Introduction
DNA vaccines are emerging as a promising vaccine modality for cancer. The DNA vaccine concept is simple: a bacterial plasmid DNA is used which contains the target antigen under the control of a strong eukaryotic promoter enabling expression in eukaryotic cells. Following injection of the DNA vaccine, the encoded antigen is expressed directly in the muscle or skin without the need to employ other expression systems [
1]. The second critical component of DNA vaccines is the ability to stimulate innate immunity through recognition of the DNA itself. Multiple sensors that recognise DNA exist in the cytoplasm in addition to the endosomal DNA sensor TLR9 [
2,
3]. While many types of DNA, mainly derived from viral sequences, can stimulate cytoplasmic DNA sensors, TLR9 recognises specific immunostimulatory sequences (ISS); hypomethylated CpG motifs [
3,
4]. These are frequently found in bacterial DNA but not commonly in vertebrate DNA and are often located within antibiotic resistance genes. The result of stimulation through these sensors is induction of inflammatory cytokines and type I interferons which are responsible for the self-adjuvanting effect of DNA vaccines [
5]. Despite the relative simplicity of DNA vaccines they are able to induce both the durable humoral immunity and cytotoxic T cells desired for cancer attack [
1].
Although DNA vaccines performed well in mice, the conditions of injection in patients differed and led to limited immunogenicity. Several approaches to improve immunogenicity of DNA vaccines have been developed [
6,
7]. An important strategy was to combine injection with in vivo electroporation (EP) which significantly improved immunogenicity by increasing cell transfection, leading to improved antigen expression and adjuvancy [
7,
8]. Robust immunological responses after DNA vaccine plus EP have been demonstrated in patients, with indication of clinical efficacy [
9,
10]. A recent phase IIb clinical trial using a DNA vaccine encoding HPV oncogenes E6 and E7 combined with EP has demonstrated regression of dysplastic processes in patients with HPV-driven cervical dysplasia [
11].
Traditionally DNA vaccines relied on plasmid (PL) DNA prepared using
Escherichia coli with an antibiotic resistance gene for selection. A subsequent multiple step purification is required followed by endotoxin removal if intended for clinical use. Recently a bacteria-free manufacturing platform has been developed to allow rapid production of novel doggybone™ DNA (dbDNA™), which is suitable for use as a DNA vaccine. The method involves an enzymatic amplification in vitro using two enzymes. Phi29 DNA polymerase is employed to rapidly amplify template DNA into concatamers and then the protelomerase TelN from bacteriophage N15 is used to cut and join the DNA concatamers into individual closed linear dbDNA™ [
12,
13]. The resulting DNA is fully functional, highly stable and contains only the minimal sequences required including the antigenic sequence, a promoter and a poly A tail but lacks bacterial sequences such as the antibiotic resistance gene. Although this is advantageous for patients’ safety the question of immunogenicity arises since the innate immune recognition could be compromised due to decreased ISS frequency. This is especially relevant for cancer antigens delivered through DNA vaccines as these are of non-bacterial origin and hence often lack ISS.
In this study, we compared the immunogenicity of a dbDNA™ vaccine (DB) targeting HPV16 derived E6 and E7 oncogenes to conventional PL delivery and look into the potential pathways involved in innate sensing of this novel DNA vaccine.
Methods
Preparation of DB and PL DNA vaccines
The HPV16 E6 and E7 sequences containing mutations that impair oncogenic potential were assembled as previously [
14] and the E6E7 fusion was cloned into the proTLx™ based PL. The proTLx™ PL consisted of the CMV promoter plus enhancer, a multiple cloning site and an SV40 late polyadenylation signal flanked by 2 telRL sequences, the site of protelomerase TelN recognition and cleavage. The PL backbone contained an ampicillin resistance gene and the pUC origin of replication. The resulting template PL was verified by sequencing and maintained in recombinase-deficient
E. coli.
The PL was purified using an endo-free maxiprep kit (Qiagen). The template plasmid was denatured using 0.1M NaOH for 5 min at 30 °C and then quenched in reaction buffer (30 mM Tris–HCl pH 7.4, 5 mM (NH4)2SO4, 30 mM KCl, 7.5 mM MgCl2, 2 mM dithiothreitol) containing 50 µM custom primer (Metabion), 4 mM dNTPs (Bioline), 200 units/ml Phi29 polymerase (Enzymatics) and 0.05 units/ml pyrophosphatase (Enzymatics). Upon mixing, the reaction was incubated at 30 °C for 30 h. Concatameric DNA was processed by addition of 2 mM TelN protelomerase (Enzymatics). Further processing was performed using 200units/ml XbaI (Enzymatics) and 200units/ml exonuclease ExoIII (Enzymatics). The digest mixture was cleaned from reaction components by addition of 500 mM NaCl/100 mM MgCl2 and precipitated using 3.9% polyethylene glycol (PEG) 8000 (Applichem). DNA was pelleted (10 min, 4500 g) and re-suspended in 20 ml 500 mM NaCl/100 mM MgCl2. The 3.9% PEG 8000 precipitation was repeated with final resuspension in water prior to ethanol precipitation to remove residual salts.
Vaccination protocols
C57Bl/6 mice were used in accordance with U.K. Home Office Guidelines, following the Animals (Scientific Procedures) Act 1986. Ethical approval was obtained from the Science Review Group in Southampton and the Animal Welfare and Ethical Review Board. At each prime or boost vaccination, the mice were vaccinated in both hind legs. DNA was diluted in 100 µl saline (50 µl containing 25 µg of DNA per injection). I.m. injections without EP were performed in the anterior tibialis and injections with EP were given in the quadriceps muscle. EP was carried out on mice anaesthetised by isofluorane, using a custom-made pulse generator from Inovio Pharmaceuticals. 10 trains of 1000 square wave pulses were delivered at a frequency of 1000 Hz, each lasting 400 µs (200 μs positive and 200 μs negative). Each train was delivered at 1 s intervals, with the electrical pulse kept constant at ± 50 mA. Booster injections were given 7 weeks after priming as this time point was optimal for generating T-cell responses with DNA vaccines (unpublished data).
Tetramer staining
50–100 µl blood was taken into a heparin containing buffer (100 U/ml Heparin, 10 mM EDTA, in PBS). Red blood cells were lysed using RBC lysis Buffer (Qiagen). After incubation with anti-CD16/32 blocking antibody (clone 93, eBioscience) PE-labelled H2-Db-E749–57 tetramer made in house was added 15 min before an allophycocyanin (APC)-labelled anti-CD8a antibody (clone 53 − 6.7, Biolegend). Samples were run on a FACS Canto I (BD Bioscience).
Tumour suppression and T-cell subset depletion
The TC-1 mouse lung epithelial cell line was cultured in complete RPMI 1640 media (10% FCS, 4.5 g/l glucose, 2 mM glutamine, 50 U/ml penicillin and 50 µg/ml streptomycin). TC-1 cells express E6 and E7 and are co-transformed with Ras [
15].
C57Bl/6 mice aged 10–13 weeks were injected s.c. with 5 × 104 TC-1 cells in 200 µl of PBS, on d 0. Three days later the mice received 50 µg E6E7 DB or PL DNA, or an irrelevant enhanced GFP (eGFP)-encoding DB, with or without EP. Naive mice were included as controls. Tumour measurements were taken every 3–4 days, and mice culled when tumours reached a mean diameter > 15 mm.
In T-cell subset depletion experiments one dose of DB vaccine was given with EP 4 weeks prior to TC-1 tumour challenge (as above). The depleting CD8 antibody (YTS169.4) or isotype control (both 700 µg per mouse, from BioXcell, West Lebanon, USA) was given i.p. 1 day before and 6 days after tumour challenge.
ELISPOT
Induction of peptide-specific CD8+ or CD4+ T cells in individual mice was assessed ex vivo using an IFN-γ ELISPOT kit or IL-4 ELISPOT kit (CD4+) (both from BD Pharmingen). Briefly, splenocytes (2.5 × 105 cells/well) were incubated in complete RPMI for 22 h with specific or control peptide, or peptide library (1 µM for each peptide). Overlapping peptide pools covering the E6 and E7 sequence consisted of 15-mers with an 11 amino acid overlap at > 95% purity (GL Biochem; sequences in Supplementary table 1). Control wells were incubated without peptide to assess background. Samples were plated in triplicate and mean values were expressed as spot-forming cells (SFCs) per 106 cells. Levels were considered positive if at least two times above background.
ELISA for IgG and IgG isotypes
Serum samples were collected 2 weeks after boosting. ELISAs were performed as described elsewhere [
16]. Briefly, Maxisorp Immuno plates (Nunc) were coated with 3 µg/ml E6 or E7 protein overnight at 4 °C. Both proteins were expressed in house in
E.coli. For E6 the pET41M-HPV16-E6_C4S expression vector was used (kind gift from Dr G. Travé, Ecole Supérieure de Biotechnologie, Bd Sébastien Brant, France) and the protein purified following the published protocol [
17]. Expression vector pGEX2T-E7 [
18] (Addgene plasmid # 13634) was kindly provided by Dr K. Munger (Tufts University, Boston, USA). E7 protein was purified using glutathione beads (GE Healthcare) following the manufacturer’s protocol. Serum antibody bound to the coated antigen was detected with HRP-conjugated anti-mouse IgG (The Binding Site), anti-IgG1 (Oxford Biotechnology), anti-IgG2b or anti-IgG2c (both from Harlan Sera-Lab) as previously [
16,
19].
Assessment of innate immune recognition of the DB DNA vaccine
HEK-Blue cells expressing human or mouse TLR9 (InvivoGen) were cultured in complete DMEM supplemented with 100 µg/ml Normocin, 30 µg/ml Blasticidin and 100 µg/ml Zeocin (all InvivoGen). HEK-Blue cells contain a secreted embryonic alkaline phosphatase (SEAP) reporter system, which generates a colorimetric change following TLR9 engagement.
Transfections were performed with Fugene (Promega), following the manufacturer’s protocol. Cells were plated at 6 × 104 cells per well in a 96-well plate in HEK-Blue detection media (InvivoGen) or complete DMEM. 300 ng of E6E7 DB or PL DNA complexed with Fugene (5 µl/well) was added directly to the cells. In selected wells human A151 (InvivoGen) or mouse 4084-F (InvivoGen) TLR9 antagonists were used at 10 µg/ml. After 42 h A625nm was measured. To confirm transfection efficiency was similar between DB and PL vaccines eGFP encoding constructs were used, with eGFP expression assessed using a fluorescent microscope.
To investigate the involvement of cytosolic DNA sensing pathways, the THP1-Blue ISG reporter cell line and the THP1-Blue ISG-KD-STING cell line which has STimulator of INterferon Genes (STING) knockdown, were used (both InvivoGen). Transfections were performed as above and 24 h later the cell supernatants were incubated with QUANTI-Blue reagent for 25 min and A625nm was measured.
Bone marrow derived dendritic cell (BMDC) generation and activation
DCs were prepared from bone marrow BM collected from the tibias and femurs of 6–8 week old mice. BM from MyD88 knock out (KO) mice was kindly provided by Dr S. Nanda (University of Dundee, UK) with the permission of Prof S. Akira (University of Osaka, Japan). BM from STINGKO mice was kindly provided by Dr J. Rehwinkel (University of Oxford, UK) with the permission of Prof J. Cambier (University of Colorado, USA). DCs were generated using the CellXVivo DC differentiation kit (R&D). Briefly, BM cells were plated at 106 cells/ml in media containing 800 IU/ml GM-CSF and 500 IU/ml IL-4, and cultured in 6-well plates for 6 days with the media changed every 2 days before using in activation experiments.
BMDCs were treated with 1 µg/ml of DB or PL DNA with 1 µg/ml lipofectamine 2000 (Thermofisher) or left untreated, as in [
20]. Transfection efficiency was checked by FACS using DB or PL DNA encoding eGFP. LPS and poly(I:C) were used as controls at a final concentration of 1 µg/ml. After 20 h supernatants were collected for ELISA or Luminex. BMDC collected by centrifugation were used for FACS to check upregulation of CD40, CD80 and CD86, and for RNA purification.
Real-time PCR (qPCR)
To evaluate gene expression RNA from 5 × 106 purified DC was extracted using the RNeasy kit (QIAGEN). RNA was reverse transcribed to cDNA using a high capacity cDNA reverse transcription kit (Thermofisher). mRNA levels of IL-12a, IL-12b, IL-10 and C-X-C motif chemokine ligand (CXCL)-10 [IFN-γ-inducible protein 10 (IP10)] were assessed by real-time PCR following the protocol for TaqMan™ Fast Advanced Master Mix (Thermofisher) and performed using an Applied Biosystems 7500 Real-Time PCR system (Thermofisher). All qPCR 6-FAM-labelled (6-carboxyfluorescein) primers were purchased from Thermofisher.
Evaluation of type I interferons in DC experiments
IFN-α in BMDC supernatants was measured using a Luminex bead assay kit (Life Technologies) using a Bio-Plex®200 system (Bio-RAD). IFN-β levels were assessed using the LEGEND MAX™ Mouse IFN-β ELISA Kit (Biolegend). Samples were diluted in assay diluent as required. Each sample was assayed in triplicate and cytokine standards supplied by the manufacturer were used to calculate the concentrations of the samples.
Statistics
The Mann–Whitney statistical test was used for T-cell responses and the antibody data, and the Mantel–Cox test was used to compare survival. For in vitro data the unpaired t test was used.
Discussion
Immunotherapy with monoclonal antibodies targeting checkpoint inhibitors has revolutionised the management of many solid cancers. However, only up to 50% of patients benefit. The challenge is to improve the chances of those who do not, and vaccination is expected to achieve this. The recent success of DNA vaccines with EP in patients with cervical dysplasia may re-open a window of opportunity for DNA vaccines in the clinical cancer setting [
11]. The development of a bacteria-free system for production of large amounts of DNA using the DB system aims to meet the growing demand for quality DNA vaccines. Here we explored the ability of this novel type of DNA vaccine to trigger the immune response. This has been addressed by analyzing induction of CD8, CD4 T cells and antibody using a DB DNA vaccine targeting the clinically relevant antigens HPV16 E6 and E7. We demonstrate that this vaccine is able to induce as effective anti-tumour immunity as a PL DNA vaccine, and that the potency of the DB vaccine is improved by EP. Interestingly, even low levels of CD8 T cells induced with one dose of DB without EP were protective. These responses will be likely boosted by the antigenic peptide presented by the tumour cells. The levels of protection induced here is at a similar level previously achieved after a single DNA vaccination [
34]. Further improvement could be achieved by boosting. Combination with immunostimulatory/checkpoint antibody is a promising approach to further potentiate DB efficacy as reported for PL [
34].
A compelling question has been innate immune recognition of the minimal DNA sequences contained in DB vaccines. PL DNA vaccines are recognised through TLR9 and a number of cytoplasmic sensors [
4,
5]. TLR9 recognises unmethylated CpG motifs in the context of XCGY sequences where X is not C and Y is not G. These ISS sequences are more frequently present in bacterial sequences than in mammalian sequences [
4]. In the case of DB vaccines immune recognition by TLR9 will depend on the frequency of ISS CpG motifs in the encoded antigen itself since the rest of the vaccine sequence is minimal. Presumably, this will be sufficient in most bacterial antigens. With self-antigens where the frequency of these motifs is low the DB vaccine may be unable to stimulate via TLR9. E6 and E7 oncogenes are not self-antigens but are of a viral origin. Although some viruses have sufficient ISS CpG motifs to stimulate TLR9 [
35] the HPV genome has the lowest CpG content among viruses with most of them in non-ISS context, in keeping with HPVs’ association with long-term stable infection [
36,
37].
Previously, little dependency on TLR9 for a PL DNA vaccine plus EP has been demonstrated in mice, with the majority of recognition performed through intracellular DNA sensing [
38]. It is possible that combination of DNA delivery with EP reduces dependency on TLR9 for immune recognition of DNA vaccines. Although PL DNA was effective at activation of adaptive immunity in mice lacking TLR9, little doubt exists about the involvement of TLR9 in recognition of bacterial DNA in both mice and humans [
4,
39]. In mice TLR9 is expressed in both plasmacytoid and conventional DC subsets, while the expression pattern differs in humans and is restricted to plasmacytoid DCs [
3,
40,
41]. Here we demonstrated that in contrast to PL DNA, there is a little involvement of either mouse or human TLR9 in recognition of the DB DNA vaccine. On the contrary, DB DNA is able to stimulate intracellular DNA recognition pathways at a similar level to PL DNA. Although no statistically significant differences of tetramer positive CD8 responses were observed between DB and PL, a trend of slightly higher responses was observed in the PL groups possibly accounted for by contribution of TLR9. These findings provide mechanistic insight into the immune recognition of this novel type of DNA vaccine.
The significance of the findings is clear; cytoplasmic recognition of DNA is not restricted to antigen presenting cells; these cytoplasmic DNA sensors are widely expressed as they contribute to the recognition of virally infected cells [
42]. The key signalling adaptor for cytosolic DNA sensing, STING, can bind directly to DNA and may participate in direct activation of the STING signalling pathway [
42]. Similar to the case for DNA recognition by Absent in Melanoma 2 protein (AIM2) and interferon-gamma-inducible protein 16 (IFI16), cyclic-GMP-AMP (cGAMP) is expressed following stimulation with cytoplasmic DNA can also bind STING [
43]. cGAMP is capable of spreading to neighbouring cells which results in stimulation of STING-dependent immune induction in other cells in which DNA is not present in the cytoplasm. These can lead to rapid immune induction without the actual need of DNA spread and can be involved in recognition of DB vaccine.
Interestingly, our in vivo antibody data showed DB with EP also induced IgG1 while PL induced mainly IgG2b and 2c isotypes. This was despite both vaccine groups inducing similar high levels of Th cells secreting IFN-γ, which induces isotype switch to IgG2 isotypes, and similar low levels of Th2 producing IL-4 which drives IgG1 isotype switch. It is likely that in the PL group skewing towards induction of IgG2 isotypes could occur through direct triggering of TLR9 expressed on B cells [
44]. Polarization of Th subsets depends on the initial cytokine milieu produced by activated DC. A strong Th1 bias has been demonstrated for DB here similarly to PL with a smaller Th2 component for both. Type I interferons induced by DB and PL through STING, as demonstrated here, are the likely contributors to Th1 polarization [
45]. Both DB and PL were able to induce IL-12 the key cytokine for Th1 induction with small amounts of IL-10 which contribute to Th2 responses [
32,
46].
Our data show here that in the absence of the ability to stimulate TLR9, the DB vaccine is able to induce a significant adaptive immune response particularly when combined with EP. Taken together, DB which represents a minimal sequence for antigen expression in vivo provides a valuable alternative to PL for applications to patients. Additional sequences including ISS CpG motifs can be easily incorporated into dbDNA™ as desired. The advantage over PL is increased patient safety due to lack of antibiotic resistant genes and endotoxin contamination. These data provide the first preclinical evaluation of DB DNA vaccine targeting cancer antigens.