14-11. The development of transplantable tumors in mice led to the discovery that mice could mount a protective immune response against tumors

The finding that tumors could be induced in mice after treatment with chemical carcinogens or irradiation, coupled with the development of inbred strains of mice, made it possible to undertake the key experiments that led to the discovery of immune responses to tumors. These tumors could be transplanted between mice, and the experimental study of tumor rejection has generally been based on the use of such tumors. If these bear MHC molecules foreign to the mice into which they are transplanted, the tumor cells are readily recognized and destroyed by the immune system, a fact that was exploited to develop the first MHC-congenic strains of mice. Specific immunity to tumors must therefore be studied within inbred strains, so that host and tumor can be matched for their MHC type.

Transplantable tumors in mice exhibit a variable pattern of growth when injected into syngeneic recipients. Most tumors grow progressively and eventually kill the host. However, if mice are injected with irradiated tumor cells that cannot grow, they are frequently protected against subsequent injection with a normally lethal dose of viable cells of the same tumor. There seems to be a spectrum of immunogenicity among transplantable tumors: injections of irradiated tumor cells seem to induce varying degrees of protective immunity against a challenge injection of viable tumor cells at a distant site. These protective effects are not seen in T-cell deficient mice but can be conferred by adoptive transfer of T cells from immune mice, showing the need for T cells to mediate all these effects.

These observations indicate that the tumors express antigenic peptides that can become targets of a tumor-specific T-cell response. The antigens expressed by experimentally induced murine tumors, often termed tumor-specific transplantation antigens (TSTAs), or tumor rejection antigens (TRAs), are usually specific for an individual tumor. Thus immunization with irradiated tumor cells from tumor X protects a syngeneic mouse from challenge with live cells from tumor X but not from challenge with a different syngeneic tumor Y, and vice versa (Fig. 14.10).

Figure 14.10

Tumor rejection antigens are specific to individual tumors. Mice immunized with an irradiated tumor and challenged with viable cells of the same tumor can, in some cases, reject a lethal dose of that tumor (left panels). This is the result of an immune (more...)

14-12. T lymphocytes can recognize specific antigens on human tumors

Tumor rejection antigens are peptides of tumor-cell proteins that are presented to T cells by MHC molecules. These peptides can become the targets of a tumor-specific T-cell response because they are not displayed on the surface of normal cells, at least not at levels sufficient to be recognized by T cells. Six different categories of tumor rejection antigens can be distinguished and examples of each of these are given in Fig. 14.11. The first category consists of antigens that are strictly tumor specific. These antigens are the result of point mutations or gene rearrangements, which often arise as part of the process of oncogenesis. Point mutations may evoke a T-cell response either by allowing de novo binding of a peptide to MHC class I molecules or by creating a new epitope for T cells by modification of a peptide that already binds class I molecules (Fig. 14.12). A special class of tumor-specific antigen in the case of B- and T-cell tumors, which are derived from single clones of lymphocytes, are the idiotypic sequences unique to the antigen receptor expressed by the clone.

Figure 14.11

Proteins selectively expressed in human tumors are candidate tumor rejection antigens. The molecules listed here have all been shown to be recognized by cytotoxic T lymphocytes raised from patients with the tumor type listed.

Figure 14.12

Tumor rejection antigens may arise by point mutations in self proteins, which occur during the process of oncogenesis. In some cases a point mutation in a self protein may allow a new peptide to associate with MHC class I molecules (lower left panel). In (more...)

The second category comprises proteins encoded by genes that are normally expressed only in male germ cells, which do not express MHC molecules and therefore cannot present peptides from these molecules to T lymphocytes. Tumor cells show widespread abnormalities of gene expression, including the activation of these genes and thus the presentation of these proteins to T cells; hence, these proteins are effectively tumor specific in their expression as antigens (Fig. 14.13).

Figure 14.13

Tumor rejection antigens are peptides of cell proteins presented by self MHC class I molecules. This figure shows two ways in which tumor rejection antigens may arise from unmutated proteins. In some cases, proteins that are normally expressed only in (more...)

The third category of tumor rejection antigen is comprised of differentiation antigens encoded by genes that are only expressed in particular types of tissue. The best examples of these are the differentiation antigens expressed in melanocytes and melanoma cells; a number of these antigens are proteins involved in the pathways of production of the black pigment, melanin. The fourth category is comprised of antigens that are strongly overexpressed in tumor cells compared with their normal counterparts (see Fig. 14.13). An example is HER-2/neu (also known as c-Erb-2), which is a receptor tyrosine kinase homologous to the epidermal growth factor receptor. This receptor is overexpressed in many adenocarcinomas, including breast and ovarian cancers, where it is linked with a poor prognosis. MHC class I-restricted, CD8-positive cytotoxic T lymphocytes have been found infiltrating solid tumors overexpressing HER-2/neu but are not capable of destroying such tumors in vivo. The fifth category of tumor rejection antigens is comprised of molecules that display abnormal posttranslational modifications. An example is underglycosylated mucin, MUC-1, which is expressed by a number of tumors, including breast and pancreatic cancers.

Proteins encoded by viral oncogenes comprise the sixth category of tumor rejection antigen. These oncoviral proteins are viral proteins that may play a critical role in the oncogenic process and, because they are foreign, they can evoke a T-cell response. Examples of such proteins are the human papilloma type 16 virus proteins, E6 and E7, which are expressed in cervical carcinoma.

Although each of these categories of tumor rejection antigen may evoke an anti-tumor response in vitro and in vivo, it is exceptional for such a response to be able to spontaneously eliminate an established tumor. It is the goal of tumor immunotherapy to harness and augment such responses to treat cancer more effectively. In this respect, the spontaneous remission occasionally observed in cases of malignant melanoma and renal cell carcinoma, even when disease is quite advanced, offers hope that this goal is achievable.

In melanoma, tumor-specific antigens were discovered by culturing irradiated tumor cells with autologous lymphocytes, a reaction known as the mixed lymphocyte-tumor cell culture. From such cultures, cytotoxic T lymphocytes could be identified that would kill, in an MHC-restricted fashion, tumor cells bearing the relevant tumor-specific antigen. Melanomas have been studied in detail using this approach. Cytotoxic T cells reactive against melanoma peptides have been cloned and used to characterize melanomas by the array of tumor-specific antigens displayed. These studies have yielded three important findings. The first is that melanomas carry at least five different antigens that can be recognized by cytotoxic T lymphocytes. The second is that cytotoxic T lymphocytes reactive against melanoma antigens are not expanded in vivo, suggesting that these antigens are not immunogenic in vivo. The third is that the expression of these antigens can be selected against in vitro and possibly also in vivo by the presence of specific cytotoxic T cells. These discoveries offer hope for tumor immunotherapy, an indication that these antigens are not naturally strongly immunogenic, and also a caution about the possibility of selecting, in vivo, tumor cells that can escape recognition and killing by cytotoxic T cells.

Consistent with these findings, functional melanoma-specific T cells can be propagated from peripheral blood lymphocytes, from tumor-infiltrating lymphocytes, or by draining the lymph nodes of patients in whom the melanoma is growing. Interestingly, none of the peptides recognized by these T cells derives from the mutant proto-oncogenes or tumor suppressor genes that are likely to be responsible for the initial transformation of the cell into a cancer cell, although a few are the products of mutant genes. The rest derive from normal genes but are displayed at levels detectable by T cells for the first time on tumor cells, as illustrated in Fig. 14.13. Antigens of the MAGE family are not expressed in any normal adult tissues, with the exception of the testis, which is an immunologically privileged site. They probably represent early developmental antigens reexpressed in the process of tumorigenesis. Only a minority of melanoma patients have T cells reactive against the MAGE antigens, indicating that these antigens are either not expressed or are not immunogenic in most cases. The most common melanoma antigens are peptides from the enzyme tyrosinase or from three other proteins—gp100, MART1, and gp75. These are differentiation antigens specific to the melanocyte lineage from which melanomas arise. It is likely that overexpression of these antigens in tumor cells leads to an abnormally high density of specific peptide:MHC complexes and this makes them immunogenic. Although in most cases tumor rejection antigens are presented as peptides complexed with MHC class I molecules, tyrosinase has been shown to stimulate CD4 T-cell responses in some melanoma patients by being ingested and presented by cells expressing MHC class II.

Tumor rejection antigens shared between most examples of a tumor, and against which tolerance can be broken, represent candidate antigens for tumor vaccines. The MAGE antigens are candidates because of their limited tissue distribution and their shared expression by many melanomas. It might seem dangerous to use tumor vaccines based on antigens that are not truly tumor-specific because of the risk of inducing autoimmunity. Often, however, the tissues from which tumors arise are dispensable; the prostate is perhaps the best example of this. With melanoma, however, some melanocyte-specific tumor rejection antigens are also expressed in certain retinal cells, in the inner ear, in the brain, and in the skin. Despite this, melanoma patients receiving immunotherapy with whole tumor cells or tumor-cell extracts, although occasionally developing vitiligo—a destruction of pigmented cells in the skin that correlates well with a good response to the tumor—do not develop abnormalities in the visual, vestibular, and central nervous systems, perhaps because of the low level of expression of MHC class I molecules in these sites.

In addition to the human tumor antigens that have been shown to induce cytotoxic T-cell responses (see Fig. 14.11), many other candidate tumor rejection antigens have been identified by studies of the molecular basis of cancer development. These include the products of mutated cellular oncogenes or tumor suppressors, such as Ras and p53, and also fusion proteins, such as the Bcr-Abl tyrosine kinase that results from the chromosomal translocation (t9;22) found in chronic myeloid leukemia. It is intriguing that, in each of these cases, no specific cytotoxic T-cell response has been identified in cultures of autologous lymphocytes with tumor cells bearing these mutated antigens. However, cytotoxic T lymphocytes specific for these antigens can be developed in vitro by using peptide sequences derived either from the mutated sequence or from the fusion sequence of these common oncogenic proteins; these cytotoxic T cells are able to recognize and kill tumor cells. In chronic myeloid leukemia, it is known that, after treatment and bone marrow transplantation, mature lymphocytes from the bone marrow donor infused into the patient can help to eliminate any residual tumor. At present, it is not clear whether this is a graft-versus-host effect, where the donor lymphocytes are responding to alloantigens expressed on the leukemia cells, or whether there is a specific anti-leukemic response. The ability to prime the donor cells against leukemia-specific peptides offers the prospect of enhancing the anti-leukemic effect while minimizing the risk of graft-versus-host disease. It is a challenge for immunologists to understand why these mutated proteins do not prime cytotoxic T cells in the patients in which the tumors arise. They are excellent targets for therapy, as they are unique to the tumor and have a causal role in oncogenesis.

14-13. Tumors can escape rejection in many ways

Burnet called the ability of the immune system to detect tumor cells and destroy them ‘immune surveillance.’ However, it is difficult to show that tumors are subject to surveillance by the immune system; after all, cancer is a common disease, and most tumors show little evidence of immunological control. The incidence of the common tumors in mice that lack lymphocytes is little different from their incidence in mice with normal immune systems; the same is true for humans deficient in T cells. The major tumor types that occur with increased frequency in immunodeficient mice or humans are virus-associated tumors; immune surveillance thus seems to be critical for control of virus-associated tumors, but the immune system does not normally respond to the novel antigens deriving from the multiple genetic alterations in spontaneous tumors. The goal in the development of anti-cancer vaccines is to break the tolerance of the immune system for antigens expressed mainly or exclusively by the tumor.

It is not surprising that spontaneously arising tumors are rarely rejected by T cells, as in general they probably lack either distinctive antigenic peptides or the adhesion or co-stimulatory molecules needed to elicit a primary T-cell response. Moreover, there are other mechanisms whereby tumors can avoid immune attack or evade it when it occurs (Fig. 14.14). Tumors tend to be genetically unstable and can lose their antigens by mutation; in the event of an immune response, this instability might generate mutants that can escape the immune response. Some tumors, such as colon and cervical cancers, lose the expression of a particular MHC class I molecule (Fig. 14.15), perhaps through immunoselection by T cells specific for a peptide presented by that MHC class I molecule. In experimental studies, when a tumor loses expression of all MHC class I molecules, it can no longer be recognized by cytotoxic T cells, although it might become susceptible to NK cells (Fig. 14.16). However, tumors that lose only one MHC class I molecule might be able to avoid recognition by specific CD8 cytotoxic T cells while remaining resistant to NK cells, conferring a selective advantage in vivo.

Figure 14.14

Tumors can escape immune surveillance in a variety of ways. First, tumors can have low immunogenicity (left panel). Some tumors do not have peptides of novel proteins that can be presented by MHC molecules, and therefore appear normal to the immune system. (more...)

Figure 14.15

Loss of MHC class I expression in a prostatic carcinoma. Some tumors can evade immune surveillance by loss of expression of MHC class I molecules, preventing their recognition by CD8 T cells. A section of a human prostate cancer that has been stained (more...)

Figure 14.16

Tumors that lose expression of all MHC class I molecules as a mechanism of escape from immune surveillance are more susceptible to NK cell killing. Regression of transplanted tumors is largely due to the actions of cytotoxic T cells (CTLs), which recognize (more...)

Yet another way in which tumors might evade rejection is by making immunosuppressive cytokines. Many tumors make these, although in most cases little is known of their precise nature. TGF-β was first identified in the culture supernatant of a tumor (hence its name, transforming growth factor-β) and, as we have seen, tends to suppress inflammatory T-cell responses and cell-mediated immunity, which are needed to control tumor growth. A number of tumors of different tissue origins, such as melanoma, ovarian carcinoma, and B-cell lymphoma, have been shown to produce the immunosuppressive cytokine IL-10, which can reduce dendritic cell development and activity. Thus, there are many different ways in which tumors avoid recognition and destruction by the immune system.

14-14. Monoclonal antibodies against tumor antigens, alone or linked to toxins, can control tumor growth

The advent of monoclonal antibodies suggested the possibility of targeting and destroying tumors by making antibodies against tumor-specific antigens (Fig. 14.17). This depends on finding a tumor-specific antigen that is a cell-surface molecule. Some of the cell-surface molecules that have been targeted in experimental clinical trials are shown in Fig. 14.18. So far there has been limited success with this approach, although, as an adjunct to other therapies, it holds promise. Some striking initial results have been reported in the treatment of breast cancer with a humanized monoclonal antibody, known as Herceptin, which targets a growth factor receptor, HER-2/neu, that is overexpressed in about a quarter of breast cancer patients. As we discussed in Section 14-12, this overexpression accounts for HER-2/neu evoking an antitumor T-cell response, although HER-2/neu is also associated with a poorer prognosis. It is thought that Herceptin acts by blocking interaction between the receptor and its natural ligand and by downregulating the level of expression of the receptor. The effects of this antibody can be potentiated when it is combined with conventional chemotherapy. A second monoclonal antibody that has promise for the treatment of non-Hodgkin's B-cell lymphoma binds to CD20 and is known as Rituximab. Ligation and clustering of CD20 transduces a signal that causes lymphocyte apoptosis. Monoclonal antibodies coupled to γ-emitting radioisotopes have also been used to image tumors, for the purpose of diagnosis and monitoring tumor spread (Fig. 14.19).

Figure 14.17

Monoclonal antibodies that recognize tumor-specific antigens might be used in a variety of ways to help eliminate tumors. Tumor-specific antibodies of the correct isotypes might be able to direct the lysis of the tumor cells by NK cells, activating the (more...)

Figure 14.18

Examples of tumor antigens that have been targeted by monoclonal antibodies in therapeutic trials. CEA, carcinoembryonic antigen.

Figure 14.19

Recurrent colorectal cancer can be detected with a radiolabeled monoclonal antibody against carcinoembryonic antigen. A patient with a possible recurrence of a colorectal cancer was injected intravenously with an indium- 111-labeled monoclonal antibody (more...)

The first reported successful treatment of a tumor with monoclonal antibodies used anti-idiotypic antibodies to target B-cell lymphomas whose surface immunoglobulin expressed the corresponding idiotype. The initial course of treatment usually leads to a remission, but the tumor always reappears in a mutant form that no longer binds to the antibody used for the initial treatment. This case represents a clear example of genetic instability enabling a tumor to evade treatment.

Other problems with tumor-specific or tumor-selective monoclonal antibodies as therapeutic agents include inefficient killing of cells after binding of the monoclonal antibody and inefficient penetration of the antibody into the tumor mass. The first problem can often be circumvented by linking the antibody to a toxin, producing a reagent called an immunotoxin; two favored toxins are ricin A chain and Pseudomonas toxin. Both approaches require the antibody to be internalized to allow the cleavage of the toxin from the antibody in the endocytic compartment, allowing the toxin chain to penetrate and kill the cell.

Two other approaches using monoclonal antibody conjugates involve linking the antibody molecule to chemotherapeutic drugs such as adriamycin or to radioisotopes. In the first case, the specificity of the monoclonal antibody for a cell-surface antigen on the tumor concentrates the drug to the site of the tumor. After internalization, the drug is released in the endosomes and exerts its cytostatic or cytotoxic effect. Monoclonal antibodies linked to radionuclides (see Fig. 14.17) concentrate the radioactive source in the tumor site. Both these approaches have the advantage of also killing neighboring tumor cells, because the released drug or radioactive emissions can affect cells adjacent to those that actually bind the antibody. Ultimately, combinations of toxin-, drug-, or radionuclide-linked monoclonal antibodies, together with vaccination strategies aimed at inducing T cell-mediated immunity, might provide the most effective cancer immunotherapy.

14-15. Enhancing the immunogenicity of tumors holds promise for cancer therapy

Although vaccines based on tumor antigens are, in principle, the ideal approach to T cell-mediated cancer immunotherapy, it may be many decades before the dominant tumor antigens for common cancers are identified. Even then, it is not clear how widely the relevant epitopes will be shared between tumors, and peptides of tumor rejection antigens will be presented only by particular MHC alleles. To be effective, a tumor vaccine may therefore need to include a range of tumor antigens. MAGE-1 antigens, for example, are recognized only by T cells in melanoma patients expressing the HLA-A1 haplotype. However, the range of MAGE-type proteins that has now been characterized encompasses peptide epitopes presented by many HLA class I and II molecules.

Until recently, most cancer vaccines have used the individual patient's tumor removed at surgery as a source of vaccine antigens. These cell-based vaccines are prepared by mixing either irradiated tumor cells or tumor extracts with bacterial adjuvants such as BCG or Corynebacterium parvum, which enhance their immunogenicity (see Appendix I, Section A-4). Such vaccines have generated modest therapeutic results in melanomas but have, in general, been disappointing.

Where candidate tumor rejection antigens have been identified, for example in melanoma, experimental vaccination strategies include the use of whole proteins, peptide vaccines based on sequences recognized by cytotoxic T lymphocytes (either administered alone or presented by the patient's own dendritic cells), and recombinant viruses encoding these peptide epitopes. A novel experimental approach to tumor vaccination is the use of heat-shock proteins isolated from tumor cells. The underlying principle of this therapy is that one of the physiological activities of heat-shock proteins is to act as intracellular chaperones of antigenic peptides. There is evidence for receptors on the surface of professional antigen-presenting cells that take up certain heat-shock proteins together with any bound peptides. Uptake of heat-shock proteins via these receptors delivers the accompanying peptide into the antigen-processing pathways leading to peptide presentation by MHC class I molecules. This experimental technique for tumor vaccination has the advantage that it does not depend on any prior knowledge of the nature of the relevant tumor rejection antigens, but the disadvantage that the heat-shock proteins purified from the cell carry very many peptides, so that any tumor rejection antigen might constitute only a tiny fraction of the peptides bound to the heat-shock protein.

A further experimental approach to tumor vaccination in mice is to increase the immunogenicity of tumor cells by introducing genes that encode co-stimulatory molecules or cytokines. This is intended to make the tumor itself more immunogenic. The basic scheme of such experiments is shown in Fig. 14.20. A tumor cell transfected with the gene encoding the co-stimulatory molecule B7 (see Section 8-5) is implanted in a syngeneic animal. These B7-positive cells can activate tumor-specific naive T cells to become armed effector T cells able to reject the tumor cells. They are also able to stimulate further proliferation of the armed effector cells that reach the site of implantation. These T cells can then target the tumor cells whether they express B7 or not; this can be shown by reimplanting nontransfected tumor cells, which are also rejected.

Figure 14.20

Transfection of tumors with the gene for B7 or for GM-CSF enhances tumor immunogenicity. A tumor that does not express co-stimulatory molecules will not induce an immune response, even though it might express tumor rejection antigens (TRAs), because naive (more...)

The second strategy, that of introducing cytokine genes into tumors so that they secrete the relevant cytokine, is aimed at attracting antigen-presenting cells to the tumor and takes advantage of the paracrine nature of cytokines. In mice, the most effective tumor vaccines so far are tumor cells that secrete granulocyte-macrophage colony-stimulating factor (GM-CSF), which induces the differentiation of hematopoietic precursors into dendritic cells and attracts these to the site. GM-CSF is also thought to function as an adjuvant, activating the dendritic cells. It is believed that these cells process the tumor antigens and migrate to the local lymph nodes, where they induce potent anti-tumor responses. The B7-transfected cells seem less potent in inducing anti-tumor responses, perhaps because the bone marrow-derived dendritic cells express more of the molecules required to activate naive T cells than do B7-transfected tumor cells. In addition, the tumor cells do not share the dendritic cells' special ability to migrate into the T-cell areas of the lymph nodes, where they are optimally placed to interact with passing naive T cells (see Section 8-6).

The potency of dendritic cells in activating T-cell responses provides the rationale for yet another strategy for vaccinating against tumors. The use of antigen-pulsed autologous dendritic cells to stimulate therapeutically useful cytotoxic T-cell responses to tumors has been developed in experimental models, and there have been initial trials in humans with cancer.

Clinical trials are in progress to determine the safety and efficacy of many of these approaches in human patients. What is uncertain is whether people with established cancers can generate sufficient T-cell responses to eliminate all their tumor cells under circumstances in which any tumor-specific naive T cells might have been rendered tolerant to the tumor. Moreover, there is always the risk that immunogenic transfectants will elicit an autoimmune response against the normal tissue from which the tumor derived.

Summary

Tumors represent outgrowths of a single abnormal cell, and animal studies have shown that some tumors elicit specific immune responses that suppress their growth. These seem to be directed at MHC-bound peptides derived from antigens that might be mutated, inappropriately expressed, or overexpressed in the tumor cells. T-cell deficient individuals, however, do not develop more tumors than normal individuals. This is probably chiefly because most tumors do not make distinctive antigenic proteins or do not express the co-stimulatory molecules necessary to initiate an adaptive immune response. Tumors also have other ways of avoiding or suppressing immune responses, such as ceasing to express MHC class I molecules, or making immunosuppressive cytokines. Monoclonal antibodies have been developed for tumor immunotherapy by conjugation to toxins or to cytotoxic drugs or radionuclides, which are thereby delivered at high dose specifically to the tumor cells. More recently, attempts have been made to develop vaccines based on tumor cells taken from patients and made immunogenic by the addition of adjuvants, or by pulsing autologous dendritic cells with tumor-cell extracts or tumor antigens. This approach has been extended in animal experiments to transfection of tumor cells with genes encoding co-stimulatory molecules or cytokines that attract and activate dendritic cells.