INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Studies over the past decade have established the importance of CD8⁺ T cells in acquired antiviral immunity. Antibodies had long been considered the sole determinants of viral vaccine efficacy, but work in the mouse model systems of murine cytomegalovirus and lymphocytic choriomeningitis virus (LCMV) proved definitively that induction of virus-specific CD8⁺ T cells, in the absence of antibodies, was sufficient to confer solid immunity against subsequent virus challenge (15, 17, 41). Similar findings have since emerged in a large number of animal models (14, 19, 35), and cytotoxic T lymphocytes (CTL) have been implicated in the response to primary human immunodeficiency virus (HIV) infection in humans (4, 10).

During a virus infection, T-cell responses are generally limited to only a few epitopes from the viral genome, and those epitopes to which responses are mounted are termed dominant. However, in the absence of a dominant epitope, the host is capable of mounting an immune response against other subdominant epitopes, which are often ignored if a dominant sequence is present. The characterization of subdominant epitopes is important, since these sequences can induce protective immune responses (7, 8, 23, 36-38). In BALB/c (H-2^d) mice the overwhelming CTL response is directed towards a single dominant epitope, RPQASGVYM (nucleoprotein [NP] positions 118 to 126 [NP_118-126]) presented by the L^d molecule (42), but in our previous DNA immunization studies we found that a minigene encoding this dominant epitope did not confer complete protection, even when fused to the polypeptide ubiquitin; in numerous experiments, 75% of the minigene-immunized mice survived lethal challenge, compared to 100% survival if full-length U-NP was used (28). One possible explanation was that NP contained additional epitopes that contributed to the protection conferred by the full-length protein. A previous careful search for subdominant LCMV epitopes in BALB/c mice had failed to locate any protective epitopes in the viral NP (38). However, that effort had used major histocompatibility complex (MHC) motif predictions and peptide-MHC binding as initial identifying criteria, so we decided to employ a biological criterionprotective immunityto screen for subdominant epitopes in LCMV NP. In this work we identify a region in LCMV NP containing two nested subdominant epitopes presented by different MHC alleles. This region, when delivered as a ubiquitinated DNA vaccine, confers strong protective immunity. In addition, we have evaluated the kinetics of dominant and subdominant CD8⁺ T-cell responses after DNA vaccination and virus infection, and we show that they differ in their expansion rates and their acquisition of lytic activity. Finally, we have evaluated the two major CD8⁺ T-cell effector mechanisms, perforin-mediated lysis and cytokine production, during an acute virus infection in vaccinated animals. We find that the acquisition of lytic capacity by CD8⁺ T cells differs depending on their epitope specificity and that lytic activity of virus-specific CD8⁺ T cells declines more rapidly than the ability to produce antiviral cytokines. These findings have implications for vaccine design and for the evolution of the immune response.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mice, cell lines, and viruses. BALB/c mice (H-2^d) were obtained from the breeding colony at Scripps Research Institute, and were used at 6 to 16 weeks of age. BALB clone 7 (BALB c17) cells, an H-2^d fibroblast line, and T2-L^d and T2-K^d cells (6, 47), which are deficient in the transporters for antigen processing (TAP) and express the designated murine MHC class I alleles, were maintained in culture with RPMI medium. Vero 76 cells were grown in medium 199 (GIBCO-BRL). All media were supplemented with 10% fetal calf serum, L-glutamine, and penicillin-streptomycin. The virus used was LCMV (Armstrong strain).

Virus titration. LCMV titration was performed on Vero 76 cells, which were plated at a density of 6.6 × 10⁵ per six-well plate, 24 h prior to titration. Tenfold dilutions of samples were made in medium 199 (GIBCO-BRL)-10% fetal calf serum and applied to the indicator cells. Following adsorption and infection for 1 h at 37°C in an atmosphere of 5% CO₂, the inoculum was withdrawn and replaced with an overlay of 0.5% sterile ME agarose (FMC Biochemicals)-1X complete medium 199. Four days later the monolayer was fixed with 25% formaldehyde in phosphate-buffered saline (PBS), the agarose plug was removed, and the monolayer was stained with 0.1% crystal violet-20% ethanol in PBS.

Construction of recombinant plasmids. Plasmid pCMV-NP, encoding the LCMV NP gene with the dominant H-2^d epitope deleted, was digested with the restriction enzymes BamHI, BglII, and BclI, and the resulting restriction fragments (I through V) were cloned in frame with the carboxyl-terminal end of the ubiquitin gene in pCMV-UbiqF1/2 or pCMV-UbiqF3. These plasmid vectors contain the mouse ubiquitin gene with (i) a Kozak initiator sequence (18), (ii) the last codon mutagenized from GGC (Gly) to GCA (Ala) to enhance delivery to the proteasome (29), and (iii) a silent point mutation (A to T) in nucleotide position 9 to disrupt the intragenic BglII cleavage site, thereby facilitating cloning. pCMV-UbiqF1/2 contains the recognition sites for BclI and BglII restriction enzymes immediately downstream of the ubiquitin open reading frame (ORF), designed to allow readthrough into frame 1 and frame 2, respectively, of the inserted fragment, while in pCMV-UbiqF3 the restriction recognition site for BglII was placed to permit readthrough into the third frame; thus, any fragment can easily be cloned in frame with ubiquitin. Two complementary oligodeoxynucleotides of 37 nucleotides were designed to encode the subdominant epitope and were synthesized with overhanging ends compatible with the BglII site: 5'-GATCATGCCATACATAGCTTGTAGAACATCGATTTAA and 5'-GATCTTAAATCGATGTTCTACAAGCTATGTATGGCAT. These oligonucleotides were hybridized and cloned in frame into the pCMV-UbiqF1/2 plasmid digested with BglII. The resulting plasmid, pCMV-UMGX, encodes region X (containing subdominant epitopes) preceded by a methionine and covalently attached to the ubiquitin gene. The plasmid pCMV-U (encoding ubiquitin alone [a negative control in our experiments]), and the plasmid pCMV-UMG34 (encoding the dominant CTL epitopes for the H-2^b and H-2^d backgrounds as a fusion with ubiquitin) were previously described (28), and plasmid pCMV-UMG4 encodes only the dominant H-2^d CTL epitope fused to ubiquitin. All the products are expressed under the control of the immediate early promoter of human cytomegalovirus using the pCMV expression vector (Clontech, Palo Alto, Calif.).

Protocol for DNA immunization. DNA purification was carried out by standard techniques using Qiagen mega-prep columns with endotoxin removal buffer. DNA was dissolved in 1 N saline, at a concentration of 1 mg/ml, and BALB/c (H2^d) mice were immunized three times, at 14-day intervals; on each occasion 50 µl (50 µg) was injected into each anterior tibial muscle, using a 28-gauge needle. As a positive control for immunity, some mice were immunized intraperitoneally (i.p.) with a sublethal dose of LCMV (2 × 10⁵ PFU).

Using CTL activity as a criterion of successful immunization. CTL activity following LCMV infection of previously nonimmune mice peaks at 7 to 9 days postinfection and declines thereafter. At days 4 and 5 postinfection, CTL activity is difficult to detect but is readily detectable at this time point in an LCMV-immune animal, in which the presence of memory cells allows an accelerated response to viral challenge. We exploited this phenomenon to determine whether immunization successfully induced lytic CD8⁺ T-cell responses. Six weeks after immunization (in this study, usually with a DNA vaccine), mice received LCMV (i.p.) and 4, 5, or 7 days later were sacrificed. Their spleens were taken and analyzed by in vitro cytotoxicity assay (below). Detectable lytic activity at early times postinfection (p.i.) indicates that the prior vaccination had successfully induced memory cells.

Measurement of CD8⁺ T-cell responses using intracellular staining for IFN-. Virus-specific and epitope-specific CD8⁺ T-cell responses can also be quantitated using intracellular cytokine staining (ICCS), an assay which does not depend on lytic activity. BALB/c mice were immunized with pCMV-U (negative control), pCMV-UMGX (subdominant epitope), pCMV-UMG34 (dominant epitope), or LCMV (positive control for immunization). Six weeks later, mice were challenged with a sublethal dose of LCMV i.p., and at 4, 5, or 7 days p.i. mice were sacrificed and 10⁶ splenocytes were plated in 96-well plates together with stimulator cells comprising 2 × 10⁵ BALB c17 cells/well, transiently transfected either with pCMV-U, pCMV-UMGX, or pCMV-UMG34. After a 6-h incubation in the presence of interleukin 2 (150 U/ml), (-mercaptoethanol, and brefeldin A (1 µg/ml); to increase accumulation of gamma interferon [IFN-] in responding cells), wells were washed and labeled with a cychrome-conjugated anti-CD8 antibody (0.25 µg/ml) for 30 min on ice. After washing, cells were permeabilized with Cytofix/Cytoperm for 20 min on ice and stained with a fluoresein-conjugated anti-IFN- antibody (0.4 µg/ml). Finally the cells were washed, fixed, and analyzed by fluorimetry. Reagents were purchased from Pharmingen (San Diego, Calif.). In some experiments the spleen cells were stimulated with BALB c17-infected targets or with peptide-loaded BALB c17, T2-K^d, or T2-L^d cells.

In vitro cytotoxicity assays. Effector cells were either (i) day 7 primary splenocytes from LCMV-infected mice (positive control for the assay) or (ii) day 4, 5, and 7 splenocytes from a previously vaccinated animal (to determine the success of vaccination [see above]). Target cells (usually BALB cl7) were (i) transfected with the different plasmids, (ii) coated with specific peptide, or (iii) infected with LCMV and then were labeled with ⁵¹Cr, washed, and incubated for 5 h with effector cells at the indicated effector-to-target ratios. Supernatant was harvested, and specific chromium release was calculated by using the following formula: [(sample release  spontaneous release) × 100]/(total release  spontaneous release).

Using peptide-coated cells to induce immunity. Spleens were taken from naïve mice and disrupted to form a single-cell suspension, and the cells were washed three times in medium lacking fetal bovine serum. Peptide was added to a concentration of 5 µg/ml, and cells were incubated for 2 h at room temperature. Cells were washed and resuspended in PBS (10⁸ cells/ml), and 100 µl was injected subcutaneously.

In vivo protection studies. (i) Protection reflected by resistance to a normally lethal LCMV challenge. Mice were inoculated with DNA or with LCMV as previously described, and 6 weeks later the animals were challenged with a normally lethal dose (20 50% lethal doses [LD₅₀]) of LCMV administered intracranially (i.c.) Mice were observed daily, and all recorded deaths occurred between days 6 and 12 following LCMV challenge.

(ii) Protection reflected by reduction in LCMV titers following nonlethal viral challenge. Mice were immunized with LCMV (i.p. as a positive vaccine control) or with DNA and 6 weeks later were challenged with LCMV by the i.p. route (2 × 10⁵ PFU). Four days later mice were sacrificed, and their spleens were harvested. A portion of each spleen was analyzed for anti-LCMV CTL activity and by ICCS (see above), and the remainder was used in virus titration; a low virus titer (in comparison to nonimmune control mice) indicates that the mouse has been successfully vaccinated.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Deletion of dominant CTL epitope from LCMV NP has minimal effect on vaccine efficacy. Antiviral CD8⁺ T cells protect BALB/c (H-2^d) mice against LCMV, and the CTL response appears almost monospecific, being directed towards the NP sequence RPQASGVYM; 94% of CTL clones (47 of 50) were specific for this dominant epitope (42). To determine if LCMV NP contained subdominant CD8⁺ T cell epitopes, we prepared plasmid pCMV-NP, which encodes full-length LCMV NP but with the sequence containing the dominant epitope (ERPQASGVYMGNLT) replaced by the sequence AGTA by using PCR mutagenesis. BALB/c mice (eight mice per vaccine group) were immunized with pCMV-NP, pCMV-NP, or pCMV (as a negative control), and 6 weeks later half of the mice (four per vaccine group) were infected with LCMV peripherally (2 × 10⁵ PFU i.p.). Four days later these mice were sacrificed, and their spleens were weighed, homogenized, and titrated for LCMV. The results are presented in Fig. 1. Mice immunized with pCMV-NP showed a 2- to 3-log reduction in virus titers compared to mice immunized with pCMV alone, in concordance with previously reported data (44-46). However, similar levels of protection were conferred by pCMV-NP, indicating the presence of other protective sequences. As a second criterion of protection, the remaining mice (four per group) were challenged i.c. with a normally lethal dose of LCMV (20 LD₅₀ i.c.). As shown in Fig. 1 (vertical grey bars and right axis), all pCMV-immunized animals succumbed, while 75% of the pCMV-NP-immunized mice survived. In this experiment, all mice immunized with pCMV-NP also survived the lethal challenge. Thus, by these two criteria, a plasmid lacking an epitope known to be overwhelmingly dominant in BALB/c mice is capable of conferring a level of antiviral protection equivalent to that conferred by the dominant epitope.

View larger version (47K): [in this window] [in a new window]   FIG. 1.   Deletion of the dominant CTL epitope has minimal effect on vaccine efficacy. BALB/c mice (eight per group) were immunized as indicated with LCMV (positive control), or with DNA vaccines pCMV (negative control), pCMV-NP, or pCMV-NP (NPd). Six weeks later mice were challenged in one of two ways. Four mice per vaccine group were given a nonlethal dose of LCMV (2 × 105 PFU i.p.), and 4 days later these mice were sacrificed and titers of LCMV in the spleens were determined. Results are shown for individual mice (open circles) as PFU per gram of spleen (left axis). The remaining four mice in each vaccine group were challenged i.c. with a normally lethal dose of LCMV (20 LD50). All deaths occurred between days 6 and 8 after the infection, and the percentage of surviving mice is shown (right axis) as vertical grey bars.

Localization of protective sequences using plasmids encoding fragments from LCMV NP. The NP ORF was cleaved with restriction enzymes, and the resulting fragments were cloned in frame with ubiquitin using the plasmids pCMV-UbF1/2 or pCMV-UbF3 (see Materials and Methods). Ubiquitin was used because we have previously shown that ubiquitination enhances sensitization of transfected target cells and improves CD8⁺ T-cell induction and in vivo protection against LCMV infection (28, 29) and against tumor cell challenge (43). The five subfragments used are diagrammed in Fig. 2. Groups of BALB/c mice (12 mice per vaccine group) were immunized with plasmids containing one of the five fragments or with pCMV-U, and 6 weeks later protective immunity was determined using the two modes of virus challenge. Four mice per group were challenged with LCMV peripherally (2 × 10⁵ PFU i.p.), and 4 days later the splenic viral load was determined. The remaining mice (eight per group) received i.c. LCMV (20 LD₅₀ i.c.). As shown in Fig. 3A (left panel) immunization with four of the five fragments resulted in reduced virus titers compared to those in mice receiving the negative-control pCMV-U, and fragment III reduced virus load by almost 2 logs. Mice previously immunized with LCMV had no detectable virus 4 days after i.p. challenge. When survival after i.c. challenge was evaluated (Fig. 3A, right panel), all mice immunized with pCMV-UbI, pCMV-UbII, pCMV-UbIV, and pCMV-UbV died between day 6 and 9 after challenge. In contrast, 75% (six of eight) of mice immunized with pCMV-UbIII survived. We conclude that NP_266-412 contains a strong protective sequence(s), which we considered most likely to be a subdominant CTL epitope(s) and which we named "X."

View larger version (10K): [in this window] [in a new window]   FIG. 2.   Fragments used to identify protective sequences in LCMV NP. The gene encoding the 558-residue NP is shown, with the dominant epitope (NP118-126) as a solid black box; this epitope is not present in NP, from which the five fragments were generated, and thus is absent from fragment I. The location of a sequence which binds to Kd (NP314-322) (see text) is shown as a grey box in NP and in fragment III.

View larger version (46K): [in this window] [in a new window]   FIG. 3.   Sequence X protects against two modes of virus challenge. (A) BALB/c mice (12 mice per group) were immunized with plasmids encoding the ubiquitinated NP fratments I to V, with pCMV-U (negative control) or with LCMV (L) (positive control). Six weeks later, four mice per group were challenged with a nonlethal dose of LCMV (2 × 105 PFU i.p.) and were sacrificed 4 days thereafter. LCMV titers in the spleens were determined, and the results for individual mice (closed circles) and the mean of each group (open triangles) are shown in the lefthand panel. Virus was undetectable in LCMV-immunized mice. The remaining eight mice in each group were challenged with a normally lethal dose of LCMV (20 LD50 i.c.). All deaths occurred between days 6 and 8 after the infection, and the percentages of mice which survived are shown in the right-hand panel. (B) A similar experiment was carried out, this time using BALB/c mice (eight mice per group) immunized with the minigene plasmids pCMV-UMGX or pCMV-UMG34 or (as a negative control) with pCMV-U. Six weeks later, four mice per group were challenged i.p., and the virus titers 4 days later are shown in the left-hand panel. The remaining mice (four per group) were challenged with a lethal dose of LCMV (20 LS50 i.c.), and the percentages of surviving mice are shown (right-hand panel).

More precise identification of the NP sequences which confer antiviral protection. Computer analyses by another laboratory had identified a nine-amino-acid sequence (PYIACRTSI) at NP_314-322 which represented a binding motif for the K^d allele (38). Although the equivalent synthetic peptide bound strongly to K^d, they found no virus-induced CTL activity against this sequence. However, since this sequence mapped in the protective fragment III, we used DNA immunization to reevaluate its biological role and to determine if this peptide corresponded to the protective region X. We have previously shown that the immunogenicity of minigene epitopes is enhanced by fusing the sequence to the cellular protein ubiquitin, which targets it to the proteasome (28). Therefore, we constructed a plasmid in which a short ORF, encoding the K^d-binding sequence preceded by an initiating methionine (MPYIACRTSI), was fused in frame with the C terminus of the ubiquitin gene; this plasmid was named pCMV-UMGX. Mice were immunized with plasmids pCMV-U (negative control), pCMV-UMG34 (positive control, encoding the dominant epitope RPQASGVYM), or pCMV-UMGX, and the biological efficacy of the resulting immunity was evaluated using the two challenge models described above. For both i.p. and i.c. challenges, the levels of protection afforded by pCMV-UMGX were similar to those conferred by the dominant epitope (Fig. 3B). Therefore, this sequence clearly is capable of inducing strong protective immune responses.

Region X induces a subdominant CD8⁺ T cell response, which expands more slowly than the response to the dominant epitope. Having demonstrated the protective efficacy of a minigene vaccine encoding the dominant epitope (41) and region X (Fig. 3), we wished to determine whether the latter sequence encoded a subdominant T-cell epitope and to compare the kinetics of the CD8⁺ T-cell expansions against dominant and subdominant regions. BALB/c mice were immunized with pCMV-NP (which contains both regions), with pCMV-NP (which lacks the dominant epitope), or with pCMV alone. Six weeks later, mice were infected with 2 × 10⁵ PFU of LCMV, and 4 to 7 days later they were sacrificed and antigen-specific CD8⁺ T-cell responses were measured by ICCS. The results are shown in Fig. 4. LCMV-specific CD8⁺ T-cell responses were not detected at day 4 p.i. in mice immunized with pCMV. As expected, a strong response against the dominant epitope is seen in these previously nonimmune mice by day 7 p.i. (~20% of CD8⁺ T cells); the response to stimulator cells transfected with pCMV-UMGX is much weaker (~2%) but represents the first direct evidence that acute LCMV infection induces a response to this region. Strong responses to region X were found in mice which had been immunized with pCMV-NP. By 4 days post-LCMV infection, ~7% of CD8⁺ cells produced IFN- when incubated with pCMV-UMGX-transfected cells, and this response peaked at day 5, with ~30% of CD8⁺ cells positive for IFN-. In contrast, splenocytes from mice immunized with pCMV-NP responded strongly to cells transfected with pCMV-UMG4 (the dominant epitope) but responded only minimally to cells transfected with pCMV-UMGX. Thus, the presence of the dominant epitope in the pCMV-NP vaccine almost completely suppresses the ability of the vaccinee to mount a response to region X, confirming the subdominant nature of this region. In addition, following LCMV infection, the expansion of CD8⁺ T cells specific for region X in pCMV-NP-immunized mice is slower than the expansion of the dominant CD8⁺ T-cell population in pCMV-NP-immunized mice; the subdominant response increases markedly between days 4 and 5 p.i., while the dominant response has peaked by day 4 p.i. Importantly, within 7 days of LCMV challenge, mice immunized with pCMV-NP had developed responses to both the dominant and subdominant epitopes (Fig. 4), and these responses were retained in the memory phase, after virus infection was cleared (data not shown). In contrast, pCMV-NP vaccinees mounted responses only to the dominant epitope following LCMV infection, and subdominant responses were barely detectable in the memory phase. The possible advantages of a "subdominant" vaccine such as pCMV-NP in protecting a host against wild-type and variant viruses are discussed below.

View larger version (24K): [in this window] [in a new window]   FIG. 4.   Following vaccination and challenge, subdominant CD8+ T-cell populations expand more slowly than dominant ones. BALB/c mice were immunized with pCMV, pCMV-NP, or pCMV-NP. Six weeks later mice were challenged with LCMV i.p. and at 4, 5, and 7 days p.i. were sacrificed (4 mice per vaccine and time point). LCMV- specific CD8+ T-cell responses were evaluated by ICCS, using as stimulators BALB cl7 cells transfected with plasmids pCMV-UMGX, pCMV-UMG4, or pCMV. The percentages of CD8+ T cells producing IFN- are shown with standard errors (error bars).

CD8⁺ T cells are stimulated better by transfected cells than by peptide X-coated cells. Others have suggested that peptide X (PYIACRTSI) is a cryptic epitope which is not presented by LCMV; when van der Most et al. used virus-amplified splenocytes as effector cells, no cytolytic responses were detected against peptide-coated cells, and peptide-coated cells could be used to induce low-level CD8⁺ T-cell responses capable of lysing peptide-coated targets but not virus-infected targets (38). In contrast, our findings provide two lines of evidence indicating that region X must be presented by LCMV in vivo. First, the effector cells induced by pCMV-UMGX can protect against viral challenge (Fig. 3) and therefore must recognize virus-infected cells in vivo; second, antigen-specific cells induced by pCMV-NP immunization can be expanded by exposure to LCMV in vivo (Fig. 4). One key difference between the present study and that previously published, was the latter's reliance on synthetic peptides to detect and induce responses. Therefore, we compared the stimulatory capacity of cells transfected with pCMV-UMGX to that of cells coated with peptide X. Two groups of effector cells were used: cells obtained 7 days after LCMV infection of naïve mice and cells harvested 5 days after LCMV infection of mice which had been immunized with pCMV-UMGX 6 weeks previously. As shown in Fig. 5, after incubation with transfected cells, ~2% of primary (day 7) CD8⁺. T cells produced IFN-; however, these CD8⁺ T cells are barely detectable after exposure to peptide X-coated cells. This reduced stimulatory capacity of peptide-coated cells also was revealed when they were incubated with the second group of LCMV-specific effector cells; the response to transfected cells was almost twice as strong as the response to peptide-coated stimulators. Thus, peptide X appears to stimulate responses less effectively than cells expressing this region endogenously.

View larger version (30K): [in this window] [in a new window]   FIG. 5.   Peptide X is more efficiently presented by transfected cells than by peptide-coated cells. Two different groups of effector cells were used, as shown on the y axis: (i) splenocytes harvested 7 days after infection of naïve mice and (ii) splenocytes harvested 5 days after infection of mice which had been immunized 6 weeks previously with pCMV-UMGX. Stimulator cells were BALB cl7 cells transfected with pCMV-UMGX or coated with peptide X (PYIACRTSI). The percentages of CD8+ cells producing IFN- are shown, with standard errors (error bars).

Region X contains two overlapping epitopes, each presented by a different MHC class I allele. We were surprised that peptide X-coated cells were less stimulatory than transfected cells, especially because our transfection controls (not shown) indicated that only ~20% of cells expressed transfected sequences, while all cells should have been successfully coated with peptide. We hypothesized that the intracellular synthesis and/or processing in transfected cells might be a decisive factor. Further inspection of the encoded sequences led us to realize that processing of region X from the virus, or from plasmid DNA, could give rise to a decameric peptide (WPYIACRTSI from the LCMV sequence or MPYIACRTSI from pCMV-UMGX) which could be presented by L^d (for which the consensus motif is a proline at position 2, and a hydrophobic C terminus). L^d-binding sequences were not analyzed in the preceding publications evaluating subdominant H-2^d epitopes (36, 38). Therefore, we decided to evaluate presentation of region X by both K^d and L^d, using two peptides, X (PYIACRTSI) and WX (WPYIACRTSI).

View larger version (36K): [in this window] [in a new window]   FIG. 6.   Region X contains two overlapping epitopes: one presented by Kd, and the other presented by Ld. Four mice were immunized with pCMV-UMGX and 6 weeks later were infected with LCMV. Seven days thereafter splenocytes were harvested and stimulated for 5 h on BALB c17 (Kd, Dd, and Ld) T2-Kd (Kd) or T2-Ld (Ld) cells precoated with peptide X (NP314-322) (PYIACRTSI), peptide WX (NP313-322) (WPYIACRTSI) or the dominant peptide D (NP118-126) (RPQASGVYM). Epitope-specific responses were evaluated by ICCS (see Materials and Methods).

Induction of X- and WX-specific responses by LCMV infection and DNA immunization. Our identification of overlapping epitopes in region X led us to reevaluate the subdominant response induced by acute LCMV infection and following DNA immunization. We already knew that, at 7 days p.i., ~2% of CD8⁺ T cells were specific for region X, although they responded very poorly to peptide X (Fig. 5). As shown in Fig. 7, this response is specific for WX; we were unable to detect cells specific for peptide X following LCMV infection, consistent with the previous failure to detect virus-induced CTL activity against peptide PYIACRTSI (38). Thus, viral infection of a naïve host induces a subdominant, but detectable, response to the L^d-restricted epitope, but none to the K^d-restricted epitope. We also evaluated the responses induced by immunization with pCMV-UMGX. At 15 days postimmunization, the peak of a DNA-induced CD8⁺ T-cell response (13), we identified responses to both epitopes (Fig. 7).

View larger version (34K): [in this window] [in a new window]   FIG. 7.   Induction of X- and WX-specific responses by LCMV infection and DNA immunization. Responses to peptides D, WX, and X were evaluated by ICCS 7 days after virus infection of naïve mice (top row) and 15 days after pCMV-U-MGX immunization (bottom row). The location of CD8+ IFN--positive cells is indicated by an ellipse, and for levels above background, the percentages of CD8+ T cells which produce IFN- are shown.

Immunization with peptide WX confers stronger immunity than peptide X. Previous studies had indicated that peptide X immunization (with adjuvant) failed to induce protective immunity (38). Our identification of nested epitopes encouraged us to reevaluate the protective efficacy of peptide immunization, using peptides WX and X. Therefore, spleen cells were coated with peptide WX, peptide X, or peptide D (see Materials and Methods) and were inoculated into mice. Three weeks later, the mice were challenged i.c. with 20 LD₅₀ of LCMV. As shown in Fig. 8, peptide WX and peptide D induced stronger protective responses than did peptide X, consistent with our observation that the immune response to WX is stronger than that mounted to X. 

View larger version (55K): [in this window] [in a new window]   FIG. 8.   Protective immunity induced by immunization with peptide-coated cells. Groups of mice (eight mice per group) were immunized with spleen cells, either uncoated (none), or coated with peptide D, WX, or X (see Materials and Methods). Three weeks later, mice were challenged i.c. with 20 LD50 of LCMV and were observed daily for 21 days. All deaths occurred between days 7 and 90. The percent surviving mice is shown.

CD8⁺ T cells specific for X and WX rapidly acquire the ability to produce cytokines in response to antigen but show a marked delay in the development of lytic activity. Clearance of LCMV is thought to be almost entirely CTL mediated, requiring perforin-dependent killing of infected cells (16, 39). We therefore measured the lytic activity of antigen-specific CD8⁺ T cells from mice immunized with pCMV-UMGX and infected with virus. BALB/c mice were immunized with pCMV-UMGX or pCMV-UMG4 and 6 weeks later were infected with 2 × 10⁵ PFU of LCMV. At days 4, 5, and 7 p.i., mice were sacrificed and antigen-specific CD8⁺ T-cell activities were measured by an in vitro cytotoxicity assay and by ICCS. Data from selected pCMV-UMGX vaccinees are shown in Fig. 9A, and a summary of all data is presented in Fig. 9B. As shown in Fig. 9A, at four days p.i. in mice previously immunized against region X, approximately 7% of CD8⁺ T cells produced IFN- in response to peptide X, but cytolytic activity was low at all effector:target ratios. One might argue that the absence of cytolytic activity merely reflected too low a number of antigen-specific T cells. However, the absence of lysis must instead result from a functional deficiency in X-specific T cells, because peptide D-specific T cells are even less frequent (~4%) in these mice at 5 days p.i., but nevertheless can lyse peptide-coated target cells. Thus, memory T cells specific for peptide X acquire the ability to produce cytokines before they develop lytic activity, while peptide D-specific cells may develop the two functions in parallel. However, X-specific cells do attain lytic function by 7 days p.i. (Fig. 9A). Next, for all of the antigen-specific CD8⁺ T-cell populations, the number of lytic units per 10⁶ antigen-specific cells was calculated as described in Materials and Methods. In mice immunized with pCMV-UMG4 (Fig. 9B), LCMV infection induced high levels of lysis and strong responses by ICCS, consistent with our previous results (28). Nevertheless, despite stable responses from days 4 to 7 as measured by ICCS, the lytic activity of this cell population was reduced approximately twofold at day 5 p.i. and again dropped twofold by day 7 p.i. Thus, the lytic capacity of these CD8⁺ T cells appears to diminish more quickly than their ability to produce cytokines. As described above, splenocytes from mice immunized with pCMV-UMGX (Fig. 9B) showed barely detectable X-specific lytic activity at 4 days p.i., despite readily detectable responses by ICCS (~7% of total CD8⁺ cells). Similarly, WX-specific cells constituted ~12% of the CD8⁺ population but displayed no lytic capacity. By day 5 p.i., the percentage of X- and WX-specific IFN--positive cells had increased, and cells responding to the dominant epitope were detectable by ICCS. Although X-specific cells eventually developed lytic activity, this was at least three- to fourfold lower than that of cells specific for WX or D. In summary, the data in Fig. 9 show that (i) lytic activity is lost earlier than the ability to produce antiviral cytokines and (ii) the peak lytic activity (measured as lytic units per 10⁶ cells) varies markedly among different populations of epitope-specific CD8⁺ T cells, from ~170 for X-specific cells to ~1,000 for cells targeted to the dominant epitope.

View larger version (31K): [in this window] [in a new window]   FIG. 9.   Delayed development of lytic activity by CD8+ T cells specific for the subdominant epitopes WX and X. BALB/c mice were immunized with pCMV-UMGX or pCMV-UMG4 and 6 weeks later were infected with LCMV. Four, five, or seven days later, mice were sacrificed; their spleens were harvested; and splenocytes were used in an ICCS using the indicated peptides as stimulators and also in a classical in vitro cytotoxicity assay using as targets BALB c17 cells coated with the indicated peptides. The number of lytic units per million epitope-specific cells was calculated as described in Materials and Methods. (A) ICCS and cytotoxicity data from selected p-CMV-UMGX vaccinees. ICCS dot plots are shown (x axis, CD8; y axis, IFN-) for CD8+ T cells responsive to peptide X or D, at the indicated times after in vivo secondary stimulation with LCMV; arrows point to their related cytoxocity assays. The day 4 cells (bottom dot plot) could produce IFN- but were very weakly lytic (filled circles); the low level of lysis was not due to a low number of X-specific cells, since fewer cells specific for peptide D (center dot plot) showed higher lytic activity (open squares). X-specific cells eventually acquired lytic activity (top dot plot and filled triangles). (B) Summary of ICCS and cytotoxicity data from all vacinees. For CD8+ T cells specific for peptides X, WX, and D, the lytic activity (in lytic units) is compared to the percentage of CD8+ T cells producing IFN-.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Although the biological functions of subdominant CD8⁺ T-cell epitopes are effectively suppressed by their dominant counterparts, there are compelling reasons for attempting to identify them. First, in several virus infections, for example LCMV and HIV, dominant epitopes undergo mutation, and subsequent immune-driven selection leads to the emergence of viruses lacking the dominant epitopes (25); in many cases, host responses develop against previously subdominant sequences in these viruses (4, 40). Second, when incorporated into vaccines, subdominant epitopes can protect against acute viral challenge (5, 36, 37). Third, subdominant responses can effect tumor immunity (24). Subdominant epitopes have been sought using a variety of methods. One of the more common is motif-based prediction, which has been quite successful but suffers from several disadvantages that have led us to pursue subdominant epitopes using functional, rather than structural, criteria. We had previously shown that the isolated dominant epitope from LCMV NP was a less-effective immunogen than the full-length protein (28), and so, suspecting the existence of other protective sequences in NP, we generated a battery of five overlapping fragments from NP (lacking the dominant epitope). Two fragments, III and IV, conferred good protection against i.p. infection; however, only fragment III conferred significant protection against i.e. challenge (Fig. 3A). These functional analyses led us to concentrate on fragment III, in which others had identified a sequence (NP_314-322) able to bind strongly to K^d but which, when administered as a synthetic peptide, failed to induce protective immunity (38). We show here (Fig. 3B) that this region X, when expressed as a ubiquitinated minigene, conferred protection against both modes of challenge and thus represents a biologically relevant epitope. The protection conferred appears similar to that provided by fragment III, suggesting that this may be the sole protective component in this region. Thus, functional analysis using plasmids which encode ubiquitin proteins appears to be a rapid and reliable means by which to identify subdominant (or dominant) epitopes, and the impressive protection conferred by pCMV-NP (Fig. 1) fragment III, or pCMV-UMGX (Fig. 3A & B, respectively) underscores the great vaccine potential of subdominant epitopes.

The strong protective capacity of pCMV-UMGX constituted unequivocal evidence that the related sequence must be presented by LCMV in vivo, but this contrasted with published data indicating that peptide X-coated cells were not strongly protective, a fact confirmed in our study (Fig. 8). Why does pCMV-UMGX DNA work better than peptide X? Our mapping studies revealed two overlapping epitopes, presented by L^d and K^d (Fig. 6). The minigene is expressed in frame with the C terminus of ubiquitin, and as a result, the UMGX gene product could be processed to generate peptide X (PYIACRTSI) or, alternatively, to generate the putative L^d-binding peptide MPYIACRTSI. The former peptide can induce X-specific CD8⁺ T cells (Fig. 5 and 7), as expected. The latter peptide induces CD8⁺ T cells which recognize the cognate peptide MPYIACRTSI (not shown) and which cross-react with the WX sequence (WPYIACRTSI) (Fig. 7). Cross-reactivity with the native viral sequence also occurs, since vaccine-induced WX-specific CD8⁺ T cells can be amplified by virus infection (Fig. 9B)presumably by exposure to the viral sequence WPYIACRTSI. This proposed cross-reactivity is consistent with the facts that (i) the T-cell receptor contact residues for peptides presented by L^d are located at positions 5 and 6, which are identical in these two sequences; and (ii) the critical L^d MHC contact residues lie at P2 and at the C terminus, so the MW change at position 1 may not alter binding to L^d (27). The ability of pCMV-UMGX to present both of the subdominant epitopes explains why more CD8⁺ T cells are stimulated by transfected cells than by cells coated with peptide X (Fig. 5). It also explains why pCMV-UMGX is a better vaccine than peptide X; pCMV-UMGX induces two populations of CD8⁺ T cells, one specific for each subdominant epitope. Virus infection of naïve mice induces a response to peptide WX, showing clearly that this sequence is presented in vivo, but little or no response to peptide X (PYIACRTSI) (Fig. 7). Immunization with pCMV-U-MGX induced responses to both WX and X, suggesting that concurrent presentation of the subdominant epitopes may be better achieved from plasmid DNA encoding ubiquitinated sequences than during virus infection. The identification of coterminal overlapping epitopes presented by different MHC class I alleles is unusual, having been described, to our knowledge, only once before (34).

We compared the expansion of antiviral CD8⁺ T-cell responses in animals which had been immunized to induce responses mainly against the dominant epitope (pCMV-NP) or mainly against the subdominant epitopes (pCMV-NP). Four related conclusions emerge from the data shown in Fig. 4. First, after vaccination to induce either a dominant or a subdominant response, subsequent expansion of primed cells is faster for cells responding to the dominant epitope than for cells specific for the subdominant sequences. However, the peak percentage of CD8⁺ T cells responding was similar for both the dominant and subdominant epitopes in the relevant vaccinees. Second, these data demonstrate the dramatic effects of dominance and subdominance after vaccination and virus challenge. Mice immunized with pCMV-NP and infected with virus mount strong responses to the dominant epitope but extremely poor responses to the subdominant sequences. In contrast, pCMV-NP vaccinees respond to both the dominant and subdominant sequences. These two DNA vaccines differ only in the presence or absence of the dominant sequence; thus; the presence of a dominant epitope in a vaccine suppresses the development of a subdominant response even after viral challenge. Third, a subdominant vaccine may be better than a standard vaccine, because the vaccinee may be protected against infection by a variant virus lacking the dominant epitope. Fourth, the data have implications for the long-term effects of vaccination against an organism which circulates in a host population. Mice immunized with pCMV-NP and then exposed to the virus will clear the infection, but their subsequent CD8⁺ T-cell memory pool is directed almost entirely against the dominant epitope. In contrast, mice immunized with pCMV-NP also recover from virus challenge, but they retain memory cells against both the dominant and subdominant epitopes. Thus, recipients of a subdominant vaccine who are then exposed to the wild-type virus will develop responses to both the dominant and subdominant epitopes; this broader CD8⁺ T-cell response may delay or diminish the emergence of CTL escape mutants.

We anticipated that the protective T cells in pCMV-UMGX vaccinees would be classical CTL and were surprised that at day 4 postinfection, no lytic activity was detectable against either X or WX (Fig. 9). The absence of lysis at this time point does not indicate a failure of the CD8⁺ cells to recognize the target cells, since soluble peptide WX or X could stimulate the effector cells, rendering them positive in ICCS assays. One might argue that the number of IFN--secreting cells (approximately 10% of all CD8⁺ T cells) is too small to allow detection in an in vitro cytotoxicity assay; however, this explanation is unlikely to be true because, at 5 days postinfection, specific lysis against the dominant epitope is detectable despite the fact that, by ICCS, only ~4% of CD8⁺ cells are specific for this sequence. In Fig. 9 the lytic capacities are presented as lytic units per 10⁶ epitope-specific CD8⁺ T cells; the data are shown in this manner to allow direct comparison of the lytic activities of all epitope-specific T-cell populations. Thus, the CD8⁺ T cells specific for X and WX have limited lytic capacity at early time points p.i. The lytic activity of the WX-specific cells does eventually increase to intermediate levels, but the X-specific lytic activity remains low throughout infection.

It appears likely that the outcome of viral challenge in vaccinated animals is decided very earlyperhaps hoursafter infection (1), and other studies have demonstrated that perforin-mediated CTL lysis is absolutely required for clearance of LCMV (16, 39). However, we show here that mice which had received a vaccine encoding only subdominant epitopes (pCMV-UMGX) were solidly protected against viral challenge (Fig. 3), despite having no detectable lytic activity 4 days after viral challenge (Fig. 9). How can these apparently disparate observationsthe importance of the very early response, the absence of lytic cells at early times, and the requirement of perforin for clearancebe reconciled? It has been previously demonstrated that CD8⁺ T cells can limit virus infection in a nonlytic manner, often via release of cytokines (e.g., IFN-, tumor necrosis factor alpha, etc.) (11, 12, 26) or other trans-acting molecules, such as those identified in the HIV system (3, 21, 22). The antiviral effects of the nonlytic cells identified here are most likely cytokine mediated; these cells, which can limit virus replication measured at 4 days postchallenge (Fig. 3), can make epitope-specific cytokine responses which are easily detected at this time point (Fig. 9). Nevertheless, the ultimate clearance of LCMV would be difficult to explain in the absence of a lytic CTL response; we show here that, after LCMV infection of mice immunized with pCMV-UMGX, a lytic response develops, although it is significantly delayed in comparison to the lytic activity seen in pCMV-UMG4 vaccinees (Fig. 9). Thus, cytolytic responses, rather than the pCMV-UMGX-induced nonlytic responses, most probably eradicate the virus. The above observations have important implications for vaccine design. CD8⁺ T-cell responses are invariably accompanied by some degree of immunopathology, and in some circumstances, this pathology is mainly perforin mediated; however, it is possible to uncouple the antiviral efficacy of the T-cell response from its immunopathological consequences (9). Therefore, in certain instances, it may be beneficial to administer a vaccine which induces CD8⁺ T cells that lack lytic activity but are able to produce cytokines in response to viral challenge.

Our data also demonstrate that the two major effector functions of CD8⁺ T cells, target cell lysis and cytokine production, are acquired and decline at different rates over the course of the antiviral immune response. The ability of CD8⁺ T cells to produce cytokines in response to antigen contact precedes their ability to lyse target cells, and lytic activity declines more rapidly than does the ability to synthesize cytokines in response to antigen (Fig. 9). This rapid decline in lytic activity, with retention of competence for cytokine production, is consistent with published data concerning CD8⁺ memory T cells. At >30 days postinfection, CD8⁺ memory T cells can initiate cytokine synthesis very quickly (20, 31, 33) but take longer to acquire lytic ability (2, 30). We suggest that the lysis^low-cytokine^competent functional phenotype of memory cells is acquired earlier than previously thought, as soon as antigen load is diminished.

In conclusion we show that DNA immunization, in concert with functional criteria, can identify epitopes which were not found using structural predictive schemes in combination with synthetic peptides. Furthermore, DNA vaccination with ubiquitinated minigenes induced responses to both subdominant epitopes, while virus infection failed to do so; the beneficial effects of ubiquitination have been demonstrated previously (28). The remarkable protective potential of subdominant epitopes mandates their further evaluation as vaccines against microbial disease and cancer.