Cytotoxic T Cell Responses to DNA Vaccination: Dependence on Antigen Presentation via Class II MHC¹

The mechanism by which direct injection of DNA into tissues induces humoral and cellular immune responses to encoded Ags has been the subject of much recent investigation. Experiments have demonstrated that gene gun-mediated delivery of DNA to skin can result in the direct transfection of dendritic-like cells that migrate to local lymph nodes (1). The role of migratory cells is further reinforced by studies in which the site of DNA inoculation is excised shortly after immunization, with little or no effect upon the resulting immune response (2, 3). Furthermore, studies using bone marrow chimeras have demonstrated that cellular responses to the DNA-encoded Ags are dependent upon bone marrow-derived cells rather than on resident host tissue cells (4, 5).

The above studies do not exclude the possibility of “cross-priming” (6) after DNA vaccination. In this process, non-APCs in the local tissue secrete Ag or peptides that are taken up by professional APCs, such as dendritic cells. Normally, exogenous Ags preferentially enter the MHC class II loading pathway. However, dendritic cells have been shown to be unique among APC for their ability to “cross-prime,” presenting exogenous Ags in association with class I (as well as class II) MHC molecules (7, 8).

To assess the role of cross-priming by secreted Ags during DNA vaccination, transfected myoblasts have been transplanted into host tissues and used to generate cellular responses (9, 10). These responses are most likely explained through a process of cross-priming via secreted peptides or proteins. Alternately, in the normal situation of DNA vaccination into muscle, dendritic cells may phagocytose apoptotic bodies derived from transfected tissue cells (as they do for virally infected cells (11)), resulting in cross-priming. However, there is no evidence that nonprofessional APC can directly prime immune responses after transfection via DNA vaccination.

Once Ag has been processed by APC such as dendritic cells, whether by direct transfection or cross-priming, T cell responses can be elicited. But what is the precise role of CD4⁺ T cells in contributing to CD8⁺ T cell responses? In general, CD8⁺ T cell responses are highly dependent upon the prior stimulation of CD4⁺ T cells (12, 13, 14). This stimulation can be circumvented by stimulation of CD40 on dendritic cells, among other means (15, 16, 17). In vitro priming of dendritic cells with class I MHC-binding peptides can also result in the production of CD8⁺, class I-restricted CTL responses (18, 19). In addition, i.v. injection of synthetic peptides can prime CTL responses, even in CD4-deficient mice (20). Thus, it has been postulated that dendritic cells can directly prime naive CD8⁺ T cells and induce CTL responses.

If dendritic cells are indeed transfected by DNA vaccination, and/or if they present MHC class I-associated peptides via cross-priming, one might assume that CD8⁺ T cell responses to DNA immunization could be induced in the absence of CD4⁺ T cell help. Indeed, some strategies for DNA vaccination have apparently been based upon this implicit assumption. For example, multiple CTL epitopes have been assembled into expression vectors and successfully used for DNA vaccination and induction of CTL responses (21, 22). On the surface, this would seem to validate the idea that class II-restricted T cell help is not required for generation of CTL responses by DNA vaccination. Alternately, this approach may succeed because of the inadvertent creation of MHC class II-binding helper epitopes via the head-to-tail assembly of CTL epitopes. Another group has used “single” T cell epitopes of up to 17 amino acids, with or without a leader sequence, as DNA vaccines to generate CTL responses to HIV gp120 and mutant p53 via gene gun immunization (23). Although these responses would be predicted to be CD4-independent, this was not directly tested. In fact, the possibility remains that MHC class II-binding epitopes were also present in their construct, potentially generating CD4-dependent T cell help.

In this study, we set out to directly test the hypothesis that CTL generation via DNA vaccination does not require class II-restricted T cell help. We did this by constructing vectors containing only MHC class I or MHC class I + class II-restricted epitopes of OVA. Our results demonstrate that, contrary to the above reasoning, class II-restricted T cell help via CD4⁺ T cells is absolutely required for efficient generation of CTL responses to OVA DNA immunization. These results have both theoretical and practical importance to the design of future minimal epitope constructs for DNA vaccination.

A vector for DNA immunization (pOVA) containing the OVA gene linked to a human Vκ leader sequence and driven by a CMV promoter has been described (24). This vector contains CpG motifs known to be important for immunogenicity of DNA vaccines (25) and generates strong CTL responses in vivo (24). The plasmid was modified by removal of the OVA gene by restriction digestion with DraIII and BamHI. The OVA gene was replaced with a sequence generated by two overlapping synthetic oligonucleotides encoding the MHC class I-restricted T cell epitope of OVA, SIINFEKL (amino acids 257–264), and resulting in restriction site overhangs to allow directional cloning into the vector backbone. A second vector was also created by insertion of overlapping synthetic oligonucleotides encoding the above SIINFEKL epitope followed by the MHC class II-restricted T cell epitope of OVA, TEWTSSNVMEERKIKV (amino acids 265–280). The correct construction of these vectors was confirmed by DNA sequencing.

Female C57BL/6 mice, 6–8 wk of age, were purchased from Harlan BioScience (Indianapolis, IN). They were injected i.m. with 100 μg of one of the above vectors, or intradermally (i.d.)³ with 50 μg of one of the above vectors in normal saline. The i.m. injections were done either once or twice (4 wk apart), with CTL assays done 4 wk after a single injection or 2 wk after a second injection. All i.d. injections were done twice, 4 wk apart, with CTL assays done 2 wk after the second injection.

CTL assays were performed as described (24). Briefly, mouse spleens were aseptically harvested, single cell suspensions made, and the cells treated with a hypotonic buffer to remove RBC. They were then incubated at a 10:1 ratio with irradiated EG.7-OVA cells (a gift of M. Bevan, University of Washington, Seattle, WA) for 6 days, followed by analysis of lytic ability using a standard ⁵¹Cr release assay. Target cells were EG.7-OVA, EL-4 (obtained from the American Type Culture Collection, Manassas, VA), or EL-4 pulsed for 1.5 h with 1 μg of SIINFEKL peptide.

Mice were given daily i.p. injections of 200 μg of mAb for 3 days, then rested for 4 days. Representative mice were then bled to quantitate depletions by flow cytometry. All mice were also immunized with DNA on this day, and received an additional 200 μg injection of mAb, which was repeated every 4–5 days for the duration of the experiment. The i.d.-immunized groups received a second DNA injection 4 wk after the first injection, with spleens being harvested 2 wk after this second injection. Representative mice were also bled at the time of splenic harvest to recheck depletions via flow cytometry. Depletions at all time points checked were always ≥95%. mAb used for depletions included: irrelevant mouse IgG1 (clone LC4, produced in our laboratory (26)); anti-CD4 (clone GK1.5); and anti-CD8 (clone 53-6.7). Anti-CD4 and anti-CD8 hybridomas were obtained from Irving Weissman (Stanford University). They were produced as ascites in Swiss nu/nu mice, and the ascites was clarified by high-speed centrifugation, sterile filtered, and titered for IgG1 concentration by ELISA.

To test whether a single CTL epitope could be used as a DNA vaccine to induce CTL responses in mice, we constructed a vector (referred to as pep DNA) expressing the SIINFEKL peptide of OVA (amino acids 257–264). This peptide was joined to a 5′ Ig Vκ leader, since this leader was used in our OVA DNA vaccination vector containing the complete OVA cDNA (referred to as OVA DNA). The OVA DNA construct was previously shown to induce strong CTL responses in mice (24). Also, we reasoned that the Vκ leader may aid in directing translated products into the endoplasmic reticulum for loading onto class I MHC molecules.

The pep DNA vaccine was given i.m. to more than 20 C57BL/6 mice over several experiments and always induced a weak or negative CTL response, as assayed on the OVA-transfected EG.7-OVA cell line (Fig. 1). By contrast, the OVA DNA vaccine generated reproducibly high CTL responses. This was true after either one or two immunizations; i.e., the pep DNA vaccine did not generate significant CTL responses even after boosting.

The pep DNA vaccine was modified by addition of a Th cell epitope from OVA, TEWTSSNVMEERKIKV (amino acids 265–280). This two-epitope DNA vaccine (called dbl pep DNA) was then compared with the pep DNA vaccine and the OVA DNA vaccine for the ability to induce CTL responses (Fig. 2). Whether assayed on EG.7-OVA targets, or SIINFEKL peptide-pulsed EL-4 targets, the pep DNA construct again induced low (Fig. 2,A) or negative (Fig. 2,B) CTL responses. However, the dbl pep DNA vaccine induced a strong CTL response. In one experiment (Fig. 2,B), the dbl pep DNA induced superior lysis compared with OVA DNA; in another experiment, OVA DNA was slightly superior (Fig. 2,A). The OVA DNA and dbl pep DNA responses were SIINFEKL peptide-specific as seen by the use of peptide-pulsed target cells. The responses were also CD8-dependent since addition of anti-CD8 mAb during the lysis portion of the assay abrogated cytotoxicity against EG.7-OVA (Fig. 2 B).

It remained possible that the CTL response to the dbl pep DNA vaccine (as well as the OVA DNA vaccine) was not mediated via MHC class II-dependent help, as suggested by the above experiment, but rather was due to some other effect of the additional sequence(s) surrounding the SIINFEKL epitope. To test the dependence on T cell help directly, mice were depleted of either CD4⁺ or CD8⁺ cells before vaccination with the OVA DNA vaccine. As seen in Fig. 3, depletion of either CD4⁺ or CD8⁺ cells completely abrogated the CTL response against either EG.7-OVA or peptide-pulsed EL-4 cells. Similar results were obtained using the dbl pep DNA vaccine, in that the CTL response to this construct was also dependent upon CD4⁺ cells (data not shown). To demonstrate that the response against EG.7-OVA was CD8-dependent, anti-CD8 mAb was added to cultures during the lysis portion of the assay, abrogating cytotoxicity against EG.7-OVA. Thus, the CTL response to i.m. OVA DNA vaccination is dependent upon MHC class II-mediated, CD4⁺ T cell help.

The skin is thought to be an efficient site for the induction of immune responses to DNA vaccination because of the large number of Langerhans cells, which can be activated to become Ag-presenting dendritic cells. It thus seemed possible that the direct transfection of these cells via i.d. DNA vaccination might be more efficient and might lead to direct stimulation of CD8⁺ T cell responses. We therefore immunized mice i.d. with either the pep DNA vaccine or the OVA DNA vaccine, and tested their CTL responses after two immunizations (Fig. 4,A). As with i.m. DNA vaccination, the pep DNA construct failed to elicit detectable CTL responses, while the OVA DNA construct generated robust CTL activity against EG.7-OVA, but not parental EL-4 cells. To test the effect of CD4⁺ cells directly in this system, depletion experiments were conducted as in Fig. 3, but mice were vaccinated twice by the i.d. route. As with i.m. DNA vaccination, the CTL response to i.d. OVA DNA vaccination was highly dependent upon the presence of CD4⁺ cells; depletion of either CD4⁺ or CD8⁺ cells before vaccination abrogated the response (Fig. 4 B). Thus, despite the increased number of potential dendritic cell targets in the skin, i.d. DNA vaccination was also found to be dependent upon CD4-mediated, MHC class II-restricted T cell help.

In this study, we tested the notion that DNA vaccination may not require CD4-mediated T cell help to induce CTL responses. This hypothesis was based on the finding that dendritic cells can be directly transfected by DNA immunization (1), as well as obtaining Ag for presentation on MHC class I via cross-priming (10). Furthermore, dendritic cells have been shown to be able to directly prime naive CD8⁺ T cell responses (18, 19, 20). However, our results clearly show that the CTL response to i.m. and i.d. OVA DNA vaccination is greatly augmented in the presence of MHC class II-mediated CD4⁺ T cell help. Thus, we would suggest that, if dendritic cells are indeed able to stimulate CD8⁺ T cell responses in the absence of CD4⁺ T cell help, this function is not efficiently stimulated by DNA vaccination.

Can dendritic cells directly stimulate CD8⁺ CTL responses? Two studies have shown that pulsing dendritic cells with a single MHC class I-restricted peptide can elicit strong CTL responses (18, 19). However, because these dendritic cells were pulsed ex vivo, they may have been subjected to stimulation with other foreign Ags from FCS-containing medium. Thus, they may present peptides via class II MHC in addition to the class I epitopes with which they were pulsed and may therefore engender CD4⁺ T cell help via these other peptides. In fact, Porgador and Gilboa (19) showed that the CTL response to peptide-pulsed dendritic cells was CD4-dependent by depleting CD4⁺ cells in vivo. This is a strong argument against the ability of dendritic cells to stimulate CD8⁺ responses in the absence of CD4⁺ help.

Another study demonstrated the ability to induce CTL by injection of synthetic peptides in vivo into mice (20). These authors found that CTL could be generated even after CD4⁺ cell depletion or in CD4-deficient mice. However, the peptides in these experiments were given in incomplete Freund’s adjuvant, which may somehow overcome the need for CD4⁺ cells. Also, the dose of peptides (100 μg given s.c.) may be significantly higher than that engendered by DNA vaccination. Thus, these results may be explained by the effect of Ag dose, and DNA vaccination may simply result in too little Ag presentation to stimulate CD4-independent CTL responses.

Other studies have pointed to the importance of CD4⁺ cells in maintaining CD8-dependent responses. Walter et al. (27) studied the CTL response to CMV in bone marrow transplant recipients receiving adoptive transfer of autologous CD8⁺ T cells. Those patients who maintained strong CTL responses over time had recovered endogenous CD4⁺ T cell help; those who did not recover sufficient CD4⁺ help failed to maintain their CTL responses due to apoptosis of the transferred CD8⁺ cells.

Some investigators have created constructs for DNA vaccination which consist entirely of CTL epitopes (21, 22). These “polytope” constructs use multiple CTL epitopes strung head-to-tail but do not include any helper epitopes. If our results can be generalized to other systems, such constructs would not be predicted to work. However, they may have been successful because the assembly of the polytope construct created neoepitopes that served to generate class II MHC-restricted help. In fact, the construct of Thomson et al. (21) contains a potential MHC class II-binding epitope, LSYIPSAEK, formed by the fusion of CTL epitopes 7 and 8. This sequence contains anchor residues L and K at positions 1 and 9, and the preferred residue I at position 4. This would be predicted to bind to I-E molecules of both H-2^b and H-2^d haplotypes, according to the data compiled by Rammensee et al. (28). The polytope construct was tested by Thomson et al. (21) in H-2^b and H-2^d mice. Thus, we would predict that this sequence, and/or others formed as part of the polytope construct, may contribute to the ability to generate CTL responses in mice of these haplotypes.

Perhaps most intriguing of previous investigations into DNA vaccination with CTL epitopes is the work of Ciernik et al. (23). These authors constructed “single epitope” DNA vaccines that induced CTL responses against HIV gp120 and mutant p53. The responses were improved by addition of an adenovirus E3 leader for targeting to the endoplasmic reticulum. How could these vaccines work without the inclusion of helper epitopes? Several possibilities include the following. 1) The 17-amino acid p53 peptide and/or the E3 leader may generate a helper epitope, and CTL responses without the leader or with a more minimal CTL epitope may be analogous to low-level responses occasionally seen with our single epitope vaccine (see Figs. 1 and 2). 2) The E3 leader may target the peptide to the ER more efficiently than our Ig Vκ leader, and this efficient targeting is required for generating strong CTL responses without CD4⁺ T cell help. 3) The use of gold particle bombardment for delivery in their study may generate different results from i.m. and i.d. injection used in our study. In any event, the dependence of CTL generation on CD4⁺ cells by Ciernik et al. was not tested, making distinguishing among these possibilities difficult. Furthermore, our results argue that single epitope DNA vaccines will not always work well, despite the optimism generated by the above studies. Our results also argue that single epitope vaccines can be greatly improved by the addition of a helper epitope.

Because of these considerations, future DNA vaccines that target class I MHC molecules should include both MHC class I and class II-binding epitopes in multiple epitope DNA vaccines.

We thank Vishal Doctor for technical assistance with DNA injections, and Ronald Levy for critical review of the work.