For much of the twentieth century, studies of the immunological response to tumours remained on the fringe of mainstream efforts in immunology. Scepticism was high concerning the existence of an immune response to cancer in humans and doubt existed concerning the applicability to humans of information derived from studies of transplantable murine tumours. A widely quoted article in the British Journal of Cancer reported no evidence of immune response to 27 different spontaneous tumours in mice and concluded that: “transplanted tumour systems...entail artifactual immunity associated with viral or chemical induction”1. Another review commenting on cancer immunotherapy concluded that: “It would be as difficult to reject the right ear and leave the left ear intact as it is to immunize against cancer”2.

Much has changed in the past 15 years, as increasing information about the molecular basis of tumour–host interactions has developed. The convergence of information resulting from basic studies in cellular immunology, along with increasing sophistication in biotechnology, which has made biologic reagents available in pharmacological amounts, has opened extraordinary possibilities for the development of effective immunotherapies for patients with cancer3. In addition, the ability to genetically modify cells involved in immunological reactions and to generate recombinant vectors containing genes encoding cancer antigens has resulted in early efforts at gene therapy of cancer.

During the past two decades, four sequential questions have characterized progress in the development of human cancer immunotherapy, discussion of which forms the basis of this review. Can immune manipulation cause the regression of established human cancers? What are the antigens involved in the immune recognition of human cancers? Can anti-tumour T cells be generated in patients by immunization with cancer antigens? What mechanisms limit cancer regression despite the in vivo generation of anti-tumour T cells?

Can immune manipulation cause cancer regression?

The first clear indication that immunological manipulations could cause the regression of established, invasive human cancers came from studies of the administration of interleukin-2 (IL-2) to humans with metastatic kidney cancer or melanoma4. IL-2, a cytokine produced by human T-helper lymphocytes, has a panoply of immune regulatory effects, including the expansion of lymphocytes following activation by specific antigen. IL-2 has no direct impact on cancer cells, which can grow unimpeded in vitro in high concentrations of IL-2. Thus, the impact of IL-2 on cancers in vivo derives from its ability to expand lymphocytes with anti-tumour activity.

The administration of high-dose recombinant IL-2 to humans was reported to mediate the regression of even bulky, invasive tumours in selected patients with metastatic melanoma, kidney cancer and non-Hodgkin's lymphoma4. These initial studies showed that 15–20% of patients with these metastatic cancers sustained an objective cancer regression (50% total reduction), and complete regression of metastatic tumour occurred in half of these patients. In another study of 409 IL-2-treated patients, 8.1% of patients with metastatic melanoma or kidney cancer achieved a complete response and 9% achieved a partial response5. With a median follow-up of 7.1 years, 82% of these completely responding patients remained in continuous, ongoing, complete regression from 3 to over 12 years from the onset of treatment (Fig. 1), and many were probably cured. Studies of 255 patients with metastatic kidney cancer6 and 270 patients with metastatic melanoma7 from 22 different institutions achieved similar results. These studies showed that this relatively simple immunological manipulation could mediate the regression of human cancer in a variety of organs and spurred intensive efforts to understand, at a molecular level, these complex immunological anti-tumour events.

Figure 1: Complete regression of a large liver metastasis from kidney cancer in a patient treated with IL-2.
figure 1

Regression is ongoing seven years later.

Which antigens are recognized in human cancers?

Multiple studies in experimental animals showed that cellular rather than humoral immune responses were responsible for the rejection of transplanted tumours or allogeneic (genetically different) tissues. With the exception of antibodies directed against growth factor receptors on cancer cells, the administration of antibodies has had little impact on the growth of solid tumours. Thus, significant effort has been devoted towards the identification of antigens recognized by human T lymphocytes8,9.

Both CD8+ cytotoxic T cells and CD4+ T-helper cells recognize antigens presented as small peptides in the groove of surface human leukocyte antigen (HLA; the human analogue of the major histocompatibility complex (MHC)) molecules. CD8+ cells recognize peptides of 8–10 amino acids in length, derived from intracellular cytoplasmic proteins, digested in proteosomes and presented via the endoplasmic reticulum on cell-surface class I HLA molecules. In contrast, CD4+ cells use a different intracellular pathway and present engulfed extracellular proteins, digested to peptides in intracellular endosomes and presented on cell-surface class II HLA molecules. Thus, the recognition of antigens by T cells involves the recognition of both peptides and specific HLA molecules. These different pathways of antigen processing required the development of separate techniques to identify tumour antigens, but all depended on the ability to generate T lymphocytes capable of recognizing human cancer cells.

Many antigens recognized by CD8+ cells have been identified by transfecting complementary DNA libraries from tumour cells into target cells expressing the appropriate HLA molecule, and then using anti-tumour T cells to identify the appropriate transfectants8,9. Alternatively, peptides eluted from the surface of human cancer cells (or from HLA molecules purified from cancer cells) can be pulsed onto antigen presenting cells (APCs) and tested for reactivity with specific anti-tumour lymphocytes10,11. Purification and sequencing of these peptides can then lead to the identification of the parent protein.

A third technique often referred to as 'reverse immunology' has been used successfully to identify whether candidate proteins, selected because of their unique overexpression on cancer cells, represent cancer antigens12. In vitro sensitization techniques are used to generate T cells that are reactive against the specific candidate antigens. If these T cells can also specifically recognize intact human cancer cells, the candidate protein is considered to be a tumour antigen. Another technique known as SEREX (serologic analysis of recombinant cDNA expression libraries)13 is based on the assumption that antibody production against a protein requires helper T cells. Diluted serum from cancer patients is used to detect proteins encoded by cancer cDNA libraries that are expressed in prokaryotes.

Because of the relative ease of generating human T cells that recognize melanomas, most human tumour antigens so far identified have been derived from this tumour type, although many antigens expressed on common epithelial tumours have also been identified. Examples of antigens recognized by CD8+ cells and presented on class I HLA molecules are presented in Table 1.

Table 1 Examples of human cancer antigens

Knowledge of class II-restricted human cancer antigens recognized by CD4+ cells has lagged behind the identification of class I-restricted antigens. Transfection of cDNA libraries into target cells using common techniques is not effective because the encoded proteins do not travel to the class II pathway. But a new technique14 involving the screening of cDNA libraries fused to genes encoding invariant chain sequences designed to guide the transfected proteins into the class II presentation pathway has the potential for wide applicability. By transfecting these fusion vectors into APCs engineered to contain the appropriate molecules required for class II presentation, many new human tumour antigens recognized by CD4+ T cells have been identified. Examples of class II-restricted cancer antigens are presented in Table 1.

There is increasing evidence of a relationship between infectious agents and the incidence of cancer15. Many of the viruses associated with oncogenesis also present proteins on the induced cancers that can serve as targets for immune attack (Table 2). Thus, the E6 and E7 epitopes on cervical cancers caused by human papillomavirus, epitopes from Epstein–Barr virus (EBV) on lymphomas, and human T-cell lymphotropic virus-1 epitopes on adult T-cell leukaemias represent a different class of cancer antigens. Immunization against these antigens might be useful in cancer therapy, and elimination of these infectious agents might also be a strategy to help prevent cancer.

Table 2 Infections agents as antigens to prevent or treat cancer*

Many different intracellular proteins are known to represent human cancer antigens. Stoler et al. estimated that about 11,000 genomic alterations occur in a cancer cell, and such genomic instability provides multiple opportunities for the development of cancer antigens either by the overexpression of individual proteins or by the expression of mutated proteins16.

Cancers of the haematopoietic system represent unique situations not shared by most cancers arising in solid tissues (the subject of this review). B lymphocytes can express unique idiotypes resulting from the gene rearrangements involved in antibody production. Because each B-cell clone gives rise to a lymphoma uniquely expressing this idiotype, it can serve as a cancer antigen. The graft-versus-host reactions in patients with leukaemia undergoing allogeneic bone marrow transplantation can be associated with graft-versus-tumour effects that can enhance the therapeutic impact of chemotherapy (see article in this issue by Appelbaum, pages 385–389). The antigens that serve as targets of this immune attack have not been clearly identified.

Much has been learned in the past decade concerning cancer antigens on solid cancers. Four general principles from these findings are presented in Box 1.

Can immunization generate anti-tumour T cells?

Cancer immunotherapies fall into either active immunization or passive transfer approaches and the identification of cancer antigens has impacted on both areas3. Passive (or adoptive) approaches involve the transfer of immune cells with anti-tumour reactivity. Landsteiner and Chase first described the transfer of delayed hypersensitivity reactions from one animal to another using cells from sensitized donors17. Early studies of tumour immunity in mice showed that specific immunity to tumours could be transferred to normal mice using lymphocytes from the spleen or peritoneal cavity of immunized donors18,19. Early studies in humans, done before the identification of human cancer antigens, involved the transfer to tumour-bearing patients of lymphokine-activated killer (LAK) cells with non-HLA-restricted ability to recognize and lyse cancer cells in vitro20. Despite the apparent success of LAK cells in treating micrometastases in experimental animals, clinical results in humans were disappointing. Techniques for growing large numbers of tumour-infiltrating lymphocytes (TILs) from resected tumours resulted in T-cell populations capable of specifically recognizing cancer antigens from about one-third of patients with melanoma21,22. TILs could be expanded to 1010–1011 cells and, when adoptively transferred along with IL-2 into melanoma patients, resulted in an objective response rate of about 35% (refs 23, 24). This objective regression rate was twice that seen with IL-2 alone and was also achieved in patients who had become refractory to treatment with IL-2 alone. In other studies25, tumour regression resulted from adoptive transfer of either fresh or cultured donor lymphocytes in patients with lymphoproliferative disorders or lymphomas following allogeneic transplantation.

The ability to successfully immunize patients against defined cancer antigens has facilitated the in vitro generation of anti-tumour T cells that can be expanded and used for adoptive therapy26. The ability to clone lymphocytes derived from single starting cells selected for their high avidity for tumour antigens, and to grow them to large numbers, is not only creating new possibilities for passive immunotherapy, but also provides a means of identifying the exact cellular characteristics that are required for mediating tumour rejection27,28,29. The genetic modification of these lymphocytes to improve their anti-tumour efficacy (for instance, by inserting genes encoding anti-tumour or chemokine receptors or genes encoding anti-tumour cytokines) is under active investigation30.

The achievements of active immunization against infectious diseases such as smallpox and polio have provided hope that cancer patients could be actively immunized against their own cancers to prevent or treat the disease. Before the identification of human cancer antigens, cancer vaccine approaches depended on immunization with either autologous or allogeneic whole cancer cells or cancer cell extracts (Box 2). But this approach is limited by the minute quantity of cancer antigenic molecules present in the intact cell. A variety of approaches to increase the immunogenicity of whole tumour cells has been attempted, including the injection of these cancer cells in a variety of adjuvants, or transducing cancer cells with genes encoding cytokines such as granulocyte–macrophage colony-stimulating factor, tumour necrosis factor or interferon-γ. Only limited evidence has been generated that these approaches can generate T cells in humans that can recognize intact tumour cells.

The identification of human cancer antigens has opened new approaches to the development of cancer vaccines (Box 2). Although often present in large amounts in the cell, epitopes from non-mutated differentiation antigens often exhibit low affinity for cell-surface HLA molecules. Mutated epitopes generally exhibit high affinity for HLA molecules, but often are derived from proteins with relatively poor expression in the cancer. Clinical trials using each of the different types of antigens will be required to determine which will be most effective in mediating anti-tumour immune effects. Multiple assays are available to assess the anti-tumour immune response of lymphocytes obtained before and after immunization. Assessment of immune status is often limited to circulating or lymph node lymphocytes rather than lymphocytes at the tumour site.

Recombinant expression of the genes encoding cancer antigens in Escherichia coli, yeast or baculovirus can result in the production of large quantities of purified cancer antigens for use in immunization, although the difficulty and expense of generating recombinant proteins that are suitable for human administration has significantly limited the application of this approach.

Many studies have used immunization with recombinant viruses that encode cancer antigens, including adenovirus, vaccinia virus and avipox31,32,33,34. But only weak generation of anti-tumour T cells has been reported using these approaches, which is perhaps due to the presence of neutralizing antibodies that exist in most humans against the envelope proteins of these viral vectors. Many current studies emphasize the use of recombinant avipox viruses, as humans have not previously been exposed to these viruses and the viruses cannot replicate in human tissue. To avoid possible immunization to viral envelope proteins, an alternate immunization approach has involved the direct injection of 'naked' DNA encoding cancer antigens into skin or muscle35. The poor efficiency in vivo of transfection of DNA has limited its value for the generation of immune responses against cancer antigens, although successful immunization against infectious agents has been reported36.

Increasing information concerning the importance of professional APCs such as dendritic cells or Langerhans cells in generating immune responses in humans has stimulated attempts to use these cells in cancer vaccines37. These attempts have used APCs pulsed with recombinant tumour antigens, tumour lysates or tumour-derived peptides, or infected with recombinant viruses or RNA. More recently, immunization with dendritic cells fused to whole tumour cells has been reported38.

T cells recognize peptides presented on the surface of tumour cells, a response that has led to immunization studies using immunodominant peptides derived from tumour antigens26. This approach has been surprisingly successful for generating high levels of circulating T cells directed against cancer antigens. The immunogenicity of peptides derived from tumour antigens has been increased substantially by altering specific amino-acid residues at positions that anchor the peptide to the appropriate HLA molecule39. Immunization with these modified peptides can result in as many as 4% of all circulating CD8+ T cells that are reactive with their own cancers. As the ability to immunize patients improves, the use of immunotherapy for the prevention of cancer recurrence in high-risk individuals represents an exciting area of clinical investigation.

When these approaches are used in the absence of cytokine administration, only sporadic instances of cancer regression result. Peptide vaccines given in conjunction with IL-2 may be capable of mediating substantially higher levels of cancer regression than administration of IL-2 alone. In one study, objective clinical responses were seen in 30–35% of patients receiving immunization with a modified peptide from the gp100 molecule (gp100:209-217(210M)) when administered with high-dose bolus IL-2 (refs 26, 40). This response rate was twice that seen in a large number of patients treated with a similar schedule of IL-2 alone5. However, the simultaneous administration of peptide plus IL-2 resulted in a decrease in circulating anti-tumour cells compared to that found in patients receiving peptide alone, possibly due to traffic of specific lymphocytes to the tumour site, with a consequent decrease of these cells in the circulation.

What mechanisms limit cancer regression?

The identification in growing tumours of TILs with the ability to specifically recognize cancer antigens and destroy tumour cells in vitro, coupled with the ability to successfully immunize patients to raise high levels of circulating anti-tumour T cells, raises a perplexing problem. Why do cancers continue to grow in the face of seemingly potent cellular anti-tumour reactions? No clear explanation for this phenomenon exists, but many hypotheses have been proposed41.

The factors limiting the therapeutic impact of anti-tumour T cells can be divided into either lymphocyte or tumour factors. Many of the T cells that are found within tumours are CD8+ cells. Experimental evidence in mice, as well as preliminary evidence in humans, suggests that the survival and effectiveness of CD8+ cells is dependent on helper factors derived from CD4+ cells27. Thus, a successful immune reaction depends on the generation of both CD4+ and CD8+ cells, each of which are stimulated by unique and separate antigens. The general technique for cloning CD4+ cells described earlier will be of value in discovering antigens that can be used to stimulate CD4+ helper reactions14.

Although T cells can be found that react against tumour, these may be present at insufficient levels to mediate tumour destruction. The immune reaction against EBV antigens in patients with infectious mononucleosis can rise as high as 40% of all circulating CD8+ cells, and such large numbers of anti-tumour T cells may be required to achieve anti-tumour effects. It is also possible that the T cells that are generated do not have sufficient avidity for tumour cells, or that the T cells that are generated do not produce the appropriate cytokines or have sufficient lytic activity. To study these phenomena, efforts at cloning T lymphocytes with especially high avidity for tumour cells or unique immunological functions will aid in the understanding of the types of immune cells that are required for successful anti-tumour immune responses27,28,29.

There are a variety of active mechanisms that may limit the effectiveness of immune stimulation. These include: active 'tolerance' of T cells resulting from the lack of expression of appropriate co-stimulatory molecules on the tumour; the active downregulation of T-cell-receptor signal transduction; the programmed cell death (apoptosis) of T cells when encountering tumour; or an active suppression by lymphocytes.

The tumour itself may be an active participant in causing immune suppression (for example, by producing local immunosuppressive factors, such as transforming growth factor-β) and there is evidence that tumours can lose expression of tumour or HLA antigens by mechanisms of immune selection. Lack of expression by the tumour of appropriate activation factors or lack of internal cellular mechanisms for apoptosis or other cell-destruction pathways may also protect the tumour cell from immune destruction.

Concluding comments

Studies of tumour immunology and immunotherapy have entered the mainstream of current studies in immunology and cancer research. The demonstration that even bulky invasive tumours can undergo complete regression under appropriate immune stimulation by IL-2 has shown that it is indeed possible to treat cancer successfully by immune manipulation. The recent discoveries of tumour antigens, and of successful means for raising anti-tumour T-cell numbers in humans by immunization, have solved some of the problems confronting the successful application of immunotherapy to the treatment of human cancer. Current studies are aimed at optimizing immunization and understanding the mechanisms used by the tumour to escape destruction.