Genetic Vaccination for the Active Immunotherapy of Cancer

INTRODUCTION

One of the most important scientific discoveries of the last ten years has been the unambiguous definition of tumor associated antigens (TAA) recognized by human T lymphocytes. In addition to representing the long-awaited legitimization of tumor immunology, this discovery has practical implications for understanding the dynamics of the immune responses against endogenously expressed antigens. TAA are processed intracellularly into short peptides that are then presented by Class I and Class II molecules of the major histocompatibility complex (MHC). An examination of the properties and nature of the constantly expanding variety of TAA is beyond the scope of this article and is addressed in other exhaustive reviews [Robbins and Kawakami, 1996; Rosenberg, 1999; Van den Eynde and van der Bruggen, 1997]. However, from an immunological point of view, which is that of the present paper, TAA can be divided into three main groups:

1) viral, mutated or aberrantly expressed antigens (CDK4, b-catenin, GnT-V, Casp8);

2) cancer/testis-specific antigens (MAGE, BAGE, GAGE, PRAME and NY-ESO-1);

3) differentiation antigens (tyrosinase, Melan-A/ MART-1, gp100, TRP-1/gp75 and TRP-2) and overexpressed antigens (Her2/neu, CEA, p53);

TAA arising from mutations of genes that are crucial for cancer pathogenesis and progression are ideal candidates for the generation of tumor-specific vaccines. However, except for the frequent mutations occurring in particular “hot spots” of some oncogenes and anti-oncogenes (i.e. RAS oncogene), the individual natures of these TAA make their use at present unlikely. Viral antigens are present in a limited group of human tumors that are closely associated with infectious agents such as cervical carcinoma (Human Papillomavirus, HPV), Burkitt’s lymphoma (Epstein Barr Virus, EBV), and hepatocellular carcinoma (Hepatitis B Virus, HBV). Shared TAA belonging to the second and third groups are the most likely choice for tumor vaccine development. The second group includes antigens shared by tumors with distinct histology and origin, but also present in the testis and placenta. These TAA should be highly immunogenic because their HLA Class I binding epitopes are normally presented only in tumors. In fact, the placenta is a temporary anatomical structure, and germ cells in the testes are not recognized by the immune system because they are isolated by anatomical barriers and lack HLA Class I surface expression [Haas et al., 1988; Uyttenhove et al., 1997]. The third group includes the antigens most frequently recognized on human melanomas, as well as the TAA that are expressed at different levels in neoplastic and normal tissues. The development of an immune response to these antigens is quite unexpected because it occurs despite the mechanisms of central and peripheral tolerance to “self” antigens.

The identification and molecular characterization of TAA has also provided the means to create cancer vaccines based on the insertion of the TAA coding sequences into suitable bacterial or viral expression vectors. The term recombinant will henceforth be used to indicate a microorganism that carries an exogenous genetic information by virtue of laboratory manipulation (i.e. recombinant technology). Recombinant vaccines are thus vaccines in which the DNA of the host vector integrates a heterologous DNA sequence coding for an immunogen. Viral and bacterial vectors that reach the cytoplasm have been exploited to deliver the antigen to the intracellular route of processing in order to be presented with Class I MHC molecules, and elicit CTL responses that were thought to be important for antitumor activity. However, recent studies have shown that a vector is not necessary for vaccine preparation. Immunization with “naked” plasmid DNA (i.e. devoid of any other structural components as proteins, lipids, or carbohydrates), and to some extent with RNA can elicit powerful cellular and antibody responses, provided that the gene is under the control of a promoter allowing gene transcription in eukaryotic cells. Compared to cell-based vaccines or cell lysates, recombinant vaccines have multiple advantages: they can focus the immune response against a single, specific TAA, and thus limit the possibility of releasing an uncontrolled autoimmune aggression against unknown antigens shared by normal tissues and tumor cells; can be manufactured on a large scale, stored, handled, and administered to patients in different clinical centers, thus avoiding the complex manipulation of tumor samples that is required to generate cell vaccines and feasible in only a few selected laboratories; and permit adjuvants, routes of inoculation, and boosting procedures to be arranged in various ways. Of course, the perfect vector does not exist, and every choice has its own liabilities. The advantages/ disadvantages of various vectors that can be employed for genetic vaccinations are summarized in (Table 1).

Viruses

Viruses are very attractive vectors because they can activate both arms of the immune response, and elicit antibody, T helper (Th), and cytotoxic T lymphocytes (CTL) in the absence of costimulation [Bachmann et al., 1998]. It is likely that their long-lasting coevolution with human beings allowed the selection of specific patterns that are recognized by the innate immune system (Pathogen-Associated Molecular Patterns, PAMP), and create an optimal immunostimulatory environment to trigger the immune response. The ultimate goal of viruses is not to kill the host, but to have enough time to replicate and spread among individuals. Viruses have developed various strategies to hamper the immune response [Alcami and Koszinowski, 2000], but the majority of human beings nonetheless can resolve a viral infection by mobilizing various innate and specific immune cells. Several viruses are potential candidates for cancer immunotherapy, which makes the choice complex. The parameters relevant to safe virus utilization include recombination with wild-type viruses, insertional and oncogenic potential, and virus-induced immunosuppression. Live attenuated viruses selected for their reduced pathogenicity could potentially undergo back mutations and revert to a virulent phenotype, thus provoking diseases or shedding among individuals in confined communities. This potential risk can be greatly reduced by the targeted deletion of virulence and host range genes in viral vectors [Paoletti et al., 1995].

Viral vectors are not equivalent in terms of the immune responses they elicit. While antibodies are thought to be the main protective mechanism against cytopathic viruses, CTL are the first line of defense against noncytopathic virus [Zinkernagel et al., 1996]. Analogously, according to their cytopathic potential different viruses or viral vectors might preferentially activate different populations of the immune system.

Poxviruses

Poxviruses are enveloped viruses with a linear double-stranded DNA genome. Large amounts of foreign DNA (up to 30 kb) can be stably inserted into the poxvirus genome by homologous recombination. Poxviruses employ a built-in transcriptional and post-translational apparatus to produce large amounts of the protein encoded by the inserted gene [Moss, 1996; Perkus et al., 1995]. Unique among the DNA viruses, poxviruses sojourn entirely within the host cell cytoplasm which virtually eliminates the possibility of an oncogenic insertion into the host genome. Poxviruses have acquired a wide array of measures to evade or counteract the immune system, but immunization with poxviruses has been one of the most successful strategies against human disease. Vaccinia virus (VV), the prototype of the Poxviridae family, was used in worldwide campaigns against smallpox; to date, no other large-scale vaccination program has had such an impact on human disease, because smallpox has been practically eradicated. The possible complications associated with VV immunization especially in immunocompromised patients have led to the development of several attenuated poxiruses with an incomplete replication cycle in mammalian cells. NYVAC was realized by deleting 18 different genes that are important for the pathogenicity, and the regulation of host range infectivity [Gonin et al., 1996]. A highly attenuated strain of VV, known as modified VV Ankara (MVA), was inoculated as smallpox vaccine in more than 120,000 recipients without causing any significant side effects [Mahnel and Mayr, 1994]. MVA replication is blocked at the stage of virion assembly [Sutter and Moss, 1992], and for this reason the MVA vectors produce recombinant proteins expressed under the control of both early and late viral promoters, thus mimicking the expression of wild-type virus. To further improve the safety of poxvirus vectors, several research groups have also engineered poxviruses which normally do not infect human beings. The Avipoxviridae family comprises viruses such as fowlpox (FPV) and canarypox, which can productively infect avian but not mammalian cells, are not crossreactive with VV [Cox et al., 1993; Perkus et al., 1995], and thus could be used in patients previously exposed to VV. Mice immunized with canarypox virus encoding p53 were protected from challenge with a tumor cell line expressing high levels of mutant p53, and protection was effective regardless of whether wild-type or mutant p53 was used for the immunization [Roth et al., 1996]. More recently, Hodge et al. [Hodge et al., 1997] evaluated the efficacy of a recombinant canarypox virus expressing human CEA alone, or in combinantion with recombinant (r)VV-CEA. An immunization schedule based on inoculation with rVV followed by recombinant canarypox virus acted better than each single immunogen in eliciting CEA-specific immune responses. Importantly, it was possible to administer multiple boosts of recombinant canarypox virus, and thus increase the T cell response and the antitumor effects, a further advantage of avipoxviruses that is precluded by VV-based vaccines.

Adenoviruses

The adenovirus is a non-enveloped, icosahedral virus containing a linear stretch of double-stranded DNA. The virus replication and assembly of virions occurs entirely in the nucleus of infected cells, and the production of complete virions results in the death of the host cell with the release of large amounts of infectious particles. Adenovirus-based vectors offer an interesting array of choices for gene transfer applications. The first-generation of these vectors was based on the deletion of the E1 and E3 regions which then accomodated the inserted foreign DNA. These vectors are defective for replication in normal cells, infect both replicating and non replicating cells, accommodate a limited amount of heterologous DNA (up to 7.5 kb), and can be grown at very high titers in 293 cells that constitutively express the E1 region product [Imler, 1995]. High titers of adenoviruses are usually necessary for immunotherapy applications. First generation adenoviral vectors have another characteristic that make them a suitable vaccine candidate: since most of the viral genes are retained, they elicit a strong cellular and humoral immune response in vivo, resulting in a transient gene expression [Gonin et al., 1996; Xiang et al., 1996]. Of course, these same properties make them poorly practical for prolonged gene transfer. This limitation led to the development of the later generation of adenoviral vectors, the fully deleted vectors or ‘helper-dependent’ vectors that are grown in host cells super-infected with a helper virus that provides in trans the viral functions and proteins required for virus assembly. Helper-dependent adenoviruses can host up to 36 kb, the space left by the deletion of the entire viral genome [Chen et al., 1997; Morral et al., 1999].

The major limitation of adenovirus use in cancer immunotherapy is the fact that they are ubiquitous and constantly infect human beings, so that the majority of normal individuals has neutralizing serum antibodies against the most common adenovirus serotypes. This pre-exisiting immune response precludes reiterated systemic administration of adenovirus-based vaccines. Moreover, liver toxicity was described following systemic administration of high titers of first generation, E1-deleted adenovirus vectors optimized for gene therapy [O'Neal et al., 1998]. These side effects, which indeed raise some concerns about adenovirus administration in patients might not restrict their use in cancer immunotherapy because the systemic delivery of high viral titers would certainly not be the favored immunization route.

Alphaviruses

The alphavirus genus include viruses like Sindbis, Semliki Forest and the Venezuelan equine encephalitis viruses whose genome is a capped, polyadenylated, single-strand of positive sense RNA enclosed in an icoshedral capside. Viral RNA molecules are self-replicating because at the 5' end they encode proteins required for replication and transcription of the viral RNA (replicase unit). Following cell infection, the replicase is translated directly from genomic RNA into a polypeptide with the enzymatic ability to cleave itself into four subunits (nonstructural proteins, nsP1-4). The complex then synthesizes the negative-strand RNA that is used as a template for the synthesis of the positive-strand, the genomic RNA. Replicase also binds to a strong internal promoter to transcribe a subgenomic RNA that encodes the structural proteins necessary for final virus assembly. In alphavirus replicon vectors, a heterologous gene replaces the genes encoding the viral structural proteins. Alphaviruses have many features that make them attractive cancer vaccine candidates: they infect a wide range of cell types from different species; they have very high levels of replication and gene expression; their cell cycle is restricted to the cytoplasm without the intervention of the host cell replication/translation apparatus; replicon genomes are not packaged and do not spread in the absence of the structural proteins; and they induce a suicidal infection, i.e. the cell harboring the virus is eventually induced to undergo apoptosis [Garoff and Li, 1998; Lundstrom, 1997]. Replicon vectors can be introduced into cells in various ways: as RNA transcribed in vitro from plasmids in which the replicon is under the control of a bacteriophage promoter; as DNA under the control of a eukaryotic promoter, usually CMV; and as virus particles obtained from virion packaging in cells where the structural proteins are provided in trans by a defective helper genome [Berglund et al., 1998; Smerdou and Liljestrom, 1999; Ying et al., 1999].

Other Viruses

Herpes simplex virus (HSV) is an enveloped double-strand DNA virus that replicates in the nucleus. Its large genome (150 kb) encodes about 80 different genes, and can accommodate about 30 kb of foreign genes. The HSV genome remains episomal after infection, thus eliminating the risk of insertional mutagenesis. Replication-incompetent viruses have been developed by deleting the immediate-early genes, by disabling the glycoprotein H, or by engineering HSV amplicons [Krisky et al., 1998; Spaete and Frenkel, 1982; Todryk et al., 1999]. These vectors have been used mainly for gene transfer, or for tumor killing due to their intrinsic oncolytic potential that can be increased by the insertion of suicide genes or cytokines. From an immunological viewpoint, HSV-based vectors raise some concern because the virus has been shown to block dendritic cell (DC) maturation, proceed to a latent state of infection in the central nervous system, and release viral proteins that interfere with complement activation and binding of immuno-globulin to Fc-receptors [Alcami and Koszinowski, 2000; Glorioso et al., 1995].

Adeno-associated virus (AAV) is a single-strand DNA virus whose replication in mammalian cells is deficient, and requires coinfection with a helper virus, such as adenovirus or HSV. AAV can integrate its genome in a specific locus on human chromosome 19. AAV-derived vectors infect both dividing and nondividing cells, resulting in prolonged transgene expression, are not toxic to infected cells, and do not induce a host immune response [Monahan and Samulski, 2000]. Their use as a cancer vaccine appears to be limited.

Overcoming Viral Vector Weaknesses

Some anticipated drawbacks of recombinant viruses were highlighted by the first clinical trials in cancer patients; indeed, many patients had neutralizing antibodies against VV due to childhood vaccination for smallpox prevention, and against Adenoviruses, which cause repeated upper respiratory tract infections throughout the life. The high serum titers of neutralizing antibody directed against adenoviral antigens are probably responsible for the absence of immunization against human melanoma antigens following administration of recombinant adenoviruses expressing the antigens MART-1 and gp100 [Rosenberg et al., 1998b]. However, strategies to overcome pre-existing immunity to recombinant viruses have been developed that could avoid the rejection of viral recombinant vectors for genetic vaccination.

The mucosal route of administration could be used to overcome pre-existing immunity to poxviruses. Whereas systemic immunization with rVV expressing HIV gp160 was totally ineffective in mice previously exposed to VV, intrarectal immunization induced specific serum antibody and strong HIV-specific CTL responses that could be detected in mucosal and systemic lymphoid tissue [Belyakov et al., 1999].

Genetic manipulation of DC in vitro is another strategy that could be used to surmount immunity in individuals previously exposed to the viruses. The immunization of mice with adenoviruses usually results in the induction of high titers of neutralizing antibodies, whose presence, however, did not affect the immunogenicity of DC infected in vitro with the recombinant adenovirus expressing the antigen ovalbumin (OVA); indeed, repeated administration of virus-infected DC elicited an anti-OVA CTL response even in mice previously infected with the empty recombinant vector [Brossart et al., 1997]. It was also shown that protective immunity against mouse melanoma “self” antigens, gp100 and TRP-2, was obtained in mice that had been preimmunized against adenovirus by using DCs transduced with an adenoviral vector encoding these antigens [Kaplan et al., 1999]. Gene transfer to different human DC subpopulations by vaccinia and adenovirus vectors was shown to be a feasible strategy for TAA loading [Di Nicola et al., 1998].

The availability of an array of non-cross-reactive viral vectors allows the administration of different recombinant vaccines encoding the same TAA; these vaccines can be alternated to boost the antitumor immune response and minimize the detrimental effects of serological and cellular responses to the vector antigen components [Wang et al., 1995a]. MVA is a very promising vector for the development of recombinant vaccines for cancer, and can be efficiently used in combination with DNA vaccines. Priming with plasmid DNA encoding Plasmodium berghei antigen followed by a single boost with a recombinant MVA expressing the same antigen induced complete protection against challenge with the sporozoite; unexpectedly, protection was abrogated when the order of immunization was reversed [Schneider et al., 1998]. DNA priming followed by MVA boosting is currently being investigated in a number of preclinical tumor models in mice, and might provide clinical benefits through the induction of high levels of CD8⁺ T cells [Ramshaw and Ramsay, 2000]; other mixed regimens associating viruses with viruses, or proteins are also being tested in clinical trials [Bonnet et al., 2000]. However, a clear-cut advantage of the DNA/MVA prime/boost regime in tumor models has yet to be shown. Indeed, when a large panel of recombinant immunogens encoding a model tumor antigen were tested to compare the best therapeutic effect on established pulmonary metastases, heterologous boosting with FPV and VV, independently of the order of inoculation, had a synergistic effect on the survival of tumor-bearing mice that was significantly higher than that obtained with DNA/FPV or DNA/VV combinations [Irvine et al., 1997].

Attenuated intracellular bacteria

Heterologous DNA sequences under the transcriptional control of a promoter active in eukaryotic cells can be loaded into attenuated bacteria, thus obtaining two major benefits: immunogenicity can be improved because bacteria possess many PAMP, some of which are known such as lipopolysaccharide (LPS) and lipotheicoic acid, and others are yet to be disclosed; the mucosal route of administration can be exploited. Strains of Shigella flexneri and Salmonella typhimurium have been attenuated by deletion of genes involved in cell wall synthesis, and used to deliver eukaryotic expression vectors in the cytoplasm of infected cells [Shata et al., 2000]. These auxotrophic bacteria can be easily transformed with eukaryotic expression vectors provided that they include the E. coli origin of replication Following infection of mammalian cells, auxotrophic bacteria undergo autolysis due to the deficient cell wall synthesis, but they find a way to release the plasmid DNA inside the cytosol. S. flexneri enters the host cell cytosol from the phagosome by means of an invasin gene product [Sizemore et al., 1995]. Similarly, gram-negative bacteria have been engineered to release the antigen via the a-hemolysin secretion system of E. coli. Conversely, the S. typhimurium strain remains in the phagosome, but nonetheless delivers its eukaryotic expression vector into the host cell cytoplasm through an unknown pathway [Darji et al., 1997]. When given orally, Salmonella-based vaccine targeted the expression of a reporter gene with high efficiency into about 20 % of splenocytes comprising both F4/80⁺ macrophages and CD11c⁺ DC [Paglia et al., 1998]. In this study, the oral vaccination induced T-cell responses and elevated serum antibody titers to the b-galactosidase (b-gal) antigen in immunized mice.

Gram-positive bacteria can also be used as plasmid DNA carriers. Listeria monocytogenes can target the cytosol of splenic APC following oral administration, and can thus convey the antigen to both the Class I MHC pathway of endogenous antigen presentation and the Class II pathway of exogenous antigen presentation [Ikonomidis et al., 1994; Shen et al., 1995]. L. monocytogenes has been attenuated by deletion of the genes responsible for either intracellular mobility, or cell-to-cell spreading after bacterial escape from the host cell phagosome. Delivery of plasmid vectors from the macrophage phagosome to the cytosol depends on the L. monocytogenes autolysis caused by the production of PactA-dependent Listeria-specific phage lysine [Dietrich et al., 1998]. Interestingly, the transferred plasmid DNA integrated into the host cell DNA at a detectable frequency.

Differences in the mechanism of immune protection between viral and bacterial vaccines have been reported, and must be addressed when choosing the vector. For example, immunization with lymphocytic choriomeningitis virus (LCMV) or recombinantL. monocytogenes expressing the LCMV nucleoprotein (NP) antigen induced a comparable increase in antigen-specific CTL precursor frequency, as assessed by the ability to lyse target cells pulsed with NP-derived peptides and by limiting dilution analysis. However, only LCMV vaccination could efficiently protect from a lethal choriomeningitis induced by intracerebral injection of LCMV, or from infection with an rVV encoding the LCMV NP gene [Ochsenbein et al., 1999]. Despite analogous precursor frequencies, the effector properties of the memory CTL generated following recombinant Listeria vaccine adminis-tration were short lived. This dichotomy between in vitro assays evaluating CTL function, and in vivo antiviral properties reflects a frequently observed paradox in immunotherapy, as discussed in greater detail below.

DNA

Following the first report that the intramuscular injection of “naked” DNA encoding the influenza NP could induce NP-specific CTL, and protect mice from challenge with heterologous influenza strains [Ulmer et al., 1993], DNA immunization has gained momentum and can now be considered one of the most rapidly evolving technologies for cancer vaccination. There are several reason for this success. DNA provides a stable and long-lasting source of antigen, and DNA immunization is a simple and efficacious way to elicit both antibody- and cell-mediated immune responses. Compared to recombinant viruses, DNA vaccines offer a number of potential advantages because they are cheap, easy to produce, and do not require special storage or handling. DNA vaccines can be constructed to express only the heterologous gene in order to induce a selective immune response, and allow repeated inoculations in the same patient. DNA vaccination has proven to be a generally applicable approach to various preclinical animal models of infectious and noninfectious diseases (reviewed in [Donnelly et al., 1997; Gurunathan et al., 2000]), and several DNA vaccines have now entered phase I/II human clinical trials. Although the clinical application of DNA vaccines is a very young practice, some trials have proved that antigen-specific CTL can be elicited against malaria and HIV proteins in human volunteers [Calarota et al., 1998; Wang et al., 1998].

The preparation of DNA vaccines requires different steps: 1) cloning of a heterologous gene under the control of a promoter active in eukaryotic cells and usually derived from the cytomegalovirus (CMV) immediate early region; 2) purification of the endotoxin-free DNA plasmid from bacteria “factories”; 3) administration of the expression vectors by direct intramuscular or intradermal injection with a hypodermic needle, or using a helium-driven, “gene gun” to bombard the skin with DNA-coated gold beads. Heterologous DNA can also be introduced into recombinant Salmonella, or Listeria strains and can thus be administered by a mucosal route, as previously discussed. In addition to these classic routes of DNA delivery, plasmid-based gene expression vectors have also been admixed with polymers and administered with a needle-free injection device, achieving high and sustained levels of antigen-specific antibodies [Anwer et al., 1999]. The route of DNA delivery is not a secondary aspect of DNA immunization because it appears to determine the type of immune response by preferentially activating different Th populations; gene gun bombardment elicits a prevalent Th2 response, while intramuscular injection induces Th1 activation, even though the form of the antigen (i.e., membrane-bound vs secreted) can also have an important effect [Feltquate et al., 1997; Pertmer et al., 1996].

The mechanism of DNA-induced immunization has not yet been fully elucidated. An exclusive role for the direct transfection of normal tissue cells has been questioned, as surgical ablation of the injected muscle within 1 minute of DNA inoculation did not affect the magnitude and longevity of DNA-induced antibody [Torres et al., 1997]. Moreover, studies with bone marrow chimeras clearly indicated that bone marrow-derived APC, either transfected by the DNA plasmid or able to capture the antigen expressed by other transfected cells, were necessary to prime T and B lymphocyte responses [Iwasaki et al., 1997]. Indeed, more recent evidence suggest that Th and CTL are activated by DC directly transfected in vivo following DNA immunization [Akbari et al., 1999; Porgador et al., 1998].

In general, the potency of naked DNA does not equal that of recombinant viruses, probably because DNA does not undergo a replicative amplification in the transfected cells, which in turns limits the amount of heterologous antigens produced. Inflammatory responses caused by DNA inoculation are more contained that those occurring during infection with viruses; for this reason, repeated inoculations of plasmid DNA, or the use of adjuvants are generally required for the induction of an optimal response.

RNA

As previously discussed, replicon-based RNA vaccines can immunize against specific TAA. However, since few TAA candidates are currently available, the number of tumors that could be treated with antigen-specific immunotherapy is circumscribed. To obviate this limitation, E. Gilboa and collaborators have developed an interesting application of genetic vaccination based on RNA. They demonstrated that splenic DC pulsed with chicken OVA mRNA, synthesized and capped in vitro, elicited OVA-specific CTL and conferred protection to mice challenged with OVA-expressing tumor cells [Boczkowski et al., 1996]. DC could also be pulsed with total, unfractionated RNA extracted from OVA-positive tumors, and obtain a comparable therapeutic effect. From these initial findings, a widely applicable strategy was derived by pulsing DC with total or polyA⁺ tumor-derived RNA isolated from poorly immunogenic mouse tumors such as melanoma B16. Using tumor-RNA-loaded DC an increase in the mean survival time was reported for mice bearing either pulmonary or central nervous system localization of this very aggressive melanoma [Ashley et al., 1997; Boczkowski et al., 1996]. Because the mRNA content of single cells can be amplified, tumor mRNA or corresponding cDNA libraries supply an unlimited source of tumor antigens that could be useful for genetic vaccination, but the stability of the RNA might represent an important drawback to this vaccine.

A polypeptide with a catalytic reverse transcriptase activity is a subunit of the telomerase ribonucleoprotein complex (TERT), a TAA recently shown to be recognized by human T lymphocytes in the context of the HLA-A2 molecule. Telomerase is reactivated in the majority of mouse and human cancers, while it is silent in normal tissues, thus possessing the optimal requirements for a truly tumor-restricted and widely shared antigen [Minev et al., 2000; Vonderheide et al., 1999]. DC loaded with in vitro transcribed mouse TERT RNA elicited CTL that recognized a mouse melanoma and a thymoma that both displayed measurable TERT activity. The antitumor protection obtained with this vaccine was efficacious in three different combinations of mouse strains and tumors, one of the few examples of the widespread impact of active immunotherapy with a single TAA vaccine [Nair et al., 2000]. However, DC transfected with RNA derived from a TERT-positive tumor provided a greater tumor protection than DC pulsed with just the TERT RNA. Moreover, anti-TERT specific CTL were not induced by tumor-RNA-loaded DC, and the immune response was tumor-specific, i.e. restricted to the very tumor whose RNA was extracted to pulse DC. These data raise some interesting points of speculation. First, TERT appears to not be an immunodominant antigen in mice (but it could be different in humans). Second, multiple unknown antigens present in the total tumor RNA extracts might elicit a poly-clonal response whose therapeutic potential is greater than an oligo-monoclonal response directed towards a single antigen like TERT. It must be considered that some of the most impressive effects on the growth of established mouse tumors were obtained by immunization with either DC pulsed with unfractionated tumor-derived peptides, or specia-lized extracellular vescicles called “exosomes” derived from DC previously pulsed with tumor lysate or TAA peptides [Zitvogel et al., 1996; Zitvogel et al., 1998]. The importance of poly-immunization in immunotherapy of cancer will be discussed below.

Genetic vaccination applied to transplantable tumors

The efficacy of genetic vaccination against experimental tumors was initially assessed using exogenous antigens such as b-gal from Escherichia coli, chicken OVA, hemoagglutinin (HA) from influenza virus, and NP from vescicular stomatitis virus. The exogenous genes were placed into tumor cells, and cloned into suitable vectors for construction of recombinant vaccines (Table 2). These artificial antigens were operationally employed as model TAA, but the relevance of their immunogenicity must be considered with some restrictions. Although model TAA are usually antigens that can elicit Th and CTL responses to some immunodominant peptides, their transduction into a tumor cell line is not sufficient per se to affect tumor growth rate and lethality. When expressed by tumor cells, these TAA are not able to induce a measurable immune response before the tumor size precludes mouse survival. However, model TAA are not true, “self” antigens as they cannot shape the T lymphocyte repertoire by causing clonal deletion because they are not present in the thymus, and peripheral tolerance can be sustained only by the tumor cells, and not by normal tissues as in the case of shared, differentiation TAA. Moreover, epitopes of TAA recognized by CTL on the surface of tumor cells could be inefficiently generated by professional APC since they possess an immunoproteasome in which some catalytic subunits of the standard proteasome are replaced by the IFN-g-induced homologs. Peptides generated from the processing of the same cytoplasmic protein can thus be different in tumor cells and in APC, as demonstrated for some human melanoma antigens [Morel et al., 2000]. For the model TAA, a complete correspondence was found for the epitopes presented by tumor cells and those generated following immunization. These main differences between model and “true”, self TAA are summarized in the following diagram.

Model TAA might be considered as surrogates of the viral antigens (i.e. EBV, HPV, HBV) expressed by some human tumors, with the limit that they interact with the immune system only following tumor implantation and not before tumor development.

	Model tumor antigen	Self tumor antigen
Epitope definition	TH and CTL	Mostly CTL
Central (thymic) tolerance	Not possible	Possible
Peripheral tolerance	Limited to that induced by same minor tumor cells	Induced by tissues sharing the antigen
Immunoproteasome-generated epitopes are identical to those presented by the tumor cells	Always	Not always
Therapy following vaccination	Observed	Rare
Tumor evasion by antigen loss	Frequent	Not deeply investigated

Perhaps the most interesting lesson on the use of cancer vaccines encoding model TAA concerns the factors determining successful therapy of established tumors. Recombinant viruses, such as VV, MVA, FPV, and adenoviruses encoding b-gal, have been used to eradicate established pulmonary metastases of an adenocarcinoma cell line expressing b-gal (reviewed in [Restifo, 1996]). The therapeutic effect could manifest on 6 day-old pulmonary metastases that are macroscopically visible, and it was optimal when immunomodulatory molecules (such as cytokines and costimulatory molecules) were inserted into the viruses, or when the recombinant viruses were coadministered with exogenous Th2-type cytokines, such as IL-2, IL-12 [Bronte et al., 1995; Chamberlain et al., 1996; Chen et al., 1996; Rao et al., 1996; Wang et al., 1995a; Wang et al., 1995b]. Surprisingly, Th2-type cytokines could also increase the therapeutic efficacy of rVV encoding a model TAA [Kaufman et al., 1999]. DNA vaccines were not equally potent in eliciting immune responses capable of destroying established tumors. Immunization with a plasmid encoding b-gal prevented the growth of pulmonary metastatic tumor, but had little or no impact on the growth of established lung metastases [Irvine et al., 1996]. A significant reduction in the number of established metastases was observed only when cytokines, such as IL-12 or IL-2, were administered following DNA inoculation. The therapeutic efficacy of cancer vaccines is frequently correlated with the induction of a strong anti-tumor response mediated by CD8⁺ lymphocytes [Bronte et al., 1995; Porgador et al., 1998]. Compared to proteins or peptides that undergo processing and presentation in association with MHC molecules, genetic vaccines must go through an additional step: the synthesis of the antigen protein by the transcriptional/ translational apparatus of the APC in the recipient organism. More than the overall rate of antigen production detected in vitro or in vivo in somatic cells, in fact, the ability of the vector to target TAA expression to DC is considered crucial for the induction of an effective anti-tumor immune response [Bronte et al., 1997; Porgador et al., 1998]. For example, in creating a rVV, a model TAA gene had to be placed under the control of a vaccinia early promoter which allowed its expression in professional APC like DC, although late promoters drove up to 50-fold more antigen production in permissive cells [Bronte et al., 1997].

Despite the efficacy of cancer vaccines in treating established tumors or preventing tumor challenge, recurrence has been observed after an initial control of tumor growth [Bronte et al., 1995; Matsui et al., 1999]. Escape from immune surveillance occurred because: a) TAA-loss variants generated by a silencing of the promoter driving the transcription of the model TAA emerged; b) a mutation in model TAA gene sequence occurred; c) there was a block at the level of mRNA expression ([Bronte et al., 1995; Carroll et al., 1997; Matsui et al., 1999], and unpublished results). Interestingly, the recurrence could develop in spite of the presence of tumor-specific CTL, and could be drastically reduced by elimination of the CD4⁺ lymphocytes that exerted a negative effect on CD8⁺ lymphocyte function [Matsui et al., 1999], thus suggesting that the initial state of antitumor immunity must be kept “alert” for some time after the initial rejection for complete tumor eradication.

Unlike model TAA, vaccination protocols targeting the immune response to self TAA have to deal with the mechanism of central tolerance that keeps the function and the number of self-reactive T lymphocytes under control, and leaves only T lymphocytes with low affinity T-cell receptors [Ashton-Rickardt et al., 1994; Hoffmann et al., 1995]. To address the immunological impact of the endogenous expression of TAA, mouse models of active immunization against melanocyte differen-tiation antigens (MDA) have been recently explored by different groups (Table 3). It is easy to understand why MDA were chosen for the mouse studies as this group of antigens comprises the TAA most frequently recognized by T lymphocytes on human melanomas.

The prominent finding of mouse studies with MDA is that peripheral tolerance is not an absolute limit since it can be broken by xenoimmunization, or immunization with an altered source of the antigen. Immunization with a rVV encoding the human gp100/pmel-17, or a plasmid DNA encoding the human TRP-1/gp75, elicited an immune response against the respective mouse homologues [Overwijk et al., 1998; Weber et al., 1998]. Moreover, tolerance to mouse TRP-1/gp75 could also be broken by administering the mouse antigen in an altered form, such as the mouse protein expressed in insect cells or produced during infection with a rVV [Naftzger et al., 1996; Overwijk et al., 1999]. Another lesson from these investigations is that different effector mechanisms are responsible for the recognition and elimination of melanoma cells. The immune response to TRP-1/gp75 is able to control tumor challenge, and even cause rejection of a small tumor burden through a complex immune network involving CD4⁺ lymphocytes, antibodies, NK1.1 cells, and Fc receptor b–chains: melanoma cells targeted by specific antibodies are likely eliminated by ADCC-mediated lysis [Clynes et al., 1998]. Importantly, the induction of an effective immune response against TRP-1/gp75 was associated with autoimmune skin depigmentation (vitiligo) caused by the destruction of the normal melanocytes [Hara et al., 1995; Hawkins et al., 2000; Overwijk et al., 1999].

The anti-TRP-1 response appears to be dominated by antibodies, one of the few examples of this kind in the literature. Even the simple passive transfer of mouse antibodies against TRP-1 to tumor-bearing mice caused rejection of established melanomas [Hara et al., 1995]. Conversely, TRP-2 antigen is the main target of mouse CTL generated following immunization with irradiated melanoma cell vaccine [Bloom et al., 1997]. Xenoimmunization with plasmid DNA encoding human TRP-2 caused tumor immunity and vitiligo in C57BL/6 mice, but unlike the TRP-1 model, CTL and not antibodies were necessary to assure tumor protection and induce autoimmunity [Bowne et al., 1999a]. Similar results were obtained by gene gun immunization or using a recombinant adenovirus expressing human TRP-2 [Steitz et al., 2000]. On the other hand, active immunization of C57BL/6 mice with the mouse TRP-2 gene generated CTL that recognized the immunodominant TRP-2 peptide but not a TRP-2-positive mouse melanoma; as expected, this response was weakly protective ([Tuting et al., 1999] and unpublished results). However, inoculation of a mouse TRP-2-encoding plasmid in cardiotoxin-pretreated muscles of F1(C57BLxBALB/c) mice was able to induce a fully protective immunity against B16 melanoma cells that required both CD8⁺ T and NK cells; moreover, eradication of established subcutaneous tumors was achieved in 50% of the mice by a single injection of a rVV encoding the mouse TRP-2 [Bronte et al., 2000b], and no vitiligo was observed in this study. The genetic background might thus have a great influence on the intensity of the anti-tumor response and the autoimmune sequelae induced by TAA-specific vaccines, and mouse MHC appear to be the most likely candidate genes. Defining the associations between genetic factors and tumor rejections might be of paramount, predictive importance for tumor-bearing patients.

The same mouse studies that provided convincing evidence of the possibility to elicit an immune response to melanoma self-antigens, also unveiled the limited therapeutic efficacy of cancer vaccines. Xenogeneic immunization with human gp100 elicited CTL cross-reactive with mouse gp100 that could treat pulmonary metastases of B16 melanoma upon adoptive transfer [Overwijk et al., 1998]. However, the capability of genetic vaccines encoding gp100 to stimulate a state of active immunity conferring protection from a lethal challenge with B16 melanoma is still contended [Hawkins et al., 2000; Schreurs et al., 1998; Yang et al., 1999] In vitro antigen re-stimulation may be necessary to expand the low numbers of gp100-specific CTL elicited by vaccination or, alternatively, a control circuit may operate in the organism to limit the proliferation and the activity of these self-reactive CTL, as revealed by studies with trangenic mice.

Other self TAA distinct from the MDA have also been analyzed in mouse studies. Vaccination with a plasmid DNA, or with a recombinant Semliki Forest virus expressing the antigen P1A induced protection from challenge with P815 mastocytoma cells, whereas a recombinant adenovirus was ineffective if not boosted with a peptide in adjuvant [Brandle et al., 1998; Colmenero et al., 1999; Rosato et al., 1997]. P1A is thought to be the mouse equivalent of the human category of cancer/testis TAA but it was recently demonstrated that it is also expressed in lymphoid tissues and in lungs, although at low levels [Sarma et al., 1999]. Immunization with a second antigen expressed by P815 (P1E), which is generated by point mutation of an ubiquitously expressed gene, methionine sulfoxide reductase, could elicit tumor-specific CTL but did not lead to tumor protection [Bilsborough et al., 1999], a confutation of the general idea that mutated, non-self TAA are more efficacious as tumor rejection antigens.

A separate chapter must be reserved for vaccination with clonotypic receptors expressed by some hematologic malignancies such as lymphomas, multiple myelomas and leukemias. The immunoglobulin and the T cell receptor (TCR) clonally present on the malignant cells constitute remarkable examples of tumor-specific antigens. Immunoglobulins and TCR contain unique portions that intervene in antigen binding (idiotypes), and can be recognized by the immune system. The idiotype of B-cell lymphoma is a weak antigen whose immunogenicity was increased by a gene-fusion approach. Composite genes including the idiotype sequence and the cytokine GM-CSF or the fragment C of tetanus toxoid were shown to protect mice from lethal challenge with mouse B-cell lymphomas [King et al., 1998; Syrengelas et al., 1996]. The fused genes thus appeared to provide the required immunological help, a strategy that could be extended to other TAA as well. Although CD4⁺ lymphocytes can be required for the priming, the tumor-protective effects of idiotype DNA vaccination can be primarily ascribed to anti-idiotype humoral immunity [Syrengelas and Levy, 1999]. Interestingly, the idiotype/tetanus-toxoid vaccine was efficacious against a myeloma that did not express the immunoglobulin on its surface [King et al., 1998]. To date, active immunotherapy of T cell malignancies with TCR has been attempted mostly with soluble proteins admixed with adjuvant revealing that adjuvants that induced a Th1-type immune response favored tumor protection but only CD8⁺ T cells were required for complete tumor prevention [Widera et al., 2000].

Transgenic Mouse Models

As seen above, transplantable tumors can be treated by genetic vaccination either at a very early stage (prevention of tumor challenge), or at a more advanced stage (therapy of established tumors and/or metastases). However, these experimental settings cannot be considered true models for prevention of spontaneously arising cancers. Moreover, the tumor-specific population cannot be tracked down in vivo, an experimental flaw that has stimulated investigation of transgenic mouse models.

In a pioneer study, transgenic mice that expressed low levels of Friend murine leukemia virus envelope protein in lymphoid cells under the control of an immunoglobulin promoter were not capable of mounting an envelope-specific response because of a central thymic tolerance. Adoptively transferred envelope-specific T cells from immunized normal B6 mice mediated the complete eradication of a Friend virus-induced erythroleukemia that over-expressed the envelope protein. The transfer did not induce detectable autoimmune damage to lymphoid tissues, suggesting that the quantitative difference in the expression of envelope protein between the tumor and normal cells might permit the generation of antigen-specific T cells capable of a selective antitumor activity [Hu et al., 1993]. The demonstration of this gap between autoimmunity and tumor immunity based on the level of gene expression was important to strengthen the idea of an immune-mediated rejection of cancers, but it had to be confirmed in more relevant models where the lymphocyte response could be induced in the same transgene-expressing host.

Transgenic mice bearing model tumor antigens in peripheral tissues, TCR recognizing these antigens, or both have been remarkably instructive for the comprehension of the mechanisms assuring tolerance to peripheral antigens. Studies in transgenic mice have shown that T cells specific for a self tumor antigen that have escaped thymic selection can still be present in the lymphoid organs and blood without exercising antitumor activity. Three main mechanisms of peripheral lymphocyte control have been described: deletion (self-reactive lymphocytes can be constantly and actively deleted in the periphery); anergy (self-reactive lymphocytes are unable to respond to peripherally expressed antigens even though they recognize them); and ignorance (self-reactive lymphocytes simply ignore the antigen). Soon after their activation, tumor-reactive CTL can also be forced to exit the tumor site and reach the secondary lymphoid organs without being able to control local tumor growth except for the first few days [Romieu et al., 1998].

A complete deletion of the self-reactive lymphocytes has rarely been described in mouse tumor models. In some reports, tumor-reactive lymphocytes can be coerced to undergo apoptosis triggered via Fas (CD95) engagement by Fas-ligand aberrantly expressed by the tumors, a mechanism that has been proposed for tumor escape but whose relevance is currently under dispute [Igney et al., 2000; Restifo, 2000]. Deletion of self reactive T lymphocytes can also be excluded as an obligate outcome during human tumor development, since melanoma-bearing patients possess tumor-reactive lymphocytes in the tumor infiltrate that can be expanded by culture in vitro with IL-2 [Rosenberg, 1999]. On the other hand, anergic T lymphocytes have been detected in melanoma patients following staining with HLA-A*0201 tetramers harbouring the peptides for MART-1 or tyrosinase antigen [Lee et al., 1999]. Two populations of melanoma-specific anergic lymphocytes have been described among the peripheral blood mononuclear cells: a classic memory/effector T cell population, and a lymphocyte population with a mixed phenotype possessing markers of both naïve and effector lymphocytes. Both populations had impaired responses, but the second cell subset was completely unable to lyse melanoma target cells, or produce cytokines in response to mitogen stimulation. This extreme form of anergy could not be reversed in vitro, and was selective for the tumor antigens, because it did not involve other CD8⁺ lymphocytes that could respond normally to antigens like EBV. Finally, lymphocytes with an a/b TCR isolated from melanoma and vitiligo patients express NK cell inhibitory receptors (KIR). These receptors are clonally expressed on the surface of the NK cells, engage Class I MHC molecules, and deliver a negative signal that restrains cytolytic activity against healthy cells. Aberrant events leading to the loss of one or more alleles of the Class I MHC molecules make normal cells susceptible to attack by NK cells. The reason for KIR presence on mature lymphocytes is currently unknown, since they are also present on a fraction of T lymphocytes that specifically recognize melanoma differentiation antigens which can be found in normal, healthy donors. It was advanced that KIR expression might be involved in down-regulating the reactivity against self-antigens, but these lymphocytes have been shown to recognize melanoma cells that have undergone a partial HLA allelic loss [Huard and Karlsson, 2000a; Huard and Karlsson, 2000b; Ikeda et al., 1997a]. Further studies in mouse models are required to understand the contribution of KIR to the maintenance of peripheral tolerance.

Like findings in patients and normal donors, tolerance to transgenes is not absolute in mice and can be breached by various treatments. Mice transgenic for a TCR specific for self-peptides presented on the MHC Class I molecule L^d rejected an L^d-positive skin graft, but not a tumor expressing the L^dmolecule despite the relative abundance of antigen-specific T cells. Tumor challenge was rejected effectively only when two conditions were satisfied: the CD8⁺ T cells were activated by immunization with a full-thickness skin graft, and the tumorcells used for challenge were genetically manipulated to express the costimulatorymolecule B7-1. Active immunization had no detectable effect once the tumor was established [Wick et al., 1997]. In one of the few examples of mice transgenic for a TCR specific for a mouse tumor antigen, P1A high affinity CTL were not eliminated from the T cell repertoire, but ignored the antigen expressed at low levels in various tissues [Sarma et al., 1999]. Transgenic, P1A-specific CTL represented a conspicuous percentage of the peripheral T lymphocytes and could lyse P1A-positive tumor cells following in vitro stimulation, but were able to restrain the growth of a challenge with the same cells only if they expressed the B7-1 costimulatory molecule, a finding analogous to that obtained by Wick et al. These data support the concept that a constant activation at the tumor site could be required for the antitumor activity of self-reactive T lymphocytes.

Other transgenic models confirmed that self-tolerance to model antigens was quantitative rather then complete, and viruses were particularly apt to overcome it. Vaccination with a rVV encoding the immunodominant H- 2K^d epitope of influenza HA in transgenic mice expressing the HA on pancreatic islet cells could activate low avidity CD8⁺ T cells that were still capable of rejecting an inoculum of tumor cells that expressed high levels of HA [Morgan et al., 1998]. Low affinity CTL against influenza NP that bound poorly to tetrameric MHC/NP epitope could also be traced in NP-transgenic mice. This oligoclonal population of CD8⁺ lymphocytes recognized and killed NP-expressing tumors, could be expanded in vivo following challenge with influenza virus, and persisted as a memory population for prolonged periods of time [de Visser et al., 2000]. Staining with MHC tetramer appears to be a useful method for distinguishing between low and high avidity, and can also be applied to isolate tumor-reactive CTL with the highest affinity [Yee et al., 1999].

Spontaneously Arising Tumors

All the models based on the transfer of transplantable tumor cell lines inevitably suffer from many important drawbacks. Challenge with a suspension of relatively few, separated tumor cells has little in common with vascularized tumors, as many human cancers appear to be at diagnosis. Mice transgenic for oncogenes or mutated oncosuppressor genes might consitute a better model for exploring the role of the immune responses in tumorigenesis. First, tumor onset and progression mimic the history of naturally developing tumors in humans. Second, tumor escape by antigen loss is likely to occur less frequently for oncogenes that are indispensable for transformation and tumor development, at least in theory. Female mice transgenic for the rat HER-2/neu p185 oncogene under the control of the mouse mammary tumor virus promoter develop atypical hyperplasia in situ that progresses to invasive lobular carcinoma affecting all ten mammary glands. Vaccination with a plasmid DNA encoding the extracellular and transmembrane portion of HER-2/neu inhibited the progression of spontaneous carcinogenesis by reducing the tumor number per mouse, and increasing the overall survival. Mice protected from carcinogenesis did not reject a challenge with a syngeneic tumor cell line derived from a lobular carcinoma, raising the possibility that the mechanisms underlying the rejection of transplantable tumors are intrinsically different from those controlling slow progression of carcinogenesis [Rovero et al., 2000]. Antibodies, but not CTL against the neu oncogene that could account for the down-modulation of membrane expression of p185 in tumor-free glands were found in immunized mice. Chronic administration of IL-12 early during mouse life was able to slow down the progression from pre-neoplastic lesions to overt carcinoma through an immune-independent process resulting in local derangement of tumor angiogenesis, with hemorragic necrosis, activation of tumor-associated leukocytes, release of cytokines (IFN-g, IP10, MIG, MCP1), and induction of nitric oxide synthase [Boggio et al., 2000; Boggio et al., 1998]. These data suggest that non-specific mechanisms of tumor rejection could be coupled with DNA vaccination to increase the preventive potential.

Simian virus 40 (SV40) large T antigen (Tag) oncogene has been widely used to create a number of transgenic mice that develop spontaneous tumors in different tissues according to the promoter driving the tissue-specific expression of the oncogene. To study therapeutic approaches to prostate carcinoma, a transgenic mouse was recently created, the transgenic adenocarcinoma mouse prostate (TRAMP) mouse, a line of C57BL/6 mice that develop a prostatic intraepithelial neoplasia within 8-12 weeks of age that progresses to adenocarcinoma with metastasis affecting 100% of the males by 24-30 weeks of age [Greenberg et al., 1995]. TRAMP mice were established using a rat probasin promoter to drive expression of the SV40 Tag to the prostatic epithelium. Prostate cancers in TRAMP mice occurs only after the normal growth and development of the prostate gland, when increasing levels of androgens direct the prostate specific probasin-controlled transgene to express the SV40 Tag. Moreover, 3 cell lines were established from a primary prostate tumor of TRAMP mice: TRAMP-C1, TRAMP-C2, and TRAMP-C3 [Foster et al., 1997]. Following the demonstration that blockade of the T cell inhibitory receptor, cytotoxic T lymphocyte antigen-4 (CTLA-4), by means of an anti-CTLA-4 antibody in combination with surgical exeresis of the primary nodule could limit the mestastatic spread of TRAMP-C2 tumor [Kwon et al., 1999], efforts were made to evaluate whether vaccination could alter spontaneous tumorigenesis in TRAMP mice as well. Treatment of TRAMP mice with a combination of an antibody blocking CTLA-4 and a GM-CSF-expressing tumor cell vaccine caused a significant reduction in tumor incidence in a 2 month follow-up; prostate tissue sections revealed a low grade tumor heavily infiltrated with leukocytes. Important in view of its possible implications in human prostate cancer treatment, the same vaccination protocol in nontransgenic mice caused an extensive prostatitis likely due to the immune response to unknown self prostate antigens [Hurwitz et al., 2000].

TRAMP mice are a suitable model to exploit the consequences of immunization with recombinant vaccines encoding specific TAA. SV40 Tag is an obvious candidate as TAA since its function is critical for tumor initiation. Recent studies have demonstrated SV40 presence and expression in many human mesotheliomas [Carbone et al., 2000] and even though its involvement in oncogenesis is still under scrutiny [Carbone et al., 1997; De Luca et al., 1997], SV40 Tag might be an interesting immunotherapeutic target since it is basically expressed by tumor cells and not normal tissues. SV40 Tag is the target of a strong CTL response in C57BL mice, in which 3 immunodominant H-2^b-restricted epitopes have been defined: epitope I (Tag_206-215), epitope II/III (Tag_223-231), and epitope IV (Tag_404-411), all restricted by H2-K^b. An immunorecessive epitope V (Tag_489-497) is restricted by H2-D^b[Mylin et al., 1995]. Analysis of Tag transgenic mice has suggested that the localization of the transgene appears to affect critically the degree of tolerance towards the CTL epitopes. SV11⁺ mice, which develop spontaneous tumors of the choroid plexus due to SV40 Tag expression, are functionally tolerant to all the immunodominant Tag epitopes, and only adoptive transfer with normal C57BL/6 spleen cells followed by immunization with Tag-expressing VV led to the generation of Tag-specific CTL and an antitumor effect [Schell et al., 1999]. On the other hand, 501 mice, which express SV40 Tag under the influence of the alpha-amylase promoter and develop osteosarcomas, are partially tolerant to the H-2^b-restricted Tag epitopes. Progression to tolerance to individual Tag CTL epitopes was age- and epitope-dependent, and tumor progression resulted in the specific loss of reactivity towards epitope IV [Schell et al., 2000]. Finally, mice that developed a progressive liver tumor expressing Tag responded to immunization with an immuno-dominant, Class I MHC-restricted Tag epitope mixed with a heterologous helper peptide by generating low-avidity CD8⁺ T cells that did not control tumor development. In contrast, transfers of higher-avidity Tag-specific CTL generated in normal mice were effective in inducing a reduction in tumor growth, suggesting that passive was superior to active immunotherapy for the treatment of spontaneous tumors [Romieu et al., 1998]. Although adoptive transfer is a valid approach (as discussed below), this conclusion is not fully justified because immunization with peptides might be less powerful in stimulating low affinity CTL than immunization with viruses or other recombinant vaccines.

Mice bearing a Tag transgene under the control of the rat insulin promoter (RIP) developed pancreatic b-cell tumors that produced insulin, and led to progressive hypoglycemia. In a double transgenic model, mice spontaneously developed pancreatic b-cell tumors that expressed the glycoprotein (GP) of LCMV. Tumor growth did not elicit CTL activity against GP, whereas LCMV infection induced an antitumor CTL response that efficiently reduced tumor growth, resulting in prolonged survival [Speiser et al., 1997]. The tumor-specific CTL response was not tolerized nor sustained even though the tumor cells continued to express MHC Class I molecules. Only repeated adoptive transfer of virus-activated splenocytes further prolonged survival, suggesting that reiterated transfer of freshly activated effector cells was required to sustain an antitumor immunity. Therefore in RIP/GP mice, as in the other models illustrated previously, CTL can be generated but their function subsides over time. Memory is a crucial issue in antitumor responses, and this study indicated that CTL recognizing TAA might be short-lived if the antigen is not presented in a suitable form.

Findings in RIP/Tag transgenic mouse model also led to questioned whether breaking tolerance is sufficient to cause tumor-specific immunity. By crossing these mice with mice transgenic for the TCR specifically recognizing Tag-specific peptide associated with Class II MHC molecules, F1 mice were generated whose T lymphocytes comprised mostly Tag-reactive CD4⁺ T cells. This abnormally high number of antigen-specific T lymphocytes was expected to control tumor development, and indeed, pre-neoplastic lesions that developed during the life of the mouse where heavily infiltrated with TCR⁺ lymphocytes. However, this severe insulitis never progressed to overt diabetes. Moreover, once tumors were established, infiltration disappeared [Ganss and Hanahan, 1998]. These data suggested that lymphocyte access to tumor masses, a critical step in the antitumor response, was obstructed. Of interest due to the possible therapeutic implications, high dose radiation was shown to render the pancreas permissive to extravasation of T lymphocytes previously activated in vitro; extravasation correlated with the appearance within the tumor nodules of high endothelial venules, specialized anatomical structures that are gates for the migration of recirculating lymphocytes to peripheral lymph nodes and inflamed tissues [Kraal and Mebius, 1997].

Effector Mechanisms for Tumor Rejection

Tumor eradication by immune effectors, in many respects, resembles the battle between somatic tissue cells and lymphocytes during autoimmune disease [Pardoll, 1999]. In these diseases, the immune response is kept under control until an unknown, precipitating event leads to a fall in the defenses, and complete tissue devastation. In nonobese diabetic (NOD) mice, and in the TCR transgenic model of diabetes, for example, the insulitis can last for months, and the transition to diabetes depends on the acquisition by the infiltrating lymphocytes of novel effector functions, that are either the consequence of the recruitment of a set of highly destructive effector cells, or a loss of local regulatory/suppressor activity [Andre et al., 1996]. Experiments of adoptive transfer of pancreas-specific T cell clones indicated that only the Th1 CD4⁺ clones were diabetogenic, suggesting that a transition from a Th2 to a Th1 lymphocyte response might be the main mechanism underlying the passage from a silent to a tissue-destructive immune response [Katz et al., 1995]. Unfortunately, few reports clearly substantiate the idea that antitumor activity as well might correlate with a prevalence of Th1, or a shift from a predominating Th2 to a Th1 response [Hu et al., 1998]. The bulk of evidence indicates that distinct cytokine-producing subsets of CD4⁺ and CD8⁺ lymphocytes can all exert antitumor effector functions. In fact, Th1 and Th2 lymphocytes derived from OVA-specific TCR transgenic mice were equally effective in eliminating tumors when adoptively transferred into mice bearing mouse A20 lymphoma cells transfected with the OVA gene. Th1 and Th2 cells used different effector mechanisms, as only Th1 lymphocytes required an LFA-1-dependent, cell-to-cell adhesion step. Both Th1 and Th2 lymphocytes required the contribution of the host CD8⁺ T cells to eliminate the tumors, and induced a long-lasting memory against A20-OVA tumor. Interestingly, memory CD8⁺ CTL could be recovered only from the splenocytes of mice cured with Th1 cell transfer, suggesting that different effector/memory cells were recruited by Th2 lymphocytes [Nishimura et al., 1999].

CD8⁺ effector T cells secreting IL-4, IL-5, and IL-10, designated Tc2 because of the analogy with Th2 lymphocytes, were also shown to contribute to pulmonary metastases clearance upon adoptive transfer by a mechanism requiring IL-4 and IL-5 secretion and the recruitment of host CD8⁺ T cells together with other leukocyte populations [Dobrzanski et al., 2000; Rodolfo et al., 1999].

The contribution of CD4⁺ T lymphocytes to the development of tumor immunity is generally thought to take place during the priming phase of the immune response, since CD4⁺ T cells are necessary for the development of fully mature CTL effectors [Ossendorp et al., 1998]. This help is mainly indirect, because CD4⁺ cells make APC that have captured, processed and presented tumor antigens in association with MHC Class I molecules (cross-priming) more efficient in activating naïve CTL precursors [Huang et al., 1994]. The interaction between CD4⁺ T cells and APC involves a number of different molecules that transmit reciprocal signals, and the most relevant molecular interaction appears to be the CD40/CD40-ligand engagement that results in the activation of both APC and T cells [Ridge et al., 1998]. It has become clear, however, that CD4⁺ T cells can have a direct antitumor activity against tumors that do not present MHC Class II molecules on their surface. Two types of experimental data support this thesis: i) the adoptive transfer of CD4⁺T cells to tumor bearing mice caused tumor rejection; ii) the depletion of CD4⁺ T cells in immunized mice before tumor challenge impaired ability to reject the challenge [Hung et al., 1998; Mumberg et al., 1999; Surman et al., 2000]. Interestingly, in some models, CD8⁺ T cells were almost completely not required for tumor rejection. It is likely that CD4⁺ T cells can induce a delayed-type hypersensitivity (DTH)-like reaction at the tumor site which recruits cells different from CD8⁺ lymphocytes like macrophages, granulocytes, eosinophils or NK cells that are instructed to destroy the tumor cells by releasing superoxide and nitric oxide [Greenberg, 1991; Hung et al., 1998; Levitsky et al., 1994]. Among the mediators of the antitumor DTH, IFN-g appears to be a key cytokine. IFN-g is produced mainly by CD4⁺ Th1 lymphocytes, but also by CD8⁺ T, and NK cells, and has a central role in immunosurveillance since mice lacking it are more prone to develop carcinogen-induced and spontaneous tumors [Kaplan et al., 1998]. IFN-g can exert a direct effect on tumor cells by rendering them more immunogenic, but may also act indirectly since expression of the IFN-g receptor in the nonhematopoietc cells of the host was necessary to control a tumor challenge upon vaccination with irradiated tumor cells [Qin and Blankenstein, 2000]. The authors postulated that tumor-specific CD4⁺cells released IFN-g that exerted a negative effect on tumor vasculature. Accordingly, it was shown that IFN-g can induce the expression of the chemokine IP10, a powerful angiogenesis inhibitor, in tumor cells [Coughlin et al., 1998]. Moreover, the antiangiogenic properties of IL-12, a cytokines IL-12 that has been widely used in mouse tumor models, can be attributed to the IFN- g-mediated release of IP-10 [Coughlin et al., 1998].

An interesting observation concerning the CD8⁺ lymphocytes is that although they have been often associated with tumor eradication, their cytolytic activity does not seem to correlate with tumor control, at least in some studies. Adenovirus-mediated transfer of IL-12 into established nodules of the P815 mouse mastocytoma caused a marked tumor regression attributable, in part, to CD8⁺ T lymphocyte activity. However, no detectable CTL activity was found in the splenocytes of the tumor-rejecting mice. By means of immunoscope analysis, a clonal expansion of tumor reactive CD8⁺ cells bearing a public T cell rearrangement was instead detected among the tumor-infiltrating lymphocytes, and in lymphocytes collected from the lymphoid tissues and blood [Fernandez et al., 1999]. Immunoscope is a highly sensitive, PCR-based method that determines the ranges of CDR3 lengths of the TCR chains displayed by complex populations of T lymphocytes. The antitumor population was thus present in tumor-rejecting mice, but lacked cytolytic properties. In search of other attributes of antitumor effectors, an IFN-_ transcript increase was found at the tumor site, but the adenovirus-mediated transfer of this cytokine did not exert a direct antitumor activity suggesting that neither was this cytokine involved in IL-12-dependent rejection. Therefore, this study suggest that it would be worthwhile to search for characteristics other than the ability to release IFN-g in tumor-rejecting CD8⁺ lymphocytes.

What is the main property of the tumor- or tissue-destructive effectors? There appears to be no simple answer to this question. Although it might seem odd, this attribute could be distinct from the parameters ordinarily tested during the immune follow-up of vaccinated mouse and patients. The search for this unknown property, this “factor X” (a cytokine, a chemokine, a membrane molecule, a combination of these?), is still ongoing and it could benefit from gene expression profiling approach. It is unlikely that the search for a single lymphocyte function, such as CTL activity or ability to release individual cytokines, would help understand the complex scenario of tumor or tissue destruction. No correlation has yet been found between clinical responses to immunotherapy and the functions of anti-tumor CTL in vaccinated patients [Marchand et al., 1999; Rosenberg et al., 1998a], with the notable exception of patients bearing hematologic malignancies treated with bone marrow transplant or IFN- a2b [Molldrem et al., 2000]. Probably, the most promising approach to the investigation of the molecular mechanisms associated with tumor rejection would be to look for genes preferentially expressed in lymphocytes infiltrating a metastasis that regresses following immunotherapy, as compared to non-regressing lesions [Kammula et al., 1999]. An example of the potential of multiple gene analysis comes again from mouse studies on a TCR transgenic model of diabetes, in which the passage from silent insulitis to overt disease is precipitated by treatment with the immunosuppressive drug cyclophosphamide [Andre-Schmutz et al., 1999]. From an examination of the profile of cytokine gene expression within the lesions, it was found that IL-18, IL-12, and TNF-a were strictly associated with the early events of self-aggression whereas, IL-1 -beta, IL-6 and IFN-g were only induced in a second phase.

Adoptive Transfer

Effector lymphocytes primed by genetic vaccination against both model- and self-tumor antigens have been expanded in vitro, and adoptively transferred back to the host to cure established pulmonary metastases [Overwijk et al., 1998; Wang et al., 1995b], a strategy that could be used in cancer patients as well. The phase of antigen-directed expansion in vitro might temporarily relieve the tumor-specific lymphocytes from the negative control exerted by the tumor, or by the mechanisms of peripheral tolerance. Anergic lymphocytes could be rescued in vitro by simple IL-2 exposure [Beverly et al., 1996]. As previously described, a rVV expressing the human gp100 TAA could break tolerance to the mouse antigen, and induce the generation of CTL that were cross-reactive with the mouse homolog but could not protect from a melanoma challenge. Conversely, in vitro stimulation of these effectors with the human gp100 derived peptide led to lymphocyte expansion and their adoptive transfer could efficiently clear the lung from established pulmonary metastases [Overwijk et al., 1998]. Adoptive transfer appears to be one of the most successful strategies in treating mouse tumors, and some formidable therapeutic results have been achieved. For example, CTL clones against adenovirus early region 1 (E1)-transformedcells propagated in vitro, and adoptively transferred into nude mice in conjunction with recombinant IL-2 destroyed subcutaneous tumor masses of E1-induced tumors up to10 cm³in volume [Kast et al., 1989]. In another recent study, the transfer of mouse splenocytes expressing a K^d-restricted transgenic TCR specific for a mutated ERK2-kinase-derived peptide to a normal recipient conferred protection from subsequent challenge with a syngeneic fibrosarcoma [Hanson et al., 2000]. The protective effect was impressive due to the very low number of TCR⁺ cells transferred (£ 3x10⁴/mouse), whereas peptide-stimulated normal splenocytes were protective against the fibrosarcoma challenge when transferred in large numbers and only in the presence of a continuous supply of IL-12 [Ikeda et al., 1997b]. However, eradication of established tumors by adoptive transfer of TCR⁺ splenocytes was possible within the time window of one week from tumor inoculation, and with the transfer of a number of tumor-specific lymphocytes 100-fold higher than in the tumor prevention setting. Interestingly, tumor- specific T cells could still reject a small tumor even in the presence of a controlateral large tumor burden, suggesting that tolerization of tumor-specific lymphocytes had not taken place, at least in the first two weeks of tumor growth [Hanson et al., 2000].

A double transgenic mouse model expressing the oncogene Tag under a rat elastase promoter and human mucin 1 (MUC1) developed spontaneous acinar cell carcinomas with underglycosylated MUC1 a situation reminiscent of many human colon and pancreatic tumors. MUC1-specific CTL developed in the transgenic mice that recognized an MHC Class I restricted epitope in the central tandem repeat region of MUC1 only when concentrations greater that 1 mM of peptide were used to pulse stimulator cells. As expected, these low-affinity CTL did not affect the development of spontanous carcinomas. However, CTL expanded in vitro in tumor-lymphocyte cultures, and transferred to the same transgenic mice suppressed the growth of a transplanted melanoma tumor expressing MUC1. These investigations evaluated several possible mechanisms of tumor escape in the spontaneously arising carcinomas but no clear picture emerged. As in other studies, the developing carcinomas were not at all infiltrated with lymphocytes, suggesting a deficit in lymphocyte homing to the tumor site [Mukherjee et al., 2000].

The optimal tumor size for rejection following adoptive transfer is another unresolved issue. It seems intuitive that a small tumor burden will be easier treated with immunotherapy approaches, but some reports deny this conclusion. Adoptive transfer of CD8⁺ T cells with a transgenic TCR specific for influenza virus NP into immunodeficient mice bearing small NP-expressing EL4 tumors caused activation of tumor-specific lymphocytes and tumor infiltration, but no rejection. In contrast to other studies, transferred tumor-specific CD8 T cells rejected a large tumor burden [Cordaro et al., 2000]. The critical parameter which determined the success of adoptive transfer appeared to be the ratio of antigen-specific effectors to the antigen-expressing tumor. Activated effector CD8⁺ T cells appeared much earlier in the presence of large tumors, a likely consequence of the amount of tumor-antigen released by dying tumor cells, and subsequent cross presentation by infiltrating APC. Interestingly, the efficacy of TCR⁺ T cells against small tumors could be improved by repeated immunizations with the NP peptide in IFA. These findings imply that treatment of minimal residual disease with adoptive transfer could fail, unless vaccination is provided following transfer.

A simple explanation for the discrepancies emerging from TCR transgenic and adoptive transfer models is not forthcoming but some considerations may be made. The transfer of lymphocytes into mice rendered immunodeficient by various treatments must be regarded with caution, because the absence of other lymphocytes can open up some homoeostatic niches that are not available in normal, immunocompetent mice (and in many patients). Small numbers of peripheral T lymphocytes expand when introduced into a T-cell deficient host, a process that can be restrained by co-transfer of other mature CD4⁺ or CD8⁺ T lymphocytes [Rocha et al., 1989]. This expansion could also result in a modification of the effector properties of transferred cells. The progeny of naïve T cells undergoing homoeostatic proliferation can transiently display effector functions, i.e. mediate CTL activity, or secrete IFN-g when stimulated directly ex vivo [Cho et al., 2000a; Goldrath et al., 2000]. TCR mouse models, moreover, may not reflect the scenario of human pathologies especially when an artificial high frequency of antigen-specific T lymphocytes is realized by the TCR transgene transfer, or when the transgenic self-antigen is a viral or exogenous protein. Furthermore, different exogenous antigens artificially expressed in peripheral tissues could act in different ways in inducing peripheral tolerance depending on the amount of antigen expressed, the accessibility to exogenous compartments, the rate and type of cell death caused by the transgene expression itself. Lastly, lymphocyte infiltration at the tumor site may be influenced by the histology and the natural development of the experimental tumors (cell lines derived from solid tumors are more apt to grow in the subcutanous space than lymphomas; sponta-neously developing tumors undergo stages of development from precancerous to cancerous lesions that cannot be reproduced with transplantable tumors).

Strategies to Improve the Strength of Genetic Vaccines

1. Adjuvants for Genetic Vaccines

Adjuvants are generally defined as any substance that enhances the immunogenic activity of a vaccine. Adjuvants, however, are also able to modify the quality of the response by selectively activating one or another T cell subset. A combination of antigens and adjuvants can give rise to a response that is different from that obtained with the antigen alone [Singh and O'Hagan, 1999]. Moreover, the molecular definition of the steps of T lymphocyte activation has generated a class of adjuvants that can affect different moments of the immune response, from the priming to the execution of the effector program of each lymphocyte population.

Natural Adjuvants

It has been advanced that cellular death, by itself and without the intervention of external agents, must be able to signal the environmental perturbation, and prompt APC for the initiation of an immune response [Matzinger, 1994]. Natural or endogenous adjuvants can thus be defined as the unknown mediators resulting from the pathological destruction of tissues which can act on APC by inducing their activation. Tumor cells constantly die during neoplastic growth and the intracellular material they release can be taken up and processed by tumor-infiltrating APC. Cross-presentation by bone-marrow derived APC is essential to prime Class I MHC-restricted CD8⁺ lymphocytes against the tumor [Huang et al., 1994], but the real nature of the cross-presenting cells has eluded attempts to characterize it. That DC or other tumor-infiltrating APC can stimulate antitumor lymphocytes ex vivo or upon adoptive transfer to a naïve mouse was recently demonstrated in mice inoculated with an adenocarcinoma engineered to express GM-CSF and CD40-L [Chiodoni et al., 1999]. DC separated from leukocytes infiltrating tumors producing GM-CSF and CD40L, in fact, were shown to be mature, able to capture tumor-derived cellular antigens through uptake of apoptotic bodies, and present them to Class I restricted CTL. In a different tumor model, a spontaneous lung carcinoma expressing OVA and the cytokine IL-3, the cross-presenting population possessed the characteristic of a tumor-infiltrating macrophage [Pulaski et al., 1996]. These apparently clear-cut examples of APC participation in TAA presentation leave some unanswered questions behind. First, tumors normally do not express high levels of GM-CSF, IL-3 or CD40-L, so it is difficult to predict how these cytokines alter the normal cross-presentation process within the tumor mass. Second, it is not clear what form of cell death is compatible with cross-presentation. Tumor immuno-genicity has been related to the necrotic death of tumor cells via induction and release of heat shock proteins that act as chaperones for TAA peptides [Melcher et al., 1998]. Nevertheless, apoptotic cells engulfed by DC were shown to stimulate the antitumor response of CD8⁺ lymphocytes both in vitro and in vivo [Hoffmann et al., 2000; Jenne et al., 2000; Rovere et al., 1998].

The engulfment of apoptotic, but not necrotic cells, by DC results in the presentation of MHC Class I-restricted peptides to CD8⁺ lymphocytes, whereas both apoptotic and necrotic cells can be an optimal source of antigen for MHC Class II-restricted, CD4⁺ lymphocytes [Albert et al., 1998; Inaba et al., 1998]. Despite the fact that both necrotic and apoptotic cells can be efficiently presented, the effect of dying cells on DC maturation is thought to be extremely different. Necrotic cells or their soluble products, analogously to bacterial molecules like LPS, can induce maturation of DC, whereas apoptotic cells lack this characteristic [Gallucci et al., 1999; Salio et al., 2000]. These findings recently led to formulate the hypothesis that immature DC that have engulfed apoptotic cells deriving from the normal tissue turn-over intervene in peripheral tolerance induction, and would exert their inhibitory function primarily because of the lack of costimulatory molecules [Steinman et al., 2000]. Albeit still not fully substantiated by in vivo data, the idea of immature “tolerizing” DC is intriguing, and could help to explain tolerance induction of self-reactive CD8⁺ lymphocytes. However, it is difficult to understand how immature DC that are incapable of expressing complexes between peptides and Class II MHC molecules on their membrane could induce tolerance of self-reactive CD4⁺ lymphocytes that is a critical step in tolerance maintenance in several models. It has been provocatively suggested that an in vitro artifact could be the simple solution to this conundrum. It does not matter whether death is necrotic or apoptotic, as only mycoplasma contamination found in some human tumor cell lines is associated with DC maturation [Salio et al., 2000]. However, alternative explanations are conceivable. The presentation of self antigens by DC would not necessarily lead to T cell activation. A recent study employed a T cell clone that recognized two Class II MHC-restricted epitopes of the self antigen human thyroid peroxidase (TPO): an immunodominant but cryptic peptide (P3) that induced full activation, and a second peptide (P4) that acted as altered peptide ligand since it was able to engage the TCR and cause lymphocyte proliferation only at very high concentrations. Immunologically competent DC derived from human monocytes pulsed with P4, or with the whole TPO protein induced anergy of the T cell clone, implying that the antigen and not the APC type could be critical in shaping the lymphocyte response [Quaratino et al., 2000]. Alternatively, a specialized population of DC, that in mice can be identified by the expression of CD8 marker, could selectively induce suppression and apoptosis of activated T lymphocytes when presenting self peptides [Grohmann et al., 2000; Kronin et al., 1996; Suss and Shortman, 1996].

Apoptosis is regarded as an immunologically neglected event, a form of death that does not cause inflammation and APC activation because it occurs under physiological circumstances such as during morphogenesis, in developing and growth-factor deprived tissues, during thymus education of T lymphocytes, and in terminally differentiated hemotapoietic cells [Bellamy et al., 1995; Cohen et al., 1992]. An opposite view was recently advanced, based on the findings that the enhanced efficacy of self-replicating RNA vaccines appeared to rely strictly on the caspase-dependent apoptotic death of transfected cells, followed by uptake and presentation by DC [Ying et al., 1999]. Like other studies, this assumption was based mainly on in vitro evidence. It was theorized that apoptosis and inflammatory signals can both be triggered by double strand RNA, an intermediate molecule produced during the intracellular biological cycle of many infectious viruses, and that viruses themselves have developed more than one strategy to halt apoptosis in host cells suggesting that they must have dealt with this mechanism of viral continence during their evolution. These investigators also advanced that the function of naked DNA vaccines as well might be related to the transfection-dependent death of tissue cells [Restifo et al., 2000].

Adjuvants Influencing the Priming Phase

Professional APC express nonclonal pattern-recognition receptors (PRR) that detect invariant molecular structures shared by pathogens of various origin (pathogen-associated molecular patterns, PAMP). One way PAMP help the immune response is by inducing DC maturation, as shown for LPS, viruses and bacteria [Banchereau et al., 2000; Banchereau and Steinman, 1998; Janeway, 1992]. Many adjuvants assist the priming phase of the immune response by inducing the recruitment and activation of APC through the engagement of one or more PRP. Bacterial and synthetic DNA can act as a ligand of PRP, and possess a built-in adjuvant activity that is restricted to a particular base context. The immunostimulatory activity of DNA vaccines, in fact, resides in the prokaryotic portion of the plasmid, which contains a CpG in the hexamer PuPuCpGPyPy [Klinman et al., 1996; Krieg et al., 1995]. The motif has been defined in greater detail, and appears to be slightly different for mice and primates: in the former, the central CpG pair can be preceded by any base except C, and followed by any base except G (5'- GACGTT-3') [Yi et al., 1998], whereas a T preceding the CpG is usually preferred in the latter (5'- GTCGTT-3') [Hartmann et al., 2000]. Differently from the eukaryotic counterpart, the prokaryotic C base in CpG islands is usually not methylated. In their unmethylated form, these hexamers stimulate monocytes and macrophages to produce different cytokines with a Th1 promoting activity including IL-12, TNF-a, and type I IFN (IFN a/b) [Halpern et al., 1996; Klinman et al., 1999] which, in turn, act on NK cells and induce the activation of their lytic machinery and release of type II IFN (IFN-g). Synthetic CpG-containing oligodeoxynucleotides can function as adjuvant independently of the plasmid backbone. They bind to an unknown, likely intracellular receptor, and activate stress-kinase and NF-kB signalling pathways in a manner similar to the LPS of gram-negative bacteria [Krieg et al., 2000]. Immunostimulatory DNA sequences chemically linked to OVA protein are strong stimulators of a Th1-like responses, and can prime the host for an OVA-specific CTL activity in CD4-deficient mice [Cho et al., 2000b]. These Th-independent CTL were highly effective against an OVA-positive tumor cell line in models of prevention and therapy. Systemic administration of CpG-motif-containing DNA in mice can activate a non-specific, innate immune response to tumors that can lead to a substantial tumor treatment. A 5-mm-diameter subcutaneous neuroblastoma was treated in 50 % of the mice inoculated daily with CpG oligodeoxyucleoties in proximity of the tumor. Rejection was mediated by activated NK, but allowed the establishment of a tumor-specific memory response since the surviving mice rejected a second tumor challenge [Carpentier et al., 1999]. NK cells and systemic IFN-g release were also responsible for the anti-metastatic effect of cationic lipid-DNA complexes injected intravenously, a response that cationic lipids alone were not able to induce [Dow et al., 1999]. This non specific effect of DNA vaccines could explain some degree of protection from tumor challenge in mice immunized with a control plasmid lacking TAA sequences [Bronte et al., 2000b]. Some constructs are likely to possess more immunostimulatory properties than others according to the content in CpG islands. A plasmid that incorporated several CpG islands in the prokaryotic ampicillin-resistance gene induced a stronger immune response compared to a second plasmid carrying the kanamycin-resistance gene which possesed none [Sato et al., 1996]. To date, it is not known whether the CpG containing sequences will have the same immunostimulatory properties when applied to the vaccination of human beings, but results of studies in primates allow a certain optimism [Davis et al., 2000; Hartmann et al., 2000; Jones et al., 1999]. Moreover, CpG motifs can definitely activate subsets of freshly isolated human DC to promote Th1 immune responses in vitro [Hartmann et al., 1999].

GM-CSF is another adjuvant that plays an important role in the priming of cellular immune responses. Professional APC were recruited and expanded by GM-CSF released locally by g-irradiated cDNA transduced tumors; these APC presented antigens released by dying tumor cells to naïve lymphocytes, resulting in a memory T cell response that effectively prevented subsequent challenge with the wild-type tumor [Dranoff et al., 1993; Levitsky et al., 1996; Thomas et al., 1998]. Local recruitment of DC together with granulocytes and macrophages was also shown to occur after inoculation of a plasmid DNA encoding GM-CSF [Bowne et al., 1999b; Haddad et al., 2000]. Following the pioneer studies with engineered tumor cell vaccines, GM-CSF has been widely used as adjuvant in conjunction with genetic and non genetic vaccines (reviewed in [Sedegah et al., 2000]), but recent reports have signaled some restrictions for its use as adjuvant. Gene modification of tumor cells with GM-CSF in the absence of g-irradiation does not generally cha