DNA fusion vaccines enter the clinic

Plasmid backbone

Bacterial DNA is known to differ from mammalian DNA in the level and methylation status of CpG-containing motifs. The development of CpG-based oligonucleotides as adjuvants, in most cases stabilized as phosphorothioate analogs, led to the concept that the adjuvant activity of plasmid backbones also relied on these sequences and that their effect was mediated via TLR9 [1]. This conclusion was modified by the observation that TLR9^−/− mice responded well to DNA vaccines [2]. It is now clear that plasmid DNA activates multiple pathways via endosomal and cytosolic sensors [3] with the outcome being stimulation of innate immunity. The pathways involved converge on induction of interferon α/β, together with activation of the inflammasome AIM2, generating IL-1β and IL-18 [4]. Clearly, double stranded DNA is a major activator of innate immunity, and DNA vaccines have a built-in adjuvant.

Tumor antigens

One of the frustrations of antibody therapy is the relative lack of target proteins expressed on the tumor cell surface and therefore susceptible to antibody attack. The majority of known target molecules are intracellular and are only displayed as MHC Class I-associated peptides. The main goal therefore has to be to induce peptide-specific CD8+ T cells able to recognize and kill tumor cells. Since ~40% of the Caucasian population carries the HLA-A0201* haplotype, the focus has been on finding peptides which bind to this, and algorithms have been developed for identification of candidate sequences within tumor antigens [5]. The other point, sometimes neglected, is that the peptide must be processed and presented by the tumor cell. Many clinical trials of peptides as vaccines have been carried out, with limited success, since responses, if induced, tended to be transient [6]. It became clear that exogenous peptides alone fail to activate effective levels of CD8+ T cells and are poor at inducing memory T cells.

T-cell help

Pre-clinical studies revealed one reason for failure was a requirement for T-cell help [7]. This can be provided either by co-injecting a CD4+ T-cell inducing peptide such as PADRE or by using long peptides, which have the advantage of both potentially activating T-cell help and also being preferentially processed by dendritic cells [8]. Our approach was always to provide high levels of T-cell help, achieved in most of our designs by fusing a gene derived from tetanus toxin (TT) to the candidate peptide sequence [9]. This is highly effective but is not the only microbial sequence which can be used. We have also explored plant viral coat proteins in this setting [10] and, although not yet in clinical trials, in our view, these have considerable potential. The three essential components of our DNA fusion vaccines are shown in Fig. 1.

For induction of CD8+ cytolytic T cells, we engineered the Fragment C of TT to produce a single domain and fused the tumor peptide sequence to the C-terminus, thereby promoting its processing and presentation (Fig. 2). We selected the domain sequence on the basis of it contributing no identifiable HLA-0201*-binding peptides that could compete with the tumor peptide [11]. Immunodominance is a real concern for tumor vaccines, where target peptides may be weak. It appears that competitor peptides able to induce high levels of CD8+ T cells can lead to killing of antigen-presenting cells before a weaker peptide can activate a response [12]. Careful selection of molecules aimed to provide T-cell help is therefore required.

Our main design named p.DOM-epitope (Fig. 2) has now been tested in ~20 models [9] with consistent success in inducing durable CD8+ T-cell responses able to suppress tumor. When compared with exogenous peptide (+PADRE and adjuvant), DNA delivery was superior [13]. The p.DOM-epitope design is also generally superior to the full length sequence ± added DOM [14] (and unpublished data). The vaccine design is also effective in inducing CD8+ T-cell responses in tolerant settings [15]. We have considered the question of whether there might be a need to induce tumor-specific CD4+ T cells, either for support for long-term CD8+ T-cell responses or to assist access to a tumor site. The pre-clinical models do not indicate that this is required, but we are developing vaccines to induce CD4+ T-cell responses for further testing. One possibility is that having a vaccine that induces a CD4+ T-cell response might have the undesirable outcome of activating regulatory T cells, and this will be investigated.

Peptide selection and testing

Algorithms and testing of spontaneous T-cell responses in patients have identified many candidate peptides with potential for immune targeting [5]. We have selected a range of these and placed them in the p.DOM-epitope format. To test immunogenicity, we have used the invaluable HHD mice which carry a transgene encoding the human HLA-A2 restriction element [9]. While assessment of responding CD8+ T-cell responses is straightforward, it is more difficult to measure the ability of such T cells to kill human targets. The reason for this is that the mouse CD8 molecule does not interact with the human HLA molecule, and T cells have to be of very high affinity to overcome the lack of this interaction. Our approach has been to transduce the HHD molecule into the human target cell, and using this approach, we have been able to demonstrate effective killing.

Our initial focus has been on a protein expressed in prostate cancer, prostate-specific membrane antigen (PSMA), using the target peptide PSMA27 [16]. A second study has been of the Wilms tumor antigen (WT1), a protein expressed by several tumors, including chronic myeloid leukemia, using two target peptides, WT1-37 and WT1-126 [13]. In all cases, CD8+ T cells were readily induced in HHD mice and these were able to kill transduced target cells [17]. For each antigen, the p.DOM-epitope design was superior to the full length design.

Vaccine delivery by electroporation

Our pre-clinical data had shown that a critical influence on DNA vaccine performance is the volume injected [18]. In mice, 50 μl is the standard volume for intramuscular injection, and this cannot be reduced to <25 μl without a dramatic fall in response. Concentration of DNA is a less important variable. It appears that the disruption of muscle fibers by the hydrostatic pressure may be the key to either more efficient transfection or activation of inflammation. If calculated on a weight for weight basis, scaling up for patients would require >100 ml to be injected, clearly not feasible, but this requirement can be overcome by electroporation. Electroporation improves antibody and CD4+ T-cell responses [19]. We confirmed this and showed that a particularly effective combination was to prime with naked DNA and boost with electroporation [18]. Several groups have developed clinically acceptable electroporators, and we have worked with Inovio to test this strategy in the clinic.

Clinical trial of patients with prostate cancer

Building on our pre-clinical data, we began testing the pDOM-epitope vaccine concept in patients with solid tumors. We targeted the HLA A0201*-restricted peptide PSMA27 in a phase I/II study, recruiting patients who had progressed after radical treatment for prostate cancer [20]. Patients were identified by rising PSA levels in the absence of radiological evidence of metastatic disease. The trial recruited in two parallel arms, in which 3× monthly DNA injections were given as an intramuscular injection, with or without electroporation (EP). In both arms, escalating doses of DNA were tested in cohorts of 5 patients: 800, 1,600, and 3,200 μg/dose in the DNA alone arm, 400, 800, and 1,600 μg in the DNA/EP arm. At 6 and 12 months, boosting was given and cross-over was permitted. The vaccine was very well tolerated, and toxicity was mainly limited to local injection site reactions [20]. Electroporation also was well tolerated in our patients and with 150 applications no safety concerns were identified, in terms of either subjective tolerance or local or systemic toxicity. In particular, no significant changes in LDH or creatine kinase were observed either immediately after, at 3 or 5 days post-injection, or at repeated delivery.

We were able to assess three aspects of the immune response: antibody and CD4+ T-cell responses against the DOM protein, which indicate performance of the vaccine in each patient, and tumor-relevant CD8+ T-cell responses against the target PSMA27 peptide. Overall, humoral IgG responses to the DOM protein were substantially boosted by EP [20]. In patients who began vaccination with DNA alone, boosting with EP was able to rescue some of the humoral responses, but not to the same level as for patients who were given DNA with EP on each occasion. In total, 21/30 (70%) of patients developed significant humoral responses. Rather dramatically, all but one of the patients (30/31 (97%) evaluable patients) mounted significant increases in T-cell responses to DOM, readily detectable ex vivo. There was a suggestion for more rapid responses post electroporation, but no clear effect of either dose or delivery strategy was identified. Ideally, a large-scale clinical trial of the effects of EP on immune responses and on clinical outcome should be carried out, and this will be important for prophylactic vaccines. For cancer, the observation of stimulation of antibody and CD4+ T-cell responses, with a possible increase in the rate of response, may be sufficient to include EP in DNA vaccine delivery.

We considered that CD8+ T-cell responses at the time points used would most likely be of the central memory subset therefore all measurements were made following 1 week of restimulation ex vivo. Overall 17/31 (55%) patients showed significant and specific responses to PSMA27 peptide in terms of IFNγ production by ELISPOT. Similar to CD4+ T-cell responses, CD8+ T-cell responses appeared more quickly in the EP arm, though overall the frequency and level of CD8 T-cell responses were comparable between delivery strategies by week 16. Dose appeared again not to affect immune responses.

In terms of clinical effects, serum PSA levels proved to be erratic and less informative than expected, perhaps in line with other studies [21], [22]. The sample size is too small to assess effects on survival; however, a trend of delayed time to next treatment is emerging, compared with a cohort of HLA-A0201-negative patients who fulfilled other recruitment criteria but were not vaccinated. Our plan is to assess the effect in a randomized study, with increased numbers of patients.

A second study with the DOM-epitope design was conducted in patients with carcinoembryonic antigen (CEA)-expressing malignancies [23]. The peptide target was the HLA A0201*- restricted epitope of the CEA molecule, CAP-1 [24]. The study allowed recruitment based on immunohistochemical expression of CEA in the tissue, irrespective of tumor type. Twenty-seven patients were recruited, 15 with end stage disease, 12 without measurable disease. For the latter group, an estimated risk of recurrence at 5 years of about 50% was the key entry criterion. Six doses of vaccine were given at 1, 2, 4, 6, 8, and 12 weeks at a single dose of 1 mg.

Overall, the volume of disease affected induction of vaccine-specific immune responses against the DOM protein: 92% of patients without measurable disease developed cellular immune responses, but only 53% of those with end stage disease. CAP-1 responses were detected in about half of the overall responders, consistent with the data from the prostate study. In contrast to the prostate study, ~50% of patients developed an unexpectedly frequent but mild and self limiting side effects, mainly affecting the bowel. Diarrhea was associated with the frequency and durability of T-cell responses to vaccine-delivered DOM protein. Minor decreases in serum CEA levels were observed only in patients with diarrhea and, importantly, survival of patients with end stage disease with diarrhea was significantly prolonged compared with those without.

CD8+ T-cell responses to CAP-1 in the blood did not correlate either with immune responses to DOM or with outcome. In contrast, pre-clinical models have shown parallelism between CD4+ T-cell responses against DOM epitopes and CD8+ T-cell responses against tumor epitopes, even in tolerant settings (14). The apparent lack of this linkage in a few patients is likely to be due to the restriction of our assays to cells in the blood, with no information on T cells in the spleen or in tissue sites. No biopsy samples were available from the patients during this study. Further studies are ongoing to explore this observation, specifically of the bowel homing properties of the CAP-1-specific T cells. Detailed clinical data are being prepared for publication.

In summary, vaccination in CEA-expressing patients, with a range of organ involvement appears to have generated an unexpected diarrheal response, possibly due to autoimmunity against gut-associated CEA. A similar outcome was seen in a different context where vaccination was combined with removal of regulatory T cells [25] or where CEA specific T cells are given by adoptive transfer [26]. Further studies are obviously required but this is an intriguing indication that, at least for this antigen, transient autoimmunity might be a sign of an anti-tumor effect.