shown). In spite of the lower expression of the mutant protein, cells infected with vvF-ΔF2 were efficiently lysed by CTL F/BCH4, thus demonstrating presentation of one or more undefined epitopes from vvF-ΔF2-infected cells to multispecific CTL F/BCH4 (Fig. 2a). It can be argued that a F8593 mimotope may be recognized by F8593-specific CTL clones present in the multispecific CTL line, or that an artificial epitope is being generated from the mutant protein. To address this issue the same targets were assayed with monospecific CTL F/F8593 and no lysis was observed (Fig. 2a). Furthermore, a multispecific CTL ΔF2/BCH4 line that did not include F8593-specific clones was generated from mice that had been primed with vvF-ΔF2 (see Table 1). Cytotoxicity assays were performed and as expected CTL ΔF2/BCH4 were unable to lyse targets exogenously loaded with F8593, but were, however, able to recognize vvF-infected targets (Fig. 2b). We therefore demonstrate the existence of one or more novel uncharacterized RSV Long F glycoprotein epitopes in the F1 chain or in the first 33 aa of the protein.

(19K): Fig. 2. Evidence of presentation of novel epitope(s) to F-specific CTL. (a) P13.1 cells were infected with vvF-ΔF2 (), an rVV which encodes a mutant F protein in which aa 34128 have been deleted. P13.1 cells infected with WR, in the presence () or absence () of F8593 peptide, were used as positive and negative controls, respectively. A cytotoxicity assay was then performed with CTL F/BCH4 and CTL F/F8593 as effectors as indicated. (b) Cytotoxicity assay performed with CTL ΔF2/BCH4 as effectors. Long-term culture CTL ΔF2/BCH4 was obtained from vvF-ΔF2-primed mice that were stimulated with RSV persistently infected BCH4 cells. Targets were P13.1 cells infected with vvF (), or WR in the presence () or absence () of F8593 peptide.

In order to identify the novel epitope or epitopes, RSV Long F protein sequence was subjected to MHC class I epitope prediction programs based on motif-refined algorithms, excluding prediction of non-canonical epitopes. Programs used were SYFPEITHI (Rammensee et al., 1999) and BIMAS (Parker et al., 1994). High-score putative epitopes predicted to be presented by K^d with both programs were further screened for correct C-terminal proteasomal cleavage in NetChop 2.0 Prediction Server (Kemir et al., 2002), and peptides corresponding to several predicted epitopes were synthesized. In some instances, length variants of the same region with similar scores were synthesized. We established our internal threshold by subjecting known viral K^d-restricted epitopes to each epitope prediction program. Presentation of peptides to CTL F/BCH4 was assessed in cytotoxicity assays performed with peptide-pulsed T2/K^d targets, which are deficient in transporters associated with antigen processing and thus have more peptide-receptive molecules available. Fig. 3(a) shows that targets pulsed with peptides F249257 and F249258 were efficiently lysed by CTL, as were F8593 peptide-pulsed targets. Again, to control that F249257/8 peptides were not being cross-recognized by F8593-specific CTL clones, the same targets were confronted to CTL F/F8593, but these were only able to recognize the F8593 peptide (Fig. 3a, second panel). Moreover, to address the possibility of a non-specific binding to other human presenting molecules expressed by T2 targets, an assay was performed with CTL F/BCH4 and peptide-pulsed T2/L^d targets and no specific lysis was observed (Fig. 3a, third panel). These results indicate that peptides F249257 and F249258 are presented by K^d to specific CTL clones. As it may be argued that targets are pulsed with high concentrations of peptide (10⁶ M), titration assays were performed to assess whether physiological concentrations of the peptides were recognized. Fig. 3(b) shows that targets pulsed with down to 10¹⁰ M peptide were efficiently lysed and that the 9- and 10-mer peptides were equally recognized by CTL F/BCH4. Reciprocal CTL lysis may account for the reduced lysis observed at high peptide concentrations. Altogether, these results demonstrate that standard concentrations of the F249257 and F249258 peptides can specifically activate RSV Long F protein-specific CTL.

(18K): Fig. 3. Recognition by F-specific CTL of synthetic peptides F249257 and F249258 predicted to be epitopes presented by Kd. (a) Human T2 cells transfected with murine MHC class I molecules Kd or Ld, as indicated, were pulsed with peptides F8593 (), F249257 (•) or F249258 (); or not pulsed with any peptide (). They were then tested in a cytotoxicity assay with CTL F/BCH4 and CTL F/F8593 as effectors, as indicated. (b) Titration cytotoxicity assay performed at a 5 : 1 E : T ratio with CTL F/BCH4 as effectors, and P13.1 cells pulsed with increasing concentrations of synthetic peptides F249257 (•) or F249258 () as targets.

Furthermore, a polyclonal monospecific CTL F/F249257 line was generated from vvF-primed mice by stimulating splenocytes with F249257 synthetic peptide. The CTL F/F249257 line was able to recognize targets pulsed with down to 10¹¹ M peptide (Fig. 4a), and to efficiently lyse vvF and vvF-ΔF2-infected targets (Fig. 4b). For the establishment of this CTL line, different F249257 peptide concentrations were tested in restimulation. The optimal concentration, 10¹⁰ M, was exactly the same as previously found optimal by us for CTL F/F8593 (data not shown).

(21K): Fig. 4. Generation and properties of the F249257 monospecific CTL line. A long-term culture of F249257-monospecific CTL, named CTL F/F249257, was generated from vvF-infected mice by stimulation with F249257 synthetic peptide. (a) Titration cytotoxicity assay performed at a 5 : 1 E : T ratio with CTL F/F249257 as effectors and T2/Kd cells pulsed with increasing concentrations of synthetic peptide F249257 as targets. (b) Cytotoxicity assay performed with CTL F/F249257 as effectors and P13.1 targets infected with vvF () or vvF-ΔF2 (), or with WR in the presence (•) or absence () of F249257 peptide, as positive and negative controls, respectively.

Importantly, this epitope was also generated in natural RSV infection of mice, as similar polyclonal monospecific CTL RSV/F249258 could be established (data not shown).

In contrast, neither recognition of peptide-pulsed target cells nor generation of monospecific CTL lines were observed with other predicted epitope peptides (F365373 and F467475, data not shown). The possibility cannot be excluded that other epitopes may exist within the deleted region, but a further predicted epitope located in the region deleted in vvF-ΔF2 represented by synthetic peptides F3140 and F3240 was also tested and found negative (data not shown). The results are summarized in Fig. 5.

(7K): Fig. 5. CTL epitopes in F protein of RSV Long. The F glycoprotein has three hydrophobic regions corresponding to the signal peptide (sp), the fusion peptide (fp) and the transmembrane region (tm). Processing by furin-like proteases at two cleavage sites (arrows) yields the F1 and F2 chains that remain linked by a disulfide bond. CTL epitopes () presented by Kd identified by us in the F protein of RSV Long are F8593 (in the F2 chain with sequence KYKNAVTEL) and F249258 (in the F1 chain with sequence TYMLTNSELL). Peptides predicted to be epitopes presented by Kd but not recognized by CTL are also indicated (). Their sequences are F3140, EFYQSTCSAV; F3240, FYQSTCSAV; F92106, ELQLLMQSTPAANNR; F365373, VFCDTMNSL and F467475, LYVKGEPII. The region deleted in the vvF-ΔF2 mutant is schematically indicated below with numbers marking the amino acids delimiting it.

Thus, we have identified a CTL epitope that is presented by K^d as a 9-mer (F249257; TYMLTNSEL) or a 10-mer (F249258; TYMLTNSELL) peptide. This epitope is located near neutralizing antibody antigenic site II (López et al., 1998), and is expected to also exist in the F protein of the A2 strain as its sequence and that of the flanking regions that may influence the efficiency of antigen processing and presentation (Del Val et al., 1991) are also conserved in at least 35 residues on either side. However, this F249257/8 epitope has not been identified in previous studies (Chang et al., 2001; Jiang et al., 2002) that used overlapping peptides in their screening. The panel of 15-mer overlapping peptides used in the identification of non-canonical A2 strain F92106 epitope did not include any peptide in which the F249257/8 epitope was complete (Jiang et al., 2002), which explains why it was not detected. Our results demonstrate that epitope prediction programs, although unable to predict non-canonical epitopes, can be powerful tools for CTL epitope identification, as discussed recently (Nussbaum et al., 2003).

Shifting immunodominance pattern of CD8⁺ T-lymphocyte responses to F8593 and F249258 after in vivo vvF infection and in CTL lines
A potential competition between the two epitopes presented by K^d and located in the F protein of RSV Long (F8593 and F249257/8) led us to examine the hierarchy of CD8⁺ T-lymphocyte responses. Factors that potentially contribute to immunodominance include efficiency of antigen processing, binding affinity of peptides to class I molecules, presence of T lymphocytes capable of responding to a given peptide/MHC complex and inhibition of responses to subdominant epitopes by CD8⁺ T lymphocytes specific for dominant epitopes (Chen et al., 2000; Yewdell & Bennink, 1999). The fact that recognition of peptide F249257 by monospecific CTL F/F249257 was better than by CTL F/BCH4 (Figs 4a and 3b, respectively), suggested that F249257 might be a subdominant determinant in the latter multispecific CTL line. Furthermore, when T lymphocyte clones specific for the immunodominant F8593 epitope were not present in the CTL line, as in CTL ΔF2/BCH4 or in CTL F/F249257 (see Table 1), targets pulsed with down to 10¹¹ M F249257 peptide were efficiently lysed (Fig. 6a). Therefore, the limit in F249257 peptide recognition by CTL F/BCH4 is imposed by F8593-specific T lymphocyte clones, not by BCH4 cells, as these are able to select more sensitive CTL when stimulating splenocytes from vvF-ΔF2-primed mice.

Table 1). (b) Cytotoxicity assay performed with CTL F/BCH4 effectors and P13.1 targets infected with vvF () or vvF-ΔF2 (); or with WR in the presence of F8593 () or F249257 (•) peptides as positive controls; or with WR () as a negative control.

Further evidence of the decreased sensitivity in epitope recognition of F249257/8-specific CTL clones present in CTL F/BCH4 came from the different recognition pattern observed at low effector to target (E : T) ratios in a cytotoxicity assay performed with multispecific CTL F/BCH4 and infected targets (Fig. 6b). At an E : T ratio of 1 : 1, 50 % or more specific lysis was only reached with targets presenting the F8593 immunodominant epitope, with both peptide-pulsed and vvF-infected cells. Differences in recognition did not result from differences in efficiency of antigen presentation from vvF and vvF-ΔF2, as Fig. 4(b) shows that both rVV were able to equally present the subdominant determinant to CTL F/F249257, a CTL line devoid of F8593-specific T lymphocyte clonotypes. In fact, recognition of vvF by CTL F/BCH4 (Fig. 6b) is higher than by CTL F/F249257 (Fig. 4b). This is not unexpected, as recognition of more epitopes by the former multispecific line than by the latter monospecific line (see Table 1) can easily account for this.

To assess whether responses to the immunodominant determinant F8593 were preventing the emergence of responses to the subdominant determinant F249257/8, we quantified CD8⁺ T lymphocytes specific for each epitope in an intracellular cytokine staining assay using peptide-pulsed cells to stimulate CTL F/BCH4 and CTL ΔF2/BCH4. Fig. 7(a) (upper panels) shows that when targets were pulsed with peptide F8593, 77 % of the CD8⁺ lymphocytes in the CTL F/BCH4 line were activated to produce IFN-γ, whereas only 9·8 and 9·4 % of the CD8⁺ lymphocytes were activated by peptides F249257 and F249258, respectively. Therefore CTL lines obtained from vvF-primed memory mice show immunodominance, as CTL responses are strongly skewed to F8593 at the expense of responses to F249257/8. Interestingly, when the immunodominant epitope F8593 was absent in the in vivo priming event that led to the generation of the CTL ΔF2/BCH4 line, 98 and 52 % activation was achieved by stimulation with peptides F249258 and F249257, respectively, showing that in the absence of the F8593 epitope, responses to the F249257/8 epitope are very efficiently generated. Thus, there is no limitation in the in vivo availability of CD8⁺ T lymphocytes capable of responding to F249257/8.

(31K): Fig. 7. Populations of CD8+ T lymphocytes responsive to F8593 and F249257/8 in CTL lines (a) and after vvF infection in vivo (b). Intracellular cytokine staining assays were performed by stimulating CTL or splenocytes with synthetic peptides, as indicated, in the presence of brefeldin A to intracellularly accumulate IFN-γ produced upon activation of CTL by interaction with peptide presented by Kd. Cells were stained with anti-CD8, fixed, permeabilized and stained for cytoplasmic IFN-γ. (a) Long-term CTL cultures CTL F/BCH4 (top panels) and CTL ΔF2/BCH4 (lower panels) were stimulated with the indicated peptides. Percentages of total CD8+ lymphocytes that stained for IFN-γ are shown in the upper right quadrant of each dot plot. Over 8000 CD8+ cells were analysed in each sample. (b) Splenocytes were obtained from mice 7 days post-infection with 107 p.f.u. vvF to study the primary response (white bars; data are the mean of seven mice from four independent experiments). The memory response was assessed in mice that received a booster of vvF 3 weeks later, and then were left resting for several months (grey bars; data are the mean of two mice from two independent experiments). The secondary response was studied in animals of the memory response group 7 days after a third dose of vvF (black bars; data are the mean of two mice from two independent experiments). Percentage of total CD8+ lymphocytes that stained for IFN-γ after activation with each peptide is represented. A mean of 50 000 CD8+ cells were analysed in each sample.

In addition, as 98 % of CD8⁺ T lymphocytes in CTL ΔF2/BCH4 responded to F249258 10-mer peptide, we suggest that no other epitope is recognized by this CTL line. Concerning endogenous antigen processing of the F glycoprotein in BCH4 cells, the fact that 10-mer F249258 is recognized by all CTL in the CTL ΔF2/BCH4 line strongly suggests that the 10-mer is probably a naturally produced K^d ligand in BCH4 cells.

Hierarchy in primary, memory and secondary CD8⁺ T-lymphocyte responses to F8593, F249257 and F249258 was studied in vivo after vvF infection (Fig. 7b). CD8⁺ lymphocytes activated with each peptide in an intracellular cytokine staining assay are observed with a frequency similar to that reported in other systems (Chen et al., 2000). Staining with tetramers K^d/F8593 and K^d/F249258 essentially gave the same results (data not shown). Low frequencies are probably due to competition with vaccinia virus determinants, which for example have previously been suggested to account for difficulty in detection of the NP₁₄₇₁₅₅ epitope following infection with an rVV encoding the complete influenza virus nucleoprotein, NP (Chen et al., 2000; Restifo et al., 1995). This has been discussed in terms of the limitations of viral vectors that elicit massive CTL responses to their own antigens.

In the primary immune response to vvF (Fig. 7b, white bars), there is no limitation for the generation of a similarly effective CD8⁺ T-lymphocyte response to either epitope, as frequencies are comparable after stimulation with either peptide, thus concluding that no significant hierarchy exists in the in vivo primary response in vvF-infected mice. However, frequencies of epitope-specific populations in the memory immune response (Fig. 7b, grey bars) increase for F8593 and decrease for F249257/8 in comparison with the primary immune response (Fig. 7b, white bars). Therefore, in the initial experimental setting for in vitro selection of CTL lines by restimulation with BCH4 cells, the F8593-specific CD8⁺ T lymphocyte population is already in advantage. This difference is probably amplified by restimulation with BCH4 cells to that observed in the CTL F/BCH4 line (Fig. 7a).

Furthermore, in a secondary immune response (Fig. 7b, black bars) the mean percentage of activated CD8⁺ lymphocytes after stimulation with peptide F8593 is almost 10-fold higher than after stimulation with peptides F249257 or F249258. This difference in the number of F8593 and F249257/8 epitope-specific CD8⁺ T lymphocytes is similar in proportion to that observed in the established in vitro generated CTL F/BCH4 line (Fig. 7a). Therefore, the CTL F/BCH4 line is representative of an in vivo secondary response to vvF infection.

Different patterns of T cell immunodominance in primary and secondary responses have been described previously (Belz et al., 2000; Crowe et al., 2003). In influenza virus primary infection, similar frequencies of T cells specific for epitopes D^bPA₂₂₄ and D^bNP₃₆₆ are observed, whereas the secondary response is dominated by T cells specific for D^bNP₃₆₆ (Belz et al., 2000). The mechanism underlying this changing pattern of immunodominance has been investigated in a recent study that proposes a model in which differential antigen presentation by dendritic and non-dendritic cells together with the capacity of T cells to perceive the antigens presented by these cells, regulate the T cell specificity pattern (Crowe et al., 2003). Along with this hypothesis, we suggest that in our experimental system professional antigen presenting cells with a similar proportion of epitopes F8593 and F249258 would activate naïve T cells during the primary immune response, whereas during the secondary immune response memory T cells would also be activated by cross-priming or by non-professional antigen presenting cells with a higher proportion of epitope F8593. This would lead to the gradual establishment of a CD8⁺ T-lymphocyte response strongly skewed to epitope F8593.

The different immunodominance pattern for epitopes F8593 and F249258 observed in primary and secondary vvF infection must be taken into account when considering single epitope vaccines. However, the finding that CD8⁺ T lymphocytes specific for either epitope are generated in vivo holds promise for the potential to generate a multispecific CD8⁺ response in natural infection or after vaccination. In conjunction with neutralizing antibodies targeting the F protein, a multispecific CD8⁺ response would better contribute to clearing RSV infection, as it would hinder the emergence of viral CTL escape mutants lacking single determinants. Definition of novel CTL epitopes thus contributes to addressing these issues in the mouse model.

We thank Drs H.-G. Rammensee (Tübingen University, Tübingen, Germany), G. Hämmerling (German Cancer Research Centre, Heidelberg, Germany) and G. Taylor (Institute for Animal Health, Compton, UK) for cell lines; and Rafael Blasco (INIA, Madrid, Spain) for pRB21 plasmid and vRB12 vaccinia virus. Recombinant human IL2 was a gift of Hoffmann-La Roche (Nutley, NJ). We are grateful to Dr J. A. López for help in initial stages of construction of rVV. Technical assistance of F. Vélez and C. Mir is gratefully acknowledged. This work was supported by grants PM99-0022 from Ministerio de Ciencia y Tecnología, 08.2/0004/2000.1 from Comunidad de Madrid, and 01/33 from Instituto de Salud Carlos III to M. D. V. and 01/24 from Instituto de Salud Carlos III to J. A. M. C. J. is the holder of a grant from Ministerio de Ciencia y Tecnología, P. L. and F. M. were supported by Comunidad de Madrid and the European Union.