T cells take aim at cancer.

E since Paul Ehrlich introduced the term ‘‘magic bullet,’’ the exquisite capacity for specificity afforded by the immune system has always underpinned its appeal as a therapeutic weapon against cancer. The first clinical validation of this principle came in the form of mAb administration, which, after a decade of skepticism, produced therapeutic successes in breast cancer and B cell lymphomas. T cell-based immunotherapy offers an even broader therapeutic potential, owing to the ability of T cells to recognize peptides derived from proteins in any cellular compartment. These peptides are produced when proteins are processed by specialized machinery; they then combine with MHC molecules, which transport them to the cell surface where the peptide– MHC complexes can be recognized by T cells. Although hundreds of experiments in rodent tumor models support the notion that tumor-specific T cells can be activated to inhibit tumor growth, direct evidence for therapeutic capacity in human cancer has been marginal. Now, two clinical studies using adoptive transfer of melanoma-specific T cells provide clear evidence for the ability of T cells to mediate antitumor activity and provide important general principles for T cell immunotherapy (1, 2). The most direct approach for analyzing the activity of tumor-reactive T cells is to grow T cell clones ex vivo with defined specificity and adoptively transfer them back into the tumor-bearing host. It is typically much more difficult to grow tumor-reactive T cells from cancer patients than to grow virus-reactive T cells. This has been thought to be because the immune system views tumors more as ‘‘self’’ than foreign; thus, T cells become tolerant to tumor antigens. Nonetheless, tumor tolerance is relative and not absolute, leaving open a potential window of immunotherapeutic opportunity (3). Melanoma has been the most popular target for T cell-based immunotherapy in part because it is much easier to grow tumorreactive T cells from melanoma patients than any other type of human cancer. These T cells have been be used to define specific antigens recognized by T cells. A surprising finding to fall out of these antigen discovery efforts is that the melanoma antigens most commonly recognized are not tumor specific, but rather are tissue-specific melanocyte antigens such as tyrosinase, MART-1 MelanA, and gp100 (4). Thus, at least in melanoma, tumor immunity is in part synonymous with tissue-specific autoimmunity. This autoreactivity is not an absolute barrier to tumor immunotherapy because many of the common cancers such as prostate cancer, breast cancer, pancreatic cancer, and melanoma arise from tissues dispensable to life. The ability to grow melanocyte-specific T cells from melanoma patients does not mean that in vivo tolerance has been broken as a consequence of tumorigenesis. Many of the T cell clones grown from these patients recognize peptides that bind poorly to their presenting MHC allele or possess T cell receptors with relatively low affinity for their cognate peptide–MHC complex. These T cells thus recognize melanoma melanocyte antigens weakly compared with typical virus-specific T cells and presumably fail to become activated in vivo, thereby ‘‘ignoring’’ melanocytes and melanoma cells. Evidence from animal models suggests that high-affinity, tumorreactive T cells are more actively tolerized than weaker low-affinity T cells by mechanisms involving deletion, anergy induction, or suppression (5). Given the emerging view that individuals with cancer contain tumor-reactive T cells that are naturally tolerant of their cancer, a central question in T cell immunotherapy is whether they can be activated and expanded to induce clinically useful antitumor responses. Even low-affinity T cells can potentially discharge their effector function if properly activated. For example, CD8 T cells that recognize peptides presented by MHC class I molecules require 100 peptide–MHC complexes plus multiple costimulatory signals to become primed; however, once activated (to so-called killer T cells or cytotoxic T lymphocytes) they can kill target cells expressing as few as one peptide–MHC complex in the absence of any additional accessory signals (6, 7) (Fig. 1). To address the in vivo activity of melanoma-specific T cells, Yee et al. (1) grew T cell clones from patients with advanced melanoma specific for two well-defined melanoma melanocyte antigens, MART1 Melan A or gp100. They initially primed T cells in vitro by using peptide-loaded dendritic cells, the most powerful stimulator of immune responses. Many of the T cell clones grown under these conditions fail to kill tumor cells even when activated, reflecting the generally low affinities of T cells remaining after tolerance has operated on the available repertoire. However, a subset of CD8 T cell clones were indeed capable of killing autologous tumor cells, confirming the existence of tumor-reactive T cells in vivo that could be activated and expanded in vitro. T cell clones capable of recognizing tumor were rapidly expanded in culture (using an anti-CD3 antibody that stimulates through the T cell receptor signaling pathway) and roughly 6 billion cells with pure clonal specificity were transferred back to patients every 2–3 weeks. The first infusion was given without any systemic IL-2, and the subsequent infusions were given with relatively low doses of IL-2 that should activate the high affinity IL-2 receptor ( chains) expressed on activated T cells. IL-2 is a T cell growth factor produced by CD4 helper T cells. A number of important insights can be gleaned from the analysis of the 10 treated patients in the Yee study. First and foremost, they demonstrated that adoptively transferred T cell clones could persist in vivo and further could traffic into tumors. T cell clones were tracked in vivo by staining cells with peptide–MHC tetramers that exhibit specific and stable binding to cognate T cells, thereby allowing flow cytometric analysis (8). The persistence of