Regulation of ligands for the activating receptor NKG2D

Abstract

The outcome of an encounter between a cytotoxic cell and a potential target cell depends on the balance of signals from inhibitory and activating receptors. Natural Killer group 2D (NKG2D) has recently emerged as a major activating receptor on T lymphocytes and natural killer cells. In both humans and mice, multiple different genes encode ligands for NKG2D, and these ligands are non-classical major histocompatibility complex class I molecules. The NKG2D–ligand interaction triggers an activating signal in the cell expressing NKG2D and this promotes cytotoxic lysis of the cell expressing the ligand. Most normal tissues do not express ligands for NKG2D, but ligand expression has been documented in tumour and virus-infected cells, leading to lysis of these cells. Tight regulation of ligand expression is important. If there is inappropriate expression in normal tissues, this will favour autoimmune processes, whilst failure to up-regulate the ligands in pathological conditions would favour cancer development or dissemination of intracellular infection.

Introduction

Cytotoxic cells express a range of activating receptors including Natural Killer group 2D (NKG2D), 2B4, NKp80 and the natural cytotoxicity receptors (NCRs) NKp30, NKp44 and NKp46.¹ The NKG2D receptor is a key activating receptor on natural killer (NK) and T cells. NK cell stimulation through activating receptors is opposed by signalling through inhibitory receptors such as the killer immunoglobulin-like receptor (KIR) molecules, NKR-P1 and CD94/NKG2(A/B), which are also expressed on cytotoxic NK cells and recognize ‘self’ major histocompatibility complex (MHC) class I molecules expressed on healthy cells. The inhibitory CD94/NKG2A and CD94/NKG2B receptors recognize the non-classical MHC class I molecule human leucocyte antigen (HLA)-E, which can be thought of as a ‘marker’ for intracellular health.^2–4 This interaction prevents the activation of the effector cell and the subsequent lysis of the cell expressing HLA-E. Loss of expression of MHC class I molecules in pathological conditions abolishes the inhibitory signal to cytotoxic effector cells. According to the ‘missing self’ hypothesis, failure to express these indicators of health allows the inhibitory signals to cytotoxic cells to be overridden by triggering of activating receptors, which leads to target cell lysis by effector cells.^5,6 However, tumour and virus-infected cells not only down-regulate or lose the expression of MHC class I molecules, but can also induce the expression of other ‘self’ molecules that are markers of cellular stress. A set of these molecules is recognized by the NKG2D activating receptor, which has been considered a dominant activating receptor, such that triggering of NKG2D can induce cytotoxic lysis of the stressed target cells expressing ligands for NKG2D, despite the expression of normal levels of MHC class I molecules.^7,8 The current consensus is that NK cell activation and cytotoxicity result from the integration of signals produced by ‘missing self’ and ‘induced self’ interactions and a shifting of the balance in the signalling strength between the inhibitory and activating receptors (reviewed by Malarkannan⁹).

NKG2D is a C-type lectin-like molecule. The archetypal C-type lectins bind carbohydrates in a Ca²⁺-dependent manner through a common structural motif, but NKG2D has no features to suggest that it binds carbohydrates. It is a homodimeric type II transmembrane glycoprotein encoded on human chromosome 12, within the NK gene complex, and in a syntenic position on mouse chromosome 6.¹⁰ Human NKG2D is expressed on NK cells, CD8⁺ T-cell receptor (TCR)-αβ T cells and TCR-γδ T cells. Murine NKG2D is also expressed on NK cells, but is only present on activated and memory CD8⁺ TCR-αβ T cells and only 25% of splenic TCR-γδ T cells. Neither mouse nor human NKG2D is normally expressed on CD4⁺ TCR-αβ T cells. Murine NKG2D has also been detected on NKT cells, some interferon-producing killer dendritic cells, and at the mRNA level in macrophages (reviewed by Coudert and Held¹⁰).

Signalling through NKG2D

Certain activating NK cell receptors, such as the NCR, associate with a transmembrane signalling adaptor molecule, which possesses an immunoreceptor tyrosine-based activation motif (ITAM). These adaptors include CD3ζ, FcεRIγ and DAP12.¹ There are two isoforms of NKG2D in the mouse, which differ in the length of their intracellular cytoplasmic tail. The shorter form is known to associate with ITAM-bearing DAP12, which transduces the NKG2D activating signal via recruitment of the ZAP-70 and Syk tyrosine kinases. However, the cytoplasmic tail of the longer form appears to prevent interaction with this adaptor.^10,11 Both forms of murine NKG2D can also associate with another adaptor molecule, DAP10, which lacks an ITAM but instead bears a YxxM motif for Src homology 2 (SH2) domain binding.¹² NKG2D engagement results in the phosphorylation of DAP10 by Src family kinases and recruitment of the p85 subunit of phosphatidylinositol 3-kinase and Grb2, and ultimately leads to calcium flux and cytotoxicity.¹¹ The short form of murine NKG2D is only expressed in activated NK cells, whilst the long form is constitutively expressed. Human NKG2D has no short form and only associates with DAP10. In both humans and mice, NKG2D signalling through DAP10 is sufficient for cell-mediated cytotoxicity (reviewed by Upshaw and Leibson¹¹).

Role of NKG2D in primary activation and costimulation

On activated NK cells, NKG2D serves as a primary activation receptor, such that NKG2D engagement alone is sufficient to trigger NK cell-mediated cytotoxicity. In contrast, NKG2D appears to act as a costimulatory receptor in CD8⁺ T cells, requiring TCR-mediated signalling for full activation of these effector cells (reviewed¹⁰). There is controversy over whether NKG2D is capable of mediating activation in the absence of other stimuli in T cells^13–16. The actual function of NKG2D may be determined by additional factors, such as the activation status of the effector cell or the cellular environment e.g. effector cell priming by exposure to cytokines (reviewed by Coudert and Held¹⁰).

Overview of NKG2D ligands

A distinctive feature of the NKG2D system is that there are multiple ligands for this receptor, all of which are distantly related homologues of MHC class I proteins. The first ligands identified were the MHC class I chain-related proteins (MIC) A and MICB, which are encoded within the human MHC (6p21.3).¹⁷ Their sequence homology to MHC class I proteins is only approximately 25%, and although they have α1, α2 and α3 domains, they do not associate with β2-microglobulin or peptide.^17,18 There are no known orthologues of the MIC molecules in the mouse MHC. However, MIC molecules are conserved in non-human primates, cattle and pigs.^17,19,20 A second family of NKG2D ligands, the UL16 binding protein (ULBP) family, is encoded outside the MHC (6q24.2–25·3).²¹ The ULBPs are also known as the retinoic acid early transcript 1 (RAET1) family, as they show sequence homology to the mouse retinoic acid early 1 (Rae1) proteins (Rae1α-ε). In total there are 10 members of the human RAET1 gene family, of which only five are expressed and encode functional proteins, which are able to bind NKG2D.^21,22 The ULBPs were named for their ability to bind the human cytomegalovirus UL16 protein²³ although it is now known that only ULBP1, ULBP2 and RAET1G bind UL16.^22,24–26 MICB has also been shown to associate with UL16.²⁷ Like the MIC proteins, RAET1E and RAET1G are type I transmembrane proteins,²² whilst ULBP1–3 are glycosylphosphatidylinositol (GPI)-linked proteins.²¹ The ULBP/RAET1 family lack an α3 domain and contain only α1 and α2 domains. Mapping close to the murine Raet1 genes on chromosome 10 are H60 and murine ULBP-like transcript 1 (MULT1), ligands for murine NKG2D for which there are no human equivalents.²⁰ The various human and mouse ligands are presented in Fig. 1, and the evolutionary relationships between the ligands are illustrated in Fig. 2.

Figure 1.

Multiple ligands for the human and mouse NKG2D receptor. Like major histocompatibility complex (MHC) class I molecules, MHC class I chain related proteins (MIC) A and MICB have α1, α2 and α3 extracellular domains and transmembrane domains. The remaining human and mouse ligands lack the α3 extracellular domain. Within the UL16 binding protein (ULBP) family, retinoic acid early transcript 1E (RAET1E) and RAET1G also have transmembrane domains, whilst ULBP1–3 are glycosylphosphatidylinositol (GPI)-linked proteins. Murine retinoic acid early 1 (Rae1) proteins are also GPI-linked to the membrane, whilst H60 and murine ULBP-like transcript 1 (MULT1) harbour transmembrane domains. The equilibrium dissociation constants (K_D) for the NKG2D–ligand interactions, as determined by surface plasmon resonance studies, are also presented (ND, not determined).

Figure 2.

Phylogenetic tree illustrating the evolutionary relationships between the different human (dark grey) and mouse (light grey) NKG2D ligands, generated using Clustalw (http://www.ebi.ac.uk/clustalw/). Accession numbers: MHC class I chain related proteins (MIC) A, NP_000238; MICB, NP_005922; UL16 binding protein 1 (ULBP1), NP_079494; ULBP2, NP_079493; ULBP3, NP_078794; retinoic acid early transcript 1E (RAET1E), NP_631904; RAET1G, NP_001001788; retinoic acid early 1α (Rae1α), NP_033042; Rae1β, NP_033043; Rae1γ, NP_033044; Rae1δ, NP_064414; Rae1ε, NP_937836; H60, NP_034530; murine ULBP-like transcript 1 (MULT1), NP_084251.

Structural and binding studies of NKG2D interactions

A key question is whether the different ligands are equivalent in their capacities to trigger NKG2D signalling. Structural and binding studies have been carried out for some of the ligand interactions.^28–32 The different ligands bind to relatively equal surface areas on the same region of NKG2D, but significant differences in affinities exist and may have functional implications.³⁰ The affinities of binding range from around 6 nm to 1 µm for the interactions that have been investigated to date (Fig. 1), and structural and binding studies suggest that the ligands can compete with each other for receptor binding.³⁰

Polymorphism of MIC genes

The MIC molecules are encoded by the most highly polymorphic human genes after the standard HLA molecules. At least 60 and 25 alleles have been identified for MICA and MICB, respectively,^20,33 and the various alleles have different affinities for NKG2D as measured by the interaction of soluble recombinant NKG2D with the surface of cells transfected with different MICA alleles.³⁴ Unlike the classical MHC class I molecules in which polymorphisms cluster in the exons encoding the peptide-binding α1 and α2 domains, polymorphisms in MICA and MICB are evenly distributed throughout the entire genes.²⁰ The functional significance of MIC polymorphisms is not yet well understood, although there is some evidence for selective pressure generating this diversity (reviewed by Stephens¹⁷). There are documented disease associations with certain alleles of MICA, including genetic linkage to Crohn's disease, type I diabetes mellitus, rheumatoid arthritis, systemic lupus erythematosus, psoriasis and Behçet's disease (reviewed by Bahram et al.²⁰), although some of these associations could arise from linkage disequilibrium with the HLA genes or with other neighbouring genes.

Role of NKG2D in the pathogenesis of tumours and other diseases

NKG2D signalling is of particular importance as deregulation of this system could have detrimental consequences. Failure to stimulate effector cytotoxic cells through this receptor in pathological conditions could lead to tumorigenesis or dissemination of intracellular infection. Conversely, inappropriate signalling through NKG2D in healthy cells could result in autoimmune disease (Fig. 3). Most of the data regarding association of the NKG2D system with autoimmune disease relate to the MIC molecules.³⁵ Aberrant expression of NKG2D ligands, which inappropriately activate NKG2D, has been linked to rheumatoid arthritis, coeliac disease and type I diabetes mellitus.³⁵

Figure 3.

Stringent regulation of NKG2D ligands is important. Failure to up-regulate the ligands in pathological situations, resulting in too little NKG2D signalling (left), can lead to the development of cancer, or the spread of infection. Conversely, inappropriate expression of the ligands, and hence too much NKG2D signalling in healthy cells (right), can lead to autoimmune disease.

NKG2D plays a critical role in the elimination of tumour cells by cytotoxic effector cells.^{7,8,23,36–41}In vitro studies have demonstrated that the expression of an NKG2D ligand is sufficient to trigger cytolysis by an effector cell expressing NKG2D.^{15,40,42–45} Furthermore, the formation of tumours could be prevented through NKG2D signalling.^46,47 Therefore, NKG2D is important both in tumour immunosurveillance to prevent tumour initiation and also in immune-mediated rejection of tumour cells to prevent tumour progression.

The significance of NKG2D signalling in protecting against infection and tumorigenesis is highlighted by the development of mechanisms by viruses and tumour cells to evade detection by this system. NKG2D is down-regulated on effector cells following chronic exposure to its ligands on tumour cells, which results in defective NK cell-mediated cytotoxicity.^46,48,49 The cell surface expression of ligands for NKG2D is also down-regulated by proteolytic shedding mediated by metalloproteinases that are secreted by tumour cells.^50–53 This results in the release of soluble forms of the ectodomains of these ligands, which have been detected in the sera of patients with a number of different cancers.^51,54 Not only does this mechanism down-regulate ligand expression on target tumour cells, but soluble MICA released by tumour cells also causes the internalization and lysosomal degradation of NKG2D, and hence a reduction in NKG2D levels on NK cells and CD8⁺ T cells.^51,55 Furthermore, transforming growth factor (TGF)-β produced by tumour cells and human and mouse CD4⁺ CD25⁺ regulatory T cells down-regulates both NKG2D and its ligands on cytotoxic effector and tumour cells, respectively.¹⁰ Viruses have also developed mechanisms to avoid alerting the immune system through NKG2D. The human cytomegalovirus glycoprotein UL16 binds ULBP1, ULBP2 and MICB, and retains these molecules intracellularly, thereby reducing cell surface expression levels and susceptibility to NK cell cytotoxicity.^24,25,27,56 Furthermore, murine cytomegalovirus proteins can also down-regulate the mouse NKG2D ligands.²⁰

Expression of NKG2D ligands in normal healthy cells

Most normal healthy tissues do not express ligands for NKG2D,^57–59 but there are exceptions to this general rule. There is developmental regulation of the ligands during embryonic development. Murine Rae1β and Rae1γ are expressed at the transcript level in the early embryo, particularly in the brain.^57,59 There is also evidence for developmental regulation during normal haematopoiesis. Ligand expression (mainly ULBP1) is detected on the surfaces of B cells, platelets, monocytes and granulocytes in peripheral blood from healthy donors, but is absent on erythrocytes, monocyte-derived dendritic cells, T cells and NK cells.⁶⁰ Bone marrow (BM)-derived CD34⁺ progenitor cells are negative for ULBP1–3 expression, but ligand expression (in particular ULBP1) is induced during normal myeloid differentiation, characterized by loss of the CD34 early haematopoietic marker and expression of the myeloid markers CD33 and CD14.⁶⁰ ULBP1 expression can also be induced by in vitro culture of CD34⁺ cells with myeloid growth and differentiation factors.⁶⁰ Furthermore, retinoic acid-induced differentiation of embryocarcinoma cells stimulated expression of Rae1,⁵⁷ consistent with the idea that NKG2D ligands are developmentally regulated. High levels of MICA, ULBP1 and ULBP3, and lower levels of ULBP2, are also detected on BM stromal cells,⁶¹ and NKG2D ligands, in particular ULBP3, have been shown to be expressed on human mesenchymal stem cells.⁶² Furthermore, murine Rae1 and H60 are expressed on BM cells (mainly myeloid progenitors) from BALB/c mice after repopulation of an irradiated host.⁶³ Interestingly, expression of these ligands appears to be strain-dependent, as C57BL/6 BM cells express little NKG2D ligand. Indeed, both single positive and double positive thymocytes from BALB/c, but not from C57BL/6, mice express ligands for NKG2D.^43,64

Both MICA and MICB are expressed on normal intestinal epithelial cells.⁵⁸ This is believed to result from stimulation by bacterial flora in the gut. At the mRNA level, there is also some evidence that the ULBPs are expressed in other healthy tissues.^23,65 Similarly, MULT1 mRNA is detected in a number of normal tissues.^29,66 However, cell surface expression levels of the ligands were not investigated in these studies, and there have been some reports that the expression of ULBP mRNA does not always correlate with cell surface expression levels in peripheral blood⁶⁰ and in tumour cell lines,²³ which suggests regulation of ULBP expression at a level other than transcription.

Up-regulation of NKG2D ligands in activated immune cells

NKG2D ligand expression can also be induced when immune cells are activated, such as when B cells are treated with concanavalin A (ConA) or lipopolysaccharide (LPS)⁴³ and when T cells are stimulated with a variety of stimuli such as phytohemagglutinin (PHA), ConA, or anti-CD3 or anti-CD28 monoclonal antibody plus phorbol 12-myristate 13-acetate.^43,67–70 MICA and MICB are induced on dendritic cells after interferon (IFN)-α stimulation⁷¹ and ULBP1 is up-regulated on the surface of normal monocytes and leukaemic blasts in response to growth factors plus IFN-γ treatment.⁶⁰ High doses of LPS induce the transcription and cell surface expression of ULBP1-3 and the cell surface expression of constitutively transcribed MICA on human macrophages, which triggers NKG2D-mediated NK cell cytotoxicity.⁷²

Expression of NKG2D ligands on tumour cells

NKG2D ligands are constitutively expressed on a wide variety of tumour cells and tumour cells lines.^36,50,51 Transformation alone is not sufficient for induction of ligand expression, as expression of a number of oncogenes (K-ras, c-myc, Akt, E6, E7, E1A or Ras V12) in murine ovarian cells did not up-regulate NKG2D ligands.⁷³ However, when ovarian cell lines transformed with K-ras and c-myc or Akt and c-myc were implanted into the ovaries of nude mice, this resulted in the formation of tumours, and high levels of NKG2D ligand were detected in cell lines established from these tumours.⁷³ This suggests that ligand expression is acquired during some stage of the process of tumorigenesis. However, there are some conflicting reports which demonstrate that E1A expression is sufficient to up-regulate mouse and human NKG2D ligands in primary mouse cells, and mouse and human tumour cell lines,⁷⁴ and that pharmacological and siRNA-mediated silencing of BCR/ABL activity or oncogene expression could down-regulate MICA and MICB cell surface expression in the K562 chronic myelogenous leukaemia cell line.⁷⁵ Loss of tumour suppressor function has also been shown to up-regulate NKG2D ligands. Rae1ε cell surface expression is enhanced on JunB-deficient cells.⁷⁶ However, absence of the tumour suppressor p53 does not induce expression of any of the murine ligands.⁷³ The constitutive expression of NKG2D ligands on some tumour cells may result from chronic activation of the ataxia telangiectasia mutated (ATM) kinase, which plays a key role in the DNA damage response pathway.⁷³

It is important to note that each of the NKG2D ligands has diverse expression patterns on tumour cells.^23,36,77 This may reflect functional differences among the various ligands. The MIC ligands are frequently found on epithelial tumours derived from various tissues, but not commonly on haematological maligancies. In general, the ULBP family displays a converse pattern of expression; they are infrequently found on epithelial tumours, but are commonly found on T-cell leukaemia cell lines and have been detected on leukaemic cells from patients (reviewed by Coudert and Held¹⁰).

Up-regulation of NKG2D ligands on infected cells

NKG2D ligand expression is readily detected on cells infected with bacteria^78,79 or viruses.^24,56 MICA expression was increased in epithelial cell lines following infection by adherent Escherichia coli,⁷⁹ in dendritic and epithelial cells after Mycobacterium tuberculosis infection,⁸⁰ and in cytomegalovirus-infected fibroblasts and endothelial cells.⁸¹ Human macrophages infected with influenza virus or Sendai virus exhibited augmented levels of MICB.⁸²

Stimuli known to up-regulate NKG2D ligand expression

Heat shock treatment up-regulates MICA on epithelial cells.⁵⁸ Indeed, the MICA and MICB promoters contain heat shock transcriptional elements similar to those found in the promoters of heat shock proteins, such as heat shock protein 70 (HSP70).⁵⁸ Oxidative stress has also been shown to increase MICA and MICB transcript levels in a colon carcinoma cell line.⁸³ Recently, a wide variety of stimuli causing genotoxic stress or resulting in DNA replication arrest, including treatments with ionizing radiation, cisplatin, aphidicolin, mitomycin C or hydroxyurea, were shown to significantly up-regulate NKG2D ligands in transformed murine ovarian epithelial cells and normal adult fibroblasts, and in human secondary foreskin fibroblasts.⁷³ The DNA damage response pathway, initiated by the ataxia telangiectasia mutated (ATM) or ataxia telangiectasia and Rad3 related (ATR) kinases, was implicated in the regulation of NKG2D ligands in response to these insults. Inhibition of ATM, of ATR and of a downstream mediator of these pathways, Checkpoint kinase 1 (Chk1), suppressed the observed up-regulation of the ligands for NKG2D in response to the different treatments,⁷³ demonstrating that these pathways play an important role in ligand regulation. Interestingly, p53, which is also a downstream mediator of the ATM/ATR pathways,^84,85 was not necessary for the induction of ligand expression in response to the various treatments.⁷³ Treatments that modify chromatin, resulting in a more open structure, have also been shown to up-regulate NKG2D ligand expression.^73,86

Molecular mechanisms triggering NKG2D ligand expression

Relatively little is known about the molecular mechanisms that trigger NKG2D ligand gene expression.^{58,68,73,77,87–89} The DNA damage response pathway is frequently activated in cells infected with viruses and in tumour cells.^73,90,91 Furthermore, the induction of MICA expression following T-cell activation was shown to involve the phosphorylation of ATM.⁷⁰ Therefore, it is possible that this pathway could play a common essential role in NKG2D ligand induction in tumorigenesis, infection and T-cell activation, as well as in other situations of cellular stress caused by DNA damage and DNA replication inhibition. Heat shock transcriptional elements^58,87,88 and nuclear factor (NF)-κB signalling^68,70 have also been implicated in the regulation of MICA. Figure 4 illustrates the variety of stimuli that can induce the expression of NKG2D ligands, and some of the regulatory pathways that have been documented to date.

Figure 4.

A variety of stimuli that can regulate NKG2D ligands (NKG2DL) have been identified. However, little is known about the molecular mechanisms underlying the regulation of these ligands. There is evidence of transcriptional regulation, but the ligands may also be regulated at a level other than transcription. The solid arrows demonstrate the major regulatory pathways that have been identified. Indirect or less well-established pathways are shown as dotted arrows (ATM, ataxia telangiectasia mutated; ATR, ataxia telangiectasia and Rad3 related; Chk1, Checkpoint kinase 1; ConA, concanavalin A; IFN, interferon; IR, ionizing radiation; LPS, lipopolysaccharide; mTOR, mammalian target of rapamycin; NF, nuclear factor; PHA, phytohemagglutinin; PMA, phorbol 12-myristate 13-acetate).

Concluding remarks and future perspectives

Some conventional therapies against cancer may already target NKG2D signalling through up-regulation of the ligands for this receptor, in addition to directly having cytotoxic and other effects.⁷³ It is clear that tight regulation of NKG2D ligands is essential to ensure effective tumour immunosurveillance and rejection, and elimination of pathogen-infected cells, whilst also preventing inappropriate killing of healthy cells, which could lead to autoimmune disease. Perhaps this can explain the evolution of the multitude of NKG2D ligands, which are distinct in terms of their tissue expression patterns, levels and kinetics of expression, and affinities for their common receptor, suggesting that they are not simply redundant in function. It is entirely conceivable that the expression of the different ligands is governed by the type and degree of cellular or environmental stress, which can ultimately direct an appropriate and adequate response through NKG2D signalling. Furthermore, this diversity may be a consequence of selective pressure exerted by viruses and tumours, which have used a range of mechanisms to evade detection by NKG2D. Future investigations directed at dissecting the mechanisms controlling the regulation of NKG2D ligands should offer a better understanding of the role of the NKG2D system in immunity and how deregulation of this system results in disease, and also the potential for using NKG2D ligands as targets for therapy.