Novel Vaccines for the Treatment of Gastrointestinal Cancers
Novel Vaccines for the Treatment of Gastrointestinal Cancers
Continuing advances in immunology and molecular biology during the past several decades have provided optimism that immunomodulatory strategies may be clinically useful in patients with cancer. Key advances have included: (1) recognition of the critical role of the antigen-presenting cell and greatly improved understanding of antigen processing and presentation, including the molecular interactions between HLA molecules and antigenic epitopes on the antigen-processing cell and the receptors on T cells, and (2) the roles of costimulatory molecules such as B7.1, ICAM-1, and LFA-3 in the induction and maintenance of an immune response. In addition, new techniques have allowed us to identify immunogenic antigenic determinants, alter their binding affinities, and evaluate the overall success of the intervention through both in vivo and in vitro assays. Carcinoembryonic antigen (CEA) is overexpressed in a large number of gastrointestinal, lung, and breast cancers. Clinical trials have established treatment protocols using viral vectors to immunize patients to CEA without producing deleterious autoimmune phenomena. By combining various vectors to include MUC-1 and/or CEA plus costimulatory molecules in a prime-and-boost regimen, we are beginning to see signs that this intervention can not only produce changes in immune function but also potentially improve clinical outcomes. Phase III studies to test these hypotheses are under way.
In the late 19th century, William Coley at Memorial Sloan-Kettering Cancer Center in New York reported spontaneous regression of sarcomas following severe bacterial infections. However, attempts to reproduce these results with bacterial extracts met with limited success. Subsequently, attention shifted from infectious particles to antigens specific for tumor cells recognized by the immune system in a manner that causes the rejection of the tumor as a foreign tissue or non-self (tumorspecific transplantation antigens). For example, in mice, immunity to methylcholanthrene- induced malignancies could be produced by injecting genetically identical animals with tumor cell lysates. This observation appeared to reflect the tumor-specific nature of the antigens induced by chemical mutagens. In the early 1960s, investigators were encouraged by reports that malignant cells from different tumors share common tumor-associated antigens.[ 4] The results of these studies, coupled with advances in immunology, led to the concept of immune surveillance- a process in which immunocompetent hosts are protected from the emergence of malignant clones, which are ostensibly identified as non-self, by elements of the innate and adaptive immune system. [ 5] These observations were exciting and provided hope that stimulating an active, specific immune response might cure cancer, but subsequent basic science and clinical studies were disappointing. For example, the concept of immunosurveillance was questioned when athymic nude mice were shown to have the same incidence of tumors as wild-type animals, and results of early clinical trials with various approaches to stimulating an active antitumor immune response were not encouraging. More recent work has greatly expanded our understanding of the interactions between elements of the immune system and cancer cells and has renewed hope that therapeutic cancer vaccines and other immunomodulatory approaches may, in fact, become valuable tools for clinical on- cologists. Cytotoxic T lymphocytes have been shown to kill HLA-matched tumor cells in vitro, and immunogeninduced upregulation of humoral immunity has been shown to be associated with improved outcomes in patients with breast cancer or melanoma.[ 7-9] The success of donor lymphocyte infusions in patients with hematologic malignancies has also provided empiric evidence that immunomodulatory therapy can be an important clinical tool in patients with advanced malignancies.[10,11] Advances in Therapeutic Cancer Vaccines It is a key tenet of tumor immunology that tumor cells express antigens and the uniqueness of these antigens makes them potential targets of innate and adaptive immune elements. Thus, it is no surprise that clinical failures of therapeutic cancer vaccines may be linked to a failure of antigen recognition by the immune system. With the advent of sensitive assay systems, investigators have been able to demonstrate the presence of naturally occurring T cells directed against tumor-associated antigens (TAAs). Gilboa has categorized TAAs that may function in tumor rejection as two types: (1) patient-specific, mutated self-antigens; and (2) shared, nonmutated self-antigens. Mutated self-antigens are the result of the genetic instability of an individual tumor and are generally incidental to tumor pathophysiology. While they would be anticipated to be potent immunogens, identification and isolation of these unique (private) antigenic determinants would be impractical. The significance of mutated self-antigens to successful immunotherapy was illustrated in a series of experiments by Boon and colleagues. These investigators were able to easily transplant cell lines from teratocarcinoma and Lewis lung carcinoma into syngeneic mice. If, however, the malignant cell line was first mutated by exposure to N-methyl-N'-nitro-Nnitrosoguanidine, development of cellline- specific mutated self-antigens elicited an immune response that protected the mice from the transplanted tumor cells.[13-15] Unlike vaccines to private antigens, vaccines to shared (public) antigens as off-the-shelf therapeutics have the potential to benefit many patients. In most instances, these shared antigens correspond to normal gene products expressed on tumor cells, so patients may be expected to be tolerant to many of them. Other shared tumor antigens, however, are fetal gene products not expressed in adult tissues or are common molecules related to malignancy and found only in tumor cells. By virtue of their limited distribution in either time or space, oncofetal or sequestered antigens, respectively, have the potential to be targets for immunotherapy. Recently reported examples of oncofetal tumor antigens include survivin, an inhibitor of apoptosis expressed during fetal development and in many types of cancer cells, and OFA, the precursor to the mature laminin receptor found in fetal cells.[16,17] In addition, some proteins have been identified as common to certain types of tumor cells but absent from normal tissues. Human telomerase reverse transcriptase may be an example of a common tumor- specific antigen since it is overexpressed in many tumors but absent from normal cells. Optimizing Antigenicity
The process of effective recognition of antigens and the development of an adaptive immune response is complex. Tumor antigens must be taken up by phagocytic cells such as dendritic cells, processed into immunogenic fragments of defined size, and assembled into complexes with the appropriate major histocompatability complex (MHC) molecule-class I for presentation to CD8 T cells and class II for presentation to CD4 T cells. These complexes are then displayed on the surface of the dendritic cells. MHC molecules derive their name from their role in transplant rejection. In response to differentiating elements released by inflammatory cells and the pathologic process itself, dendritic cells differentiate into potent antigen- presenting cells and migrate to lymphoid tissue, where they encounter the T cells. T cells become activated when their antigen-specific T-cell receptor binds the antigen-MHC complex presented on the surface of the antigen-presenting cell. T cells only bind an antigen-MHC complex if the complexed antigen is recognized by its T-cell receptor (cognate antigen). The ability of an antigen to elicit an immune response is partially a function of the number of cognate reactive (vs tolerant) T cells and the binding affinity between the responsive T cell and the antigen-presenting cell. However, the potency of an immune response also depends on other factors such as the interactions between the MHC complex and CD4 and CD8 molecules on T cells and the generation of costimulatory signals between the antigen-presenting cell and T cells. Costimulatory molecules are expressed on the surface of antigen-presenting cells and are recognized by corresponding receptors on T cells. Tumor cells generally lack costimulatory molecules and, therefore, function very poorly as antigen-presenting cells. Moreover, when antigen is presented in the absence of costimulation, T cells tend to become anergic, which may be a mechanism for protection of normal cells against an autoimmune response and also is a proposed mechanism of tumor escape. The most widely studied costimulatory interaction is through B7.1 (CD80) on antigen-presenting cells and the molecule CD28 on T cells (Figure 1). For the T cells to survive, proliferate, and form memory T cells, additional costimulatory interactions are required, such as the interaction between CD40 on the activated antigenpresenting cell and CD40 ligand on the T cell. Thus, to optimize antigenicity, it is reasonable for therapeutic cancer vaccine developers to employ strategies similar to those observed in vivo-ie, to maximize the immunologic response by presenting antigen in the context of costimulatory molecules. The importance of costimulatory molecular signals may have been reflected in the observations by Coley more than a century ago that immune responses against weak antigens require additional signals (eg, such as those released during infection). In Vitro Quantification of Immune Responses
While experiments can be conducted in syngeneic animal models or with standardized cell lines, it is difficult to quantify responses of patients to novel immunotherapies. Although two patients may both have the same type, grade, and stage of solid tumor, they may differ in age, comorbidities, tumor antigenicity, tumor suppressor gene status, HLA type, dendritic cell function, T- and B-cell numbers and function, response to chemotherapy, and so forth. Consequently, it is important to develop quantitative methods to evaluate the magnitude and durability of immunologic responses to cancer vaccines. In vivo assays have been available for some time and include delayed-type hypersensitivity skin reactions, depigmentation of the tumor in the case of melanoma vaccines, and quantification of tumor infiltrating lymphocytes. In addition, a variety of new in vitro analytic tools have been developed. These in vitro methods have been important in the continuous improvement of vaccine strategies and in the selection of strategies to move to clinical trials. Ultimately, an in vitro assay may serve as a surrogate indicator of therapeutic vaccine efficacy if the measured immune responses can be shown to correlate with clinical outcome. These in vitro assay systems include enzyme-linked immunospot (ELISpot) for the measurement of antigen- specific T-cell responses, flow cytometric measurement of intracellular cytokine production, and MHCpeptide tetramer analysis. ELISpot employs antibodies in a sandwich assay with signal amplification to enumerate cytokine-secreting cells. When automated, the results are reliable and reproducible. Cloned peptide-MHC complexes can be stabilized and tetramerized with streptavidin labeled with phycoerythrin. When incubated with mononuclear cells, the ability of tetramers to stain and activate CD8 T cells is strongly dependent on binding of CD8 to the same class I molecule engaged by the T-cell receptor. Peptide-specific T cells can then be enumerated in a flow cytometer. All three methods are very sensitive, but the peptide-MHC tetramer analysis appears to be the most sensitive. It is also the most technically demanding assay and, at this time, is not sufficiently robust for broad clinical application. Development of CEA-Based Vaccines in GI Cancers CEA as a Target of Immunotherapy
Carcinoembryonic antigen (CEA), reported in 1965, is one of the first oncofetal tumor antigens to be discovered. It is normally expressed in fetal colon. Although its expression is restricted in adult life, CEA is present in saliva, feces, serum, colonic mucosa, and fluid from colonic lavages in adults. CEA is overexpressed in a high percentage of adenocarcinomas, particularly those of endodermal origin (eg, non-small-cell lung, stomach, colon, rectum, and pancreas). As such, it is considered a shared (public), nonmutated oncofetal self-antigen and its concentration in the circulation is widely used as a serologic marker of malignancy. CEA is a member of the immunoglobulin superfamily with a molecular weight of 180 to 200 kDa. The gene for CEA codes for a 70-kDa protein and is located on chromosome 19. The difference in molecular weight between the gene product and the final antigen is secondary to extensive post-translational glycosylation.[ 24] In normal colonic epithelium, CEA is localized to the luminal surface-an arrangement that suggests that it contributes to spatial orientation of colonocytes and that it may also function to preserve the adult gut's mucosal barrier. In tumor cells, however, CEA is irregularly distrib- uted throughout the cell membrane. As an intercellular adhesion molecule, CEA may contribute to the formation of metastasis. There is a correlation between serum CEA in patients with cancer and the incidence of hepatic metastases, but this could also simply reflect tumor burden. Due to its association with malignancy and wide distribution in human tumors, CEA has the characteristics of an ideal target for vaccine therapy. However, there are several potential problems that must be resolved. Since CEA is also normally expressed in adult life, the immune system is normally tolerant to CEA. The tolerance develops in utero with deletion of autoreactive clones of T cells through a process of negative selection within the thymus. Small numbers of CEA-reactive T cells may still reach the periphery. In general, tolerance is induced in these clones by some combination of clonal inactivation, clonal deletion, and/or cytokinedependent suppression and immune deviation. Thus, strategies employed to overcome tolerance to CEA might overcome tolerance in these clones and elicit an immune response that would not only target the tumor but could also induce autoimmune side effects. Experimentally, tolerance to CEA can be overcome without induction of adverse autoimmune events. Investigators have vaccinated mice transgenic for human CEA and generated anti-CEA antibodies, immunoglobulin class switching, TH1-type CEA-specific CD4+ cells, and CD8+ cellular cytotoxicity.[25,26] CEA Vaccine Strategies: Peptide Modification
As described above, T cells only recognize antigen after it has been processed into fragments (epitopes) composed of 8 to 10 amino acids (CD8+ cells) or longer (CD4+ cells) and presented on the surface of the antigen-presenting cell complexed with MHC class I or II molecules, respectively. T-cell epitopes from tumor antigens such as CEA have been identified by either identifying the target of an existing T-cell response or, more commonly, using modeling to predict the binding affinity of a particular amino acid sequence to a specific MHC class I or II molecule. At present, a number of CEA-associated peptides have been identified and have been used to generate T-cell responses. The majority have been restricted to the (MHC class I type) HLA-A2, the supertype present in 40% to 50% of humans.. HLA-A3- restricted peptides have also now been identified and shown to be immunogenic.[ 28] While less attention has been focused on MHC class II-restricted peptides for numerous reasons, evolving understanding of the role of CD4+ cells in modulating antigen-specific immune responses has led to increased emphasis on these epitopes. As an antigen, CEA is weakly immunogenic. It is reasonable to assume that at least part of the explanation for the low immunogenicity of native CEA peptides is their low affinity for MHC molecules. Therefore, to develop a more effective therapeutic cancer vaccine, strategies have been employed to increase the immunogenicity of CEA. If, in fact, binding affinity is the problem, amino acid substitutions of key residues that anchor the peptide to the MHC molecule might be used to increase the peptide's affinity for MHC molecules. This strategy has been employed effectively to improve T-cell responses to several melanoma antigens. Another successfully used approach to enhance the T-cell response is to modify the residues recognized by the T-cell receptor. The HLA-A2- restricted 9-amino acid peptide, CEAassociated peptide 1 (CAP-1), was modified by replacing the asparagine at position 6 with an aspartic acid. In in vitro assays, this modified peptide, CAP-1-6D, was more effectively recognized by T cells than the native CAP-1 peptide and significantly upregulated the production of granulocyte macrophage colony-stimulating factor (GM-CSF) and interferon- gamma by the target cell. Gene array studies demonstrated that, compared with the native CAP-1 peptide, CAP-1-6D stimulates the expression of different genes and gene clusters in CD8+ T cells than the native peptide. Expression of lymphotactin and granzyme B were increased many times over the levels in CD8+ T cells stimulated with CAP-1. Lymphotactin recruits T and NK cells into the region, and granzyme B is transferred from cytotoxic T lymphocytes to target cells during the process of granule- mediated cytotoxicity.[33,34] Viral Vectors for Antigen Delivery
Recombinant viruses have been used for both gene therapy and as vectors for immunotherapy. When used as immunotherapeutics, genes for a tumor antigen can be introduced into a virus that directs the expression of the antigen and stimulates a strong immune response to the specific tumor antigen. In the development of vaccine therapy, the choice of a vector is as important as the selection of the tumor antigen. Potential viruses include adenoviruses, retroviruses, and poxviruses, and each has its advantages and disadvantages. For example, while the adenoviruses are effective vectors and have been widely used in experimental studies, they infect nondividing cells and require repetitive dosing because their genes are only transiently expressed. In addition, adenoviruses generate an immune response or enhance the immune response in the approximately 80% of humans who already have neutralizing antibodies to these viruses.[ 37] Retroviruses integrate into the host genomes and, therefore, bring the risk of insertional mutagenesis.
- Poxviruses- Poxviruses such as the vaccinia virus are particularly well suited as vectors for therapeutic cancer vaccines. They are easily engineered, accommodate large inserts of foreign DNA, allow post-translational modification of expressed proteins, replicate accurately without helper viruses, stimulate potent immune responses, and have been extensively tested in humans as smallpox prophylaxis.[ 24] Recombinant vaccinia (rV) viruses that express human CEA have been developed. Incorporating the gene for CEA into the virus involves inserting the nucleic acids for the gene's start site immediately adjacent to the transitional start site of a vaccinia gene promoter.
Following transfection, the rV- CEA infected cells express a protein product recognized by monoclonal antibodies to CEA. CEA transgenic mice, which express human CEA as a self-antigen and are immunologically unresponsive to CEA administered as a whole protein, respond to immunization with rV-CEA with TH1-type CEA-specific CD4+ responses and CEA peptide-specific cytotoxicity. Immunization with rV-CEA protected these mice from challenge with CEA-expressing tumor cells. Vaccination with rV-CEA is also immunogenic in humans. Following vaccination with the viral construct, previously immuno-unreactive peripheral blood lymphocytes from patients with metastatic carcinoma were able to mount a cytolytic T-cell response to specific CEA epitopes in vitro. Vaccinia virus has many properties desired in a vector. Although its intense immunogenicity potentially could limit its repeated use, this concern is not clinically significant since current strategies only employ it as a vector to prime the immune system.[ 41] Vaccinia proteins are highly immunogenic; consequently, it is likely that vaccinia can only be administered once or twice before a patient will develop high titers of neutralizing antibodies that will limit expression of recombinant genes. To circumvent this problem, researchers have studied potential alternative vectors such as attenuated vaccinia strains or avian pox viruses such as the canarypox virus.[42,43] An advantage of canarypox virus is that it replicates only in avian species. When administered to humans, it expresses its transgene products for 2 to 3 weeks in the infected mammalian cells but does not replicate and is unable to infect other cells, thus minimizing the likelihood that an enhanced immune response will neutralize the vaccine. The protective and antitumor activity of canarypox virus has been demonstrated in preclinical models. [ 41] The first phase I trial of a fowlpox recombinant CEA vaccine was conducted in 1999 by Marshall et al. Although no objective antitumor response was observed in patients with measurable disease during the trial, vaccination resulted in statistically significant increases in cytotoxic T-cell precursors specific for CEA. In preclinical studies, Hodge et al showed that a combination vaccination strategy with rV-CEA administered first followed by fowlpox-CEA could generate a more vigorous Tcell response than either vaccine alone. This approach was subsequently validated in a phase I study in patients with advanced CEA-positive malignancies. The study confirmed that immunization was most effective when rV-CEA was the primer and the fowlpox-CEA used to boost the immune response. In addition, this study showed that further increases in cytotoxic T cells could be produced by administering GM-CSF (Leukine) with subsequent vaccinations.
Dendritic Cells and CEA
Dendritic cells are the most potent type of antigen-presenting cells and are essential to prime the adaptive immune response. Although rare in the peripheral blood, dendritic cells can be harvested from the blood and expanded in vitro. Alternatively, administration of the hematopoietic growth factor flt3 ligand can dramatically mobilize these cells into the peripheral blood, increasing numbers by as much as 20-fold. When exposed for short periods to high concentrations of antigen, dendritic cells are said to be "pulsed." Pulsing has been used to generate more effective antigen-presenting cells. Dendritic cells pulsed with an HLA class I-restricted CEA peptide epitope have been shown to generate cytotoxic T lymphocytes specific for CEA.[49,50] In a phase I study of patients with advanced CEA-positive malignancies, Morse et al administered autologous dendritic cells pulsed with CEA mRNA. There were no significant treatment-related toxicities. Of 24 evaluable patients, 1 showed a complete response, 2 had minor responses, 3 exhibited stable disease, and 18 had progressive disease.