|Home | About | Journals | Submit | Contact Us | Français|
A growing body of research addresses how pathways dysregulated during tumorigenesis are linked to innate immune responses, which can contribute to immune surveillance of cancer. Innate components of the immune system localized in tissues are believed to eliminate early neoplastic cells, thus preventing or delaying the establishment of advanced tumours. This Review addresses our current understanding of the mechanisms that detect cellular stresses that are associated with tumorigenesis, and that culminate in the recognition and, in some cases, the elimination of the tumour cells by natural killer (NK) cells and other lymphocytes that express NK cell receptors.
Cancers arise in a multi-step process that involve the dysregulation of oncogenes, tumour suppressors and pro-apoptotic signals 1. Genetic and epigenetic alterations lead to the alterations in cellular growth cycles, differentiation programmes and cell death pathways1. Studies of colon cancer have served as a paradigm for considering the stepwise alterations that culminate in metastatic cancer 2 (Figure 1). Initial genetic changes result in hyperproliferation of initially normal cells, and subsequent oncogene mutations and epigenetic changes underlie the development of increasingly dysplastic pre-cancerous adenomas or other precancerous lesions2 . These initial changes activate key tumour suppressors, such as p53[G], that suppress the cell cycle and may induce cellular senescence or apoptosis2. Mutations of the corresponding tumour suppressor genes or other genes in the relevant pathways are correlated with the appearance of cancer. Other mutations accumulate that influence angiogenesis, migration and metastasis. Although the order of occurrence of these various mutations may vary in different types of cancer or even different instances of the same type of cancer, evidence suggests that homozygous mutations in p53 usually become predominant relatively late in tumorigenesis 2, highlighting the fact that early events in tumorigenesis activate p53 and create selective pressure for loss of this key tumour suppressor.
In addition to these largely cell-intrinsic barriers to tumorigenesis, evidence has accumulated for cell extrinsic barriers to tumour development, some mediated by the immune system. Some cancers are linked to infectious agents, and in some of these cases transformation depends on direct infection of the pre-malignant cell 3. Examples relevant to human disease include cervical carcinoma, some lymphomas and Kaposi’s sarcoma[G]. In these instances, the transformed cell may express non-self antigens encoded by the pathogen that can be targeted by T and B cells, just as may occur in any infection. Other cancers arise by spontaneous genetic and/or epigenetic changes. In these tumours, self antigens are sometimes overexpressed and can be targeted by the adaptive immune system due to their unnaturally high abundance, but in other cases adaptive immune responses are not readily detected and may not have a major role in tumour suppression 4. In those instances, tumour suppression may involve the innate immune system. In general, adaptive immune responses are initiated by signals that are associated with inflammation caused by innate immune responses4, and this is probably also true for responses to tumours as well. Therefore, the initial innate immune response to tumours may be decisive in determining whether immune surveillance is effective. It must be emphasized that many immune responses to tumours are not only non-protective, but paradoxically promote cancer. Important research in that area has been reviewed elsewhere and will not be addressed in detail here4, 5.
Natural killer (NK) cells are one important component of the innate immune response to tumours. NK cells are lymphocytes that differ from B and T cells in that they use numerous receptors, none of which are encoded by genes that undergo rearrangement. Each NK cell expresses several stimulatory and inhibitory NK cell receptors that can function with some independence, enabling NK cells to separately target cells that increase or decrease their expression of various ligands6. Many of the NK cell inhibitory receptors are specific for MHC class I molecules, which are expressed by normal cells but often lost from infected cells or tumour cells7. Ligands for NK cell stimulatory receptors are usually poorly expressed by healthy cells but are increased in expression on ‘unhealthy’ cells, such as transformed, infected or stressed cells 6, 7. NK cell activation is controlled by the balance of stimulatory and inhibitory signaling incurred when target cell engagement occurs and the various receptors engage their ligands 6–8. Hence, some normal cells display stimulatory ligands, but fail to be killed by NK cells because their MHC class I molecules engage inhibitory receptors that counteract stimulatory signaling. Increased expression of stimulatory ligands by a target cell can overcome inhibitory signaling in the NK cell, resulting in target cell lysis. Loss of inhibitory MHC ligands can also result in target cell lysis. Both occur in the context of disease, sometimes in the same cell8 . In addition to their role in NK cells, some stimulatory NK cell receptors are also commonly expressed by subsets of T cells, where they are believed to provide an innate signal that enhances T cell activation9.
Recent evidence suggests that the host immune response to tumour cells is in some cases linked to specific events that are associated with cellular transformation and tumorigenesis. In this article, we focus on the mechanisms that link oncogenic stress to innate immune stimuli, in particular the stimuli that act through NK cell receptors and the cells that express them.
Immune surveillance of cancer is a concept that was for some time considered discredited, but has received strong experimental support in the past ten years or so (Table 1). Before reviewing the evidence, it is useful to summarize the experimental systems employed in such studies.
Many past studies of the immune system’s role in cancer have relied on tumour transplant models, typically using tumour cell lines that are implanted subcutaneously in recipient animals. Although this approach is useful, it suffers from the fact that implanted tumours involve cell lines that clearly escaped immune surveillance in the animal from which they were derived and may also be aberrant due to prolonged culturing. Furthermore, they are usually implanted in ectopic sites in relatively large numbers, and fail to recapitulate the earliest stages of transformation and tumorigenesis. As a result, the physiological relevance of observations showing immune destruction of implanted tumours is often questioned. Alternative approaches include studies of fully spontaneous cancer, cancer induced by transgenic expression of oncogenes and carcinogen-induced cancer (Table I). These approaches are considered more reliable for probing the role of the immune system, because the tumours arise in the normal tissue site, typically from a single cell, and proceed through the various stages of tumour development. Nevertheless, they vary in their difficulty of use and other shortcomings as summarized below.
Cancer that develops in normal untreated mice is presumably the most faithful mouse model of human cancer, and has been studied in some analyses of immune surveillance. However, the very low incidence of fully spontaneous cancers in most mouse strains and the heterogeneity of cancers that arise are drawbacks for these studies. Despite the difficulties, a key report by Shankaran et al. showed increased incidence of spontaneous lung and intestinal carcinomas in 129/Sv mice that lacked all T and B cells due to mutations in recombination-activating gene 2 (Rag2)10[G]. Interestingly, mammary carcinomas were not detected in RAG2-deficient mice, but arose in mice lacking RAG2 as well as STAT1 (signal transducer and activator of transcription 1), a component of the interferon (IFN) receptor signalling complex, suggesting a role for innate immune responses in the control of mammary tumours. A general and robust role of perforin [G] in immune surveillance of cancer was demonstrated with perforin knockouts in two mouse strain backgrounds 11–14. TRAIL (TNF-related apoptosis-inducing ligand), another mediator of cytolysis has also been implicated in responses against cancer 11–13. These studies suggest a role for both the adaptive and innate immune responses in the control of fully spontaneous tumours.
Genetically engineered mice that overexpress oncogenes or have altered tumour suppressor function develop spontaneous cancers that mimic the development of natural malignancies in most respects. The tumours arise and progress autochtonously (at their natural sites), including the earliest stages, and are likely to reflect the natural interactions between the tumour and the immune system in initially normal tissues. In most of these models, most or all animals develop tumours of the same type, and in some cases tumours develop with predictable kinetics, allowing systematic analysis of the various stages of cancer. A drawback of some of these models is their high penetrance, such that tumors are repeatedly initiated in the same host and can overwhelm the immune response. A few relevant models will be summarized here.
In the TRAMP (transgenic adenocarcinoma of the mouse prostate) mouse model the rat probasin promoter directs the expression of the SV40 virus early genes[G] (large T and small t antigens) to the prostate epithelium of adult mice. Prostate tumours developing in these mice are infiltrated by leukocytes, including CD8+ T cells specific for tumour-associated antigens 15, 16. Other cell types that infiltrate the tumours include NK cells, γδ T cells[G] and NKT cells[G] (P. Savage, personal communication). Liu et al. recently demonstrated that γδ T cells could suppress high-grade prostate tumours in the TRAMP model17. Furthermore, another recent study demonstrated that the immunoreceptor natural killer group 2, member D (NKG2D) contributed to the control of the high-grade, aggressive form of carcinomas that develop in some of these mice 18. As discussed later, NKG2D is a stimulatory receptor expressed by NK cells and some T cells, suggesting that one or both of these cell types is involved in the control of aggressive prostate adenocarcinomas.
The Eμ-Myc transgene model consists of the Myc oncogene expressed under the control of the Eμ immunoglobulin heavy chain enhancer and the Myc promoter. Eμ-Myc transgenic mice that lack T and B cells due to a mutation in Rag1 succumb to preB lymphomas a few weeks earlier than their Rag1+ littermates, suggesting that mature T (or B) cells may limit the development of these tumours19. Furthermore, lymphomagenesis was also accelerated in Eμ-Myc mice that lacked the gene encoding NKG2D, suggesting a role for NKG2D either in promoting NK- or T cell-dependent elimination of lymphomas18. All Eμ-Myc transgenic mice eventually develop lymphomas, perhaps because the initiation of new tumours finally overwhelms immune system control 18, 20.
Mutations of the gene encoding the tumour suppressor p53 occur spontaneously in most naturally arising cancers. p53 deficiency removes a key barrier to tumorigenesis and contributes to genomic instability 21. Mice with homozygous germline mutations in p53 (p53−/− mice) develop mainly lymphomas of thymic origin, whereas p53+/− mice develop both disseminated lymphoma and non-lymphoid tumours, mainly sarcomas. Studies demonstrate that tumour incidence is higher in p53-deficient mice that are also deficient in IFNγ, the IFNγ receptor and/or STAT1 22. IFNγ-insensitive p53−/− mice developed a broader spectrum of tumours compared to mice lacking p53 alone 22. In addition, iNKT cells have been shown to contribute to the control of tumour development in p53+/− mice23 and both perforin- and TRAIL-mediated apoptotic pathways participate in this response. 12, 13.
Tumours arising as a result of chemical carcinogenesis arise autochtonously and proceed through the stages of tumorigenesis. Carcinogen-based models, although useful, have been criticized because the potent carcinogens may induce local tissue inflammation that has the potential to alter experimental outcomes 24. On the other hand, an advantage to these models compared to some transgenic models is that the carcinogen can be titrated so that tumor incidence is limiting.
Among the most commonly studied models of mouse tumorigenesis is the induction of fibrosarcomas in mice treated subcutaneously or intradermally with the carcinogen methyl-cholanthrene (MCA). Genetic studies showed an increased incidence of MCA-induced fibrosarcomas in mice with defects in perforin 25, 26, the IFN signalling pathway 10, 22, 27, or the TRAIL mediated cytotoxicity pathway 28, 29. Both αβ and γδ T cells were implicated in the surveillance of MCA-induced fibrosarcomas 30, 31. Among αβ T cells, NKT cells32 and possibly CD4+ T cells 10 were found to be most important. Additionally, NK cells play a critical role as direct effectors of tumor lysis and possibly as collaborators with iNKT cells 33, 34 . By contrast, CD8-deficient mice exhibited no defect in the control of MCA-induced tumours 25.
Another widely used carcinogen model is the induction of skin cancer involving the tumour initiator and promoter combination DMBA (dimethylbenz(a)anthracene) and TPA (12-O-tetradecanoyl phorbol 13-acetate). Mice deficient in skin-associated γδ T cells exhibited a higher incidence of DMBA/TPA-induced tumours 30. CD8+ T cells, by contrast, may promote tumorigenesis in this model under some conditions 35.
As reviewed elsewhere36, evidence supports a role for the adaptive immune system in immune surveillance, indicating a role for T or B cell receptors. Here we focus on receptors used by NK cells and in some cases shared by T cells that have defined antigen specificities and have been implicated in tumour surveillance. Among these receptors are NKG2D, NKp46, NKp30, NKp44, NKp80, 2B4 and DNAM1 (DNAX accessory molecule 1). Several of the receptors seem to mediate “induced self recognition”, meaning that the ligands are encoded by the host’s genome, are poorly expressed by normal cells, and are upregulated in stressed or diseased cells8. This phenomenon is best characterized in the case of the NKG2D receptor. For this reason, and because its role in tumour surveillance has been examined extensively, we focus our discussion on NKG2D and its ligands, after introducing some of the other key NK cell receptors.
These four stimulatory receptors are collectively called natural cytotoxicity receptors (NCR) 37–40 although NKG2D and other receptors also have important roles in the recognition and cytotoxicity of tumour cells by NK cells as discussed below. The four receptors have been well studied in human NK cells, but only one of them, NKp46, has been characterized in mice. In humans, NKp46, NKp30 and NKp80 are expressed by all NK cells, whereas NKp44 is only expressed by activated NK cells 41, 42. Blocking one or more of these receptors with antibodies often inhibits killing of tumour cell lines by human NK cells in vitro 37–39. Furthermore, mice with a targeted mutation in NKp46 were impaired in their ability to reject a transferred lymphoma cell line that expressed ligands for that receptor 43.
A surprising diversity of unrelated ligands has been reported for NCRs, including viral haemagglutinins (NKp46 and NKp44) 44, 45, heparan sulphate proteoglycans (NKp30 and NKp46) 46 and the nuclear factor HLA-B-associated transcript 3 (BAT3) (NKp30) 47 and the activation-induced C-type lectin (AICL) (NKp80) 48. Additional studies will be needed to determine the importance of these various ligands for NCRs.
2B4 is a member of the SLAM (signalling lymphocyte activation molecule)-related family of receptors, which is expressed by all NK cells, γδ T cells, a subset of CD8+ T cells and all human CD14+ monocytes 49. The unique ligand for 2B4 is CD48, also a SLAM-related receptor, which is expressed by all haematopoietic cells. 2B4 can reportedly function as either a stimulatory or inhibitory receptor, depending on the splice isoform that is expressed, the identity of the signalling adaptor molecule it associates with, and the extent of cross-linking of the receptor 50–52. 2B4 has a role in rejecting tumours that express its ligand, CD48 53.
DNAM1 is an adhesion molecule that is constitutively expressed by most NK cells, T cells, macrophages and dendritic cells 54, 55. Ligands identified for DNAM1 include nectin 2 (also known as CD112) protein and CD155 (also known as the poliovirus receptor, PVR) 56, 57. These ligands are often expressed by tumour cells and can activate or enhance tumour cell lysis in vitro 58, 59. Recent studies showed that DNAM1-deficient mice have reduced capacity to reject certain tumour cells in vivo and to limit formation of carcinogen-induced tumours in vivo 34, 60.
NKG2D is a lectin-like type II transmembrane homodimer that has received considerable attention in light of evidence for its role in immune responses in the context of cancer, infection and autoimmunity. NKG2D is expressed by virtually all NK cells and activated CD8+ T cells, and subsets of γδ T cells and NKT cells 61, 62. In certain conditions, NKG2D is also expressed by CD4+ T cells, at least in humans63–67.
A surprisingly large number of different NKG2D ligands have been identified, all of which are related self proteins that are similar to MHC class I molecules (Box 1). However, normal cells typically do not express the ligands well, whereas they are often specifically upregulated in cancerous or stressed tissues (Table 2).
NKG2D ligands are self proteins that are related to MHC class I molecules, although they differ from MHC class I molecules in that they do not present molecular cargo and fail to bind β2-microglobulin 134. The ligands include MHC class I polypeptide-related sequence A (MICA) and MICB in humans 62, which have no mouse homologues, and the ULBP (cytomegalovirus UL16-binding protein) or RAET1 (retinoic acid early transcript 1) ligand families that exist in both humans and mice135–138. Each human or mouse strain has the capacity to express approximately 5–10 different NKG2D ligands. Most normal cells, however, do not express significant levels of NKG2D ligands on the cell surface. By contrast, most tumour cell lines express one or more NKG2D ligand 39, 62, 135, 136. Furthermore, many primary human tumours express NKG2D ligands 117, 139, 140. Similarly, expression of the NKG2D ligands RAE1 or MULT1 (murine ULBP-like transcript 1) was observed on primary lymphomas generated in Eμ-Myc mice 18, 141 and on primary adenocarcinomas generated in TRAMP mice 18. Ligand expression is also induced in cells infected with certain pathogens 142.
Evidence has accumulated showing that NKG2D has an important role in immune surveillance of tumours (Figure 2). NKG2D-dependent elimination of tumour cells expressing NKG2D ligands has been well documented in vitro 39, 62, 68, 69 and in vivo in tumour transplant experiments 70, 71. In humans, specific NKG2D gene polymorphisms were associated with susceptibility to cancer72.
The most direct evidence supporting a role for NKG2D in tumour surveillance came from analysis of tumour incidence in gene-targeted mice that lack NKG2D, and also carrying transgenes that elevate the incidence of specific cancers 18. In TRAMP mice, the incidence of a highly aggressive form of prostate adenocarcinoma was markedly increased when the mice were also deficient for NKG2D18. Similarly, in mice carrying the Eμ-Myc transgene that causes B lymphoma, onset of lymphoma was accelerated by 7 weeks if the mice were also deficient for NKG2D18. It has not yet been established whether NKG2D-dependent control of tumours in these models is mediated by NK cells or one or more type of NKG2D-expressing T cell, or by both NK cells and T cells.
The action of NKG2D in tumour surveillance in the TRAMP model was also suggested by the finding that many of the aggressive adenocarcinomas that arose in NKG2D-deficient mice expressed one or more of the NKG2D ligands, whereas similar tumours that arose in NKG2D-proficient mice generally lacked expression of NKG2D ligands. A likely explanation is that aggressive prostate adenocarcinomas tumours that arose in NKG2D-proficient mice were subjected to NKG2D-mediated immune surveillance, resulting in selection for variant tumour cells that had lost expression of the ligands (Figure 2). In contrast to this pattern in TRAMP mice, the B lymphomas that arose in Eμ-Myc mice commonly expressed NKG2D ligands whether or not the mice were NKG2D deficient. These data suggested that evasion of NKG2D-mediated surveillance occurs in some cases despite continued expression of NKG2D ligands, consistent with the fact that many primary tumours in normal animals or humans express NKG2D ligands.
A role for NKG2D in surveillance of skin cancer was suggested by the increased expression of transcripts for NKG2D ligands in carcinogen-treated skin samples 30. Interestingly, transgenic mice that constitutively expressed high levels of NKG2D ligands, which results in dampened NKG2D function, exhibited an increased incidence of carcinogen-induced cutaneous malignancies 73. These data were not definitive in implicating NKG2D in cutaneous immune surveillance, however, because the NK cells in these transgenic mice exhibited defects in eliminating tumour cells that lacked NKG2D ligands, as well as those that expressed NKG2D ligands.
NKG2D deficiency did not result in the increased incidence or severity of some types of cancer, including a late arising, less malignant form of adenocarcinoma in TRAMP mice or fibrosarcomas induced by MCA18. The latter observation was surprising in light of the contrasting findings of an earlier study that used repeated injections of NKG2D-specific antibody to block the receptor74. It is possible that sustained engagement of NKG2D by antibodies impairs both NKG2D-dependent and -independent NK cell functions, as has been reported for sustained engagement of NKG2D by its ligands 73, 75–77. In any case, the observation that the incidence or severity of tumours was unaffected by NKG2D deficiency in some models suggests that these types of tumours readily evade NKG2D-dependent immune surveillance, fail to express NKG2D ligands at a sufficiently early stage, or are otherwise sequestered or insensitive to immune destruction. Possible mechanisms of evasion include upregulation of inhibitory MHC class I molecules to counteract the higher levels of stimulatory ligands, and/or loss of distinct stimulatory or adhesion ligands by the tumor cells. In contrast, the DNAM1 receptor has been implicated in immune surveillance of MCA-induced sarcoma suggesting its action is not readily evaded in this system34.
Considering the capacity for cancers to evolve in the host and the heterogeneity of oncogenic mechanisms that operate in different cancer models, it is not surprising that the NKG2D system, or any other system for that matter, is ineffective in impeding disease in certain cancer models. A detailed account of why NKG2D is ineffective in these specific models is not yet available, but it is notable that in some systems NKG2D ligands are expressed on the surface of tumor cells, but fail to promote tumor rejection, whereas in other cases NKG2D ligands are lost from the tumor cells. In the first case, it is likely that the response of NKG2D+ lymphocytes is suppressed or avoided, whereas in the latter case, tumor variants that lack expression of ligands may arise readily. Why some tumors may extinguish ligands more readily than other is unknown, but could be related to the fact that the milieu of activated oncogenes, mutated tumor suppressors and other mediators varies in tumors of different origins.
Tumorigenesis is a complex process involving the dysregulation of many cellular pathways, in many cases resulting from mutations in oncogenes and tumour suppressor genes. Recognition of cancer cells by innate immune cells must depend on their ability to distinguish these dysregulated cells from normal cells. Many of the distinctive features of cancer cells, such as proliferation, invasiveness and repression of cell death pathways, are features that are exhibited by normal cells in other contexts. Other features are more specific for diseased cells, although untransformed stressed cells may also be similarly affected. For a host defense system against cancer to be effective, the system in the cancer cell that alerts the immune system most probably involves signalling pathways that process several types of information that collectively, but not individually, identifies the cell as a cancer cell.
Features that distinguish cancer cells from most normal cells are numerous. To mention only a few, mutations that activate oncogenes probably occur very early in the development of most tumours and provide persistent proliferative signals. Oncogene-induced signals activate tumour suppressors by at least two mechanisms, one involving p19ARF (encoded by the CDKN2A gene) and the other involving the DNA damage response[G], which is activated after DNA damage is sensed by the ATM and ATR protein kinases (Box 2). Activated p19ARF and the DNA damage response can each independently induce p53 expression and other mediators that arrest the cell cycle and can induce cellular senescence (Box 3) or apoptosis when persistently activated 78, 79.
The DNA damage response protects the genome by facilitating repair of minor DNA damage in cells, and functions as a key barrier to tumorigenesis (Fig. 1). Rapid DNA replication in the context of oncogene activation is thought to result in the disruption of DNA replication forks (“replication stress”) and accompanying DNA breaks, both of which can trigger the DNA damage response by activating key sensors in the DNA damage response pathway, the protein kinases ataxia-telangiectasia mutated (ATM) and ATM and RAD3 related (ATR) 143. ATM and ATR initiate a cascade that ultimately induces cell cycle arrest and DNA repair functions. Depending on the cell type and other factors, prolonged or severe activation of the DNA damage response results in activation of apoptotic programmes or cellular senescence programmes. The DNA damage response therefore helps to preserve the integrity of the genome or eliminate severely damaged cells.
Cellular senescence, wherein tumour cells survive for a time in an irreversibly senescent state, is associated with a blockade in cellular proliferation as well as a specific programme of gene activation that results in secretion of several pro-inflammatory cytokines and chemokines. Interestingly, many of these modulators paradoxically promote tumour cell growth, and studies suggest that the senescent state can, in some cases, promote tumorigenesis 144. On the other hand, abundant evidence has accumulated that senescence is a barrier to tumorigenesis 78, 79, 145, by inhibiting tumour cell growth, inducing cell death and activating immune responses that help to eliminate cancer cells.
Mutations in tumour suppressor genes, or the pathways that activate tumour suppressors, enable continued cellular proliferation of nascent tumour cells, but dysregulate DNA replication and repair in a manner that ultimately results in instability of the genome and the accumulation of chromosomal abnormalities 80, 81.
Hence, at early stages of tumorigenesis, cells may have early warning signs such as activation of p19ARF, the DNA damage response and tumour suppressors, and activation of the gene programme that is associated with cellular senescence. At late stages the cells have other defects such as genomic instability. Furthermore, certain other stress pathways are commonly activated in cancer cells, including the heat shock response82 and the unfolded protein response[G]83. The roles of some of these pathways as warning signals that trigger anticancer immune responses is discussed below.
The role of the DNA damage response as a key barrier to tumorigenesis was highlighted by studies on human tissues showing that early pre-neoplastic lesions in the breasts, lungs, bladder and colon of patients show chronic activation of the DNA damage response, manifested by phosphorylation of ATM, CHK2 (checkpoint kinase 2 homologue) and histone γ-H2AX, another marker of DNA damage response activation 84, 85 (Figure 1). Other studies show that transgenic expression in mice of the proto-oncogene MYC leads to DNA damage, consequent activation of ATM and p53, and apoptosis of affected cells in vivo 86, 87. DNA damage induced by MYC may result from production of increased reactive oxygen species, independent of cell cycle entry, as well as from induction of rapid DNA replication 88. ATM activation under these conditions provides a barrier to tumorigenesis, because deficiency of Atm increases MYC-induced tumorigenesis in mice 86, 87. These data are consistent with the fact that mutations in ATM in humans, which cause the syndrome ataxia telangiectasia when present in the homozyogous state, increase the incidence of lymphoma, leukaemia and breast cancer 89.
A link between the DNA damage response and anti-tumour immune responses was initially established in studies of NKG2D ligands 90 (Table 2). Genotoxic stress that activates ATM or ATR can induce NKG2D ligand expression (RAE1, MULT1, H60a, MICA and ULBPs) on the surface of relatively normal cultured cells, including fibroblast cell cultures 8, 90. Induction of NKG2D ligand expression was prevented by inhibiting, or knocking down, the expression of ATR or ATM, depending on the nature of the genotoxic stress 8, 90. Most established tumour cell lines express NKG2D ligands constitutively, and knockdown studies established that constitutive ATM or ATR activation had an important role in maintaining constitutive ligand expression by those cells 8, 90. These studies indicated that the display of NKG2D ligands on tumour cells is mediated in part by an activated DNA damage response. This upregulation of NKG2D ligand expression is accompanied by increased levels of the corresponding mRNA transcripts, but it has not been determined whether this is due to increased transcription or alterations in mRNA processing.
The DNA damage response also has a role in inducing the expression of ligands for a distinct NK cell receptor, as shown by a study showing that CD155, a ligand for DNAM1, was induced by DNA-damaging drugs on the surface of multiple myeloma cells in an ATM/ATR-dependent manner 57. Finally, the DNA damage response directly regulates expression of the death receptor DR5, a ligand for TRAIL 91. Engagement of DR5 by TRAIL induces apoptosis, and evidence indicates that TRAIL is expressed by NK cells and T cells and functions as an important effector molecule in tumour surveillance by these cells 28, 92.
Cellular senescence has been linked to immune mediated tumor elimination mechanisms by a study of liver tumours generated from transformed cells that initially lacked p53 expression. When p53 expression was subsequently switched on, the tumours underwent growth arrest, exhibited features of cellular senescence and were gradually eliminated by NK cells and other infiltrating cells 78. These studies suggested that senescence associated with p53 activation in tumour cells is in some cases associated with the activation of immune responses that destroy cancer cells or their pre-malignant counterparts.
How the senescence programme is linked to anticancer immune responses has not been worked out in detail, but studies suggest that numerous ligands that stimulate immune responses are upregulated in senescent cells, including intercellular adhesion molecule 1 (ICAM1), a ligand for the adhesion receptor lymphocyte function-associated antigen 1 (LFA1) that has an important role in NK cell activation, NKG2D ligands such as MICA and ULBP2, and the DNAM1 ligand CD155 78, 93 (Table 2). As already noted above, the expression of some of these ligands is induced by the DNA damage response. Hence, it remains unclear whether the induction of these immune-stimulating ligands is related to the senescence programme itself, which normally takes several days to establish, or the DNA damage response, which may be involved in establishing the senescence programme. Notably, the induction of NKG2D ligand expression in cultured cells by DNA-damaging agents did not require p5390, whereas tumour senescence and NK-cell-dependent elimination of senescent tumour cells was induced by p53 reactivation in vivo 78. Therefore, it seems probable that both p53-dependent and p53-independent processes linked to tumorigenesis might regulate the sensitivity of tumour cells to NK cell-mediated elimination in vivo.
It has long been known that the heat shock response is activated in many forms of cancer 82. Mice deficient in heat shock transcription factor 1 (HSF1), which functions as a key inducer of heat shock response, were less susceptible than wild-type mice to cancer in experimental models, suggesting that the heat shock response is hijacked by tumours to enhance their survival94. Linking antitumour immune responses to the heat shock response may therefore function to target a response that is otherwise beneficial to tumours, thus limiting the damage.
Studies in cultured human epithelial cells implicated the heat shock response in transcriptional activation of the human MICA and MICB genes95 (Table 2). The promoter regions of both genes contain HSF1 binding elements, which were necessary for transcriptional activation 96. Distinct elements in the promoters were required to support transcription in virus-infected cells or proliferating cells, suggesting that there was some independence in the control of MICA and MICB transcription under different conditions.
So far, transcriptional activation has not been observed in the case of other NKG2D ligand families, including human ULBP and RAET1 proteins and mouse RAET1, MULT1 and H60 proteins. In the case of MULT1, however, the heat shock response nevertheless has an important role in regulating cell surface expression of the protein at a posttranscriptional step (Table 2). In fibroblast cells and other cells, the cytoplasmic tail of MULT1 is subject to ubiquitylation that targets the protein for destruction in lysosomes 97. Exposure of these cells to heat shock reversed the ubiquitin-dependent destruction of MULT1, resulting in a marked increase in levels of the protein at the cell surface. Interestingly, ultraviolet (UV) irradiation of fibroblasts had the same effect 97. Although UV irradiation induces the DNA damage response, the stabilization of MULT1 protein by UV irradiation was independent of the DNA damage response. UV irradiation alters several other pathways in cells, and it is possible that its effects on MULT1 stabilization are exerted through a pathway that overlaps with the heat shock response.
MULT1 expression provides an example of combinatorial regulation by distinct cancer-associated stress pathways, the DNA damage response 90, and the heat shock response 97. Coupling MULT1 expression to multiple stress pathways, in this case operating at different stages of biogenesis of the molecule, may represent a paradigm for mechanisms that restrict the display of immune-activating ligands to seriously diseased cells.
Evidence suggests additional modes of regulation of stimulatory ligands that activate NK cells. In F9 embryocarcinoma cells, the transcription of Raet1 genes was induced by retinoic acid98, which was the basis for the first identification of these genes. Retinoids exert growth suppressive effects on normal cells and tumour cells and have been considered promising agents for cancer therapy99. However, the generality of the role of retinoids in regulating NKG2D ligands remains unclear.
The transcription of Raet1e was inhibited by the transcription factor JUNB in cell lines and mouse tissues100. These findings are of particular interest in light of the evidence that JUN family transcription factors show complex regulation in conditions of stress and injury101.
NKG2D is expressed by all γδ T cells that reside in the mouse epidermis 68. These T cells participate in the destruction of cutaneous malignancies and wound healing102, 103. One NKG2D ligand, H60c, is expressed selectively in the epidermis 104, 105. Engagement by H60c displayed on keratinocytes, by NKG2D on epidermal γδ T cells, is essential for triggering γδ T cell activation and lysis of the keratinocytes, suggesting that NKG2D functions as a key co-stimulatory receptor in the activation of epidermal T cells 105. Interestingly, expression of H60c is upregulated in wounded skin, and cultured keratinocytes upregulate H60c but not other NKG2D ligands, suggesting that H60c may be the main NKG2D ligand involved in cutaneous immune surveillance105 (Table 2). The mechanisms that upregulate H60c in diseased and stressed tissues are not yet known. Although human and mouse cutaneous T cells differ in many respects, it is interesting that one of the human NKG2D ligands, ULBP4, is also selectively expressed in human skin106.
Regulation of NKG2D ligands by microRNAs was reported for MICA and MICB. In human cell lines, microRNAs that target the 3’ untranslated regions of the MICA and MICB transcripts inhibited steady state MICA and MICB expression 107. Stressing the cells amplified MICA and MICB expression to an extent that overcame the block imposed by the microRNAs. Whether these microRNAs are themselves regulated in normal tissues by stress remains to be established. However, it was of interest that certain tumours overexpressed the microRNAs, which may serve as a mechanism whereby tumours evade immunosurveillance 107.
An additional, important mechanism of regulation of NKG2D ligands is through shedding from the cell surface, which has been reported for MICA, MICB, ULBP2 and ULBP4 108–111 but has not been documented in the case of the mouse NKG2D ligands. Shedding of NKG2D ligands is thought to be mediated by ADAM (a disintegrin and metalloproteinase) family metalloproteinases 112–114, and is assisted by ERP5 (endoplasmic reticulum protein 5; also known as PDIA6) 115. The presence of cell free NKG2D ligands in the serum of patients with late stage cancer 109, 110 was correlated with reduced levels of NKG2D at the surface of NK cells and T cells and decreased function of these cells 108, 116, 117, and could therefore be a mechanism that enables tumours to escape immune surveillance 118. . Shedding of NKG2D ligands is also seen in cases of autoimmune disease where there is no “evasion” process66 suggesting that shedding may occur spontaneously in the case of cells that express ligands. However, it is also possible that the extent of shedding is increased as tumors progress, perhaps reflecting selection for immune-escape variants.
In the environment of a tumour, numerous events are thought to have an impact on the expression and function of immunoreceptors, including NKG2D and its ligands. In principle, these influences may increase or decrease NKG2D-dependent immune surveillance. For example, as noted earlier, inflammation associated with tumorigenesis can, in some cases, increase tumour growth and suppress protective antitumour immune responses 23, and it is possible that inhibition of NKG2D function is one example of how such inhibition occurs. It is also conceivable that in some tumour environments, NKG2D signalling induces a pro-tumour programme. Although the roles of these regulatory events in promoting or inhibiting antitumour immune responses remain uncertain, they must be accounted for when considering the role of NKG2D in tumour immunity.
Cytokines present in the tumour environment are one important potential determinant of NKG2D ligand expression. RAE1 was downregulated following exposure to transforming growth factor-β (TGFβ)114, which might therefore be one of the mechanisms of immunosuppression mediated by this cytokine. Surprisingly, IFNγ and IFNα downregulated H60a but not RAE1 or MULT1 expression by sarcoma cell lines 119. Furthermore, the expression of MICA and, in some cases, ULBP2 by melanoma and glioma cell lines was downregulated by IFNγ treatment 120, 121. These findings were surprising because IFNs generally promote antitumour immune responses.
The expression and function of NKG2D itself is also influenced by the tumour environment. Pro-inflammatory cytokines that are involved in proliferation and survival of NK cells and T cells such as IL-2 and IL-15 stimulate NKG2D expression and potentiate its cytotoxic and IFNγ secretory function 122, 123. Conversely, the immunosuppressive cytokine TGFβ secreted by several tumours and found in the serum of cancer patients124 can directly induce NKG2D downregulation when secreted in the tumour environment or as a membrane-bound cytokine on regulatory immune cells, or when present in tumour-derived exosomes125, 126. Downregulation of NKG2D expression also occurs in response to another cytokine that favours tumour growth, macrophage migration inhibitory factor (MIF)127.
In addition to cytokines, it is probable that NKG2D ligands expressed in the tumour environment can adversely affect NKG2D function. Sustained engagement of NKG2D by its ligands in vivo compromises the capacity of NK cells to attack tumours. Persistent engagement by RAE1 73, H60a 128 and MICA 75, or by soluble forms of MICA and MICB 108, is known to induce NKG2D internalization and subsequent degradation. In some cases, such persistent NKG2D engagement impairs NK cell functions more broadly, even inhibiting responses to NK-cell-sensitive target cells that lack NKG2D ligands73, 76. These findings raise the possibility that when tumour growth outpaces protective responses, NKG2D and other NK cell functions are ultimately inhibited as a result of persistent stimulation.
As a result of the signaling associated with tumorigenesis, many developing tumour cells display cell surface ligands that engage activating receptors expressed by NK cells and in some cases T cells. Some of the same signaling pathways that activate this extrinsic response are responsible for activating intrinsic tumor suppressor mechanisms such as p53-induced apoptosis and senescence. However, there are likely some differences in the specific mediators of the two types of responses, as suggested by the finding that induction of NKG2D ligands by the DNA damage response occurs in cells lacking p53 90.
As exemplified by studies of the NKG2D ligand MULT1, anti-tumor responses dependent on NK receptors are in some cases regulated by cooperation of distinct stress pathways that act at different levels of biogenesis of the immune-activating ligands. Combinatorial regulation by distinct stress pathways presumably helps to prevent inappropriate responses to normal cells. On the downside, however, these features may provide multiple targets for mutational inactivation of immune recognition of tumor cells.
An interesting question is whether the ligand-induction mechanisms linked to tumor suppressor pathways evolved specifically for anti-tumor responses, or play a more general role in disease responses, such as in responses to infections. In considering this question, it is often argued that because cancer is predominantly a disease of older individuals who are presumed to be post-reproductive, natural selection cannot directly select for anti-cancer mechanisms. It should be kept in mind, however, that this argument applies equally to intrinsic tumor suppressor mechanisms and immune-based tumor suppressor mechanisms. In light of this, some of the assumptions behind this argument may be questioned. The various barriers to tumorigenesis presumably work together to ensure that cancer is usually delayed to a later stage of life.
Immune-based anti-tumor mechanisms have the potential to complement, as opposed simply supplement, intrinsic barriers in at least two respects. First, the triggers of the immune-based mechanisms may in some cases be operative in cases where intrinsic tumor suppressors are lost from tumor cells, such as in the example above where the induction of NKG2D ligands by the DNA damage response occurs in cells that have lost the p53 tumor suppressor 90. Secondly, the innate immune system responses induced by these mechanisms may in some cases promote powerful adaptive immune responses that have the capacity for immune memory and sustained systemic protection.
Although NK cell receptor-dependent responses to tumours can be protective to the host, some tumours may avoid detection by down-regulating or shedding ligands, as suggested by studies of NKG2D ligands. The mechanisms whereby ligand expression is extinguished are incompletely understood and must be addressed in future studies. In some cases, it is probable that loss of ligand expression results from mutations in developing tumour cells that inactivate the same pathways that induce ligand expression in the first place, whereas in other cases the cells may acquire mutations in the ligand genes themselves. It is also likely that although expression of ligands for NK receptors is often protective against cancer, in some cases the tumor can evolve to exploit ligand expression to prevent a protective immune response. Desensitization of the NKG2D receptor as a result of shedding of NKG2D ligands is probably one example, but there are likely to be others.
A thorough understanding of the pathways involved in immune system activation by tumors may enable the design of therapeutic drugs. In cases where tumors acquire mutations that disable pathways that induce immune system ligands, a useful approach could be to design drugs that bypass the missing steps or re-induce ligands by another mechanism. It is possible, for example, that the efficacy of some chemotherapy drugs such as 5-FU and cisplatin may be related to increasing NKG2D ligand expression by tumour cells via activation of the DNA damage response 8, 129. In addition, proteasome inhibitors and/or histone deacetylase inhibitors represent promising candidate for enhancing expression of NKG2D ligands130, 131. Given the likelihood that ligand expression is induced by the synergistic action of several signals associated with the cancerous state, such interventions may be quite selective in targeting tumour cells as opposed to normal cells. Where inhibitory cytokines have a role in suppressing ligand expression on tumour cells, blockade of cytokine action may serve as a useful therapeutic intervention. As an alternative approach, although NKG2D ligands are absent from some tumour cells in vivo, vaccines comprised of cells expressing tumour antigens and NKG2D ligands may nevertheless be effective at inducing effective adaptive immune responses against tumours in some cases 70, 132.
On the other hand, many if not most advanced cancers continue to express ligands for NKG2D and other activating receptors on NK cells with impunity (Figure 2). Tumour escape in these instances may be due to impaired functioning of NK cell receptors due to inhibitory cytokines in the microenvironment, in which case cytokine blockade may be beneficial. Shedding of ligands, such as NKG2D ligands, are thought to repress protective immune responses, suggesting that antibodies that systemically block or remove the shed proteins may be therapeutic111. In other cases, persistent stimulation of the NK cells via NKG2D and possibly other activating receptors may result in anergy. A deeper understanding of how such persistent stimulation inactivates NK cell receptors and NK cell activity may provide approaches to reverse the unresponsive state of NK cells. Finally, in cases where NKG2D ligands continue to be expressed by tumour cells, they represent an inviting target for therapeutic antibodies designed to eliminate ligand-expressing tumour cells.