Targeting Carcinoembryonic Antigen with DNA Vaccination: On-Target Adverse Events Link with Immunologic and Clinical Outcomes

The goal of harnessing the immune system and provoking antitumor immunity is now beginning to be realized, and recent regulatory approval of the first immunotherapeutics for prostate cancer (1) and melanoma (2, 3) heralds a new and exciting phase in the field. A number of vaccines appear to confer clinical benefit in early-phase studies (4–7), with many agents now in phase III testing. Checkpoint blockade is taking its place in the mainstream management of cancer, but after enormous excitement, it is also becoming clear that only a subgroup of patients benefit (8). If immunologic visibility of the tumor is reflected in the presence of tumor-infiltrating lymphocytes and links to survival (9, 10), activating preexisting antigen-specific T-cell responses could be an effective approach to improving clinical outcomes. Exploiting antigen-specific responses to tumor antigens that are shared between different patients and cancer types is one option, provided that antigen remains visible to immune attack and is not eliminated by selective pressures within the tumor.

Carcinoembryonic antigen (CEA) is an immunoglobulin-like molecule involved in cell adhesion (11). In healthy adults, CEA is primarily found at low levels on the surface of colonic mucosa. Following transformation, CEA is expressed in 70% of all human cancers (12), making it an attractive target for immunotherapy (13). Different immunotherapeutic strategies targeting CEA have been tested clinically; vaccines using peptides (14, 15), dendritic cells (DC; refs. 16–18), plasmid (19, 20) or viral vector delivery (21–26), and engineered T cells (27) have been assessed in phase I/II clinical trials and demonstrate varying clinical efficacy.

Our approach has been to develop DNA vaccines to stimulate antitumor immunity. To overcome the weak immunogenicity of tumor-associated antigens, we have linked an MHC class I–restricted peptide to a potently immunogenic domain (DOM) from fragment C (FrC) of tetanus toxin (28). By exploiting the nontolerized CD4⁺ T-cell repertoire for tetanus, DOM stimulates CD8⁺ T-cell immune responses via linked T-cell help with induction of cytotoxic CD8⁺ T cells against the vaccine-encoded MHC class I–restricted peptide in wild-type and HLA-A2 transgenic mice (28). The fusion vaccines are able to stimulate CD8⁺ T cells in tolerant mice, induce T-cell memory, and protect against tumor challenge, with early clinical data supporting the translation into humans (29).

In this exploratory phase I/II study, we use DNA fusion vaccination to target the HLA-A*0201–restricted CAP-1 peptide from CEA (30) in patients with CEA⁺ malignancies.

In this multicenter, nonrandomized, two-arm phase I/II study we examined the safety, immunogenicity, and clinical effects of our DNA vaccine encoding the DOM–CAP-1 fusion gene (28) in patients with CEA-expressing cancers (EudraCT number: 2004-001932-21; Cancer Research UK protocol PH1/099). Regulatory and ethical approval was obtained from the UK Medicines and Healthcare Regulatory Authority, the national ethics committee responsible for the conduct of studies with genetic vaccines (GTAC), and the local research ethics committee at each center. Written informed consent was provided by each patient prior to any study-specific interventions. The vaccine was produced to GMP by the NHS Blood and Transplant, Clinical Biotechnology Centre, University of Bristol (Bristol, United Kingdom). The study was sponsored by Cancer Research UK.

Eligibility criteria included a CEA-expressing malignancy confirmed by central immunohistochemical review of archived tumor; ≥18 years of age; HLA-A*0201 positivity; World Health Organization (WHO) performance status of ≤1; completion of oncological treatments >8 weeks prior to consent; lymphocyte count of >1.0 × 10⁹/L; platelet count of >50 × 10⁹/L, with normal clotting; no clinical or immunologic signs of autoimmunity, specifically inflammatory bowel disease; absence of systemic immunosuppressants, such as steroids; and adequate contraception in place. Patients with radiologically measurable metastatic disease after exhaustion of standard treatment options and in progression were entered into arm-I. Patients in radiological complete remission after surgical clearance with or without adjuvant chemotherapy, but at a high risk of recurrence (estimated at recruitment to be ≥50% at 5 years), were recruited to arm-II. Patients receiving ≥3 or ≥1 dose(s) of vaccine were evaluable for immunogenicity and toxicity, respectively.

DNA vaccine (1 mg/dose) was delivered intramuscularly at weeks 0, 1, 2, 4, 8, and 12 (29, 31); on-study evaluation was up to week 64 or clinical/radiological disease. At WHO grade ≥3 toxicity, vaccination ceased; at grade 2 toxicity, further vaccination was delayed until recovery to ≤1 (CTC v2.0). Imaging by CT was at baseline, weeks 16, 24 (arm-I only), and 64/off-study visit, or as clinically required, with assessment using RECIST v1. Regular blood collections for full blood count, biochemistry, serum CEA, and immunologic evaluation were performed (baseline, weeks 1, 2, 4, 8, 12, 16, 20, 24, 28, 32, 40, 52, and 64). Autoimmune profiles were measured at 3 monthly intervals. Clinical assessment and co-medications were documented at each visit. Time to progression (TTP) and overall survival (OS) were recorded from date of consent to date of event or censor (December 31, 2012).

CAP-1 peptide presentation was assessed on HLA-A*0201⁺ renal control, normal colonic, and malignant colorectal tissue (6). HLA peptide pools from shock-frozen tissue samples were immunoprecipitated with BB7.2 antibody (32). Peptides were analyzed by on-line nano liquid chromatography: nanoAcquity UPLC system with 25-cm capillary columns (internal diameter 75 μm) filled with 1.7 μm BEH130 C18 particles (Waters). The acetonitrile (in 0.1% formic acid) gradient consisted of 1%–13% for 10 minutes, 13%–26.5% for 140 minutes, and 26.5%–34.5% for 40 minutes, with a 300 nL/minute flow rate. Analysis was on an LTQ-Orbitrap Velos with a nanoelectrospray ion source (Thermo Fisher Scientific). CAP-1 peptide detection was performed by data independent acquisition in CID mode, including an isotope (¹³C₆/¹⁵N)–labeled CAP-1 peptide. Resolution was 30,000 for the Orbitrap; fragment spectra were recorded at low resolution in the LTQ. Data analysis was performed using Skyline (33).

Immunologic responses were assessed using validated assays (34–36); T-cell responses were reported to MIATA guidelines (http://www.miataproject.org; Supplementary Table S1; ref. 37).

Antivaccine humoral responses were measured in the serum of patients (triplicate) by separate ELISAs against recombinant DOM and FrC protein (31, 34). A response to the DOM helper sequence was assigned when a significant (P < 0.05) increase over prevaccination baseline (week 0) was detected at single [+] or multiple [++] time points; an increase of ≥ 2-fold over baseline was not mandated.

Antivaccine cellular responses were determined by ex vivo IFNγ ELISPOT assay (34–36); PBMCs (4 × 10⁵) were stimulated (triplicate) with recombinant FrC protein (20 μg/mL; ref. 34), CAP-1 peptide (YLSGANLNL, 10 μg/mL; Protein Peptide Research UK) or control for 20 hours at 37°C and 5% CO₂ (Supplementary Table S1). PBMCs were also cultured in vitro for 8 days with CAP-1 peptide (10 μg/mL) or control in the presence of IL2 (20 IU/mL; day 3 and 6); IFNγ secretion was measured by ELISPOT following restimulation with peptide (10 μg/mL; Supplementary Table S1; refs. 28, 38).

CAP-1–specific CD8⁺ T cells were identified using CAP-1 HLA-A*0201 tetramer (5 μg/mL; Protein Core Facility, University of Southampton, Southampton, United Kingdom) plus LIVE/DEAD Aqua dye (Invitrogen) and anti-CD3/CD8/CD4 staining at selected time points (Supplementary Table S1).

CAP-1–specific T cells from patient #108 were expanded in vitro for 8 days in the presence of CAP-1 peptide (5 μg/mL), IL2 (10 IU/mL), and IL7 (5 μg/L), stained with specific tetramer or control (CMV-pp65/A2; Beckman Coulter), and single cell sorted into a 96-well V-bottom plate containing NIH-3T3 carrier cells on a BD FACSAria Flow Cytometer (BD Biosciences; ref. 39). RNA was isolated from single T cells, reverse transcribed to cDNA, and amplified by SMART-based 5′-RACE PCR. Full-length T-cell receptor (TCR) V(D)J regions were amplified with degenerate primers covering all functional Vα and Vβ genes in combination with Cα- and Cβ-specific primers (39) and cloned into pST1-TCRα/β-2βgUTR-A(120) for in vitro transcription.

Genomic DNA was extracted from 10-μm paraffin sections after dewaxing and proteinase K digestion using a Qiagen DNA Extraction Kit (Qiagen). DNA (100 ng) was amplified by PCR (duplicate) with TCR Vβ29-1 CDR3-specific primers (forward: 5′-CTGCTCCTTCTCCTGGGACTAGGCT-3′, reverse: 5′-TGGAGGGGTAAACCGTCCCTGTCC-3′) for 37 cycles at 94°C/30 seconds, 64°C/30 seconds, and 72°C/60 seconds; amplification products were TA-cloned and sequenced.

This study was conducted in accordance with the principles expressed in the 1964 Declaration of Helsinki and reviewed and approved by the Medicines and Healthcare Regulatory Authority, the Gene Therapy Advisory Committee, and the Local Research Ethics Committee. Written informed consent was obtained from all individual participants included in the study.

Statistical analyses were performed with GraphPad Prism software, v6.0a (GraphPad Software, Inc.) and IBM SPSS Statistics, v22.0 (IBM Corp). Univariate and multivariate analyses were performed in SPSS. Significance was assessed by two-sided, nonparametric Wilcoxon signed-rank test or Mann–Whitney test. Spearman rank correlation coefficient was used to test the association between two ranked variables. Distributions of time to event data were estimated using the Kaplan–Meier method and compared using Mantel–Cox log-rank testing and Cox regression analysis.

Twenty-seven patients were recruited and evaluable for toxicity and immunologic responses; 15 patients had advanced disease (arm-I), and 12 patients were in radiological complete remission (arm-II; Table 1). Ten patients completed the study; 17 patients progressed prior to week 64 (Fig. 1A). The vaccine was well tolerated; toxicities were mainly grade 1 and resolved without intervention. A high frequency of diarrhea was observed with 23 episodes reported for 13 patients (48%); of these, 6 patients were in arm-I and 7 patients in arm-II. Diarrhea was reported in patients with bowel, lung, and breast cancer (Table 1), most frequently during the initial 8 weeks on study (Fig. 1B). Onset following nearest vaccination [median 12 (1–194) days] and duration [4.5 (1–95) days] were variable.

No RECIST responses were observed; patient #101 had a 50% decrease at a single metastatic site at week 16. All patients with advanced disease progressed and died during the study/follow-up period, with median TTP 119 (56–392) days and median OS 391 (62–1,058) days (Fig. 1A). Outcome was significantly better in patients without measurable disease at trial entry (arm-II, log-rank P < 0.001 for progression and survival); 4 patients progressed [median TTP 501 (235–1441) days] and 3 died [median follow-up 1,696 (766–2,393) days for all patients].

Serum CEA was detectable in 22 (81%) patients at baseline, including all patients with advanced disease (arm-I) and 7 of 12 patients without measureable disease (arm-II). Median CEA was significantly greater in the serum of patients from arm-I at baseline [36.9 (0.8–2,385) μg/L] compared with those from arm-II [1.6 (1–53.3) μg/L, P = 0.021; Table 1]. Baseline CEA was elevated (>9 μg/L) in 12 patients and became elevated in 4 further patients while on study. Transient decreases in CEA were observed in 11 of 22 (50%) patients from both study arms (Table 2).

Humoral and cellular responses to DOM from FrC demonstrated successful vaccine delivery with an overall response rate of 60% (arm-I) and 100% (arm-II; Table 1). In patients with advanced disease, DOM responders had a better outcome than nonresponders (log-rank P = 0.012). Six patients that did not respond to the DOM helper sequence had significantly higher baseline CEA [median 371.5 (7.2–2,385) μg/L] than DOM responders [median 25.6 (0.8–426) μg/L, P = 0.026; Table 1]. DOM-specific antibody responses peaked between weeks 8 to 16 (Fig. 2A), while cellular response kinetics were variable (Fig. 2B). Mean anti-DOM humoral responses were significantly higher in patients from arm-II (P = 0.020; Fig. 2C). The difference in magnitude of cellular anti-DOM responses between study arms was not significant (Fig. 2D).

We confirmed CAP-1 presentation by mass spectrometry on frozen samples from HLA-A*0201⁺ normal colon (n = 7), primary colorectal tumor (n = 9), and metastatic colorectal tumor (n = 4), but not on normal kidney tissues (n = 4, negative control; Fig. 3A–C).

Overall, CAP-1–specific T-cell responses were detected in 10 (37%) patients postvaccination, including 3 of 15 (20%) patients with advanced disease (arm-I) and 7 of 12 (58%) patients without measureable disease at trial entry (arm-II). Low levels of IFNγ-secreting cells (<50 spots forming cells/10⁶ PBMCs) were observed by ex vivo ELISPOT in 1 of 27 patients (#108: weeks 12, 20, 24, 28, and 32), increasing to 9 of 27 (33%) patients after in vitro culture (Table 1). Responses were of varying magnitude and were observed more frequently in patients from arm-II; 3 of 15 (20%) patients with advanced disease generated CAP-1–specific T-cell responses detectable by postculture ELISPOT compared with 6 of 12 (50%) patients without measureable disease. Ex vivo tetramer staining confirmed CAP-1–specific T cells postvaccination in patient #108 (Supplementary Fig. S1A) as well as patients #114, #203, and #213; CAP-1–specific responses could also be detected postculture by ELISPOT and/or tetramer staining in all cases (Table 1). CAP-1–specific T cells expanded (6.5- to 35.6-fold) in vitro with a good correlation between cultured ELISPOT and ex vivo tetramer data (Spearman rank r = 0.735, P = 0.003; Supplementary Fig. S1B). For patient #202, CAP-1 tetramer⁺ T cells were evident following culture but could not be detected directly ex vivo.

Of the 18 patients that progressed and died during the study/follow-up period, 5 patients generated a CAP-1–specific response. We observed a median OS for CAP-1 responders of 730 (181–2,035) days compared with 528 (62–1,479) days in nonresponders (Supplementary Fig. S2); this difference did not reach significance.

Globally, vaccine-specific immune responses were significantly more frequent in patients without measureable disease (arm-II) compared with patients with advanced disease (arm-I; P = 0.037; Supplementary Fig. S3). This was also true for CAP-1–specific immune responses (P = 0.049; Supplementary Fig. S3).

We rescued paired TCRα/β chains from 24 single CAP-1 tetramer⁺ CD8⁺ T cells from patient #108 after antigen-specific in vitro expansion (Supplementary Fig. S4A). Eleven TCRs belonged to 2 different clonotypes (Supplementary Table S2); Vα12-2−J28−C paired with Vβ29-1−D1−J1-6−C1 represented the dominant clonotype (n = 10). CAP-1 specificity was confirmed by ELISPOT following transfection of primary CD8⁺ T cells with full-length TCRα/β–encoding in vitro–transcribed mRNA (Supplementary Fig. S4B). Insufficient material was available for testing of expanded, patient-derived CAP-1–specific T cells against CEA⁺ HLA-A*0201⁺ tumor cells. The CDR3β sequence of the dominant TCR clonotype was detectable by RT-PCR in the blood at CAP-1 tetramer⁺ time points, as well as from genomic DNA isolated from paraffin-embedded primary tissue resected 18 months prior to vaccination (data not shown).

Of the 18 patients who progressed and died during the study/follow-up period, 7 patients reported diarrhea. We observed a longer median OS for patients experiencing diarrhea [766 (149–1,058) days] compared with those that did not [391 (62–2,035) days; Supplementary Fig. S5]; this difference did not reach significance. For patients with advanced disease, median OS was 809 (149–1,058) days if the patient experienced diarrhea and 272 (62–1,016) days if diarrhea was not reported (Fig. 3D). In arm-II, 1 patient with diarrhea and 2 without progressed and died (Supplementary Fig. S6A). Multivariate analyses did not add any further information.

The median baseline CEA level in patients experiencing diarrhea [12 (1–53.3) μg/L for 11/13 patients with detectable CEA at baseline] was not significantly different to that of patients who did not report diarrhea [62 (0.8–2,385) μg/L for 11/14 patients; Table 1]. Diarrhea was associated with transient drops in CEA to below baseline levels (Table 2); decreases in CEA were significantly more frequent in patients with diarrhea (26 events, n = 10) compared with patients without (1 events, n = 1, P < 0.001) and identified a group of patients with advanced disease (arm-I) with a better OS [log-rank P = 0.008; HR = 0.14; 95% CI, 0.03–0.71; P = 0.017; Fig. 3E). Three patients who were in remission at trial entry (arm-II) subsequently progressed and died during the follow-up period, and of these, 1 patient showed transient drops in CEA (Supplementary Fig. S6B). Diarrhea and transient decreases in CEA were significantly associated (Spearman rank r = 0.710, P < 0.001; r = 0.722, P = 0.002, arm-I vs. r = 0.714, P = 0.009, arm-II).

All 13 patients who reported diarrhea responded immunologically. Diarrhea was observed in 13 of 22 (59%) patients with antivaccine responses, including 6 patients from arm-I and 7 from arm-II (Table 1). Anti-DOM cellular responses were of significantly greater magnitude (P < 0.001; Fig. 4A) and frequency (P = 0.004; Fig. 4B and C) in patients with diarrhea; no effect on humoral responses was observed.

We conducted an exploratory phase I/II trial to test the safety and efficacy of an anti-CEA DNA fusion vaccine encoding DOM-CAP-1 in patients with CEA-expressing cancers. The vaccine was safe and well tolerated. Mild bowel-related toxicity was observed in 48% of patients and was associated with better clinical and immunologic outcomes. Diarrhea was significantly associated with transient drops in CEA and identified a group of patients with advanced disease with an 86% reduction in risk of death; decreases in CEA also appeared to link to improved survival for patients without measureable disease at trial entry, but with only three events the difference did not reach significance.

TCR rescue from CAP-1–specific T cells showed that CAP-1 can be recognized in the tissue spontaneously and that CAP-1–specific T cells can be expanded by DNA vaccination and detected in the circulation. Although we do not directly show effector function in the bowel mucosa, the demonstration of CAP-1 peptide presentation in both benign and malignant tissue makes it plausible that antigen-specific tissue recognition underpins our observations of both diarrhea and CEA drops as vaccine-mediated, on-target effects. In support, when targeting prostate-specific membrane antigen in 30 patients using an otherwise identical vaccine design, no diarrhea was reported (29, 31). As gastrointestinal effects were not restricted to patients with colorectal cancer, it seems unlikely that diarrhea is a reflection of previous surgical or cytotoxic manipulation of the bowel.

Dose-limiting gastrointestinal autoimmune pathology has been reported previously in association with effective CEA-targeted immunotherapy in preclinical (40) and human (27, 41) studies, demonstrating that CEA must be targeted with caution. Other trials of CEA-targeted vaccines have reported gastrointestinal side effects, however, without relating diarrhea events to clinical benefit (14, 16, 19), or interpretation was confounded by coadministration of chemotherapy (15). Conversely, some CEA vaccine trials do not report bowel-related adverse events (17, 18, 20–24), and this may reflect differences in vaccine potency.

Our vaccine was developed to exploit T-cell help from an undeleted repertoire against the foreign antigen DOM from tetanus (28, 42) to overcome tolerance to tumor-associated antigens; this can be achieved preclinically (28). We demonstrate here that if tolerance against CAP-1 exists in patients, it can be overcome relatively easily by vaccination. Conditioning of the vaccine site through tetanus-derived T-cell help has recently been assessed in a study in patients with glioblastoma (43) with previous exposure to tetanus where it was associated with improved outcomes. We are using linked T-cell help, which may be more efficient than separate antigens. Of note, in the glioblastoma study, the intended vaccine target pp65 from CMV is a xenogeneic sequence, and the bar for induction of anti-pp65 T-cell responses may be lower than for CEA (43).

Our data further demonstrate that clinical context affects vaccination outcome. Where previous studies were conducted in patients with advanced disease, tumor load may have adversely impacted measurable effects through systemic and local immunosuppression. This also implies that adverse immune effects, which must be linked to an intact immune effector function, could be missed (44–47). We demonstrate CAP-1- and DOM-specific responses of significantly lower frequency and magnitude in arm-I, the latter offering a measurable insight into the global loss of immune competence in patients with advanced epithelial cancers. It is uncertain whether this would be further confounded by ongoing or previous cytotoxic treatment (48). The observed immunologic differences were not due to differential presence of regulatory T cells in the blood, although other suppressive populations were not assessed. Functional assessment also helps to evaluate the immunologic context in which a particular vaccine approach is tested, and this should be considered in future vaccine studies.

Our vaccine was able to evoke CAP-1–specific T-cell responses in the blood of 37% of patients, with a good correlation between ex vivo CAP-1 tetramer⁺ CD8⁺ T cells and CAP-1–specific IFNγ-producing cells in ELISPOT postculture, linking specificity to functionality. Our data are consistent with previous studies, where IFNγ secretion by CAP-1–specific T cells in vitro correlated with clinical benefit (16, 20, 21, 24). A recent study (49) argued that CAP-1 is not efficiently processed and presented by CEA-expressing tumor cells. We demonstrate CAP-1 presentation on HLA-A*0201⁺ MHC class I in both CEA-expressing tumors and normal bowel mucosa, supporting CAP-1 as a suitable target for immunotherapy. Our data further suggest that the assessment of diarrhea offers a more sensitive predictor of outcome than measuring circulating CAP-1–specific immune cells. Diarrhea was more frequent (48%) than the presence of CAP-1–specific T cells in the peripheral blood and may result from the homing of CAP-1–specific T cells to the site of antigen expression. It is tempting to speculate that the circulating immune cells are not the same as those with effector function in the tissue, as has been observed in patients with melanoma (50).

In summary, we demonstrate a link between tumor load and the ability to respond to vaccine-specific immune stimulation, that is, immune competence. Where vaccination induced CAP-1 responses, we observed longer TTP and OS, although this did not reach significance. More than half of the patients experienced diarrhea. This in turn linked to immunologic and clinical outcomes, in particular, to transient decreases in CEA that allowed the identification of a group of patients with measurable disease who fared much better. Our data suggest a breaking of peripheral tolerance to the self-antigen CAP-1 by expansion of antigen-specific CD8⁺ T cells and further argue that these T cells can home to the tissue where the antigen is located and visible to T-cell attack. Furthermore, our data argue that shared tumor-associated antigens can contribute to successful immune attack and should continue to be considered as important targets for immunotherapy. In this particular clinical context, diarrhea can be used as a biomarker predicting clinical and immunologic efficacy of CEA-targeting therapies and offers an extra dimension to the assessment of circulating immune cells. A larger, ideally randomized, study is required to confirm our observations. Combination studies to increase vaccine potency, for example by using anti-PD1 antibodies, are in development.

T. Weinschenk has ownership interests (including patents) in Immatics Biotechnologies. P. Simon is listed as a co-inventor on a patent on a method for providing antigen-specific T-cell receptors that is owned by BioNTech Cell & Gene GmbH and TRON, Translational Oncology at the University Medical Center of the Johannes Gutenberg University. U. Sahin is an employee of and has ownership interests (including patents) in BioNTech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.J. McCann, A. King, D.I. Jodrell, F.K. Stevenson, C.H. Ottensmeier

Development of methodology: K.J. McCann, A. Mander, A. Cazaly, L. Chudley, J. Stasakova, S.M. Thirdborough, A. King, S. Halford, U. Sahin, F.K. Stevenson, C.H. Ottensmeier

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.J. McCann, A. Mander, A. Cazaly, L. Chudley, J. Stasakova, S.M. Thirdborough, A. King, E. Buxton, S. Halford, A. Bateman, A. Oapos;Callaghan, S. Clive, A. Anthoney, D.I. Jodrell, T. Weinschenk, P. Simon, C.H. Ottensmeier

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.J. McCann, A. Mander, A. Cazaly, A. King, S. Halford, T. Weinschenk, P. Simon, G.J. Thomas, F.K. Stevenson, C.H. Ottensmeier

Writing, review, and/or revision of the manuscript: K.J. McCann, A. Mander, A. Cazaly, A. King, E. Buxton, C. Edwards, S. Halford, S. Clive, A. Anthoney, D.I. Jodrell, T. Weinschenk, U. Sahin, G.J. Thomas, F.K. Stevenson, C.H. Ottensmeier

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Mander, A. Cazaly, E. Buxton, C. Edwards, C.H. Ottensmeier

Study supervision: A. King, C. Edwards, S. Halford, C.H. Ottensmeier

Other (manufacturing of the plasmids for therapy and regulatory support): P. Lloyd-Evans

The authors thank Scott Harris of the Medical Statistics Department, University of Southampton, for advice on the statistical evaluation and interpretation of this study.

This study was supported by a clinical trials grant from Cancer Research UK and in part by the NIHR Experimental Cancer Medicine Centre, Southampton, and by the Southampton Welcome Trust Clinical Research Facility.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.