Correspondence to: Dr. Natalia Savelyeva, Genetic Vaccine Group, Cancer Sciences Unit, Southampton General Hospital, Southampton SO16 6YD, United Kingdom, Fax: +44-2380-795152, E-mail: email@example.com
The cancer/testis antigen NY-ESO-1 contains an immunodominant HLA-A2-binding peptide (SLLMWITQC), designated S9C, an attractive target for vaccination against several human cancers. As cysteine contains a reactive SH, the oxidation status of exogenous synthetic peptide is uncertain. We have designed tolerance-breaking DNA fusion vaccines incorporating a domain of tetanus toxin fused to tumor-derived peptide sequences (p.DOM-peptide), placed at the C-terminus for optimal immunogenicity. In a “humanized” HLA-A2 preclinical model, p.DOM-S9C primed S9C-specific CD8+ T cells more effectively than adjuvanted synthetic peptide. A DNA vaccine encoding the full NY-ESO-1 sequence alone induced only weak S9C-specific responses, amplified by addition of DOM sequence. The analog peptide (SLLMWITQL) also primed peptide-specific CD8+ T cells, again increased by DNA delivery. Importantly, T cells induced by S9C-encoding DNA vaccines killed tumor cells expressing endogenous NY-ESO-1. Only a fraction of T cells induced by the S9L-encoding DNA vaccines was able to recognize S9C and kill tumor cells. These data indicate that DNA vaccines mimic posttranslational modifications of SH-containing peptides expressed by tumor cells. Instability of synthetic peptides and the potential dangers of analog peptides contrast with the ability of DNA vaccines to induce high levels of tumor-lytic peptide-specific CD8+ T cells. These findings encourage clinical exploration of this vaccine strategy to target NY-ESO-1.
NY-ESO-1 is a cancer/testis antigen, expressed in a wide range of cancers but not in normal somatic cells. It appears to induce spontaneous antibody and T-cell responses in patients, even in advanced disease. This natural immunogenicity has encouraged exploration of vaccination strategies to amplify immunity to effective levels. A major goal for attack on intracellular antigens is induction of high levels of peptide-specific CD8+ T cells able to lyse the tumor target. NY-ESO-1 has the advantage of a clear HLA-A2-binding immunodominant peptide, SLLMWITQC (S9C) sited at positions 157–165, offering an attractive candidate for a specific vaccine, especially as there appears to be an available T-cell repertoire for this peptide in patients.[3-6] Initially, there were attempts to vaccinate patients with the peptide, and a small study of 12 patients with progressive NY-ESO-1-expressing tumors reported induction of T-cell responses, particularly in seropositive patients. However, the alternative approach of an NY-ESO-1 recombinant protein vaccine was then investigated, delivered with the ISOMATRIX adjuvant, mainly in patients with melanoma. Vaccination induced antibody, CD4+ T cells and CD8+ T cells, with some suggestion of an improved clinical outcome. Curiously, peptide recognition by the induced CD8+ T cells differed from that of the spontaneous response, with little recognition of the S9C peptide. This pattern was also noted in a vaccine trial, again using NY-ESO-1 recombinant protein, delivered with Montanide and CpG.[9, 10] Although CD8+ T cells were induced in half the patients, the peptides recognized were mainly at positions in the centre of the protein (positions 81–110). Even when S9C-specific T cells were detected, they appeared to have decreased tumor reactivity. In a parallel trial of a similar NY-ESO-1 protein/adjuvant combination, there appeared to be rather more responses against S9C, evident in nine of 14 patients, especially after prolonged vaccination.
More recently, recombinant vaccinia and fowlpox vectors expressing NY-ESO-1 have been used as vaccine vehicles for patients with either epithelial ovarian cancer or melanoma. Again, antibody and CD4+ T-cell responses were induced and, in melanoma patients, CD8+ T-cell responses, with possible clinical benefit. In terms of S9C peptide, priming of CD8+ T-cell responses, detectable after expansion in vitro, occurred in some patients (6/25), and an increase in pre-existing responses occurred in others (7/25). Importantly, if S9C-specific cells were found, they were capable of lysing target NY-ESO-1-expressing tumor cells. Thus, induction of S9C-specific T cells appeared variable, but remains an important goal.
One problem with S9C is that it contains cysteine at an anchor position. Cysteine (Cys) is a reactive amino acid because of its free SH moiety, which tends to undergo oxidation to SS at neutral pH. It will react with free Cys under physiological conditions (cysteinylation) and can dimerize depending on local conditions. This instability could affect clinical performance and might explain differing results. To avoid this problem and increase binding to MHC Class I molecules, analog peptides have been explored, with Cys substituted by a more stable amino acid such as valine, alanine or leucine. Peptide analogs not only form more stable complexes when bound to HLA molecules but also can be more efficient at expanding CTL lines in vitro. The question for analog peptides is whether the induced T cells can kill target tumor cells. Unfortunately, with the possible exception of Melan A/Mart-1, T cells raised against a whole range of analog peptides designed to target tumors, including those from gp100, CEA, survivin and NY-ESO-1, failed to achieve this at significant levels.[16, 17]
We have developed a novel DNA fusion vaccine encoding the first domain (DOM) of the fragment C (FrC) of tetanus toxin (TT) fused at the C-terminus to candidate MHC Class I-binding epitope sequences. This p.DOM-peptide design provides high levels of CD4+ T-cell help from the undamaged anti-TT repertoire, and the placement of the tumor-derived epitope appears to confer an advantage in priming peptide-specific CTLs. Our vaccine design has been demonstrated to induce epitope-specific CTLs against tolerized antigens. In the clinic, a p.DOM-peptide vaccine incorporating an epitope from a prostate cancer antigen ± electroporation induced CD8+ T-cell responses in the majority of patients. We reasoned that a DNA vaccine encoding S9C would deliver the peptide with posttranslational modification similar to tumor cells, thereby overcoming the difficulties associated with this otherwise attractive epitope. Using “humanized” HLA-A2 transgenic (HHD) mice, we show that this is the case.
Material and Methods
DNA vaccines and peptides
The p.DOM plasmid containing the sequence encoding the first domain of TT FrC has been described. Four additional DNA vaccines were constructed: two containing DOM linked to either the HLA-A0201-restricted NY-ESO-1 epitope S9C or to its analog S9L, and two containing the full-length NY-ESO-1 sequence either with (p.DOM-NYESO) or without DOM (p.NYESO) (Fig. 1). The full-length vaccines were constructed as follows: cDNA was obtained from human testis tissue, and NY-ESO-1 was amplified. To fuse NY-ESO-1 to DOM, primers containing the sequence coding for the final amino acids of DOM followed by a sequence coding for the linker AAAGPGL and the first amino acids of NY-ESO-1 were designed and the vaccine constructed. The epitope vaccines were constructed similarly using primers containing the sequence coding for the last amino acids of DOM followed by the sequence coding for the peptide. DNA vaccines were prepared and evaluated as described.
The HLA-A0201-restricted S9C (SLLMWITQC) peptide, its analog S9L (SLLMWITQL), the MHC Class II-restricted p30 (TTFNNFTVSFWLRVPKVSASHLE) and PADRE (KSSAKXVAAWTLKAAA; where X = cyclohexylalanine) peptides were synthesized commercially and supplied at more than 95% purity (PPR, Southampton, UK). Peptides were resuspended in dimethyl suphoxide (DMSO) at 10 mM and stored at −20°C.
Mice and vaccinations
HHD mice have been described elsewhere. As endogenous H-2b class I expression has been disrupted, these mice present antigens in the context of HLA-A2. HHD mice aged 6–12 weeks were immunized i.m. as described previously. Briefly, each mouse received a total of 50 μg DNA vaccine injected into two sites in the quadriceps. For peptide vaccination, 50 μg of either S9C or S9L peptide mixed with 100 μg PADRE peptide was emulsified in Montanide (ISA 50 V2, Seppic, Valbonne, France) and 100 µl of the vaccine was injected s.c. into each of two sites at the flank. Mice were vaccinated at day 0 and spleen cells were harvested at day 14. Experiments involving animals were conducted according to U.K. government license regulations and were approved by the University of Southampton's ethical committee. For cytotoxic assays, mice were primed and then boosted by DNA injection/electroporation at day 28. Splenocytes were harvested 7 days later.
Enzyme-linked immunosorbent spot (ELISPOT) analysis was performed using the BD ELISPOT Set (ELISpot Set, BD Pharmingen, San Diego, CA) for murine IFN-γ, as described previously. Briefly, viable mononuclear cells from individual mice were incubated (2.5 x 105 cells per well) in complete medium for 18 hrs with S9C or S9L peptides at varying concentrations (10−5–10−9 M) to assess CD8+ T-cell responses or with 10−6 M either p30 or PADRE peptides to evaluate CD4+ T-cell responses. An irrelevant H-2Db binding peptide was included as a control. Samples were plated in triplicate, and the mean reading expressed as spot-forming cells (SFC) per 106 cells. The reducing agent tris(2-carboxyethyl) phosphine hydrochloride (TCEP Pierce Biotechnology, Rockford, IL), which has been shown to prevent cysteine-containing synthetic peptides from forming disulfide bonds, was included (200 μM) during all incubations with the S9C peptide. The values two times greater than controls were considered positive.
Target cells for cytotoxic assays
The prostatic epithelial cell line derived from transgenic adenocarcinoma mouse prostate (TRAMP) was maintained in Dulbecco's modified Eagle's medium (DMEM) high glucose plus 10% FCS, 25 U/ml penicillin-streptomycin, 5 μg/ml insulin (Sigma-Aldrich, St.Louis, MO) and 10−8 M dihydrotestosterone (Sigma-Aldrich).
Lentivector preparations expressing a FLAG-tagged NY-ESO-1 cDNA under the control of the spleen focus forming virus promoter (SFFV) were prepared by cotransfection of plasmids pMDG and p8.91 into 293T cells as described.[27, 28] Supernatants containing lentivectors were collected, concentrated 100-fold by ultracentrifugation and titrated by Q-PCR, as previously described. TRAMP cells were transduced for 24 hrs with lentivector using a multiplicity of transduction of 100 according to the titre as assessed by a quantitative polymerase chain reaction assay (Q-PCR). Two days after transduction, NY-ESO-1 expression was assessed by flow cytometry using a FITC-labeled anti-FLAG antibody (Sigma-Aldrich). The cells were cloned by limiting dilution to isolate single-cell clones to achieve homogeneous expression. The homogeneous expression in individual clones was confirmed by immunofluorescent microscopy.
Invitro cytotoxicity assay
To assess peptide-specific cytotoxic T lymphocyte (CTL) responses, vaccinated mice were sacrificed on day 7 after boosting and single-cell suspensions made from individual spleens in complete medium. Splenocytes were washed and resuspended at 3 x 106 cells/ml in 15 ml complete medium, together with recombinant human IL-2 (20 IU/ml; Perkin-Elmer, Foster City, CA) and either S9C or S9L peptide (1 μM). Following 6 days stimulation in vitro (37°C, 5% CO2), the lytic ability of vaccine-induced CD8+ T cells was assessed using a standard 5-hr chromium release assay. Target cells included TRAMP cells transduced with HHD alone (TRAMP-HHD+) or with HHD and NY-ESO-1 (TRAMP-HHD+NYESO+), or TRAMP-HHD+ cells pulsed with S9C peptide.
ELISPOT data were analyzed using the Mann–Whitney test for nonparametric data. Data from lysis of target cells were pooled and expressed as mean ± standard deviation and compared using a two-tailed t-test. Groups were considered significantly different when p < 0.05.
Comparison of DNA delivery (p.DOM-S9C) with synthetic S9C peptide in priming CD8+ T-cell responses
HHD mice were vaccinated with the DNA vaccine (p.DOM-S9C) or with synthetic S9C peptide (plus PADRE peptide and Montanide). To compare specific responses, a short restimulation (18 hr) with peptides, followed by ELISPOT was used. A representative experiment comparing the frequency of primed IFNγ-producing S9C-specific CD8+ T cells at day 14 shows significantly higher levels (median = 80) following DNA delivery with low but measurable levels (median = 30.9) induced by the S9C peptide (p = 0.006) (Fig. 2). This difference was reproducible in all of three experiments. In contrast, CD4+ T-cell responses to the p30 epitope within the DOM sequence and to the PADRE synthetic peptide were comparable, indicating good coinduction of CD4+ T-cell help by both vaccines (Fig. 2). The control p.DOM vaccine induced a response to p30 but no S9C-specific response.
Efficacy of the analog S9L peptide, delivered via DNA or as synthetic peptide, in priming specific (S9L) and cross-reactive (S9C) CD8+ T cells
A similar comparison of efficacy of DNA delivery with synthetic peptide was made with the analog peptide, S9L, known to display a higher affinity for HLA-A2. Responses were high overall, with DNA vaccination (p.DOM-S9L) again superior (median = 409) to peptide (median = 105; p = 0.016) (Fig. 3ai). Comparable levels of CD4+ T cells were induced by both vaccines.
The S9L vaccines also induced lower but significant levels of S9C-reactive T cells (Fig. 3aii). In fact, the frequencies of primed cross-reactive T cells induced by p.DOM-S9L were similar to those induced by p.DOM-S9C (Fig. 3aii). Thus, the cross-reactive portion of the S9L response able to recognize S9C reaches the level induced by the specific vaccine, but is accompanied by a high level of non-cross-reactive S9L-specific CD8+ T cells. An estimate of this unwanted response was made by comparing the ratio of the S9L-specific response to the S9C-specific response induced by either p.DOM-S9L (5.1:1) or by p.DOM-S9C (1.6:1). Clearly, the specific vaccine induces a more S9C-focused T-cell population. This experiment was repeated twice with similar results obtained.
In addition to the level of unwanted T cells, there is also the question of avidity for homologous or cross-reactive peptides. Avidity includes the binding of peptide to HLA-A2, which is designed to be high for the analog peptide, and the binding to the presented peptide of the induced T cells. We measured the quality of the T cells induced by the p.DOM-S9L vaccine by first assessing the interaction with homologous S9L. As shown in Figure 3b (left-hand curve), apparent avidity was high with 50% occupancy at 0.02 µM peptide concentration. In contrast, the T cells induced by the p.DOM-S9C vaccine showed a lower avidity for S9L (∼x3), requiring 0.06 µM peptide (p = 0.008). The ability of the T cells induced by either vaccine to bind to S9C was comparable (right-hand curves). Taken together, these results show that there is no gain in using the analog peptide to induce S9C-reactive T cells. The analog appears instead to induce high levels of unwanted T cells with high avidity for the analog peptide.
Cytotoxicity of T cells specific for either S9C or S9L against target cells expressing endogenous NY-ESO-1
Vaccine-induced CD8+ T cells were expanded in vitro using the homologous peptides and the ability of expanded cells to lyse target cells analyzed. The target cells were transduced with HLA-A2 (HHD) and either coated with soluble S9C peptide or transduced with full-length NY-ESO-1. The p.DOM-S9C vaccine or the synthetic S9C peptide vaccine each induced CD8+ T cells able to kill both sets of target cells. Peptide-coated cells were killed very efficiently as expected but cells expressing endogenous NY-ESO-1 were also killed. Representative data from individual mice are shown in Figure 4a for DNA or synthetic S9C peptide vaccines, and pooled data from six to eight individual mice are shown in Figure 4b. The pooled data indicate a mean level of 35% ± 5% killing of the target cells with endogenously expressed NY-ESO-1 for p.DOM-S9C and 28% ± 6% (p < 0.05 for both) for S9C peptide vaccines at effector:target ratio of 60:1. This difference was reproducible in all of two experiments.
The S9L-containing vaccines induced a proportion of T cells able to recognize S9C and kill peptide-coated target cells, although less efficiently than the S9C vaccines (Figs. 4c and 4d). The S9C-specific population was clearly not expanded efficiently by the S9L peptide. The expanded S9L-specific cells were completely unable to kill target cells expressing endogenous NY-ESO-1 (p > 0.05). This was reproducible in all of two experiments.
Comparison of the ability of DNA encoding either full-length NY-ESO-1 (±DOM) or peptide (p.DOM-S9C) to prime S9C-specific CD8+ T cells
A DNA vaccine encoding the full NY-ESO-1 sequence (p.NYESO) could be advantageous because it would not be restricted to a subset of patients of defined HLA type. We therefore compared the ability of a full-length DNA NY-ESO-1 vaccine with the p.DOM-S9C vaccine in priming S9C-specific CD8+ T cells. The full-length sequence was able to prime only a weak response (median = 29), whereas p.DOM-S9C induced a high level of response (median = 322, p = 0.001) (Fig. 5a). Fusion of the DOM sequence to the full-length NY-ESO-1 (p.DOM-NYESO) improved performance (median = 75, p = 0.019). Although the trend was for p.DOM-peptide to induce higher levels of response, the difference did not reach significance (p = 0.0541) (Fig. 5a). The induced CTL responses induced by p.DOM-NYESO vaccine were able to kill tumor cells expressing endogenous NY-ESO-1 (Fig. 5b). Both data sets were reproduced in all of two experiments
NY-ESO-1 was one of the first cancer-testis antigens identified by SEREX technology. It is a relatively small molecule (∼18 kDa) containing an immunodominant peptide, S9C, able to bind to HLA-A2. These features, and its expression by a wide range of cancers, led to numerous attempts to induce immunity, often focusing on S9C-specific CD8+ T-cell responses. The importance of S9C is underscored by the fact that it appears to be the only HLA-A2-binding peptide constitutively expressed by tumor cells at levels suitable for cytotoxic attack by CD8+ T cells. However, there is a complication with S9C relating to potential instability because of the presence of cysteine, with an SH vulnerable to oxidative damage and dimerization.
This posttranslational modification has implications beyond NY-ESO-1 because ∼14% of T-cell epitopes potentially contain Cys residues. Analog peptides have been designed to replace Cys with Leu, Ser or Ala but this approach raises questions about T-cell recognition. Although it is possible to optimize anchor residues, the effect on T-cell recognition is less predictable. There is the added complication of conformational constraints on peptides via interactions with flanking residues from the peptides themselves. Tests with individual T-cell lines are useful but may not be readily applicable to polyclonal T cells.
An easier strategy is to allow the antigen-presenting cell to process the vaccine-delivered peptide naturally. This might occur using full-length injected protein but lysosomal enzymes are not the same as proteasomal/endoplasmic reticulum enzymes, and exogenous proteins are generally less efficient at inducing CD8+ T-cell responses than intracellular antigens. DNA vaccines should overcome this problem by delivering antigen directly to the desired processing pathways. However, a clinical trial of a DNA NY-ESO-1 vaccine failed to induce significant antibody, CD8+ T cells or clinical benefit. CD4+ T cells were induced but immunity was apparently suppressed by activation of regulatory T cells. One reason for this could be a lack of functional T-cell help, known to be required for breaking tolerance likely to exist in patients, and for activating memory CD8+ T cells. Our design includes the DOM sequence which does both. In our preclinical study, tolerance does not exist, but we were able to show strong priming of S9C-specific CD8+ T cells, especially with the p.DOM-S9C vaccine, but also with full-length sequence, p.DOM-NYESO. Although NY-ESO-1 should be foreign in mice, and therefore induce some T-cell help, addition of DOM appears to amplify this. It is likely to be essential in tolerant patients.
Following vaccination with p.DOM-S9C, CD8+ T cells able to kill tumor cells expressing endogenous NY-ESO-1 were induced. The analog S9L peptide vaccines could also induce cross-reactive T cells able to target S9C, but the cost was induction of high levels of irrelevant cytolytic T cells, which could have off-target activity. Even the Melan A analog peptide, which does not contain an SH group, and is a rare exception in mimicking the natural peptide very efficiently, also induces non-cross-reactive “irrelevant” CD8+ T cells. For NY-ESO-1, Ala or Ser substitution for Cys would be better mimics than Leu, but there seems to be no need for this because the S9C-encoding DNA vaccine is highly efficient. In general, DNA vaccines are superior to peptide vaccines and the encoded peptide appears to be posttranslationally modified in a comparable way to that from endogenous full-length protein. DNA vaccines encoding full-length NY-ESO-1 are effective if the DOM sequence is included, and are only marginally less effective than p.DOM-S9C, but they would have the advantage of wide application. Although it would not directly attack intracellular antigens, antibody is also likely to be induced by the full-length NY-ESO-1 sequence, but was not measured in our study aimed to compare induction of CD8+ T cells. Clinical trials are the next step and could build on the experience already obtained with p.DOM design, including the addition of electroporation to enhance effectiveness in large animals and in patients.[20, 38, 39]
The authors thank the Leukaemia and Lymphoma Research, UK for financial support to J.C.P., J.R. and N.S. D.E. was funded by a Miguel Servet Fellowship from the Instituto de Salud Carlos III, Spain. They are also grateful to Lynsey Block for secretarial assistance.