ABSTRACT
The phenomenon whereby the host immune system responds to only a few of the many possible epitopes in a foreign protein is termed immunodominance. Immunodominance occurs not only during microbial infection but also following vaccination, and clarification of the underlying mechanism may permit the rational design of vaccines which can circumvent immunodominance, thereby inducing responses to all epitopes, dominant and subdominant. Here, we show that immunodominance affects DNA vaccines and that the effects can be avoided by the simple expedient of epitope separation. DNA vaccines encoding isolated dominant and subdominant epitopes induce equivalent responses, confirming a previous demonstration that coexpression of dominant and subdominant epitopes on the same antigen-presenting cell (APC) is central to immunodominance. We conclude that multiepitope DNA vaccines should comprise a cocktail of plasmids, each with its own epitope, to allow maximal epitope dispersal among APCs. In addition, we demonstrate that subdominant responses are actively suppressed by dominant CD8+ T-cell responses and that gamma interferon (IFN-γ) is required for this suppression. Furthermore, priming of CD8+ T cells to a single dominant epitope results in strong suppression of responses to other normally dominant epitopes in immunocompetent mice, in effect rendering these epitopes subdominant; however, responses to these epitopes are increased 6- to 20-fold in mice lacking IFN-γ. We suggest that, in agreement with our previous observations, IFN-γ secretion by CD8+ T cells is highly localized, and we propose that its immunosuppressive effect is focused on the APC with which the dominant CD8+ T cell is in contact.
Microbial genomes usually encode many potentially immunogenic sequences which can be presented to the host immune system, but the host immune response usually is limited to only a few epitopes, which are termed “dominant.” This constraint on host immunity, where the response to one microbial sequence predominates while responses to other, “subdominant” sequences are minimal, is termed immunodominance. Immunodominance affects both T-cell and antibody responses; only the CD8+ T-cell response is studied here. The number of epitopes which induce dominant CD8+ T-cell responses varies somewhat depending on the microbe and the host background but does not appear to be determined by the size of the microbial genome. For example, the genome of lymphocytic choriomeningitis virus (LCMV) is about 20-fold smaller than that of cytomegalovirus and about 200-fold smaller than that of the bacterium Listeria monocytogenes, but in all cases, infection results in strong CD8+ T-cell responses to only ∼1 to 3 epitopes (8, 9, 17, 25, 30, 39). Immunodominance is frequently defined as a hierarchy of responses to different epitopes. However, a key feature of immunodominance--which is often overlooked—is that removal of a dominant sequence permits the host to respond more strongly to the previously subdominant epitopes (1, 23), and any mechanistic explanation of immunodominance must address how the presence of a dominant epitope limits the responses induced by subdominant epitopes. The effects of immunodominance occur not only during microbial infections but also during vaccination, severely restricting the immune responses induced by conventional vaccines. A vaccine which could induce responses to all epitopes, dominant and subdominant, might be more efficacious than conventional vaccines; indeed, subdominant CD8+ T-cell epitopes can induce protective immunity in several virus and tumor models (13, 24, 28, 32-34). Therefore, a full understanding of the mechanisms which underlie immunodominance is not merely of academic interest and may permit the design of vaccines which induce much broader immune responses.
Using the LCMV model system, we (40) and others (30) have shown that the CD8+ T-cell response on the BALB/c background is almost monospecific, with >95% of the responding CD8+ T cells focused on the dominant epitope NP118-126 (RPQASGVYM) on the viral nucleoprotein (NP). Minor responses to two subdominant epitopes on the viral glycoprotein (GP) have been detected (34), and by using plasmid DNA immunization, we have recently identified an additional subdominant epitope on NP which, when delivered as a minigene vaccine, can induce protective immunity as effective as that induced by the dominant epitope (28). In addition, dominant and subdominant LCMV epitopes have been identified on the H-2b background (30, 33, 36). Previous analyses of immunodominance have been based on viral or bacterial infections and have provided much enlightening data; however, interpretation is often complicated by the reliance on live, replicating agents, whose tropism and titer (and thus, antigen load) can vary, often in response to the very immune responses being measured. DNA vaccines recapitulate many of the aspects of virus infection (for example, endogenous synthesis of antigens), but they are simpler and are incapable of replication. Furthermore, we have recently shown that CD8+ T-cell responses to DNA vaccines can be detected directly ex vivo (i.e., in the absence of any exposure to virus) and peak at ∼15 days postimmunization (3, 18). Thus, in this study, we have combined virus infection and DNA vaccination to dissect immunodominance. We show that immunodominance affects DNA vaccines, and we describe a vaccine which circumvents immunodominance, inducing concurrent responses to dominant and subdominant CD8+ T-cell epitopes. Furthermore, we show that prior immunization targeted to one dominant epitope can dramatically alter the dominance hierarchy upon subsequent viral infection; vaccine-primed dominant CD8+ T cells actively suppress the development of CD8+ T-cell responses to all other epitopes—even normally dominant epitopes—upon virus infection. We show that immunodominance is much less profound in the absence of gamma interferon (IFN-γ), allowing us to propose a general mechanism for the phenomenon; we suggest that IFN-γ secreted by the dominant CD8+ T-cell population suppresses the development of CD8+ T-cell responses to other epitopes.
MATERIALS AND METHODS
Cell lines and viruses.The BALB cl7 (H-2d) and MC57 (H-2b) fibroblast cell lines were maintained in RPMI medium (Sigma) supplemented with 10% fetal calf serum, l-glutamine, and penicillin-streptomycin. The virus used was LCMV (Armstrong strain).
Mice.Mice used were BALB/c (H-2d), C57BL/6 (H-2b), and transgenic knockout mice on both major histocompatibility complex (MHC) backgrounds which lacked IFN-γ (GKO mice). All mice were obtained from the breeding colony at The Scripps Research Institute and were used at the age of 6 to 16 weeks.
Recombinant plasmids.All the products are expressed under the control of the immediate-early promoter of human cytomegalovirus by using the pCMV expression vector (Clontech, Palo Alto, Calif.). Plasmid pCMV-NP encodes the full-length NP gene from LCMV (43). Plasmid pCMV-NPΔ is similar but has the dominant H-2d epitope (NP118-126, RPQASGVYM) deleted. pCMV-UMG4 contains a minigene encoding the dominant epitope (sequence given above), and pCMV-UMGX contains a minigene expressing region X (sequence, PYIACRTSI), recently identified as a subdominant epitope (28). Finally, pCMV-U-MG4X and pCMV-UMGX4 contain minigenes encoding both the dominant and subdominant sequences in the order indicated. All four minigene plasmids contain the mouse ubiquitin gene with (i) a “Kozak” initiator sequence (21), (ii) the last codon (ubiquitin residue 76) 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. Plasmid pCMV-U (encoding ubiquitin alone; a negative control in our experiments) has been described previously (27).
DNA immunization and LCMV infection.DNA purification was carried out by standard techniques using Qiagen (Valencia, Calif.) Megaprep columns. DNA was dissolved in normal saline (0.9% [wt/vol] NaCl), at a concentration of 1 mg/ml, and mice were immunized by injection of 50 μl (50 μg of DNA) into each anterior tibial muscle by use of a 28-gauge needle. In mice which were coimmunized with two different plasmids, the plasmids were inoculated separately, into different hind limbs. Mice infected with LCMV received 2 × 105 PFU intraperitoneally.
Measurement of antigen-specific CD8+ T-cell responses by using intracellular cytokine staining (ICCS) for IFN-γ and TNF-α.At the indicated times postimmunization or postinfection, mice were sacrificed, splenocytes were prepared, and 106 splenocytes were plated in 96-well plates together with the indicated peptides representing the following epitopes: for the H-2d background, NP118-126 or NP313-322 (dominant and subdominant epitopes, respectively); for the H-2b background, NP396-404, GP33-41, or GP276-285 (dominant epitopes). After a 6-h incubation in the presence of interleukin-2 (150 U/ml), 50 μM β-mercaptoethanol, and brefeldin A (1 μg/ml, to increase accumulation of IFN-γ or tumor necrosis factor alpha [TNF-α] in responding cells), cells were washed and then labeled with a cytochrome-conjugated anti-CD8 antibody (0.25 μg/ml) for 30 min on ice. After a wash, cells were permeabilized with Cytofix/Cytoperm for 20 min on ice and then stained with a fluorescein-conjugated anti-IFN-γ (0.4 μg/ml) or anti-TNF-α (0.8 μg/ml) antibody. Finally, the cells were washed, fixed, acquired on a FACScan flow cytometer, and analyzed by using CellQuest software.
RESULTS
Immunodominance during LCMV infection and DNA immunization.Mice were infected with LCMV, and 7 or 50 days later, spleens were harvested and splenocytes were analyzed by ICCS using peptides as stimulators. The peptides used were peptide D (representing the dominant epitope, RPQASGVYM) and peptide WX (WPYIACRTSI), a recently identified subdominant epitope (28); both epitopes are presented by Ld. As shown in Fig. 1A, during acute infection, the dominant response comprised ∼38% of CD8+ T cells, and a further 2.1% of CD8+ T cells responded to the subdominant sequence. In the memory T-cell population, the proportions of dominant- and subdominant-epitope-specific cells were 16 and 0.5%, respectively. To determine whether the CD8+ T-cell responses induced by DNA vaccines showed similar patterns of immunodominance, mice were immunized with pCMV-NP, a plasmid encoding the full-length LCMV NP, containing both the dominant and subdominant epitopes described above. As shown in Fig. 1B, 15 days later a strong response could be detected against the dominant epitope, but none could be detected against the subdominant epitope, suggesting that immunodominance also affects DNA vaccines.
Immunodominance during LCMV infection and DNA immunization. BALB/c mice were immunized as shown; at the indicated times they were sacrificed, and ICCS assays were carried out using as stimulator cells BALB cl7 cells coated either with the dominant-epitope peptide NP118-126 (peptide D) or with the subdominant-epitope peptide NP313-322 (peptide WX). Uncoated cells were included as controls. Percentages of CD8+ cells positive for IFN-γ are shown (data from representative mice). (A) Mice were infected with LCMV intraperitoneally and were sacrificed 7 or 50 days later. (B) Mice were inoculated with plasmid DNA and were sacrificed 15 days later.
The effects of immunodominance on a DNA vaccine can be circumvented by epitope separation.CD8+ T-cell epitopes can be recognized when encoded by short open reading frames, which we termed “minigenes” (26, 37), and can be expressed in concert when linked together in a “string-of-beads” construct (38). Although it was first demonstrated using recombinant vaccinia viruses, the minigene approach can be applied to DNA immunization (3, 14, 27, 44). Furthermore, we have recently shown that the CD8+ T-cell responses induced by minigene DNA vaccines can be detected directly ex vivo by using ICCS (3). Therefore, to allow us to further investigate the induction of CD8+ T-cell responses specific for the dominant or the subdominant epitope, we generated minigene plasmids encoding either the dominant epitope alone (pCMV-UMG4) or the subdominant epitope alone (pCMV-UMGX). BALB/c mice were immunized with these plasmids, individually or in combination; 15 days later, they were sacrificed and their splenocytes were analyzed directly ex vivo by using ICCS. Remarkably, very similar responses were induced by the dominant (UMG4) and subdominant (UMGX) vaccines; in both cases, slightly more than 1% of CD8+ T cells were epitope specific (Fig. 2). Therefore, the lower response to the subdominant epitope seen in infected mice and in pCMV-NP-immunized mice may not result from a lower frequency of naïve T cells specific for this sequence. Furthermore, as shown in Fig. 2 (bottom panels), the similarity between the dominant and subdominant responses was maintained even when both plasmids were administered in combination to individual mice. Thus, immunodominance can be largely circumvented by the simple expedient of physically separating the sequences, even if both vaccines are administered simultaneously.
Immunodominance can be overcome by epitope separation. Mice were immunized with the indicated plasmid DNAs, individually or in combination. Fifteen days later, mice were sacrificed, and ICCS assays were carried out by using as stimulator cells BALB cl7 cells coated either with peptide NP118-126 (peptide D) or with peptide GP313-322 (peptide WX). Uncoated cells were included as controls. Percentages of CD8+ cells positive for IFN-γ are shown (data from representative mice).
The efficiency of antigen presentation is one factor determining immunodominance.It was possible that the presence of the dominant epitope prevented the processing or presentation of the subdominant epitope and that epitope separation circumvented this problem. To determine the relative efficiencies of presentation of the dominant and subdominant epitopes, and to evaluate any effect of one epitope on the other, we transfected cells with a variety of plasmids and assessed their abilities to present the epitopes by using them as stimulator cells in an ICCS assay. The indicator cells in the assay were populations of epitope-specific CD8+ T cells. To generate the epitope-specific CD8+ indicator cells, mice were immunized with pCMV-UMG4 or with pCMV-UMGX, both of which induce epitope-specific responses (see Fig. 2), and 6 weeks later were infected with LCMV to expand the number of responding cells. Five days later, cells were harvested and incubated with stimulator cells which had been transfected, LCMV infected, or peptide coated (Fig. 3). CD8+ T cells from mice that had received the pCMV-UMG4 vaccine showed strong responses to LCMV-infected cells and to cells transfected with pCMV-NP, indicating that, as expected, the dominant epitope is presented by cells expressing full-length NP; they also responded to cells coated with peptide D or transfected with pCMV-UMG4 or pCMV-UMG4X (a plasmid in which both epitopes are arranged in tandem in a single open reading frame). In contrast, cells from pCMV-UMGX vaccinees did not respond strongly to LCMV-infected cells or to cells transfected with pCMV-NP or pCMV-NPΔ (both of which constructs contain the subdominant epitope). The weakness of this response cannot be attributed to a low frequency of subdominant epitope-specific CD8+ T cells in this indicator cell population, since many CD8+ T cells produced IFN-γ after incubation with stimulator cells which had been coated with peptide WX or transfected with pCMV-UMGX or pCMV-UMG4X. The fact that the subdominant epitope is less efficiently presented from a full-length protein than from pCMV-UMGX may reflect an effect of the sequences flanking the epitope, which differ between the full-length and minigene constructs; effects of flanking residues on epitope processing have been reported previously (6, 16). Whatever the reason, the subdominant epitope is indeed presented less efficiently than the dominant epitope from full-length NP, which may in part explain the dominance hierarchy. However, the subdominant sequence must be presented from full-length NP to a significant extent, because (i) subdominant epitope-specific CD8+ T cells are readily detectable after virus infection (Fig. 1A), (ii) the responses to full-length NP by pCMV-UMGX vaccinees, although low, are significantly above background (Fig. 3), and (iii) immunization with pCMV-UMGX confers solid protection against virus challenge, confirming that virus-infected cells must present the epitope in vivo (28). Although a difference in antigen processing or presentation provides an attractive explanation of why dominant responses are stronger than subdominant responses, it is unlikely to be the general explanation for immunodominance, for two reasons. First, the cell surface abundance of an epitope does not directly correlate with its immunogenicity (10, 15, 35). Second, the presence of an immunodominant epitope on one viral protein can suppress the immune responses induced by epitopes on other viral proteins, and it is difficult to see why removal of a dominant epitope from one protein would increase the processing of an epitope on a different protein.
The subdominant epitope is not efficiently presented from full-length NP. BALB/c mice were immunized with pCMV-UMG4 or pCMV-UMGX and 6 weeks later were infected with LCMV. Five days postinfection, mice were sacrificed, and splenocytes were harvested and used in an ICCS assay. Stimulators used were BALB cl7 cells either infected with LCMV, coated with the dominant (D) or the subdominant (WX) peptide, or transfected with either pCMV-NP (full-length protein), pCMV-NPD (full-length NP from which the dominant epitope has been deleted), pCMV-UMG4 (dominant minigene), pCMV-UMGX (subdominant minigene), pCMV-UMG4X (both minigenes), or pCMV-U (negative control). Four separate experiments were carried out, and the means ± standard errors of the means are shown.
Competition for MHC binding does not adequately explain immunodominance.Since both the dominant and subdominant epitopes studied above are presented by Ld, the dominant epitope might competitively inhibit binding of the subdominant epitope to the MHC molecule, explaining why separating the epitopes led to an increased subdominant response. To test this hypothesis, we made a plasmid (pCMV-NPΔ) which encoded full-length NP from which the dominant epitope had been deleted, and we determined whether or not cells transfected with this plasmid were more effective at stimulating subdominant-epitope-specific T cells. As shown in Fig. 3, the percentage of subdominant cells responding to pCMV-NPΔ was only very slightly greater than the proportion responding to cells transfected with pCMV-NP. Thus, if competition for binding to Ld occurs, it does not have a profound effect in this case. Other considerations also suggested that competition for MHC binding is unlikely to provide a general explanation for immunodominance. For example, the presence of a dominant epitope can inhibit the response to a subdominant epitope which is presented by a different MHC allele; indeed, all three dominant LCMV epitopes on the H-2b background are presented by Db, but these epitopes successfully prevent the development of subdominant Kb-restricted responses (33).
Dominant CD8+ T cells actively suppress the development of subdominant responses during virus infection.A general theory of immunodominance must encompass the observations that a dominant epitope can exert its effect on epitopes on different proteins, and on epitopes presented by different MHC class I alleles. We considered the possibility that the suppressive effects of a dominant epitope were not mediated by the epitope per se, but instead by the CD8+ T-cell response induced by that epitope, which might actively suppress the development of subdominant CD8+ T-cell responses. One testable prediction of this hypothesis is that the response to a subdominant epitope would be down-regulated by preexisting CD8+ T cells specific for the dominant epitope. To test this idea, mice were immunized with either or both of the individual minigene plasmids pCMV-UMGX and pCMV-UMG4, and 6 weeks later, they were infected with LCMV. CD8+ T-cell responses against the dominant and subdominant epitopes were quantitated at 0, 4, 5, and 7 days postinfection (Fig. 4). In mice immunized with pCMV-UMG4 alone, 1.3% of CD8+ T cells were specific for the dominant epitope (Fig. 2); following virus infection, these cells rapidly expanded, reaching ∼30% of CD8+ T cells by 4 days postinfection (Fig. 4A). In these mice, the WX-specific response was below the level of detection, suggesting that the accelerated NP118-specific response resulted in even more effective suppression of the subdominant response than occurs in unvaccinated animals. Mice primed with pCMV-UMGX alone (Fig. 4B) mounted a strong response to the normally subdominant epitope, which developed more slowly than did the dominant response in the pCMV-UMG4 vaccinees but peaked at ∼40% of all CD8+ T cells. Mice which had been primed with both plasmids had similar levels of memory cells specific for the dominant and subdominant epitopes (see dot plots in Fig. 2, bottom panels), but, most strikingly, the subdominant cells underwent only a limited expansion following LCMV infection (Fig. 4C); the exponential expansion between days 5 and 7 (seen in mice immunized with pCMV-UMGX alone) failed to occur in the doubly immunized mice. Overall, the response to the subdominant epitope was 80% lower than that in mice which had been immunized with pCMV-UMGX alone (Fig. 4; compare panels B and C). The only difference between these two groups of mice was the presence or absence of NP118-specific memory cells; therefore, we conclude that, in the doubly immunized mice, the expansion of the DNA-induced WX-specific memory cells was inhibited by the rapid virus-driven expansion of the dominant memory T cells.
Prior immunization against the dominant epitope inhibits development of the subdominant response during subsequent virus infection. BALB/c mice were immunized with pCMV-UMG4 (A), pCMV-UMGX (B), or both plasmids (C). Six weeks later, the mice were infected with LCMV intraperitoneally, and 4, 5, and 7 days later, mice were sacrificed (four mice per time point). Dominant (•)- and subdominant (○)-epitope-specific CD8+ T-cell responses were evaluated at each time point by ICCS after stimulation with NP118-126 or NP313-322, respectively. Average percentages (± standard errors of the means) of CD8+ cells positive for IFN-γ at each time postinfection are shown.
IFN-γ plays a key role in determining immunodominance.The above data suggested that immunodominance is determined, at least in part, by active suppression of subdominant responses by dominant CD8+ T cells. We considered an obvious explanation for the effect noted in Fig. 4C—that the accelerated expansion of dominant cells led to the more rapid clearance of virus and therefore reduced antigen stimulation of subdominant cells. However, no significant differences in viral titers between singly immunized and doubly immunized mice were seen; virus was cleared by 4 days postchallenge in both vaccine groups (see below), confirming our published finding that the subdominant vaccine pCMV-UMGX is as effective as a vaccine encoding the dominant epitope (28). For several reasons we considered it possible that IFN-γ might be an important regulatory molecule. IFN-γ is secreted mainly by T cells and natural killer (NK) cells. Although best known for its antiviral effect, it also regulates several aspects of the immune response, including the activation-induced cell death of T cells which follows viral infection. T cells from mice incapable of responding to IFN-γ are hyperproliferative and are less susceptible to activation-induced cell death (22). Therefore, we evaluated immunodominance in IFN-γ knockout (GKO) mice. Immunocompetent BALB/c mice and their GKO counterparts were immunized with pCMV-NP, and 6 weeks later they were infected with LCMV. Responses to the dominant and subdominant epitopes were evaluated at various times postinfection by using ICCS (identifying antigen-specific effector cells by TNF-α production rather than by IFN-γ production). Representative data obtained 22 days after LCMV infection are shown in Fig. 5A. The ratio of dominant to subdominant CD8+ T cells in normal BALB/c mice is ∼14:1, similar to that shown in Fig. 1. However, in the absence of IFN-γ, the ratio was markedly reduced, to less than 2:1; this change resulted mainly from the ∼6-fold increase in frequency of CD8+ T cells specific for the subdominant sequence. Detailed kinetics of the dominant and subdominant responses in the presence and absence of IFN-γ are shown in Fig. 5B. Responses to the dominant peptide D were similar regardless of the IFN-γ status of the mice, but at all time points examined the responses to the normally subdominant peptide WX were greatly increased in GKO mice. Thus, IFN-γ appears to play an important part in establishing immunodominance.
Reduction in immunodominance in mice deficient in IFN-γ. BALB/c IFN-γ+/+ and BALB/c GKO mice were immunized with pCMV-NP. Six weeks later, mice were challenged with LCMV intraperitoneally, and they were sacrificed 4, 5, 7, 15, 22, and 24 days postchallenge (three mice per group). At each time point, epitope-specific CD8+ T-cell responses were evaluated by ICCS assay for the individual mice by using peptide-coated cells as stimulators. Since IFN-γ production could not be used as an indicator of T-cell responsiveness in GKO mice, we used TNF-α as the indicator in this ICCS assay. (A) Data from one mouse of each strain at 22 days postinfection. The epitope-specific CD8+ T-cell response as a percentage of total CD8+ T cells is shown. (B) For each time point postinfection, responses to the dominant and subdominant epitopes (left and right panels, respectively) are shown as percentages of total CD8+ T cells. Each data point is the average for three immunocompetent (•) or GKO (○) BALB/c mice ± the standard error of the mean.
IFN-γ also exerts its effects in H-2b mice and can even suppress normally dominant CD8+ T-cell responses.The final series of experiments relied on H-2b mice, for two reasons. First, we wanted to ensure that the effect of IFN-γ was not restricted to a single mouse strain or MHC haplotype. Second, three dominant epitopes (NP396-404, GP33-41, and GP276-285) are present on the H-2b background, allowing us to evaluate the effect (if any) of IFN-γ on responses to these epitopes. C57BL/6 mice and their GKO counterparts were immunized with pCMV-NP, and 6 weeks later they were infected with LCMV. As above, ICCS assays were carried out at several time points postinfection, and data from days 15 and 22 postinfection are shown in Fig. 6. After virus infection, the C57BL/6 vaccinees mounted strong responses to the dominant NP396 epitope (against which they had been primed by the DNA vaccine), but responses to the two (normally dominant) GP epitopes were barely detectable. Thus, it appears that, in immunocompetent mice, the vaccine-induced priming of NP396 memory cells results in the suppression of development of the normally dominant GP-specific responses upon subsequent virus infection. This effect requires IFN-γ, because GKO mice primed with pCMV-NP could mount strong responses to all three dominant epitopes after virus infection; the responses to the GP33 and GP276 epitopes in GKO mice were 10- and 20-fold greater, respectively, than those in immunocompetent mice.
IFN-γ also regulates dominance in H-2b mice and can suppress even normally dominant CD8+ T-cell responses. C57BL/6 mice (immunocompetent or GKO) were immunized with pCMV-NP. Six weeks later, mice were challenged with LCMV intraperitoneally; 15 or 22 days postchallenge, they were sacrificed. At both time points, epitope-specific CD8+ T-cell responses were evaluated by ICCS, using TNF-α production as the indicator. Responses were measured in splenocytes of three individual mice at each time point, and representative data are shown. Stimulator cells were either uncoated or coated with a peptide representing one of the three normally dominant epitopes: NP396-404, GP33-41, and GP276-285. Data shown are gated on CD8+ T cells, and numbers indicate the percentages of CD8+ T cells that are TNF-α positive.
In all of the experiments for which results are shown in Fig. 5 and 6, the mice had been immunized with pCMV-NP prior to infection, and we anticipated that as a result, virus would be quickly eradicated. However, we considered it very important to determine virus titers in these mice because, if virus replication were higher and/or prolonged in mice lacking IFN-γ, the increased response of subdominant CD8+ T cells might be attributable to increased viral load. Therefore, viral titers were determined in the spleens of all mice for which results are shown in Fig. 5 and 6. Among the mice for which results are shown in Fig. 5B, a very low level of virus (∼102 PFU per g, just above the level of detection) was present in two mice (one normal, one GKO) at 4 days postinfection. In all other mice at this time point, and in all mice at all later time points, LCMV was undetectable. Thus, the different patterns of immunodominance are not likely to be due to antigen load.
DISCUSSION
Immunodominance has been defined as the presence of a hierarchy of epitope-specific responses; an epitope which induces a strong response is defined as dominant, whereas epitopes inducing weak responses are defined as subdominant. However, early observations on epitope dominance by Allan and Doherty showed that, in the absence of a normally dominant Ld-restricted response, strong responses developed to normally subdominant epitopes presented by the Kd and Dd MHC class I alleles (1). This study revealed a key feature of immunodominance: the removal of a dominant response allowed the enhancement of subdominant responses. Thus, immunodominance is not merely a fixed hierarchy of responses to a variety of epitopes; central to the phenomenon is the observation that the presence of a dominant epitope leads to the suppression of responses to subdominant epitopes. The mechanism underlying this suppression was the focus of the present study.
Many factors may contribute to immunodominance (reviewed in reference 11). First, the intrinsic efficiency of antigen presentation might play a role. For example, the subdominant epitope WX is more effective at stimulating T cells when it is expressed as a ubiquitinated minigene than when it is expressed from full-length NP (Fig. 3), suggesting that the subdominance of this response during virus infection may be determined, in part, by the poor presentation of this epitope in infected cells. Alternatively, peptides might compete for binding to the MHC molecule, and the epitope which forms the most stable association might be dominant; this hypothesis could explain how a dominant epitope could actively suppress a subdominant response, by minimizing that epitope's presentation. Removal of the dominant epitope could reduce competition, allowing more-efficient cell surface presentation of the subdominant sequence, with resulting enhancement of the CD8+ T-cell response. This mechanism could contribute to immunodominance when both epitopes were presented by the same MHC allele; however, it cannot operate in all cases of immunodominance, because a dominant epitope presented by one MHC allele can interfere with responses to an epitope presented by a different MHC allele. Furthermore, immunodominance cannot be explained solely on the basis of the cell surface density of peptide-MHC complexes, because an epitope's immunogenicity does not directly correlate with its cell surface abundance (10, 15, 35). An additional mechanism which contributes to some cases of immunodominance (12) is a genetic deficit (a “hole”) in the T-cell receptor (TcR) repertoire; perhaps naïve cells carrying TcR specific for a dominant epitope outnumber those carrying TcR specific for subdominant sequences. We have not evaluated the T-cell repertoire in our studies, but mice immunized with minigene plasmids mount similar responses to both epitopes (Fig. 2), suggesting that the naïve TcR repertoires are similar for the dominant and subdominant epitopes. Thus, we conclude that the dominance hierarchy of the peptide D-specific and peptide WX-specific responses in infected mice and in pCMV-NP vaccinees (Fig. 1) is unlikely to derive from underrepresentation of WX-specific naïve T cells. Therefore, other explanations for immunodominance must be sought.
Other laboratories have suggested that a major determinant of immunodominance is the competitive interaction between epitope-specific T cells for access to antigen-expressing cells (APCs) (19, 42), and in agreement with this model, we show here that immunodominance can be overcome by expressing the dominant and subdominant epitopes on separate APCs. Mice were immunized with pCMV-UMGX or pCMV-UMG4, and 14 days later, the vaccine-induced CD8+ T-cell responses were evaluated. As shown in Fig. 2, the responses to the dominant and subdominant epitopes are equivalent, suggesting that immunodominance can be circumvented by the simple maneuver of epitope separation. If this is correct, one would predict that strong responses to both epitopes could be induced even when the dominant and subdominant epitopes were expressed simultaneously, as long as they were presented by different APCs; this expectation was borne out in mice coinjected with pCMV-UMG4 and pCMV-UMGX, in which very similar CD8+ T-cell responses were induced concurrently to both epitopes (Fig. 2, bottom panels). Taken together, the data in Fig. 1 and 2 are consistent with the idea that immunodominance requires coexpression of epitopes on APCs and can be overcome by epitope separation.
However, during a normal virus infection, dominant and subdominant epitopes cannot be artificially separated in this manner; thus, during infection, an APC which expresses a subdominant epitope most likely would also express a dominant epitope. Therefore, we hypothesized that, if dominant CD8+ T cells suppress subdominant responses, immunodominance might be magnified following virus infection of mice which had been vaccinated to induce dominant CD8+ memory T cells. We tested this hypothesis in mice which had been coimmunized with pCMV-UMGX and pCMV-UMG4. These mice had easily detectable, and quantitatively similar, responses to both epitopes when assayed after DNA immunization (see Fig. 2, bottom panels), but following virus infection, only the dominant cells expanded rapidly and dramatically; the subdominant population expanded very minimally (Fig. 4C). The poor expansion of the subdominant cells cannot be explained by inadequate processing or presentation of the subdominant epitope in LCMV-infected cells because, in mice immunized with pCMV-UMGX alone, virus infection drove a rapid and strong expansion of the WX-specific cells (Fig. 4B). Therefore, we suggest that, after LCMV infection of the doubly immunized mice, the rapidly expanding NP118-specific CD8+ T cells acted on APCs to suppress the expansion of the WX-specific cells; as a result, at 7 days postinfection, virus-driven expansion of subdominant CD8+ T cells was ∼80% lower in doubly immunized mice than in mice which had been immunized with pCMV-UMGX alone. As noted above, since APCs are unlikely to express a subdominant epitope in the absence of a dominant epitope during normal virus infections, dominant CD8+ T cells will invariably be able to suppress subdominant responses in this manner. This hypothesis explains why removal of a dominant epitope from a microbe results in increased responses to previously subdominant epitopes; in the absence of suppressive dominant cells, the subdominant CD8+ T-cell responses can develop more completely. These observations have important implications for vaccination in general, and for subunit vaccination in particular. Rather than combining as many epitopes as possible in a single vaccine construct, our findings suggest an alternative approach. A DNA vaccine should comprise a mixture of plasmids, each encoding only a short fragment of a pathogen; this would maximize epitope dispersal among APCs and thus would induce broad immunity against numerous epitopes, dominant and subdominant.
How might the dominant CD8+ T-cell response act on APCs to suppress the development of subdominant CD8+ T cells? CD8+ T cells have two basic effector functions: target cell lysis and cytokine secretion. Since immunodominance can be overcome by epitope separation (Fig. 2), and because APCs seem to be central to the effects of epitope dispersal (42), we inferred that the suppressive effect of dominant CD8+ T cells must be rather tightly focused on epitope-expressing APCs. Several studies have indicated that LCMV-specific CD8+ T cells can exert immunosuppressive effects, perhaps by APC lysis (2, 5, 7), and therefore we initially hypothesized that immunodominance might be mediated by lysis of epitope-expressing APCs. However, experiments with perforin-deficient mice revealed minimal changes in CD8+ T-cell immunodominance (data not shown). Next, we evaluated immunodominance in mice deficient in IFN-γ. Virus infection of naïve GKO H-2d mice resulted in a very strong response to the normally subdominant epitope; this was ∼6-fold stronger than that in IFN-γ-positive mice (Fig. 5A) and was maintained at all time points evaluated postinfection (Fig. 5B). Furthermore, the effect of IFN-γ was not limited to the H-2d MHC background or to the individual Ld-presented epitopes; it occurred also in C57BL/6 (H-2b) mice (Fig. 6). LCMV infection of naïve C57BL/6 mice induces CD8+ T-cell responses to three dominant epitopes: NP396, GP276, and GP33. However, if these immunocompetent mice were first immunized with pCMV-NP and then infected with LCMV, they mounted strong responses to the NP396 epitope but failed to mount marked responses to either of the GP epitopes; in essence, the prior NP immunization transformed the normally dominant GP epitopes into subdominant sequences. In contrast, in pCMV-NP-immunized IFN-γ-deficient H-2b mice, virus infection induced strong CD8+ T-cell responses against all three dominant epitopes; thus, IFN-γ was required to permit the NP-specific response to suppress the development of the two GP-specific populations. We considered the possibility that this suppression might be attributable to the more rapid clearance of virus by the IFN-γ-positive mice, with concomitant reduction of antigen load, resulting in a weak response to GP. However, virus was eradicated equally quickly by the vaccinated GKO mice, which mounted strong responses to GP; this suggested that the absence of GP-specific responses in C57BL/6 mice resulted not from low antigen loads but instead from IFN-γ-mediated suppression by the faster-developing NP-specific CD8+ T-cell response. Therefore, even normally dominant responses (in this case, those to GP) can be suppressed by rapidly expanding effector CD8+ T cells, and this suppression requires IFN-γ. In summary, IFN-γ suppresses the development of some CD8+ T-cell responses, thus rendering them subdominant; in the absence of IFN-γ, the subdominance is largely relieved. Harty and coworkers have previously reported a change (approximately twofold) in immunodominance in mice lacking IFN-γ (4). Their work relied on infectious agents, and the authors' interpretations were recently challenged; on the basis of mathematical modeling, it was claimed that the improved subdominant responses in mice lacking IFN-γ could be explained by delayed clearance of the infectious agents used and the consequent prolongation of antigen stimulation (41). Here, we provide strong evidence that IFN-γ is indeed important in determining immunodominance, and we show that prior DNA immunization magnifies the suppressive effects of this cytokine in normal mice; following infection of vaccinated BALB/c or C57BL/6 mice, subdominant responses were suppressed 6- to 20-fold in comparison to those in mice lacking IFN-γ. In addition, we provide two lines of evidence which suggest that the effect of IFN-γ on immunodominance is not mediated by changes in antigen load. First, the viral titers, and the rates of virus clearance, were similar in normal and IFN-γ-deficient mice, and second, no change in immunodominance was observed in perforin-deficient mice, in which viral loads remained high for a prolonged period.
The suppressive effect of dominant CD8+ T cells must be highly focused, since it can be overcome simply by separating the target epitopes. How can a soluble mediator such as IFN-γ act in such a localized manner? We have previously shown that IFN-γ secretion by antigen-specific CD8+ T cells is exquisitely sensitive to contact between the TcR and the peptide-MHC complex; the CD8+ T cells initiate IFN-γ production within minutes of antigen contact and terminate synthesis immediately following antigen disengagement (31). We suggest that this tightly regulated control of IFN-γ production causes the cytokine's actions to remain extremely localized, thus permitting dominant CD8+ T cells to exert their immunosuppressive effects only on APCs with which they are in direct contact. The precise mechanism by which IFN-γ imposes its suppressive effect is unknown, but it is intriguing that a recent publication showed that T cells of high functional avidity can induce antigen loss from the surfaces of APCs (20).
ACKNOWLEDGMENTS
We are grateful to Annette Lord for excellent secretarial support and to Sascha Nussbaum for critical reading of the manuscript.
This work was supported by NIH grant AI-27028 and by FIS grant 01/1443 from the Spanish Ministerio de Sanidad y Consumo.
FOOTNOTES
- Received 2 November 2001.
- Accepted 28 January 2002.
- ↵*Corresponding author. Mailing address: Department of Neuropharmacology, CVN-9, The Scripps Research Institute, 10550 N. Torrey Pines Rd., La Jolla, CA 92037. Phone: (858) 784-7090. Fax: (858) 784-7380. E-mail: lwhitton{at}scripps.edu.
↵† Manuscript 12304-NP from The Scripps Research Institute.
REFERENCES
- American Society for Microbiology
