2. NKG2D-Mediated Immune Response
The NKG2D system operates primarily in NK, CD8
+ αβ T cells and γδ T cells. Although NK cell activation is controlled by the relative balance of inhibitory and activating signals, ligation of the NKG2D receptor by any NKG2DL is sufficient to trigger lytic synapse formation
[12][13] and degranulation
[14], overriding concomitant inhibitory cues
[7]. NKG2D-triggered NK cell activation is further enhanced when the LFA-1 and 2B4 receptors interact with their respective ligands present on target cells
[15]. Signalling via NKG2D can also enhance NK cell-mediated antibody-dependent cell-mediated cytotoxicity (ADCC)
[16], including that promoted by therapeutic monoclonal antibodies such as rituximab
[17]. In advanced cancer, circulating numbers of NKG2D-expressing NK cells are commonly reduced
[18][19], and the cells often have impaired cytotoxic activity
[18][20]. Conversely, the presence of tumour-infiltrating NK cells is associated with improved outcomes in several cancer types, including breast cancer
[21], gastric carcinoma
[22] and neuroblastoma
[23].
Unlike NK cells, CD8
+ T cells cannot be activated fully by NKG2D ligation alone. Instead, NKG2D functions as a co-stimulatory molecule in these cells, facilitating enhanced cytokine release upon T cell receptor stimulation
[24][25]. Although NKG2D is not normally expressed by CD4
+ T cells, NKG2D
+ CD4
+ T cells can accumulate in cases of chronic inflammation or cancer
[3]. In cancer, these cells demonstrate immunosuppressive properties mediated by soluble Fas ligands and other immunosuppressive cytokines
[3].
γδ T cells constitute around 5% of all T cells. The dominant subtype found in blood expresses a Vγ9Vδ2 T cell receptor, and these cells can receive both activating (e.g., cytotoxicity-promoting) and co-stimulatory signals via NKG2D
[26]. Tissue-resident γδ T cells of the δ1 subset also express NKG2D and undergo activation when exposed to NKG2DL
[27]. Although MICA is one of the human NKG2DL, it is also directly recognised by the T cell receptor found on some δ1 γδ T cells
[28][29]. Human γδ T cells of both δ1 and δ2 subtypes are commonly found in solid tumours, and their presence is generally associated with a more favourable prognosis
[30][31].
3. Expression of NKG2DL in Healthy Tissues
NKG2DL are typically present at low levels under homeostatic conditions, except in gastrointestinal and glandular epithelia where they are constitutively expressed
[32][33]. In this context, expression is predominantly intracellular
[34][35] and may change upon exposure to gut flora
[36]. ULBP5 (RAET1G) isoform 1 is also highly expressed intracellularly in the anterior pituitary gland
[37]. ULBP1 has been found in B cells and monocytes
[38], while MICA and ULBP3 are present in bone marrow stromal cells
[39]. The activation of T cells or the cytokine-mediated stimulation of monocytes and dendritic cells may also promote NKG2DL upregulation
[40].
The expression of NKG2DL is subject to multiple forms of control at the level of epigenetic regulation
[41], transcription, alternative mRNA splicing, post-transcriptional regulation (e.g., by microRNAs
[42]), regulation of subcellular location (e.g., cytoplasmic versus cell surface) and release of soluble forms, either by cleavage or in exosomes
[7][43]. These regulatory pathways are considered in greater detail below. Tight regulation of NKG2DL expression is believed to be necessary in order to prevent autoimmunity
[44]. Nonetheless, given this complexity, it is perhaps unsurprising that the expression of NKG2DL at the mRNA and protein levels does not always concur
[35].
4. Induction and Regulation of NKG2DL Expression in Cancer
NKG2DL can be upregulated in response to a range of factors operating during malignant transformation.
During the process of malignant transformation, DNA damage
[45], activation of heat shock proteins
[46] and oxidative stress
[47] all occur. Remarkably, all of these factors stimulate NKG2DL expression, emphasizing the strong link between cancer and the presence of NKG2DL. In a seminal study, Gasser et al. demonstrated that the activation of the DNA damage response (DDR) triggers the production of NKG2DL
[48]. The DDR is mediated by two key protein kinases—ataxia telangiectasia mutated (ATM) and ataxia telangiectasia RAD3-related (ATR)—and it is activated upon sensing double-stranded DNA breaks and stalled replication
[49]. Importantly, this pathway is further amplified by radiotherapy and by multiple cytotoxic chemotherapy agents, including 5 fluorouracil, cisplatin, gemcitabine, temozolomide and vincristine
[48][50][51][52][53][54][55]. The DDR also causes the activation of P53, which in turn stimulates the transcription of ULBP1 and ULBP2 but not MICA/B
[56]. Another cytosolic DNA-sensing pathway, the stimulator of interferon genes (STING) pathway, was also shown to upregulate Rae-1 expression in mice
[57]. Inhibition of the STING pathway decreased Rae-1 expression in lymphoma cells and reduced their sensitivity to NK-mediated lysis
[58].
Several additional factors of potential relevance to cancer also upregulate NKG2DL expression. Early studies showed that promoter heat shock elements regulate MICA and MICB expression
[32]; consequently, the abundance of these ligands is strongly enhanced in some settings by heat shock/cell stress
[28]. Moreover, the ubiquitin-dependent degradation of the murine NKG2DL Mult-1 is reduced in response to heat shock or ultraviolet radiation, providing a precedent for the post-translational regulation of NKG2DL expression
[59].
It is perhaps unsurprising that certain oncogenic pathways have also been implicated in the induction of NKG2DL expression. The activation of the BCR/ABL oncogenic pathway has been linked to increased NKG2DL expression in chronic myeloid leukaemia
[60], while c-myc overexpression has been implicated in NKG2DL upregulation in both lymphoma
[61] and AML
[62]. Mutant ras can also promote the upregulation of NKG2DL in a manner that at least partially depends on PI3K
[63].
Expression of MICA/B is also upregulated by oxidative stress
[64][65] in an Erk-dependent manner
[64], while p38 MAPK (mitogen-activated protein kinase) can also stimulate NKG2DL expression in some circumstances
[55]. The combination of oxidative stress and Akt activation has recently been implicated in the ability of an antifungal agent (ciclopirox olamine) to increase NKG2DL expression by leukaemic cells
[66].
Uncontrolled receptor signalling constitutes another cancer-associated process that increases NKG2DL expression. Illustrating this, the co-expression of the HER2/HER3 heterodimer resulted in the enhanced expression of MICA/B in breast cancer cell lines
[67]. In both cases, PI3K signalling was implicated. Moreover, heightened EGF receptor activity has also been linked to NKG2DL upregulation
[68].
A further broad stimulus to NKG2DL expression is cellular senescence
[69], which is considered to be an emerging hallmark of cancer
[70]. Adding complexity, senescent tumour cells may also increase NKG2DL shedding, favouring immune escape
[71].
A number of transcription factors have been implicated in the regulation of NKG2DL expression. In the mouse, E2F transcription factors which promote cell cycle progression can direct the transcriptional upregulation of Rae-1 family members
[72]. A similar process was inferred in the human system by virtue of reduced MICA/B and ULBP2 expression in serum-starved HCT116 cells
[72]. It is also notable that E2F is a direct phosphorylation target of the ATM and ATR kinases mentioned above
[40]. The KLF4 transcription factor also has been linked to the expression of MICA in acute myeloid leukaemia (AML)
[73]. ULBP1 transcription is triggered by the ATF4 transcription factor, which is induced in response to nutrient deprivation, the unfolded protein response and oxidative stress
[74]. However, MICA/B expression may be inhibited by the unfolded protein response under some circumstances, once again demonstrating the complex and context-dependent nature of NKG2DL regulation
[75].
Metabolic rewiring is another distinctive feature of cancer
[45]. Once again, NKG2DL expression is influenced by cancer-associated metabolic factors such as altered glycosylation
[76][77].
Chronic inflammation is a key underpinning factor in the progression of many human cancers
[78]. A number of inflammatory cytokines have been implicated in the control of NKG2DL expression. These include TNF-α and IL-18, both of which can upregulate ULBP2 levels in leukaemic cells
[79]. NKG2DL are also upregulated by Toll-like receptor stimulation
[80]. On the other hand, interferon (IFN)-γ has been shown to reduce NKG2DL on some tumour cell types, acting via STAT1
[81], microRNA induction
[82] and MMP9 cleavage
[83]. Similar inhibitory effects have been attributed to IFN-α
[84], although there are also reports of NKG2DL upregulation in response to this cytokine
[83]. Interleukin (IL)-6 and its downstream mediator, STAT3, have also been implicated in the downregulation of NKG2DL on tumour cells
[85][86][87]. The effects of IL-10 on NKG2DL expression are complex: it downregulates MICA and upregulates MICB expression on melanoma cells
[88] and increases NKG2DL levels on macrophages
[89]. Transforming growth factor (TGF)-β downregulates the expression of NKG2DL on some tumour cell types
[90][91]. Once again, however, this may not be a universal effect since the induction of tumour-associated epithelial-to-mesenchymal transition (EMT) by TGF-β may either upregulate
[92] or downregulate
[93] NKG2DL in a context-dependent manner.
Tumour cells can also influence NKG2DL expression on stromal cells. Illustrating this, the release of lactate dehydrogenase 5 by glioblastoma cells induces NKG2DL expression on monocytes which in turn causes NKG2D downregulation on NK cells
[94]. Moreover, tumour-associated immune infiltrates and fibrovascular structures are commonly positive for NKG2DL, particularly the membrane of endothelial cells
[95].
Despite the frequency with which NKG2DL are expressed in transformed cells, levels found in malignant stem cells may be reduced or absent
[96][97]. In the case of AML stem cells, this reduction could be overcome using PARP (poly-ADP-ribose-polymerase 1) inhibitors
[97]. Similarly, both ULBP1 and ULBP3 are repressed in glioma stem cells that contain mutations in isocitrate dehydrogenase (IDH) genes
[98]. Nonetheless, other studies have confirmed that NKG2DL remain expressed on cancer stem cells in some settings (including glioma stem cells
[99]) and contribute to their susceptibility to NK cell-mediated killing
[100][101][102].
Finally, it should also be noted that NKG2D itself is also subject to cytokine-mediated regulation with increased expression in response to IL-2, IL-7, IL-12, IL-15 and type 1 interferons
[103]. By contrast, the reduced expression of NKG2D has been linked to IL-21 and TGF-β exposure
[103].
5. Tumour Evasion of NKG2D-Mediated Immune Surveillance
To counteract the above, cancers have evolved various mechanisms to evade NKG2D-dependent immune surveillance. Epigenetic repression of NKG2DL expression is mediated by several pathways, including histone deacetylation, enhancer of zeste homolog 2 and DNA methylation
[104]. The cleavage of NKG2DL from the surfaces of tumour cells is also an important regulatory mechanism. Soluble NKG2DL are usually present at low levels in the circulation of healthy individuals. However, levels may be highly increased in cancer patients, reaching ng/mL concentrations on some occasions
[105]. The release of soluble tumour-associated NKG2DL provides a potent mechanism to downregulate NKG2D on intratumoural CD8
+ T cells and peripheral blood mononuclear cells (including NK cells)
[106][107][108]. Elevated serum levels of soluble NKG2DL have been linked to worsened patient outcomes in several cancer types
[109][110][111], although, in some cases, adverse prognosis is not directly linked to NKG2D downmodulation
[112]. Moreover, patients who develop autoantibodies against MICA following anti-CTLA4 immunotherapy benefitted from a reduction in soluble MICA, restoration of NK and CD8
+ T cell function and enhanced tumour lysis and dendritic cell cross-presentation
[113].
There are two major pathways by which soluble NKG2DL are generated in cancer. First, NKG2DL undergo cleavage by ADAMs (a disintegrin and metalloproteinases) 10 and 17 and MMPs (matrix metalloproteinases), enzymes that are commonly increased in cancer
[7][114][115]. NKG2DL may also be secreted within exosomes
[7][116] or extracellular vesicles (EVs) that also contain pro-apoptotic molecules such as the TRAIL (tumour necrosis factor-related apoptosis-inducing ligand) and the Fas ligand
[117]. The relative contribution of cleaved and vesicle-derived soluble NKG2DL remains poorly characterized
[118]. Membrane-spanning NKG2DL are primarily shed following proteolytic cleavage, while GPI-anchored NKG2DL mainly undergo release via EV. Nonetheless, both transmembrane- and GPI-anchored ligands may be found in EVs in various systems
[9].
Intracellular retention of NKG2DL is another potential mechanism used to evade immune surveillance. MICA may be retained within the endoplasmic reticulum in some tumour types, in a manner that could be reversed using the proteasome inhibitor bortezomib
[119]. Intracellular retention of NKG2DL may also be promoted by a number of CMV proteins
[120] and by CEACAM1
[121].
A third recently described mechanism by which tumour cells can reduce MICA and MICB expression involves neddylation
[122]. This entails the addition of a ubiquitin-like protein known as neuronal precursor cell-expressed developmentally downregulated protein (NEDD) 8, leading to protein degradation.
A further factor that can influence the outcome of the interaction between NKG2D and its ligands is trogocytosis, a process involving the acquisition of membrane and membrane proteins from other cells during cell-to-cell interactions. Both T cells
[123] and NK cells
[124] can trogocytose NKG2DL from other cell types, leading to varied outcomes including enhanced NK cell activation or failure of immune surveillance owing to the death of these cells via fratricide. More recently, the transfer of NKG2DL via EVs has been demonstrated in a multiple myeloma model. Once again, a double-edged outcome can be envisioned whereby cross-dressing of tumour cells may passively sensitise them to NKG2D-dependent elimination, but NKG2D downregulation and the sensitisation/fratricide of immune effector cells could thwart immune surveillance
[9].
6. Role of the NKG2D/NKG2DL Axis in Animal Models of Cancer
As indicated above, the NKG2D/NKG2DL system is believed to play a critical role in the elimination of premalignant cells before they progress into clinically detectable tumours
[125]. In agreement with this, when NKG2DL are expressed on a range of malignant cell types, they facilitate tumour rejection in vivo
[126][127].
The role of NKG2D in tumour immune surveillance is strongly supported by the fact that NKG2D-deficient mice are more susceptible to spontaneous tumour development in the TRAMP (transgenic adenocarcinoma of mouse prostate) and Eµ-myc lymphoma transgenic model systems
[128]. In the TRAMP model, but not in the Eµ-myc model, tumours arising in NKG2D-sufficient mice were depleted of NKG2DL when compared to those in NKG2D-deficient mice. This highlights a divergence in mechanisms by which NKG2D immune surveillance is bypassed in both models. Adding further complexity, although antibody-mediated NKG2D neutralisation promotes enhanced sarcoma formation in response to the chemical carcinogen 3-methylcholanthrene (3-MC)
[129], NKG2D deficiency did not phenocopy this effect. Indeed, there was a small trend in the opposite direction in this more slowly evolving tumour type
[128].
The maintained expression of NKG2DL is associated with a number of autoimmune and chronic inflammatory disease states, including rheumatoid arthritis, inflammatory bowel disease, type 1 diabetes, demyelinating conditions and coeliac disease
[130]. It is noteworthy in this respect that chronic inflammation is a pathological process that can also facilitate malignant transformation
[78]. Chronic activation of NKG2D can also accelerate tumourigenesis under some circumstances. Illustrating this, the sustained transgenic expression of an NKG2DL in vivo in a mouse model led to widespread NKG2D downregulation
[131], reduced NK cell cytotoxicity mediated via NKG2D
[132] and alternative receptor systems
[133], sustained NK cell IFN-γ production
[132] and enhanced susceptibility to chemically induced squamous cell carcinoma formation
[131]. In a similar vein, the onset of diethylnitrosamine-induced hepatocellular carcinoma (HCC) was delayed in NKG2D-deficient mice
[134].
7. Pharmacological Regulation of NKG2DL Expression
NKG2DL can be regulated by a wide spectrum of pharmacological agents. Upregulation of one or more ligands has been attributed to azacytidine, trichostatin A, vitamin D3, bryostatin, all-trans retinoic acid (ATRA), proteasome inhibitors, arsenic trioxide, multiple chemotherapy agents (see Section 4 above), decitabine, multitargeted tyrosine kinase inhibitors, inosine pranobex, nutlin-3a and histone deacetylase (HDAC) inhibitors such as sodium valproate and trichostatin A
[51][52][54][55][98][135][136][137][138][139][140][141][142][143][144]. Mechanistically, HDAC inhibition leads to increased MICA expression at least in part via the transcription factor KLF4
[73]. Clinically relevant NKG2DL upregulation on malignant cells has been demonstrated in patients with AML following treatment with ATRA or valproic acid-containing chemotherapy regimens
[145]. Moreover, in patients with pancreatic cancer who received neoadjuvant gemcitabine, MICA was expressed on 85% of tumours, in contrast to 36% of cases in the placebo-treated control group
[146]. Gemcitabine also has the additional ability to reduce levels of soluble ULBP2 released by pancreatic cancer cell lines
[147].
A further approach that may be used to increase the tumour cell surface expression of NKG2DL involves the inhibition of shedding of these ligands. Illustrating this, antibodies targeted against the α3 domain of MICA hindered the shedding of this ligand in addition to MICB
[148]. As a result, NK-mediated anti-tumour activity was boosted in a number of tumour model systems. Alternatively, degradation of MICA and MICB may be inhibited using pharmacological inhibitors of neddylation
[122]. Furthermore, as indicated above, the expression of NKG2DL on AML stem cells could be achieved using PARP inhibition
[97].
By contrast, NKG2DL may also be downregulated using pharmacological interventions. Proteasome upregulation has been linked with the downregulated expression of ULBP1
[149]. Estradiol has been reported to either suppress
[150] or stimulate NKG2DL expression
[151] accompanied by enhanced ADAM 17-mediated cleavage
[152]. Activation of the unfolded protein response in hepatocellular carcinoma cells also reduced the expression of MICA/B in a manner that was partially alleviated using proteasome inhibition
[153]. Downregulation of NKG2DL in breast cancer cell lines has also been attributed to the anaesthetic agent sevoflurane
[154]. Inhibition of BRAF with vemurafenib led to reduced MICA and ULBP2 expression by melanoma cells
[155]. Rapamycin has also been linked to NKG2DL downregulation in AML
[156]. Finally, the commonly used uricosuric agent allopurinol inhibited the upregulated expression of NKG2DL induced by genotoxic stress in a manner that was dependent on the inhibition of xanthine oxidoreductase
[157]. In keeping with this, uric acid generated as a consequence of DNA damage and purine catabolism promoted MICA/B expression
[158].