To evaluate the specificity of HLA-DR1-restricted CD4 T cells elicited in response to the A/New Caledonia/20/99 influenza virus strain in DR transgenic mice, we first established conditions for T-cell priming in response to increasing doses of infectious virus introduced intranasally in PBS, an infection route meant to mimic infection in humans. Lymphoid cells were isolated from spleens and lymph nodes of infected mice at 7 to 8 days after the introduction of the virus, and CD4 T cells were purified from the cell population by the depletion of CD8 and MHC class II protein-positive cells. To evaluate the antigen specificity of influenza virus-primed CD4 T cells, we used ELISPOT assays, which allow the direct ex vivo quantification of antigen- or peptide-reactive lymphocytes without any extended in vitro culture (21
). Because early experiments indicated that, in our hands, IL-2 production was more sensitive than IFN-γ production for the detection of influenza virus-specific CD4 T cells (Fig. ), IL-2 ELISPOT assays were used to enumerate antigen-reactive CD4 T cells for the remainder of the study.
FIG. 1. Comparison of results from IL-2 and IFN-γ ELISPOT assays for the enumeration of influenza virus-specific CD4 T cells and the titration of the virus in mice. (A) Mice were infected intranasally with 100,000 EID50 of A/New Caledonia/20/99 virus. (more ...)
Shown in Fig. are the results of ELISPOT assays following the titration of the virus in vivo in the DR1 transgenic mice, with the infecting dose introduced intranasally in PBS as EID50 of virus grown in fertilized chicken eggs and recovered in allantoic fluid. Priming in the infected mice was assessed by quantifying IL-2 immunospots produced by CD4 T cells isolated from DR1 transgenic mice 7 days postinfection in response to virus-infected syngenic splenocytes (Fig. , left) or fibroblasts transfected with the gene encoding HLA-DR1 (Fig. , right). These experiments indicated that the A/New Caledonia/20/99 H1N1 strain of influenza virus, which has never been adapted to grow in mice, shows good infectivity in mice and primes T cells when introduced at relatively low doses (<2,000 EID50/mouse). The highest dose tested, 200,000 EID50, was not lethal to the animals, and all mice appeared healthy at 7 to 10 days postinfection, independent of the dose of virus used for infection. For the remainder of the assays, 100,000 EID50 per animal was used to infect mice and prime influenza virus-specific CD4 T cells.
To assess the peptide specificity of the influenza virus-specific CD4 T cells, synthetic overlapping 18-mer peptides, offset by 7 amino acids, representing the entire translated sequence of HA expressed in the A/New Caledonia/20/99 virus were used. The HA0 precursor protein of HA is more than 550 amino acids in length, and therefore, more than 80 peptides were synthesized in order to span its entire sequence. Because of the high number of peptides in the peptide scan, we used a peptide-pooling strategy (100
) to initially identify individual HA peptides to which CD4 T cells respond. Shown in Fig. is the layout of the matrix with the number of the individual peptide shown. Figure shows the results of an HA-specific ELISPOT analysis of the peptide pools, with the results presented as the number of IL-2 spots detected per 106
CD4 T cells after subtracting the number of background spots (<20). Pools displaying positive reactivity (Fig. , , , , , , , 8R, 9R, , , , , , , and 9C) each contained candidate peptides, while negative pools (Fig. , , 10R, , and 8C) were presumed to have no or very weak CD4 T-cell epitopes. Peptides within these negative pools were eliminated from further consideration. Peptides belonging to more than one strongly positive pool were considered to be candidates for the most dominant T-cell epitopes and were tested as single peptides in IL-2 ELISPOT assays. For these studies, we used an additional step of purification to be certain that the CD4 T cells detected were indeed restricted to DR1. The repertoire of DR1 transgenic mice is thought to be largely restricted to the human class II molecule expressed by a transgene, but the mice do express low levels of an endogenous class II protein (I-Af
) that is thought to poorly select and activate CD4 T cells (110
). Although the T-cell population was completely depleted of CD8 T cells prior to the addition of T cells to the ELISPOT assay, the depletion of class II protein-positive cells was only about 90% complete (Fig. ). It was therefore conceivable that some of the peptide epitopes detected in the pool arrays were restricted to the endogenous I-Af
class II molecule expressed in the B10.M host. Therefore, in all of the subsequent analyses, we adopted an even more rigorous purification strategy to eliminate any endogenous APC present among the spleen cells from the infected donors. Spleen cells were depleted of APC and CD8 T cells by treatment, as before, with a cocktail of specific anti-CD8, anti-class II molecule, and anti-B-cell antibodies and complement. CD4 T cells were further purified from the residual viable cells by using a positive selection strategy involving preparative flow cytometry with limiting concentrations of an anti-CD4 antibody that was found not to inhibit T-cell responses (data not shown). Figure shows that this strategy led to the isolation of CD4 T cells with more than 98% purity. We found in separate experiments that the use of higher concentrations of the anti-CD4 antibody GK1.5 blocked the antigen-dependent IL-2 production, attesting to the functionality of the CD4 T-cell-class II molecule interactions in the HLA-DR1 transgenic mice.
FIG. 2. Peptide screening by the peptide-pooling matrix method. Mice were infected intranasally with 100,000 EID50 of A/New Caledonia/20/99 virus. The number of IL-2-producing CD4+ splenocytes was determined 8 days later by 18-h coculture with DAP-3 DR1-positive (more ...)
FIG. 3. Purification of CD4 T cells for ELISPOT analyses. Splenocytes were harvested from infected mice, and a portion was reserved for staining (top row, prepurification). The remainder of the cell set was depleted of class II protein-positive and CD8 cells (more ...)
FIG. 4. Representative results from a single-peptide screen demonstrate the diversity of the primary CD4 response to influenza virus HA. Mice were infected intranasally with A/New Caledonia/20/99 influenza virus, and the number of IL-2-producing CD4 splenocytes (more ...)
FIG. 5. Confirmation of DR1 restriction of CD4 T cells. Mice were infected intranasally with A/New Caledonia/20/99 virus, and the number of IL-2-producing CD4 T cells from spleens was determined 8 days later by 18-h in vitro stimulation with DAP.3-5.3.1 DR1-positive (more ...)
FIG. 7. Sequence alignment of immunodominant peptides in HA. Shown is the ClustalW alignment of HA protein sequences from the New Caledonia/20/99 (H1N1), Canada/720/05 (H2N2), and Hong Kong/156/97 (H5N1) strains of influenza A virus. The regions conserved among (more ...)
FIG. 6. Summary of empirically observed immunodominant HA peptides and comparison with peptides predicted to be presented by HLA-DR1 by algorithms available on the Web. The HA sequences of 18-mer overlapping peptides are shown as a ladder. The underlined peptide (more ...)
HA-derived peptides were tested individually in several subsequent IL-2 ELISPOT assays using purified CD4 T cells isolated from mice infected 8 days previously. The results of a sample IL-2 ELISPOT assay with 21 single peptides are presented in Fig. . These results show the typical range in reactivity observed in independent assays, with several peptides, such as peptide 23 and peptide 63, eliciting the greatest number of DR1-restricted T cells while other peptides elicited intermediate (e.g., peptide 14), low (e.g., peptide 27), or undetectable (e.g., peptide 1) responses. We then sought to formally test the MHC class II restriction of the candidate peptides identified through the course of experiments using individual peptides. CD4 T-cell ELISPOT assays with the individual candidate HA peptides were performed in the presence of DR1-positive fibroblasts that expressed DR1 from transfected genes as their only MHC class II molecule, with endogenous MHC class I molecules being derived from the strain of origin of the fibroblasts (C3H [H-2k]). Untransfected fibroblasts were used as negative-control APC. The results of this experiment, presented in Fig. , show that all of the provisionally identified HA-derived CD4 T cells were indeed restricted to the HLA-DR1 class II molecule. Even the relatively minor specificities detected, exemplified by peptides 11, 27, and 45, were restricted to HLA-DR1. None of the influenza virus-specific CD4 specificities identified in our studies were detectable in uninfected mice (data not shown).
After comprehensively identifying the HA-derived peptides recognized by CD4 T cells, we asked how well these empirically defined epitopes compared to those selected by algorithms presently used to predict class II epitopes. Three programs available on the Web (reviewed in references 9
, and 84
) to predict the binding of peptides to MHC molecules, ProPred (1
), SYFPEITHI (31
), and Rankpep (16
), which have been used previously to predict T-cell epitopes, were used to scan the entire HA sequence for peptides predicted to bind to HLA-DR1. These programs implement different algorithms to predict peptide binding to HLA-DR1. Each algorithm identified approximately 12 to 15% of the peptides (e.g., 9 to 11 out of a total of more than 75 peptides) as candidate immunodominant peptides. These predicted peptides are highlighted in Fig. , which shows a ladder representation of each peptide contained in the HA peptide set. The empirically defined peptide epitopes are indicated in this figure as well, with the most immunodominant, the intermediate, and the minor peptides denoted. The criteria used for grouping the peptides are indicated in the legend to Fig. . Table shows the results of multiple independent experiments with each of the identified peptides, allowing for significant confidence that the relative immunodominance depicted in Fig. is reproducible.
Reproducibility of responses to individual peptides tested by ELISPOT analysesa
Several interesting observations emerge from the analyses shown in Fig. . First, different peptides were predicted by each of the three different algorithms, with only two 9-mer peptides (in peptide 1 and peptide 48) selected by all three. Four peptides were selected by at least two algorithms, but the algorithms selected different peptides within the 18-mer peptides. More than 10 9-mer peptides were selected by only a single algorithm, and of the algorithms that were tested, Rankpep appeared to give results that were the most discordant from those from the other two. The other remarkable observation made from this type of analysis is that many of the empirically defined epitopes were not predicted by any of the three algorithms. Two of the four most immunodominant peptides (contained within peptide 37 and peptides 63 and 64) and four of the six subdominant peptides were not predicted to be immunodominant by any of the algorithms that were used to scan the HA sequence. This result suggests that presently available algorithms are as yet inadequate for predicting peptides that will be the focus of the CD4 T-cell response, at least to influenza virus.