IPP, isopentenyl pyrophosphate; KAR, killer-activating receptor; KIR, killer-inhibitory receptor; PBL, peripheral blood lymphocytes; TCR, T-cell receptor; TIL, tumor-infiltrating lymphocytes(s).
T lymphocytes are classified into 2 subsets based on their T-cell receptor (TCR) expression. The vast majority of T cells expresses an αβ TCR heterodimer. These αβ T cells recognize antigenic peptides presented by MHC class I (for CD8+ T cells) or MHC class II molecules (for CD4+ T cells). Concepts of cancer immunotherapy are mostly concerned with activation of these MHC-restricted αβ T cells. Until recently, a numerically small subset of T cells, which expresses an alternative TCR composed of a CD3-associated γδ heterodimer, has received far less attention as a potential agent in cancer therapy. These γδ T cells share with αβ T cells certain effector functions such as cytokine production and potent cytotoxic activity but recognize different sets of antigens, usually in a non-MHC-restricted fashion. Different subsets of human γδ T cells recognize stress-inducible MHC class I-related molecules frequently expressed on epithelial tumor cells or phosphorylated metabolites which can be generated by tumor cells. In line with this, many tumor cells are highly susceptible to γδ T-cell mediated lysis. In our article, we summarize the available evidence for a contribution of human γδ T cells in tumor defense and discuss potential strategies for the immunotherapy of tumors based on the endogenous activation and/or adoptive transfer of tumor-reactive γδ T lymphocytes. © 2004 Wiley-Liss, Inc.
T lymphocytes recognize antigen via the heterodimeric T-cell receptor (TCR) molecule, which is noncovalently associated with the CD3 molecular complex. Most T cells carry a TCR that is composed of an αβ chain heterodimer. These T cells, referred to as αβ T cells, are commonly subdivided on the basis of the surface markers CD4 and CD8. In the peripheral blood of healthy adults, approximately 2/3 of αβ T cells express the CD4 antigen, while approximately 1/3 carries CD8. CD4+ αβ T cells recognize antigenic peptides, which are presented in the context of major histocompatibility complex (MHC) class II antigens by cells specialized for antigen-presentation such as dendritic cells.1 CD4+ cells are mostly helper T cells, that support a variety of immune functions by the production of cytokines. With respect to tumor defense, it is mainly the so-called Th1 cells, which secrete IL-2 and interferon-γ (IFN-γ), cytokines particularly required for the activation of CD8+ cytotoxic αβ T cells (CTL).2 CD8+ αβ T cells recognize peptides presented by MHC class I molecules.1 Because the vast majority of nucleated cells, including tumor cells, express MHC class I molecules, CD8+ αβ CTL can kill tumor cells, which naturally present tumor antigen-derived peptides in their MHC class I molecules.3
Besides CD8+ CTL, other lymphocytes with potent cytolytic effector function contribute to the immune defense against tumors. Natural killer (NK) cells are CD3-negative cytotoxic cells, which spontaneously recognize and kill certain tumor cells without prior sensitization. The activity of NK cells is tightly controlled by a set of activating and inhibitory NK receptors (killer activating receptors, KAR, and killer inhibitory receptors, KIR), which interact (or do not interact in the case of defective expression) with MHC class I molecules on the target cell.4
Substantial evidence suggests that γδ T cells represent an important additional player in the immune system's arsenal of effector cells with potential anti-tumor activity. γδ T cells account for 5–10% of CD3+ peripheral blood T cells but constitute a dominant fraction of T cells at other anatomical sites such as the intestinal epithelia. γδ T cells differ from conventional αβ T cells in several aspects. First, most γδ T cells lack CD4 or CD8 antigens and hence display a “double-negative” phenotype, although a sizeable fraction expresses CD8. The absence of CD4 or CD8 expression on the majority of circulating γδ T cells is well in line with the lack of MHC restriction in antigen recognition of this T-cell subset.5 Second, the germ-line encoded TCR repertoire of γδ T cells is strikingly small when compared to the large TCR repertoire of αβ T cells. In humans, there are only 6 expressed Vγ genes and a similarly small number of Vδ genes. Even more striking is the restricted usage of the few available gene segments at different anatomical sites. In the peripheral blood of healthy adult individuals, there is a clear preponderance of cells expressing Vγ9 paired with Vδ2, which can account for 50 up to 95% of all peripheral blood γδ T cells.6 In contrast, the intraepithelial γδ T cells predominantly express Vδ1 in association with various Vγ elements.7 The dominance of Vγ9Vδ2 cells among peripheral blood γδ T cells in the adult is not genetically determined but apparently results from the exposure to environmental Vγ9-specific ligands (such as bacterial phosphoantigens; see below) during childhood.8 The reason for the different anatomical localization of subsets of γδ T cells is not yet precisely understood. From a teleological point of view, it can be anticipated that T cells homing to a particular anatomical localization are specialized to recognize via their TCR ligands or antigens that are specifically encountered at this anatomical site. In line with this assumption, intraepithelial human γδ T cells expressing the Vδ1-encoded TCR have been found to specifically recognize stress-inducible MHC class I-related molecules on epithelial cells.9 Homing of lymphocytes is a complex process that is regulated by the interaction of homing receptors and adhesion molecules with their corresponding ligands, but also by the expression of chemokine receptors and their interaction with the locally produced chemokines. In this regard, it is of interest that the major subsets of human γδ T cells identified on the basis of their differential Vδ gene usage also express different patterns of chemokine receptors.10, 11
A major difference between αβ and γδ T cells concerns the range of antigens or ligands that are recognized by the respective TCRs. While αβ T cells recognize peptides bound to MHC class I or class II molecules (and thus display MHC-restricted recognition), most γδ T cells see their ligands without the involvement of classical MHC antigens.5 Murine γδ T cells have been found to recognize MHC class Ib-related antigens such as T22, certain viral glycoproteins and heat-shock proteins such as hsp60, as well as poorly defined antigens on stressed epithelial cells (see reference 5 for an overview). The 2 major subsets of human γδ T cells, i.e., Vδ2 and Vδ1 cells, recognize unique ligands which are not seen by human αβ T cells. Originally, it was observed that Vγ9Vδ2 T cells dramatically expand in vitro when unfractionated peripheral blood lymphocytes were stimulated with extracts of Mycobacterium tuberculosis and certain other microorganisms.12, 13 The molecular nature of the bacterial products stimulating human Vγ9Vδ2 γδ T cells has been characterized in great detail in recent years. Quite surprisingly, γδ-stimulating ligands are small molecules that comprise microbe-derived metabolites of the isoprenoid biosynthesis pathway and alkylamines.14, 15 The most potent ligands for human Vγ9Vδ2 γδ T cells are phosphorylated intermediates of the so-called nonmevalonate (or Rohmer) pathway of isoprenoid biosynthesis, which is used by many bacteria including Mycobacterium tuberculosis but not by eukaryotic cells.16, 17, 18 These so-called phosphoantigens are specifically recognized by Vγ9Vδ2 T cells and do not require presentation by classical or nonclassical MHC molecules.19 The recognition of such phosphoantigens depends on a particular topology at the surface edge of the pocket structure of the Vγ9Vδ2-encoded TCR.20 Synthetic analogues of bacterial phosphoantigens have been generated, which retain potent stimulatory activity for Vγ9Vδ2 T cells and thus can be used for large-scale expansion of these cells in vitro.17
The second most frequent subset of human γδ T cells (preferentially located within the intestinal epithelia) expresses Vδ1 in association with various Vγ gene segments. These γδ T cells recognize stress-induced MHC class I chain-related genes MICA and MICB.9 MICA and MICB are also ligands for the natural killer cell receptor NKG2D, which can be expressed on Vδ1 and Vδ2 γδ T cells and delivers an activating signal.21, 22 Therefore, the reactivity of Vδ1 γδ T cells towards MICA/MICB-expressing cells is dictated by signals generated from TCR- as well as from costimulatory NKG2D-dependent ligand recognition.23 Some human γδ T cells also recognize MHC-related CD1 molecules, specifically CD1c. Since CD1 molecules may have evolved to present lipid antigens to T lymphocytes, it is not unlikely that CD1c-reactive human γδ T cells in fact recognize self-lipid presented by CD1c.24 Additional antigens, which are reportedly recognized by human Vδ1 γδ T cells, include cytomegalovirus glycoprotein (in the absence of MHC presentation) and staphylococcal enterotoxin B (SEB) but not SEA.25, 26
The functional capacities of γδ T cells include cytokine production and potent cytotoxic effector activity. Human Vγ9Vδ2 T cells rapidly produce proinflammatory cytokines including interferon-γ (IFN-γ) and tumor necrosis factor-α (TNF-α) upon recognition of bacterial phosphoantigens.27 Under appropriate culture conditions, however, these cells can also be instructed to produce Th2 cytokines such as IL-4.28 Moreover, some γδ T cells can produce additional cytokines that are not commonly made by the conventional αβ T cells. Thus, intraepithelial murine γδ T cells can produce keratinocyte growth factor-1 (also termed fibroblast growth factor-7), which contributes to wound repair,29 and human γδ T cells have been found to produce fibroblast growth factor-9.30 These results suggest that γδ T cells have a distinct role in local immune surveillance and tissue repair.31
Activated γδ T cells are highly active killer cells. They express perforin and granzymes and use this pathway to kill mycobacteria-infected macrophages,32, 33, 34, 35, 36 but they also kill microorganisms via the production of the anti-bacterial peptide granulysin.37 Granulysin has been shown to induce apoptosis in tumor cells,38 and hence this mechanism might also contribute to the cytotoxic anti-tumor activity of γδ T cells, by virtue of granulysin-induced apoptosis of tumor cells. In addition, γδ T cells also explore the FasL-Fas pathway to induce apoptosis in Fas-expressing target cells.
An obvious approach to investigate the potential contribution of a particular lymphocyte subset to anti-tumor immune responses is the analysis of the subset distribution of tumor-infiltrating lymphocytes (TIL). Such studies have provided evidence for a potential role of γδ T cells in the immune response to tumors of epithelial origin. Choudhary and co-workers39 analyzed the tumor-infiltrating lymphocytes of a patient with recurrent renal cell carcinoma. They observed a high proportion of γδ T cells within short-term cultured tumor-infiltrating lymphocytes. More importantly, they found a preponderance of a particular Vγ3Vδ1-expressing γδ T-cell clone in 3 consecutive renal carcinomas that developed in this patient over a period of 3 years. This γδ T-cell clone lysed the autologous tumor cells as well as several allogeneic renal carcinomas but not other tumor cells or nonmalignant cells, suggesting a specific role in the immune response to the autologous renal carcinoma. Apparently, the potential immune response of γδ T cells to renal carcinoma is not restricted to the Vδ1-expressing subset. Thus, Kobayashi and colleagues40 noticed an increase in the proportion of Vγ9Vδ2 T cells in the peripheral blood of renal cell carcinoma patients when compared to healthy control subjects. The analysis of the TCR junctional diversity suggested an oligoclonal expansion of Vγ9Vδ2 T cells in the peripheral blood of these patients. In their studies, Kobayashi et al.40 also observed elevated proportions of γδ T cells among TIL, but not in all cases was there a clear correlation between increased proportions of γδ T cells in the peripheral blood vs. TIL. In line with a selective infiltration of γδ T cells into the tumor, Olive and co-workers41 also observed a significantly higher TCRδ gene expression in renal cell carcinomas compared to normal kidney, and they confirmed differences in the TCR Vγ and Vδ repertoires of TIL vs. PBL γδ T cells, again suggesting a tumor-dependent shaping of the local γδ TCR repertoire. γδ T cells with cytotoxic activity against the autologous tumor cells are present among the TIL also in other tumors of epithelial origin. Maeurer and colleagues42 isolated γδ T-cell clones from the TIL of patients with primary and metastatic colorectal cancer, which lysed autologous and allogeneic colorectal cancer cells but also a range of epithelial tumor cells of renal and pancreatic origin. Similarly, γδ T cells with potent cytotoxic activity against the autologous tumor have been isolated from the tumor tissue of lung cancer patients.43, 44 In contrast to many epithelial tumors, B-cell lymphomas do not seem to be preferentially infiltrated by γδ T cells, despite substantial sensitivity of many B-cell lymphomas towards γδ T-cell-dependent lysis and experimental evidence for a role of γδ T cells (see below). Thus, Kuriyama et al.45 found that the proportion of tumor-infiltrating γδ T cells in these patients was not different from that of lymph node γδ T cells in reactive hyperplasia patients. However, increased relative and absolute numbers of γδ T cells were present in the blood of a subgroup of these patients.
While increased proportions of γδ T cells within tumor-infiltrating lymphocytes suggest a role for these cells in tumor defense, no definitive conclusions can be drawn from such studies. Experimental models can be designed, however, to ask whether increased tumor rates occur in the absence of γδ T cells. In fact, there is clear evidence from studies with gene-targeted TCRγ−/− mice (which completely lack all γδ T cells) for an indispensable role of epithelial γδ T cells to control cutaneous malignancy. In their studies, Girardi et al.46 clearly demonstrated increased tumorigenesis in TCRδ−/− mice exposed to chemical carcinogenesis (phorbolester treatment) or inoculation of a squamous cell carcinoma line. In the absence of γδ T cells, there was also an increased incidence of progression of papillomas into malignant carcinomas. More recently, the role of γδ T cells in tumor defense was confirmed when TCRδ−/− mice were inoculated with B16F10 melanoma cells or treated with the carcinogen methylcholanthrene.47 Additional evidence for a prominent role of γδ T cells in the immune surveillance of spontaneous B-cell lymphomas was recently provided by Street and co-workers.48 These authors observed a high incidence of spontaneous disseminated B-cell lymphomas in gene-targeted mice lacking β2-microglobulin and perforin. By inoculating these tumors into a variety of gene-targeted, mutant or lymphocyte subset-depleted mice, they established that NK cells and γδ T cells (but not αβ T cells) efficiently rejected the tumors. Taken together, these different lines of evidence all point to the significance of γδ T cells in the immune surveillance and immune defense of tumors.
Several attempts have been made to identify the tumor antigens that are recognized by tumor-reactive human γδ T cells. Among the well-characterized ligands are the stress-inducible MHC class I-related MICA/MICB molecules that are expressed on many epithelial tumors and can be recognized by Vδ1-expressing human γδ T cells.9, 49 As already mentioned, MICA is also a ligand for the activating NK receptor NKGD2, which is expressed on Vδ121 but also on at least some Vδ2 γδ T cells (our own observations). The simultaneous interaction of tumor-expressed MICA with the TCR and NKGD2 on γδ T cells might result in a synergistic mode of activation,21 which, however, can be further modulated by the coexpression of inhibitory NK receptors.50 Immobilized recombinant MICA has been used to selectively activate and expand human Vδ1 γδ T cells in vitro, which retained potent cytotoxic activity towards MICA-expressing epithelial tumor cells.51
Apart from stress-inducible MHC-related molecules, heat-shock proteins (hsp) have been repeatedly implicated as tumor-expressed ligands for human γδ T cells, mainly on the basis of antibody blocking experiments. Hsp have thus been identified as target antigens on epithelial tumor cells52, 53 but also on Daudi lymphoma cells.54 However, a role for hsp as the γδ-activating ligand on Daudi lymphoma cells has been recently questioned by the results of Gober et al.,55 based on the observation that Daudi cells produce mevalonate metabolites that are recognized by Daudi-reactive Vγ9Vδ2 T cells. In these experiments, the γδ-stimulatory capacity was abrogated when Daudi cells were treated with mevastatin, an inhibitor of the enzyme HMG CoA reductase that is essentially required for the generation of isopentenyl pyrophosphate (IPP) from HMG Co A. On the other hand, Daudi cells transfected with HMG CoA reductase (thus leading to augmented IPP production) exerted enhanced stimulatory activity on γδ T cells.55 Furthermore, enhanced accumulation of intracellular IPP due to aminobisphosphonate-mediated inhibition of the IPP-degrading enzyme farnesyl pyrophosphate synthase also increased the γδ-stimulatory activity of Daudi cells. Together, these results indicate that some tumor cells – through alterations in their mevalonate metabolic pathway – might produce sufficiently high amounts of the pyrophosphate IPP to trigger γδ T-cell activation.55 It is quite possible, however, that additional antigens play a role. As reported by Schnurr et al., dendritic cells pulsed with apoptotic pancreatic tumor cells were able to activate IFN-γ production in a large fraction of Vγ9-expressing γδ T cells.56 Under these conditions, it is unlikely that IPP accumulated in tumor cells would be transferred to dendritic cells by the uptake of apoptotic tumor cells.
In view of the efficient lysis of many tumor cells by human γδ T cells, it is tempting to exploit the efficacy of a γδ T-cell-based immunotherapy of cancer in an in vivo model. However, since the above described phosphoantigens are not seen by mouse γδ T cells (due to the lack of a Vγ9Vδ2 homologous TCR), there is no easy way to address this issue in a murine system that would allow to extrapolate its significance to the human situation. Therefore, several investigators have addressed this issue using an adoptive transfer model where human γδ T cells and tumor cells are adoptively transferred into severe combined immunodeficiency (SCID) mice. Such human-SCID mouse chimeric models have been successfully used for the evaluation of anti-cancer therapies.57 In this model, Malkovska et al.58 showed that Daudi-activated human γδ T cells as well as Vγ9Vδ2 T-cell clones significantly reduced the in vivo growth of Daudi but not of Raji lymphoma (which are not lysed by Vγ9Vδ2 T cells). Similarly, the growth of human nasopharyngeal carcinoma in nude mice could be arrested by adoptively transferred allogeneic γδ T cells obtained from healthy donors.59 Notably, this model is also suitable to analyze the effect of γδ T cells on the growth of autologous tumor xenografts. Lozupone and colleagues60 observed significant anti-tumor activity of γδ T cells on autologous melanoma growth. Efficient inhibition of tumor growth required repeated administration of the tumor-reactive autologous γδ T cells, which in this case expressed Vδ1 rather than Vδ2.60 In our own human-SCID mouse experiments, we also noted a significant inhibitory effect of adoptively transferred human Vγ9Vδ2 T cells on the in vivo growth of MEWO melanoma and PancTU1 pancreatic carcinoma. In agreement with published reports,60 optimal anti-tumor activity required the repeated transfer of Vγ9Vδ2 T cells. Under these conditions, survival of tumor-inoculated mice was prolonged from 28 d to 87 days in the case of MEWO melanoma and from 23 to 48 days in the case of PancTU1 pancreatic carcinoma (to be submitted). Taken together, the results of several human-SCID mouse xenograft tumor models all support the hypothesis that human γδ T cells exert potent anti-tumor activity in vivo. Therefore, protocols need to be developed to explore a potential anti-tumor activity of human γδ T cells in clinical studies, based on today's knowledge of the mechanisms of γδ T-cell activation, expansion and ligand recognition.
In principle, 2 strategies can be envisaged to explore the potential anti-tumor activity of γδ T cells in the clinical setting: i) in vivo application of γδ-stimulating ligands (e.g., phosphoantigens), thereby inducing in vivo expansion and/or activation of γδ T cells; and ii) ex vivo isolation of γδ T cells, followed by in vitro large-scale expansion and adoptive transfer of γδ T cells into the patient. As to the former strategy, 2 approaches can be considered, i.e., the application of previously identified highly potent phosphoantigens or the application of already licensed drugs that are in clinical use for other indications. Synthetic phosphoantigens have been designed (such as bromohydrin pyrophosphate) which stimulate Vγ9Vδ2 T cells in vitro at pico- to nano-molar concentrations and thus are prime candidates for potential application in vivo.17. Such phosphoantigens not only activate γδ T cells17, 61 but also enhance their cytotoxic activity against tumor cells.62, 63 However, an interesting alternative results from the observation that aminobisphosphonates, which are therapeutically used to treat osteoporosis, also exert potent stimulatory activity on human Vγ9Vδ2 T cells in vitro and in vivo, even though at higher (i.e., micromolar) concentrations.64, 65 Like for microbial and synthetic phosphoantigens, the recognition of aminobisphosphonates by human Vγ9Vδ2 T cells is TCR-dependent, and various aminobisphosphonates differ in their γδ-stimulating activity.66 It appears, however, that there are differences in the requirement for antigen presentation between phosphoantigens and aminobisphosphonates. The “presentation” of phosphoantigens requires the continuous presence of the ligand; therefore, it is not sufficient to “pulse” antigen-presenting cells (APC) with the phosphoantigens. This is in contrast to the aminobisphosphonates where preincubation of APC with the aminobisphosphonates is sufficient to trigger γδ T-cell activation.67 In addition to a possible direct recognition by γδ T cells of aminobisphosphonates on the surface of APC, these substances most likely also exert indirect activation of γδ T cells. Aminobisphosphonates are potent inhibitors of farnesyl diphosphate synthase, an enzyme promoting the metabolism of the mevalonate pathway intermediate IPP.68, 69 The intracellular accumulation of IPP, a potent activator of Vγ9Vδ2 T cells,70 resulting from aminobisphosphonate-mediated inhibition of farnesyl diphosphate synthase, is likely to contribute to the selective activation of γδ T cells by these substances.56 In addition to their well-documented antiresorptive activity, which forms the rationale for their usage to treat osteoporosis and malignancies where patients have extensive bone resorption,71, 72 aminobisphosphoantes also exert direct antitumor effects in vitro through the inhibition of tumor cell adhesion and the induction of apoptosis.73 Therefore, it is a highly attractive proposal to explore the potential of aminobisphosphonates for the immunotherapy of various cancers on the grounds that these drugs might not only activate and expand tumor-reactive γδ T cells but also might have direct antitumor activity. The proliferative response of human γδ T cells to phosphoantigens (and aminobisphosphonates) is strictly dependent on the exogenous supply of cytokines such as interleukin-2 (IL-2), because γδ T cells themselves do not produce sufficient amounts.74 It is thus absolutely essential to combine aminobisphosphonates (and phosphoantigens) with the application of IL-2 in immunotherapeutic studies aimed at in vivo activation of γδ T cells. This, however, does not impose any difficulty since IL-2 is a licensed drug for the therapy of renal cell carcinoma.75 The first pilot study based on the aminobisphosphonate pamidronate in combination with low dose IL-2 has been performed in patients with low grade non-Hodgkin lymphoma or multiple myeloma.76 In a subgroup of 3 of 9 patients selected on the basis of a pre-existing in vitro responsiveness to pamidronate, a partial remission was observed.76 In another study where aminobisphosphonate zoledronic acid was given to cancer patients with bone metastases, Dieli and colleagues observed enhanced in vitro IFN-γ production in response to IPP 3 months after treatment, indicating the in vivo differentiation of Vγ9Vδ2 T cells towards IFN-γ producing cells.77 These promising results warrant further clinical studies to determine whether aminobisphosphonates in combination with IL-2 is an effective immunotherapy for certain types of cancer. Such studies will also reveal whether the application of aminobisphosphonates or phosphoantigens together with IL-2 can per se activate and mobilize sufficient numbers of tumor-reactive γδ T cells as to exert clinically measurable responses. An alternative to the in vivo activation of γδ T cells by aminobisphosphonates or phosphoantigens plus IL-2 would be an adoptive immune therapy based on the ex vivo isolation and large-scale in vitro expansion of γδ T cells. Again, phosphoantigens can be used to expand Vγ9Vδ2 T cells to very large cell numbers in vitro. This can be also achieved with alternative protocols based on the anti-CD3 or anti-TCR MAb-induced cellular activation. The activation-induced cell death occurring in these in vitro cultures might be prevented by the “protection” with agonistic anti-CD2 MAb and/or the addition of anti-apoptotic cytokines such as IL-15.78 Such protocols allow for the rapid generation of sufficiently high numbers of γδ T cells from cancer patients that retain their potent cytolytic activity against various tumor cells. An obvious disadvantage of any cell-based immunotherapy is the requirement for a GMP facility and the stringent requirements by regulatory authorities. Nevertheless, adoptive immunotherapy with γδ T cells will enter clinical trials in advanced cancers refractory to other therapies and with proven susceptibility to γδ T-cell-mediated cytotoxicity such as renal cell carcinoma.
Only recently, γδ T cells have attracted much interest in the field of tumor immunology due to recent insights into their role in tumor development and their potent antitumor activity. On-going and future studies will have to address clinically important issues such as the optimization of protocols based on aminobisphosphonates or phosphoantigens plus IL-2, the identification of tumor entities that are susceptible to this type of immunotherapy and the efficacy of an adoptive transfer of tumor-reactive γδ T cells. It is well possible that a combination of both approaches is required for optimal results. In any case, there is hope that the current excitement about the possible application of γδ T cells will turn into new and effective therapies for tumors that cannot be treated successfully with available protocols.