Traditional vaccines, aimed at protecting populations from infection, were developed with little knowledge of immunological mechanisms. The measure of suc-cess was a reduction in infection rate with minimal side effects. Modern preventative vaccines, required for more difficult infectious organisms, still face this measure, but can build on expanding immunological knowledge and new technology. A dramatic example is the success of the recombinant virus-like particle preventative vaccine against human papillomavirus1.
An added ambition is now to use the immune system to target persistent infections and cancer. For cancer, there is already evidence that passive immunotherapy is clinically effective. Two clear examples are the suc-cess of antibody treatment (for example, the anti-CD20 antibody for B-cell tumours2 or the anti-HER-2 (also
known as ERBB2) antibody for a subset of breast can-cer3) and the effectiveness of cellular attack on tumour
cells, evident from the ability of transferred CD8+ T cells
to attack tumour cells through graft-versus-leukaemia activity following allotransplantation4,5. The goal of
active vaccination is to induce these immune effector pathways and to establish immunological memory that is able to maintain continuous surveillance against emergent cancer cells.
One problem is that, apart from in virus-associated cancers, potential antigens are usually only weakly immunogenic. It is difficult to generalize about the nature of the target antigens, especially as genetic and proteomic probings of cancer cells are continually revealing new candidates. So far, investigators have focused on targets
which are either tumour-specific or tumour-associated antigens. The former are ideal and include idiotypic anti-gens of B-cell tumours6 and the so-called cancer-testis
antigens, which are effectively expressed only by cancer cells, and, in some cases, by multiple cancer categories7.
Tumour-associated antigens include the lineage-specific molecules that are also expressed by the normal cell of origin8. Examples under intensive investigation are the
antigens of melanoma, prostatecancer and other epithe-lial cancers. For some of these, effective immunity could damage the corresponding normal cells, demonstrating the tightrope which must be walked between cancer suppression and autoimmunity.
Although tumour antigens are generally weakly immunogenic, they can nevertheless induce a low level of spontaneous immunity. Although difficult to investigate, there are clear clinical examples in which this response is beneficial for controlling tumours9,10. However, in other
cases the spontaneous response can lead to tolerance. This natural regulatory system limits and controls immune activity, but can present a block on attempts to activate and maintain immunity by vaccination. Peripheral toler-ance is mediated in part by regulatory T cells, which can be found in tumour infiltrates11. Cautious attempts to
counter their negative influence on immunity to cancer, while avoiding unwelcome autoimmunity, are already being made in the clinic using specific immunomodula-tory drugs or antibodies12,13. Perhaps as a consequence of
some level of spontaneous immune recognition, tumour cells can downregulate expression of major histocompat-ibility complex (MHC) class I and target antigens and
Genetic Vaccine Group, Cancer Sciences Division, University of Southampton School of Medicine, Southampton General Hospital, Southampton, SO16 6YD, UK. Correspondence to F.K.S. e-mail: [email protected] doi:10.1038/nrc2326 Graft versus leukaemia Following allotransplantation of bone marrow or blood stem cells from a healthy donor into an MHC-matched patient, donor T cells may recognize peptides on patient leukaemia cells that result from polymorphic differences between the two individuals, resulting in beneficial immune attack.
Idiotypic antigen Individual antigenic determinants from the variable regions of the Ig heavy and light chains are referred to as idiotopes; the sum of the individual idiotopes is referred to as the idiotype.
DNA vaccines: precision tools
for activating effective immunity
against cancer
Jason Rice, Christian H. Ottensmeier and Freda K. Stevenson
Abstract | DNA vaccination has suddenly become a favoured strategy for inducing
immunity. The molecular precision offered by gene-based vaccines, together with the
facility to include additional genes to direct and amplify immunity, has always been
attractive. However, the apparent failure to translate operational success in preclinical
models to the clinic, for reasons that are now rather obvious, reduced initial enthusiasm.
Recently, novel delivery systems, especially electroporation, have overcome this
Cancer-testis antigens (CTA). Cancer testis antigens are encoded by germline genes but are expressed only in tumours and in male germline cells. As the latter express no MHC molecules, the CTA are effectively tumour-specific.
Tolerance
The process that ensures that B- and T-cell repertoires are biased against self-reactivity, reducing the likelihood of autoimmunity.
Regulatory T cells (TReg). Regulatory CD4
+ T cells serve to limit immune responses, thereby protecting against autoimmunity. There are two main types, termed natural for those developing in the thymus and adaptive for those arising in the periphery following infection and possibly cancer.
often secrete immunosuppressive molecules to defend themselves against attack14. Tumours can create a
tolero-genic environment which spreads to draining lymph nodes and can enhance regulatory T-cell activity. The hurdles to successful reversal of tolerance and induction of effective immunity are becoming clear and vaccines must incorporate elements to overcome them15.
For vaccination to be a realistic option, patients should ideally be in remission from their cancer and retain ade-quate immune capacity. Even in this setting, it is clearly essential for a vaccine to be powerful and this requires activation of both innate immunity and specific immune effector responses. DNA fusion gene vaccines offer the opportunity to incorporate multiple genes encoding a range of immunostimulatory molecules, either within the vaccine vector or by a separate vector. A vast array of molecules able to modulate immune responses can be delivered. They include chemokines to attract antigen- presenting cells (APC)16,17, activating cytokines18,
co-stimulatory molecules, APC-targeting antibodies and molecules to manipulate antigen presentation and/or processing19. Virus-derived molecules reported to
facili-tate cell-to-cell protein transfer, such as VP22(REF. 20), are also attractive, although the mechanism(s) of action remain unclear21. Molecules designed to circumvent
toler-ance are desirable but, as the role of the immune system is to respond to pathogens, our approach has been to use foreign antigens for activation.
Tumour antigens can be presented by DNA in a suit-able molecular form, ranging from full-length sequence, required to induce an antibody response, to short MHC class I- or II-binding peptides to optimize induction of
T-cell responses. Once induced, the effector antibody or T cells must find their target, a relatively simple require-ment for ‘liquid’ tumours such as leukaemia, but more challenging for the less accessible solid tumours.
The translational leap from preclinical models has required new technology, and novel delivery systems are now bridging this gap. The combination of DNA vaccination delivering specifically encoded antigen, together with refined objective assays of specific immune responses in patients, is allowing assessment of vaccine performance in pilot clinical trials. Larger trials of this apparently safe vaccine vehicle will then determine whether immune responses are linked to clinical efficacy. This Review will focus on DNA vaccine designs aimed to harness immunological knowledge, derived in part from preclinical models, which are now facing the ultimate test of performance in patients with cancer.
DNA fusion gene vaccine designs
DNA backbone. The basic features of DNA vaccines are described in BOX 1. The simple plasmid backbone cou-pled with the technology of gene manipulation facilitates incorporation of genes, which are then expressed by cells transfected in vivo. Although the transfection process is inefficient and varies with the target tissue and means of delivery, sufficient DNA is generally taken up to prime the immune response22. Importantly, the backbone of
bacterial DNA includes sequence motifs that stimulate innate immunity, creating an inflammatory milieu for triggering the adaptive immune response23(BOX 1).
Intracellular routing of expressed antigen. There is a clear opportunity to incorporate genes into DNA vac-cines that will influence the intracellular routing of anti-gen and thereby affect immune outcome. The common assumption is that antigen targeted to the endoplasmic reticulum (ER), where folding and initial glycosylation occur24, will be more efficient at inducing an antibody
response, whereas antigen targeted to the cytosol will gain direct access to the proteasomal degradation machinery for induction of peptide-specific CD8+ T-cell
responses. The leader (signal) sequence guides expressed protein to the ER and secreted protein follows this route, so it is useful to include a leader sequence if the aim is to induce antibody. However, there is also more surpris-ing evidence that guidsurpris-ing protein to the ER by ussurpris-ing a leader sequence can promote induction of CD8+ T-cell
responses25,26, probably owing to retrograde transfer of
protein from the ER to the cytosol27. In view of this, we
have opted to incorporate a leader sequence into all our constructs28. There have been attempts to target antigen
to a range of intracellular sites, including the proteasome or the lysosomal–endosomal pathway, but results have been varied29. One problem is that it is vital to
under-stand how the encoded antigen gains access to the APC. From muscle sites, cross-presentation is the likely major route, possibly through apoptotic cell debris30. Even from
skin sites, in which a gene gun can deliver DNA directly into Langerhans cells, allowing direct presentation(FIG. 1) of antigen, cross-presentation from keratinocytes might still be a major route31.
At a glance
• Preventative vaccination against infectious organisms has had a dramatic effect on public health. Therapeutic vaccination against cancer is more challenging but, armed with new immunological insight and genetic technology, aims similarly to harness the power of the immune system, in this case to destroy or suppress tumour cells. • Passively transferred antibodies and T cells are clearly able to attack human cancer
cells in vivo and are included in treatment protocols for some cancers. Active
vaccination would generate these effector pathways, together with immunological memory that is able to continuously detect and remove any emergent cancer cells. • Tumour antigens are being rapidly revealed, and can be expressed on cell surfaces or,
more commonly, as peptides in association with the major histocompatibility complex class I (or II) molecules. DNA vaccines can be designed to activate antibody and/or T-cell responses, providing focused immune attack on selected antigens. • DNA vaccines offer a precise but flexible strategy for delivering antigens to the immune system, and additional sequences encoding molecules to manipulate outcome can be included. The problem of translating success in preclinical models to patients seems to be overcome by using electroporation, which dramatically improves performance and is now in clinical trials for prostate cancer.
• The key to bypassing immune tolerance and activating high levels of anti-tumour antibody or cytotoxic T cells lies in inducing CD4+ T-cell help. Sequences derived
from microbial antigens can be incorporated into anti-tumour DNA vaccines, a strategy which mobilizes help for anti-tumour responses from the large non-tolerized anti-microbial repertoire.
Cross-presentation DC Keratinocyte
Myocyte
DNA vaccine
Presentation of vaccine-encoded peptides for CD4+ and CD8+ T-cell recognition
Vaccine uptake by DC at injection site Vaccine uptake by
non-DC at injection site
Indirect route Direct route Innate immunity
The first line of host defence against invading pathogenic organisms until an adaptive pathogen-specific immune response is able to develop. This is a multi-component system that includes various barriers to infection, such as physical barriers (for example, skin), physiological barriers (for example, stomach pH) and inflammatory barriers (for example, release of anti-bacterial serum proteins).
Leader (signal) sequence Leader (or signal) sequences are hydrophobic domains which are cleaved from synthesized membrane-bound or secreted proteins during transfer into the endoplasmic reticulum through the Sec61 channel. For DNA vaccines, the leader sequence is placed at the 5′ end of the encoded antigen.
Cross-presentation The uptake, processing and presentation by professional APCs of antigen that has been acquired from an exogenous source into the MHC class I pathway for induction of CD8+ T-cell responses.
Direct presentation The conventional pathway of processing and presentation of antigen into the MHC class I pathway from an endogenous intracellular source.
Antigen delivery to professional APC. DNA vaccines are commonly delivered by simple intramuscular injection, in which the expressed protein forms a depot of antigen. There may be some secretion of protein that can gener-ate CD4+ T cells and antibodies. Transfected muscle cells
clearly express antigen and act as a target for immune effector cells. Surprisingly, they can also upregulate expression of MHC class I and co-stimulatory molecules, with production of cytokines and chemokines, raising the question of a continuing role in immune activation32.
However, for induction of high-level immunity, transfer to APC occurs by cross-presentation33 and we have
con-firmed this exogenous indirect route in vivo using our vac-cines34,35(FIG. 1). This route is expected to induce antibody
and CD4+ T-cell responses but induction of CD8+ T-cell
responses was initially thought to be disfavoured, being usually sourced from intracellular antigen. However, a recently described process whereby engulfed phagosomal material can fuse with ER-derived vesicles allows access of exogenous antigen to the MHC class I loading com-partment36. It seems that dendritic cells (DC) might be
particularly suited to this pathway owing to pH-mediated limits on phagosomal degradation37,38. If apoptotic vesicles
from transfected muscle cells behave similarly, this would be one explanation of the efficient induction of CD8+
T-cell response following intramuscular DNA injection33.
Until recently it has been difficult to determine the relative levels of cross-presentation or direct presentation of anti-gen following DNA injection. However, it is now possible to manipulate vaccine design to restrict presentation of the same antigen to either cross-presentation or to direct presentation from APC34. The immune outcome differs in
level and kinetics, illustrating the potential for optimizing vaccine performance (see below).
The importance of CD4+T-cell help. CD4+ T helper
(TH) cells can be described as the orchestrators of the immune response, and are clearly vital for the induc-tion and maintenance of immune memory39–42. They
can also turn against the immune response by acting as regulatory T cells. The general inadequacy of short peptides as vaccines to induce CD8+ T cells could be
partly explained by the fact that they do not induce accompanying TH cells. For highly immunogenic tumours, anti-CD40 antibodies can apparently substitute for TH cells43, but CD40 is not the only activating
mol-ecule44 and, in a tolerant setting, CD40 activation alone
appears to be insufficient to prime CD8+ T cells45,46.
The same reliance on TH cells is evident from the rapid death of adoptively transferred human CD8+ T-cell
clones against HIV47. Even the expansion of CD8+
T cells during homeostatic proliferation depends on TH cells for protection against apoptosis48. It appears that
Box 1 | Features of DNA vaccines
Plasmid backbone
DNA vaccines consist of bacterial plasmids into which specific sequences are incorporated. Gene expression is commonly driven by the cytomegalovirus immediate early promoter and its adjacent intron A sequence, which ensure high transcription efficiency. Other elements include a transcription termination signal and a prokaryotic antibiotic resistance gene. Although many vectors are available, a consensus optimized vector is emerging as potentially the most acceptable for human use129. However, further modifications, such as the minimalistic, immunogenically defined gene expression
(MIDGE) vectors, are being explored, both to increase performance and to remove non-essential backbone sequences130.
Stimulation of innate immunity
Initially, the ability of the backbone of bacterial DNA to stimulate the innate immune system was an unforeseen but welcome observation. DNA appears to act as a pathogen-associated molecular pattern able to stimulate cells through Toll-like receptors (TLRs). Conserved molecular patterns are shared by large groups of microorganisms. In addition to bacterial DNA, these include lipopolysaccharide and double-stranded RNA. TLRs have an important role in the recognition of microbial components by the immune system131.
A crucial sequence element that is relatively common in bacterial DNA but rare in mammalian DNA is the hypomethylated CpG dinucleotide-containing motif which binds to TLR9(REF. 132). Stimulation of a range of TLR9-expressing cells, including B cells and dendritic cells (DC), then leads to a cascade of activation, proliferation and differentiation of natural killer cells, T cells and monocytes/macrophages. An industry has now developed aimed at using synthetic CpG phosphorothioate oligonucleotides as adjuvants for a range of different vaccines133. One anticipated problem for human application was that,
in contrast to mice, TLR9 is not expressed by myeloid DC but only on plasmacytoid DC. However, DNA vaccines perform well in Tlr9–/– mice, indicating that this pathway is not their only route to stimulation of innate immunity134.
Ig variable region genes B-cell lymphoma
Fragment C of tetanus toxin
DNA fusion vaccine
Fusion protein: strong immunogen
Weak immunogen Strong immunogen scFv
Anti-Id response
VH VL Dom 1 Dom 2
VH VL Dom 1 Dom 2
Anti-CD40 antibodies Agonistic anti-CD40 antibodies mimic the natural ligand (CD154) and stimulate DC to activate T-cell immunity.
Tetanus toxin This is produced by the bacterium Clostridium tetani. Treatment with formaldehyde inactivates the toxin to produce tetanus toxoid, a non-toxic but extremely immunogenic derivative that is used as a vaccine in both adults and children.
T-cell help
Activated antigen-specific CD4+ T helper cells (T
H) have a major role in linking and coordinating innate and adaptive immune responses. TH cells help B cells to produce antibodies and maintain circulating memory B cells; they also help stimulate and maintain both CD4+ and CD8+ T cell responses.
programming of CD8+ T cells for secondary expansion
is influenced by CD4+ T cells, with possible
involve-ment of interleukin 2 (IL2)49. In the absence of T H cells,
the ‘helpless’ CD8+ T cells undergo activation-induced
cell death on secondary stimulation50. By contrast,
once induced, CD8+ memory T cells rely on IL15, but
appear not to require continuing antigen-specific TH support51.
We have incorporated these principles into our DNA vaccine designs. It is unlikely that autologous tumour antigens are capable of inducing significant TH responses, so any induced CD8+ effector T cells might be doomed
to be helpless. To overcome this, we have used sequences derived from tetanus toxin encoded within the same vec-tor as the tumour antigen, so that TH cells can be engaged from a large anti-microbial repertoire to help immune responses against the tumour antigen. Depending on the target antigen, the fusion gene vaccines are designed to provide help for B cells to produce antibody, or to provide help to program CD8+ T cells for killing tumour cells
(see below). This concept of providing T-cell help within a DNA vaccine has also been developed using other non-self antigens as fusion partners, such as green fluorescent protein52, plant viral coat proteins53 or Pseudomonas
aeruginosa exotoxin54. Although the mechanisms of
action of these fused molecules may vary, they are all non-self antigens and potentially capable of activating the desired T-cell help. Another strategy for amplifying
immunity is to fuse immune-targeting molecules to antigens19,55. Although these are of clear interest in
them-selves, if they too are non-self antigens then they could have the beneficial side effect of activating TH cells.
The same principle of engagement of TH cells might be the explanation for the ability of DNA vaccines that encode xenoantigens (that is, human antigens in preclini-cal murine models, and vice versa for clinipreclini-cal trials) to induce immunity against the corresponding autologous antigens56,57. This approach is showing potential benefit
in trials of dogs with malignant melanoma, for which a vaccine encoding human tyrosinase now has conditional approval56,58. However, the ability of the foreign
xeno-geneic sequences to generate TH-activating peptides is difficult to predict as it is likely to depend on MHC class II-binding sequence differences between the xeno-geneic and synxeno-geneic versions. The hurdle for generation of CD8+ T-cell responses seems to be at induction. Once
established, ‘helped’ T cells should be maintained and be able to attack tumour cells without additional help. However, the option of repeated vaccination, and of using judicious vaccine design to induce tumour-specific CD4+
T cells, potentially required for interferonγ-mediated access to solid tumours59,60, is available.
Induction of tumour-specific antibodies. Passive antibod-ies aimed against tumour cell surface molecules are already showing clinical value61. One of the earliest antibodies to
reveal the efficacy of the approach was against the idio-typic (Id) immunoglobulin (Ig) expressed by lymphoma cells. Polyclonal62 and monoclonal63 anti-Id antibodies
showed clinical effects but the technical demands for raising patient-specific antibodies dampened enthusiasm. DNA vaccines avoid this problem as Id determinants can be expressed by using the variable region genes, either as whole Ig64 or as single chain Fv (scFv) (FIG. 2). However,
the scFv sequence alone was poorly immunogenic65,66.
Following fusion of a sequence derived from tetanus toxin (fragment C (FrC)) to the scFv, the poor performance was dramatically improved, with induction of high levels of anti-Id antibody and protection against lymphoma66,67 (FIG. 2). The principle of using FrC-specific TH cells to drive antibody responses against weak tumour antigens is well-founded in immunology and clearly operates when using DNA vaccines (FIG. 3). In this case the presence of the leader sequence allows secretion of the Id–FrC fusion pro-tein, but there is also likely to be cross-presentation of the antigen26,66. The important point is that the high numbers
of TH cells induced against FrC are able to help the anti-Id B cells because the latter co-present FrC peptide (FIG. 3). For this reason the scFv and FrC have to be linked66,67.
The FrC-specific CD4+ T cell is stimulated by antigen
that is presented by the DC, becomes activated and then recognizes the FrC peptides that are being presented by the anti-Id B cells. By CD40–CD40ligand interaction and cytokine production, help for anti-Id antibody production is then provided. The strategy can be regarded as a trick to make the immune system respond to a weak antigen. Not only are TH cells required for priming the B-cell response, but they appear to be necessary for maintenance of memory B cells68.
DC
p.scFv-FrC DNA fusion vaccine
B cell
T-cell help for B cells Anti-Id
Tumour-specific antibody FrC peptide
TCR
CD4+ T cell
Inadequate or tolerized Skin cell
Muscle cell
CD4+ T cell Secretion
FrC Id
Cross-presentation
FrC Id
Id-specific
FrC-specific
Single chain Fv (scFv). A recombinant polypeptide sequence consisting of the variable regions from the immunoglobulin light and heavy chains. These are usually separated by a short linker sequence to allow the variable regions to fold and assume the native conformation of the antigen binding site of the immunoglobulin from which they are derived.
Fragment C
(FrC). A non-toxic yet highly immunogenic polypeptide corresponding to the ~50 kDa carboxy-terminal portion of the heavy chain of tetanus toxin. It has a two-domain structure and the amino-terminal domain (DOM) contains a ‘helper’ peptide (p30) that has been shown to bind to a wide range of both murine and human MHC class II haplotypes, leading to CD4+ T-cell activation.
The principle of amplification of immunity by fusion to FrC sequence, or other non-self antigens, is also applicable to DNA vaccination against a range of tumour antigens69, and the general design is illustrated
in FIG. 4a. The scFv–FrC vaccine has now been tested as naked DNA in patients with follicular lymphoma, with a promising anti-Id response detected (see Clinical test-ing section). Responses against FrC were also seen in the majority of patients. This is encouraging, as naked DNA alone is now known to perform sub-optimally in human subjects, and exciting improvements in delivery technology are now available (see below).
Induction of tumour-specific cytotoxic T cells. Cytotoxic CD8+ T cells (CTL) are of obvious importance for
seek-ing out and killseek-ing tumour cells. In fact, the majority of tumour antigens are intracellular and can only be targeted by recognition of peptides that are bound to MHC class I molecules on the cell surface. Unlike anti-body, which persists over time, CTL that are induced by infection or vaccination expand but then the majority die, leaving only a small memory population ready for the next challenge. The goal of vaccination is therefore to induce the effector CTL and to maintain a memory population able to expand as the tumour emerges. This requires TH cells70, and we have again engaged FrC from
tetanus toxin to provide this71. The mechanism by which
TH cells support induction of CD8+ and CD4+ T cells is
likely to be by ‘licensing’ of DC70 by interacting ligand–
receptor pairs and local cytokines that are provided by the large population of FrC-specific TH cells (FIG. 5).
Licensed DC presenting peptides from both FrC and tumour antigen can then survive in an active state and resist attack by CTL72, facilitating priming and boosting
of the response to weak tumour antigens.
However, another immunological principle has to be considered: natural CTL responses to infection tend to focus on few immunodominant peptides73,74. This
phe-nomenon might be due to a need to restrict the range of potentially dangerous cross-reactive CTL, which could attack autoantigens. Immunodominance tends to be most evident at the time of boosting, when dominant CTL competition appears to limit responses to subdominant epitopes. This phenomenon is likely to affect the ability of polyepitope ‘string-of-beads’ vaccines to induce a broad CD8+ T-cell response against all constituent peptides.
Unless all are equally powerful, the CTL response would be expected to focus on the immunodominant epitope(s), and this has been observed in mouse responses75 and in
human cells76. However, the polyepitope design remains
attractive for induction of antibody or CD4+ T-cell
responses. In terms of our fusion gene vaccines, if MHC class I-binding peptides were present in the FrC mol-ecule, they could compete with tumour-derived peptides. To avoid this, we have minimized FrC to a single domain (DOM 1) that is still able to provide T-cell help but less able to produce competitive peptides35,77. This design can
incorporate full-length tumour antigen fused to DOM 1 so that potential target epitopes can be processed and pre-sented from the full tumour sequence. A vaccine encod-ing full-length antigen would be suitable for all human MHC class I and II haplotypes and might also induce antibody. The design is illustrated in FIG. 4b and is cur-rently under investigation. However, there appears to be an added advantage if the target peptide is repositioned from the backbone sequence77(FIG. 4c).
The observation that suppression of tumour growth requires high levels of CTL is no surprise78. To achieve
this, and to give the tumour-derived peptide an even greater chance to dominate the CD8+ T-cell response79,
we positioned the peptide-encoding sequence at the 3′ end of the FrC DOM sequence35. This pDOM–epitope
design (FIG. 4c) induces high levels of peptide-specific CTL in a wide range of models35,77,80,81. It is capable of
breaking tolerance, being able to induce CD8+ T-cell
responses in male mice against the male HY antigens80.
Currently, a pDOM–epitope design that encodes a pep-tide from prostate-specific membrane antigen is in clini-cal trials for patients with prostate cancer. The power of this design is impressive but there are two potential disadvantages. First, in common with all peptide-based vaccine designs, it is applicable only to patients with a specific MHC class I haplotype (HLA-A*0201), which represent ~40% of the Caucasian population. However, as peptide binding to alternative class I haplotypes is being defined, it is now feasible to extend the approach more widely. Second is the potential for tumour escape by loss of expression of the target peptide. One strategy to overcome this is to use a second vaccine incorporating another target epitope from the same or a different pro-tein. This would be delivered at a separate site, thereby avoiding negative selection by the immunodominant
Figure 3 | Pathway of provision of T-cell help for B cells. Following injection of the DNA vaccine encoding the idiotypic (Id) single chain Fv–fragment C (FrC) fusion protein, there is secretion of Id FrC, which is taken up by the Id-specific B cells. The fusion protein is also cross-presented to dendritic cells, which then present both Id-derived and FrC-derived peptides. The repertoire of Id-specific CD4+ T cells is
inadequate. By contrast, the FrC-specific CD4+ T-cell repertoire is large and these
Antibody, CD4+ T cells
Antigen-specific outcome:
High levels of CD8+ T cells Target MHC
I-binding peptide a pAntigen–FrC
b pDOM–antigen
c pDOM–epitope
Antibody, CD4+ and CD8+ T cells DOM 1 DOM 2
Surface or secreted
antigen
Surface or intracellular antigen
DOM 1 DOM 1
‘Licensing’ of dendritic cells Licensing of dendritic cells by activated CD4+ T cells is essential for the generation of effective CD8+ T cell responses and possibly also for CD4+ T cell responses. Multiple interacting molecular pairs are likely to be involved, including CD40–CD40 ligand, CD28– CD80/86 and OX40–OX40 ligand.
peptide. It would allow activation of immunity against subdominant peptides, where tolerance may be less. The ability of our various vaccine designs to induce specific immune pathways is shown in FIG. 4.
Comparing cross-presentation with direct presentation. The pDOM–epitope vaccines appear to deliver peptide to the APC by cross presentation34. It was of interest
therefore to compare the immune outcome of rerouting peptide presentation with a direct pathway of presenta-tion from within the APC. To investigate this we created a new vaccine (pDUO) in which the TH-activating sequence from the full-length DOM sequence was reduced further to the promiscuous MHC class II-binding peptide, p30 (FIG. 6a). We then placed it in a position where it can only be expressed following direct transfection (FIG. 6a,b). To achieve this, the p30 sequence was inserted into the invariant chain, replacing the MHC class II-associated invariant chain peptide (CLIP)82. The T
H-dependent
response to pDUO is then restricted to direct presenta-tion34. Availability of both pDUO and pDOM–epitope
vectors then allowed the comparison of the performance of peptide-specific vaccines delivered by either direct-presentation or cross-direct-presentation pathways (FIG. 1). The ability to prime the CD8+ T cell response differed
markedly between the two vaccines (pDOM–epitope versus pDUO), with direct presentation inducing a slower but considerably higher-level response than cross- presentation34. The reason for this is unclear but could
reflect antigen level and persistence, crucial factors in controlling immune responses. Interestingly, combining the two vaccines in a prime–boost strategy, in either order, amplified induction of CD8+ T cells dramatically34.
Humanized mouse models for testing CTL induction. The balance between a study of vaccine performance in mouse models and early clinical testing is difficult to maintain. Our decision has been to use mice only to establish principles and then to move quickly into pilot
trials in patients. For the important CTL response, mice expressing the human MHC class I HLA-A2 (REF. 83) molecule (HHD mice) (BOX 2) provide an essential model for testing the ability of DNA vaccine designs to induce responses against target human epitopes. Although there is generally no tolerance in the mice, in some cases mice do express the antigen. If they do not, they can be crossed with transgenic antigen-expressing mice to explore whether the vaccines can perform in a tolerant setting. Testing in these mice can establish only the feasibility of raising CD8+ T cell
responses. In our view, there is no way to predict clinical effectiveness in human subjects without direct testing.
Vaccine delivery and electroporation
The promise of DNA vaccines has been overshadowed by an early relatively weak performance in primates, including humans. Various strategies have been explored to overcome this, including particle-mediated delivery or gene gun, both generally administered through the skin84. This has the advantage of requiring only low doses
of DNA and is particularly attractive for prophylactic vaccines against infectious diseases, for which simple and rapid delivery is important. In several clinical trials it appears promising, with a recent trial against influenza virus inducing seroprotective levels of antibody84. There
have been fewer studies of this approach for cancer vac-cines and comparison of delivery systems is urgently needed. For the intramuscular route, a major reason for the weak performance in large animals has now been revealed. It was already clear from mouse models that the volume of injection was crucial. Injection of 50 µl into mouse muscle generates strong immune responses that are gradually lost on volume reduction85. Presumably
the hydrostatic pressure causes sufficient local damage to increase transfection and contribute to inflammation. Scaling up from 50 µl in mice to human subjects would be unacceptable, so a new approach was required.
Clues to what this might be were derived from prime–boost strategies in human subjects using a vaccine against malaria86. In one study, responses
primed with naked DNA could not be boosted with naked DNA. By contrast, boosting with the same gene encoded within a viral vector was able to induce signifi-cant responses86,87. This led to a range of prime–boost
strategies with different viral or bacterial vectors88. Two
factors are likely to operate at the time of boost: first, the level of antigen, presumed to be higher when viral vec-tor is used; and, second, activation of a local inflamma-tory response, which should aid immune activation89,90.
Although this approach has merit for prophylactic vac-cination, it might be less suitable for cancer vaccines, for which repeated vaccination is likely to be necessary to maintain continuous immune pressure on an emergent tumour. The problem arises from the accompanying immunity that is generated against the viral proteins, which blocks vector delivery91–93. Immunity against
some viral vectors, such as those based on vaccinia, might in fact be pre-existing, but, even if not, it will be induced. There is already evidence in patients from a trial of a melanoma vaccine that immune diversion to vector components can occur94.
Figure 4 | DNA fusion vaccine designs based on help from fragment C. Fragment C of tetanus toxin contains two linked domains (DOM1 and DOM2). a | For the induction of antibody and/or CD4+ T-cell responses against tumour, full-length
fragment C is fused to the 3′ end of the tumour antigen sequence. b | For induction of CD8+ T cells against tumour, DOM2, which contains potentially competitive major
histocompatibility complex (MHC) class I-binding peptides, is removed. DOM1, containing a known major histocompatibility complex class II-binding peptide (p30), is retained and fused to the 5′ (or 3′) position of the tumour antigen sequence. c | For induction of high levels of CD8+ T cells against single tumour peptides, DOM1 is fused
DNA fusion vaccine
DC licensing
CD8+ Anti-tumour effector T cells
Priming
CD40
CD40L
OX40 CD28 OX40L
CD80 CD86
MHC II–peptide DC
TCR CD4+
FrC-specific CD4+ T cell
Immunodominant peptides The natural human CTL response to an antigen tends to focus on only a few immunodominant peptides. The precise mechanisms that establish this hierarchy remain unclear but the presentation of epitopes to naive CTL is controlled at several points, including processing efficiency, capacity to bind to MHC class I molecules and the stability of the bound complex on the cell surface.
To avoid this, physical methods to amplify delivery offer an alternative, and liposomes, microparticles, poly-mers, nanoparticles and tattooing are all under investiga-tion95,96. We have focused on electroporation (EP), for
which there is a clear path to the clinic. For vaccination, electrical stimulation of skeletal muscle with a pulse generator is applied immediately after intramuscular injection of DNA97–99. The procedure increases antigen
expression, presumably by increasing transfection effi-ciency, and is accompanied by local tissue injury and inflammation100. The outcome is a dramatic
enhance-ment of humoral and cellular immune responses85,101–103.
Importantly, EP reverses the failure of low-volume injection to induce immunity in mice85,104, and increases
responses in large animals, including rhesus macaques. A combination of DNA priming with EP boost could generate antibody responses comparable to those that are induced by protein in Complete Freund Adjuvant, and also amplified CTL responses104. Although the high
transfection efficiency does increase integration levels, these remain low105,106. EP may provide a prime–boost
combination equivalent to that observed using viral vec-tors, and it is now undergoing testing in the clinic using a DNA vaccine for patients with prostate cancer. Although uncomfortable, repeated EP has been accepted by our patients without the need for general or local anaesthesia and with no apparent long-term ill effects. EP appears to have surmounted the hurdle required to translate DNA vaccination into the clinic, and multiple trials are in progress in infectious diseases and cancer.
Clinical testing
Early clinical testing remains the crucial step for deter-mining whether DNA-vaccine strategies evaluated in the laboratory merit further development. Traditional clinical trials, based largely on chemotherapy, involve a stepwise escalation, through phase I to phase IV, to
monitor toxicity and efficacy107. However, the evident
safety of DNA vaccines has led to a relaxation of the requirements to assess autoimmunity, persistence and integration by both the United States Food and Drug Administration and the national competent regulatory authorities in Europe, unless new delivery strategies are used. Although DNA vaccines encoding autoantigens might present a problem if high levels of immunity can be induced, the real question for the moment is efficacy rather than toxicity. This tends to blur the phase I and phase II boundary.
In terms of clinical setting, one problem is that drug toxicity studies are commonly carried out in end-stage cancer patients and then slowly moved into earlier dis-ease. This is inappropriate and potentially misleading for vaccine studies, as high tumour load and a failing immune capacity guarantee a poor response. It might be one reason for the apparently weak performance of immunotherapeutic trials in patients with metastatic melanoma108. A similar setting has been used for some
of the early trials of DNA vaccination. Perhaps partly owing to this, only limited immune responses have been reported so far in a range of cancer patients (reviewed in REF. 109). However, a prime–boost approach using an adenoviral vector expressing prostate membrane-associated antigen induced antibody responses in 86% of vaccinated patients110. TABLE 1 summarizes the
published trials of DNA vaccines in cancer to date. Although DNA vaccines are well-tolerated, the vari-ations in design, delivery and clinical targets make an overall assessment difficult. There are indications of immune responses in melanoma and prostate can-cer patients, and, in some cases, suggestive clinical benefit. Interestingly human papillomavirus-derived antigens appear to be effective cancer-associated immunogens111. At this stage, it is desirable to continue
to optimize design and delivery to improve immune outcome to the level required to be effective against tumour in vivo. In fact, more recent trials are beginning to produce encouraging data112. However, evaluation of
the immunogenic potential of various designs requires patients with adequate immune capacity. Now, many groups design clinical trials where ‘first-into-man’ studies are undertaken in patients with low-volume disease and with conserved immune status.
One attractive setting is of patients on a watch-and-wait policy before any treatment, although biologically these tumours may differ from those requiring imme-diate treatment. A more common setting is in patients during complete remission after standard adjuvant treat-ment (chemotherapy, radiotherapy, antibody treattreat-ment or a combination) to whom vaccination can be given after immunological recovery113. It might also be given in
parallel to standard treatments114 to exploit an apparently
preferential depletion of suppressor cells, as suggested for cyclophosphamide (regulatory T cells)12,115 and
gem-citabine (myeloid suppressor cells)116,117. Importantly,
local radiation treatment might also open the tumour bed for access by CD8+ T cells118.
Selection and coordination of patient groups is obviously aided by biomarkers which reflect disease
Figure 5 | T-cell help for induction of anti-tumour T-cell responses. Vaccine-encoded fusion protein is expressed by dendritic cells (DCs), and major histocompatibility complex (MHC) class II-binding peptides from fragment C (FrC) activate the large repertoire of anti-FrC CD4+ T cells. Pairs of receptor–ligands (matched colours) interact
Antigen-specific outcome:
high levels of CD8+ T cells Target MHC I-binding peptide
MHC I-binding epitope Leader sequence Invariant chain (Ii)
MHC II-binding peptide (p30) replacing CLIP in Ii DNA vector backbone
Promoter sequence pDUO
pDUO
Nucleus a
b
Transfected dendritic cell
MHC I and tumour peptide
Plasma membrane
MHC II and helper peptide (p30)
MHC I MHC II–Ii
ER
MHC I–peptide
Help
MHC II–Ii MHC II–peptide
CD8+ T cell
Tumour-specific
CD4+ T cell
FrC-specific Ii + MHC II-binding peptide (p30)
load and behaviour. One example is in prostate cancer: following initial therapy, rising serum levels of prostate-specific antigen herald incurable disease. This allows a semi-quantitative assessment of the ability of a vaccine to reduce or reverse disease progression.
Haematological malignancies offer an opportunity for vaccination after chemotherapy followed by stem-cell transplantation to reconstitute immune capacity. Transplantation is usually autologous, but if an allogeneic donor is used, an exciting possibility is to vaccinate the healthy donor before cell transfer.
Immune capacity. Given the complex effects of disease and treatment on the immune system, there is an urgent need to evaluate immune capacity in different settings. Available studies in patients with malignancy have focused on immune responses after stem-cell transplan-tation119 and even here there is no definitive consensus
on when immune reconstitution is sufficient for vaccina-tion with standard vaccines.
In fact, in spite of the success of prophylactic vac-cination against infection in human subjects, there have been rather few studies of the kinetics and level of
Figure 6 | pDUO: a novel vaccine design that induces CD8+ T-cell responses by direct presentation. a | Vaccine design. The pDUO vaccine design incorporates two strong viral promoters within the plasmid cassette, derived from cytomegalovirus and Simian virus 40, each of which drives expression of a separate vaccine-encoded antigenic peptide sequence: first, a tumour-derived major histocompatibility complex (MHC) class I-binding peptide with a leader sequence; second, a fragment C (FrC)-derived MHC class II-binding peptide (p30) within the invariant chain (Ii) sequence.
b | Intracellular route of processing and presentation of the peptides from pDUO. Following transfection of the DC
in vivo, the tumour-derived peptide binds to MHC class I in the endoplasmic reticulum (ER) and is then expressed at the
cell surface. In parallel, the fragment C-derived peptide, p30, binds to MHC class II together with the Ii chain. The latter is then cleaved leaving the bound p30 peptide. CD4+ T cells that are specific for p30 can then help the CD8+ T cells that
are specific for the tumour peptide. This vaccine generates CD8+ T-cell responses that are helper T cell-dependent and
helper T cells can only be generated following direct transfection of antigen-presenting cells: the p30 peptide binds to the class II molecule by the ER route. Therefore, the co-delivered CD8+ epitope can only generate immunity following
direct transfection of antigen-presenting cells at the injection site34. This contrasts with the pDOM–peptide design,
which uses the cross-priming route35(FIG. 1). The route of priming appears to influence the outcome in terms of the
cellular responses against conventional vaccines120. To
address this, and to gain insight relevant for measuring responses to DNA vaccination, we undertook a study of CD4+ T-cell memory responses to tetanus toxoid
vacci-nation in normal subjects121. Together with the expected
positive response, we detected surprisingly dynamic fluctuations. This is apparently due to bystander stimu-lation of memory T cells against other antigens during the response to the injected antigen121. It explains the
anecdotal observation that infection with influenza can cross-stimulate the cellular response to tetanus toxoid122.
We need to be aware of these phenomena when planning assessments of immune responses to new vaccines.
Clinical trials of DNA fusion gene vaccines. Our first trial (UK-007, see Gene Therapy Clinical Trials Worldwide in Further information) was carried out in patients with follicular lymphoma in clinical remission following chemotherapy. Using individual DNA idio-typic scFv–FrC fusion vaccines (FIG. 2), we delivered naked DNA alone in 25 patients by intramuscular injection. The goal was to determine the safety, dose, immunogenicity and kinetics of response. Vaccination raised no safety concerns. Although patients were lymphopenic, most responded to the vaccine with no evidence of a dose effect between 500–2,500 µg/dose. Responses against FrC were found in 72% of patients. Cellular and humoral responses developed over a period of several months, with most responses seen by week 16. This appears to be slower than the response that is seen with conventional protein vaccines121,123,124.
However, data from rhesus macaques suggest similar slow kinetics125. We evaluated anti-idiotype responses in
16 patients and found that 6 out of 16 (38%) were able to make cellular and/or humoral responses. Clinical effects are more difficult to determine in relatively indolent fol-licular lymphoma and with small sample size, but are being assessed.
Although significant, responses were relatively low, confirming the need to increase performance in human subjects109. A promising strategy is EP, which in
pri-mates increases not only the level but also the breadth of response126. A number of companies are developing
clinical devices for this purpose and with different designs of the needle arrays that are used for delivery, voltage, pulse shape and pulse frequency. We are using a device (INOVIO AS, San Diego, California, USA), which employs a two-needle array and injects DNA during insertion of the needles. We are currently testing our DNA fusion vaccine that is specific for a peptide from prostate-specific membrane antigen in patients with prostate cancer at biochemical failure in a col-laborative clinical trial with the Royal Marsden Hospital. Surrey, UK (UK-112, see Gene Therapy Clinical Trials Worldwide in Further information). The trial has two arms, one using naked DNA alone, the other with DNA EP, with cross-over from DNA to DNA EP in some cases. Delivery of plasmid DNA with the ELGEN electropora-tion device has been well-tolerated and acceptable even with repeat dosing. The study is close to completing recruitment, and data from the two lowest dose levels already suggest that EP enhances antibody and CD4+
T-cell responses against the DOM 1 sequence. Preliminary analysis of CD8+ T-cell reactivity against the
prostate-specific membrane antigen target peptide using cultured ELISPOT for interferon γindicates significant responses in 3 out of 3 patients so far (J.R., C.H.O. and F.K.S., unpublished observations).
An important development since the first study has been the validation of our key immunological assays for clinical testing. As more assays become available and better defined, it is important to attempt standardiza-tion between centres. We are participating in a multi-centre collaboration of the Cancer Immunotherapy Consortium (see Further information), which aims to standardize cellular immune assays in European aca-demic institutions. Multiple assays are now being used in current trials. Clinical assessment includes measurement of PSA serum levels, computed tomography, magnetic resonance imaging and bone scans, full-blood count, biochemical monitoring and autoimmune profile.
Conclusions and future direction
The ability to manipulate genes has transformed medi-cal science, and it is tempting to use the same knowledge and technology to develop gene-based vaccination. This goal has united experts in cancer with those in infec-tious disease who share the task of activating immu-nity against difficult targets. In fact, the two currently licensed DNA vaccines are aimed against infectious organisms in horses127 and fish128. However, for
clini-cal trials of vaccines against cancer, initial enthusiasm turned to frustration with an apparent failure to trans-late promising vaccine designs from preclinical models into human subjects. The problem lay with delivery of DNA, and might now be solved by EP, which is a known way of increasing transfection in vitro and is now suc-cessfully applied in vivo. There is still much to do in terms of optimizing vaccine design, improving immune
Box 2 | Humanized mice
HHD mice express a hybrid human HLA-A2 transgene, comprising the α1 and α2 sequences of human with the α3 transmembrane and intracytoplasmic domain of mouse H-2Db (also known as H2-D1) class I molecules, covalently linked to human β-2
microglobulin (B2M). HHD mice lack mouse major histocompatibility complex (MHC) class I expression, owing to disruption of mouse H-2Db
and B2m genes, so thymic
education occurs on the human class I background83. This is advantageous as it
prevents bias in favour of H-2-restricted cytotoxic T-cell (CTL) responses, which are often observed in HLA-A2-transgenic mice in which normal murine MHC class I expression is intact. However, HHD mice are not strong responders, owing to a reduced peripheral CD8+ T-cell count (approximately fivefold) compared with normal mice and
lower-than-normal levels of expression of the transgenic HHD molecules, so the vaccine has to be efficient. However, this model does allow testing of human vaccines designed to induce HLA-A2-restricted CTL responses.
There is the disadvantage that the mouse CD8 protein does not interact with human MHC class I α3: therefore, once it is activated, it is difficult to test the ability of the mouse CTL to kill human tumour cells. However, this can be circumvented by transducing target cells with hybrid HHD and/or tumour antigen. In our hands this model has been ideal to test induction of CTL that is restricted by the HLA-A2 molecule and to assess their ability to kill target human tumour cells in vitro. HHD-expressing
recruitment and activation and selecting appropriate target antigens. Our main focus has been to modify the molecular form of the antigen and to activate T-cell help to reverse tolerance and induce high levels of immunity. However, there are many other strategies emerging from immunological knowledge that merit investigation. In fact, the bottleneck is likely to lie in
new and expanding regulatory restrictions, which slow the essential pilot clinical trials aimed to determine efficacy. The key is close linkage between laboratory and clinic, together with commercial vision. Only these three components together will lead to the rapid testing of new vaccine designs and delivery strategies which can effectively suppress cancer.
Table 1 | Clinical trials of DNA vaccination
Disease Encoded gene Delivery Phase and patient
group
Outcome Refs
Immunological Clinical
Melanoma Gp100 1 mg DNA per dose, 4
doses, monthly, i.m. (12 patients), i.d. (10 patients)
Phase I, 22 patients, progressive measurable metastatic disease. (6 pre-vaccinated in other studies)
13 evaluable, no cellular responses to vaccine.
1 clinical PR 135
Melanoma HLA B7 MHCI 2, 9, 90 mg DNA per liposome, 3 doses, biweekly, i.l.
Phase I, 10 patients, stage IV disease, progressive
Enhanced TIL cytotoxicity Some response of injected lesions, one patient CR after infusion of TIL from modified tumour
136
Melanoma Tyrosinase epitopes 0.2, 0.4, 0.8 µg DNA per dose, 4 doses, biweekly, i.n.
Phase I, 26 patients, stage IV disease
11/24 immune
responders, 6/24 skin test positive
No clinical responses, unexpectedly long survival
137
Melanoma Mouse tyrosinase and human tyrosinase
0.1, 0.5, 1.5 mg DNA per dose, 6 doses, three-weekly, i.m. (biojector)
Phase I, 18 patients, stage III (n=16) disease
in remission, or slowly progressive stage IV (n=2)
7/18 detectable CD8+
T-cell responses
6/7 responders alive at median follow-up of 42 months, 6/11 non-responders
112
Melanoma, adjuvant
MART1 0.1, 0.3, 1 mg DNA per
dose, 4 doses, 6-weekly, i.m.
Phase I, 12 patients No immune responses detected
NE 138
Melanoma, adjuvant
Polyepitope vaccine, (DNA Mel3), MVA boost
1 mg DNA per dose or 5 x 107 pfu MVA per
dose, 4 doses, biweekly, i.m. (DNA), i.d. (MVA), DNA/DNA/MVA/MVA or MVA x 4
Phase I, 14 patients, in remission
Melan-A26-35 CD8+ T-cell
responses: 2/6 DNA/MVA after MVA boost, 4/7 after MVA only
NE 139
Colorectal cancer
CEA and HBsAg (dual expression)
0.1, 0.3, 1, 2 mg DNA per dose, 3 doses, 3-weekly, i.m.
Phase I, 17 patients, metastatic disease
Antibody responses against HBsAg, 4/17 proliferative responses to rCEA
No clinical responses
140
Prostate cancer
PSMA extracellular domain and CD86 (dual expression or separate), Ad5 boost
0.1-0.8 µg DNA per dose or 5 x 108 pfu Ad5 per
dose, weekly, i.d. (+/-GMCSF)
Phase I/II, 26 patients (16 with metastases)
86% anti-PSMA humoral responses following prime–boost, majority DTH-like response to plasmid at injection site
4 PSA responders 141, 110
Prostate cancer
PSA 0.1, 0.3, 0.9mg DNA per
dose, 5 doses, 4-weekly, i.m. or i.d. (GMCSF, IL2)
Phase I, 9 patients 4/6 Elispot (cultured) 2/8 PSA responses (1 transient)
142, 143
Anal dysplasia
HPV 16E7 string of epitopes
50-400 µg DNA per microparticle per dose, 4 doses, three-weekly, i.m.
Phase I, 12 patients 10/12 Elispot responses (ex vivo)
3 partial histological responses
111
B-cell lymphoma
Idiotypic vaccine, VH or VL linked to xenogeneic (mouse) Ig constant region
0.2, 0.6, 1.8 mg DNA per dose, 3 monthly doses i.m., then 1.8 mg DNA i.m. or i.d. with biojector; then with GMCSF
Phase I, 12 patients, in remission
7/12 xenogeneic responses (humoral, cellular), 1/12 anti-Id response
NE 64
CEA, carcinoembryonic antigen; CR, complete remission; DTH, delayed-type sensitivity; GMCSF, granulocyte–macrophage colony-stimulating factor; HBsAg, hepatitis B surface antigen; HLA, human leukocyte antigen; HPV, human papillomavirus; i.d., intradermal; Id; idiotypic; Ig, immunoglobulin; i.l., intralesional; IL2, interleukin 2; i.m., intramuscular; i.n., intranodal; MHCI, major histocompatibility class I; MVA, modified vaccinia Ankara; NE, not evaluable (patients in remission at time of vaccination); pfu, plaque-forming units; PR, partial response; PSA, prostate-specific antigen; PSMA, prostate-specific membrane antigen; rCEA,
