Natural Killer Cells in Liver Disease

Abstract Natural killer (NK) cells comprise one of the most abundant immune cell populations in human liver and the nature and functions of these cells have been a focus of recent interest. Here, we consider the possible roles of NK cells in diverse liver diseases, concentrating on data from patient studies. NK cells can be protective, killing virally infected and cancerous cells in the liver and limiting fibrosis by eliminating hepatic stellate cells. However, they can also be deleterious, contributing to pathology in viral hepatitis by killing hepatocytes and downregulating virus-specific T-cell responses. It has recently emerged that a large fraction of hepatic NK cells constitute a distinct liver-resident subset and we highlight the need to distinguish between circulating and liver-resident NK cells in future studies. There is also a need for further investigation into how NK cells are influenced by the liver microenvironment and what scope there is to harness their immunotherapeutic potential.


Natural Killer Cells
Natural killer (NK) cells are immune cells of the lymphoid lineage that recognize and kill virally infected and malignant cells. NK cells also modulate the immune response by their production of cytokines, particularly interferon gamma (IFNγ), and by their interactions with other immune cells, notably dendritic cells and T cells. Unlike T and B cells, NK cells do not undergo antigen receptor rearrangement, but instead recognize their targets using a range of germ-line encoded receptors. Therefore, in contrast to T and B cells, NK cells make a rapid and robust response without the requirement for prior sensitization. [1][2][3] Natural killer cells were first described as recognizing targets that lack major histocompatibility complex (MHC) class I: the "missing self" response. They do this by expressing a variety of inhibitory receptors with a broad specificity for MHC class I, notably the inhibitory killer immunoglobulin-like receptors (KIRs) and CD94-NKG2A. When these are not engaged, the inhibitory signal is lifted leading to target cell killing. However, some cell types constitutively express low levels of MHC class I, and some virally infected cells evade recognition by NK cells by encoding molecules that mimic MHC class I. These phenomena limit the utility of target cell recognition based entirely on "missing self." More recently, it has become appreciated that NK cells also express several activating receptors whose engagement promotes killing. One such receptor is NKG2D, whose ligands, MICA, MICB, and the ULPBs, are upregulated on stressed and transformed cells, providing another way for NK cells to recognize their targets. NK cells also express a group of activating receptors known as "natural cytotoxicity receptors" (NCRs), which include NKp30, NKp44, and NKp46. The ligands of these receptors are not yet well-defined, but they are known to recognize virally infected and cancerous cells. Binding to antibody-coated cells via the Fc receptor CD16 allows a subset of NK cells to mediate antibody-dependent cell-mediated cytotoxicity (ADCC). The balance between ligation of activating and inhibitory receptors determines the outcome of the interaction between an NK cell and a potential target, although the precise threshold for NK cell activation can be modulated by the cytokine milieu. Selected NK cell receptors and their ligands in humans are summarized in ►Table 1.
Having recognized a target, an NK cell may kill it either by releasing cytotoxic granules or by ligating target cell death receptors. Cytotoxic granules contain perforin and granzymes. Perforin forms a pore in the target cell membrane allowing granzymes to enter the cytosol and induce apoptosis. Therefore, the ability of NK cells to degranulate and their expression of perforin and granzymes are often used as proxy measures for their cytotoxicity. NK cells may also promote target cell death via tumor necrosis factor (TNF) receptor superfamily molecules expressed by the target cell. Some NK cells express TNF superfamily molecules, such as TRAIL or Fas ligand, which bind to these receptors, triggering apoptosis in the target cell.
The functions of circulating NK cells then are well-established, and the mechanisms by which they achieve target cell recognition, killing and cytokine production are well-understood. However, many nonlymphoid organs are enriched in NK cells compared with the blood. The NK populations in these organs differ phenotypically from blood NK cells; in some cases, notably in the uterus, these organ-specific NK cells are thought to have specialist physiological functions in their home tissue. 4,5 The liver contains a large number of NK cells, and these have recently been the focus of intense research interest.

NK Cells in the Liver
Natural killer cells have long been recognized to be extremely abundant in human liver, ordinarily representing between 30% and 50% of liver lymphocytes. [6][7][8] Their abundance has fueled interest in their potential role in liver disease. The specialized hepatic circulation and surrounding parenchymal and nonparenchymal cells are important local influences on the features and functions of hepatic NK cells. The extensive narrow-lumen sinusoidal vasculature is lined by a fenestrated endothelium, allowing direct contact of intravascular lymphocytes with hepatocytes and stellate cells within the space of Disse. 9 Liver NK cells were originally described as "pit cells" that were noted to be in close contact with sinusoidal endothelial cells and Kupffer cells; 10 the latter have subsequently been shown to shape NK cells by differential cytokine production following Toll-like receptor signaling. 11 Human and mouse livers both contain large NK cell populations with an "immature" phenotype. 7,8,12 In mice, these immature NK cells are now known to represent a separate, liver-resident population that is identified by its expression of CD49a and is less cytotoxic, but has a greater capacity for TNFα and granulocyte-macrophage colony stimulating factor production than circulating NK cells. 13,14 In the light of these findings, recent attempts have been made to define an equivalent NK cell population in human liver. Some human livers contain a small CD49a þ NK cell population, 15 but these differ from CD49a þ NK cells in mouse liver in several respects and may not be equivalent. 16 More promisingly, all human livers seem to contain a large NK cell population that is phenotypically distinct from NK cells found in the blood (►Table 2). [17][18][19][20] Experiments performed in a transplant setting have shown that these cells are unable to exit the liver and are long-lived; they are therefore considered liver-resident. 19 The liver-resident NK cells appear less mature than circulating NK cells, on the basis of their expression of the maturation marker CD56, the terminal differentiation marker CD57, and by their expression of various NK cell receptors. [17][18][19][20] They express the C-type lectin CD69, which is associated with Table 1 Selected natural killer cell receptors and their ligands tissue-residence, 5 as well as the chemokine receptors CCR5 and CXCR6, which can mediate liver homing and residency by recognition of their ligands CCL3, CCL5, and CXCL16, which are present in hepatic sinusoids. [17][18][19] Liver-resident NK cells express lower levels of perforin and granzyme B than their circulating counterparts, are less cytotoxic, and less able to produce proinflammatory cytokines. [17][18][19] This is perhaps not unexpected because it is in line with the immunotolerant nature of the liver, but it does suggest that liver-resident NK cells do not have the same function as circulating NK cells. The large number and tissue residence of these cells perhaps point to a tissue-specific function, although it is not yet clear what this might be. There is some evidence from mice that liver-resident NK cells are able to mediate immune memory responses to haptens and viral vaccinations. 21,22 This suggests the intriguing possibility that liver-resident NK cells in humans may be "memory" cells, a function that would be consistent with their longevity. 19 Regardless of their function, the discovery of these liverresident NK cells has several implications for the interpretation of older studies into human NK cells in liver disease. First, it should be noted that they express CD69. In the past, this was considered a marker of immune cell activation, but is now known to also be associated with tissue-resident cells. Therefore, studies which report an increase in the proportion of CD69 þ intrahepatic NK cells could be interpreted either as demonstrating activation of circulating NK cells, an expansion of the resident compartment, or even a combination of the two. Furthermore, the existence of liver-resident cells means that the behavior of NK cells isolated from blood will not necessarily represent the responses of all the NK cells present in the liver. Indeed, circulating and liver-resident NK cells may later be shown to have different, even opposing, roles in liver disease.

Natural Killer Cells in Liver Fibrosis
The liver is exceptional in its regenerative capacity, but repeated destruction and regeneration results in fibrosis, which can progress to cirrhosis. Fibrosis occurs largely as a result of extracellular matrix production by activated hepatic stellate cells. It is a process common to all chronic inflammatory diseases of the liver and, as a result, any role for NK cells in fibrosis is likely to be relevant to a broad range of liver diseases.
Several studies in mice have shown that chemicallyinduced liver fibrosis is exacerbated in the absence of NK cells, suggesting that NK cells control fibrosis. This is a result of the greater susceptibility of activated hepatic stellate cells to NK cell killing, an effect that may be mediated by their reduced expression of MHC class I, 23 increased expression of ligands for NKG2D, 24,25 NKp46 26 or receptors for TRAIL, 27 or some combination of these.
Consistent with the findings in mice, NK cells in humans also have a role in controlling liver fibrosis. Human peripheral blood NK cells are able to kill hepatic stellate cells in culture and this is dependent on NKp46, 26 NKG2A, 28 NKG2D, TRAIL and FasL. 29 NK cells that are cytotoxic, 30 and specifically more effective at killing hepatic stellate cells, 29 are associated with a lower degree of fibrosis in chronic hepatitis C. Furthermore, increased NK cell infiltration into the liver is associated with reduced fibrosis in primary sclerosing cholangitis. 31

Natural Killer Cells in Viral Hepatitis
One of the major causes of liver disease is viral hepatitis brought about by the unrelated hepatitis viruses A, B, C, and E. Each of these is considered below, although the majority of NK cell studies have focused on Hepatitis B and C. 32

Hepatitis B
In acute hepatitis B virus (HBV) infection, there is an early expansion of NK cells 33,34 but their antiviral potential is temporarily impaired, 35 coinciding with peak viremia and a surge in interleukin-10 (IL-10), which has suppressive effects on NK cells. Only when viral load, and therefore IL-10 levels, decrease, do NK cells upregulate TRAIL and regain the ability to produce pro-inflammatory cytokines.
Studies of NK cell frequency and function in chronic hepatitis B (CHB) have been complicated by the heterogeneity of patients cohorts, comprised of varying proportions in the different clinical phases. Even the nomenclature and diagnostic criteria of these four phases (currently usually known as immunetolerant, immune-active, immune-control and reactivation), based on markers such as eAg, viral load, and alanine aminotransferase (ALT), are controversial. [36][37][38] NK cell frequency in the periphery and the liver is highest in immunotolerant patients, while immune-active patients have a reduction in NK cells. 39,40 There is some disagreement on the precise way in which activating and inhibitory NK cell receptors are altered in the immunotolerant phase, but the consensus seems to be that there is a general shift toward inhibition. In immune-active patients, the consensus points to peripheral and intrahepatic NK cells expressing more activating markers such as NCRs, TRAIL, CD69, and HLA-DR, and fewer inhibitory receptors, [41][42][43][44] although one study reported increased expression of the inhibitory receptor NKG2A by peripheral NK cells in immune-active patients. 45 Consistent with a generalized activation, NK cells from immune-active patients have an increased ability to degranulate and produce cytokines compared with immune-tolerant patients and healthy controls. These changes might be attributable to increased proinflammatory cytokine production in the liver of immune-active patients. 40 A broad consensus from all studies is that NK cells in CHB have a functional dichotomy, with selective impairment in cytokine production, but relative preservation of cytotoxicity. [42][43][44]46 IFNγ is key to the noncytolytic immune control of HBV replication without excessive liver damage, so this reduction in IFNγ production is particularly significant. 47 The selective defects in NK cell function could be recapitulated by IL-10 treatment in vitro, while blockade of IL-10 and transforming growth factor beta (TGFβ) led to restoration of IFNγ production in both circulating and intrahepatic NK cells from patients with CHB. A role for these immunosuppressive cytokines in shaping the NK cell functional profile is in line with their preponderance in the liver and increased expression in CHB. 46 Additional mechanisms proposed to contribute to the impaired NK cell effector function in CHB are their reduced levels of activating receptors 48 and impaired activation by plasmacytoid dendritic cells. 49 Patients in the eAg-negative reactivation phase experience temporary increases in viremia and liver damage, known as "hepatic flares," associated with increased serum interferon-α (IFNα) and IL-8 levels, together with enhanced NK cell activation and TRAIL expression. This cytokine combination promotes TRAIL-mediated killing by upregulating TRAIL on NK cells, and altering the balance of death-inducing versus decoy TRAIL receptors on hepatocytes. Human NK cells are able to kill both hepatoma cells and primary human hepatocytes in a partially TRAIL-dependent manner. These findings provided a paradigm in which TRAIL þ NK cells, which are predominantly liver-resident, 18 eliminate virally infected hepatocytes and propagate immune-mediated liver damage. 50 Although NK cells can play a protective role by eliminating virally infected hepatocytes or activated hepatic stellate cells, their potential to regulate immunopathology is complicated by their interaction with antiviral T cells (►Fig. 1). 51 NK cells selectively kill HBV-specific CD8 þ T cells rather than T cells of other specificities; this susceptibility is partly determined by upregulation of the death-inducing receptor TRAIL-R2 by CD8 þ T cells during HBV-specific responses. Expression of TRAIL-R2 is greatly enriched on intrahepatic CD8 þ T cells, to a degree determined by viral load. The in vivo relevance of this is underscored by the finding that antiviral T-cell responses from HBV-infected livers can be rescued following overnight TRAIL blockade. 52 This finding was supported by several studies in murine models showing a vital rheostat function for NK cells in regulating T-cell responses in persistent viral infections. 53,54 Another example of NK cell/T-cell crosstalk in viral hepatitis is the finding that MICA is expressed on HBVspecific and intrahepatic CD4 þ T cells. This first demonstration of uninfected human T cells expressing an NKG2D ligand was supported by in vitro and ex vivo data suggesting that this could drive NK cell cytotoxicity. 55 Further results pointing to a role for NK cells in negatively regulating HBV-specific T cells came from a study of patients treated with nucleoside/nucleotide analogs (NUCs). The NUC treatment reduces NK cell activation and TRAIL expression; 46 reversion to a quiescent NK cell phenotype is associated with restoration of HBVspecific CD4 þ T cells. 56 Interferon-α is another mainstay of treatment for CHB, providing an opportunity to study the in vivo impact of this cytokine on NK cells. Administration of therapeutic pegylated-IFNα to cohorts with either eAg-positive or eAg-negative CHB led to an increase in serum levels of the pro-proliferative cytokine IL-15, accompanied by a cumulative expansion of CD56 bright NK cells expressing TRAIL, NKp46, and HLA-DR. 57,58 Whereas NUCs were unable to reverse the defective NK cell IFNγ production characteristic of CHB, 44,46,56 this could be restored by pegylated IFNα. 57,58 Surprisingly, pegylated-IFNα was also able to induce sustained changes in NK cell responsiveness to sequential treatment with NUCs. 58 These prolonged in vivo changes in NK cells are in line with emerging data on adaptive features of NK cells. 3

Hepatitis C
An early indication of the importance of NK cells in the outcome of hepatitis C virus (HCV) infection came from genetic studies, revealing that patients expressing the inhibitory receptor KIR2DL3 together with its ligand HLA-C1 had an increased chance of spontaneous or treatment-induced HCV clearance. 59,60 Similarly, HCV-exposed uninfected individuals were more likely to have KIR2DL3-expressing NK cells with enhanced activation and effector function. [61][62][63] Several studies have reported associations between NK cell receptor expression and function with outcomes in HCV infection, including disease activity as measured by ALT and aspartate aminotransferase, 43,64 viral load 43,65-67 and outcome of treatment. 66,68 Broadly speaking, higher NK cell cytotoxicity and expression of activating receptors are associated with viral clearance and good responses to treatment, whereas higher expression of inhibitory receptors is linked to poor responses. Therefore, activated NK cells seem to be beneficial in the control of HCV.
In chronic HCV, the total frequency of NK cells and the proportion of the more cytotoxic CD56 dim subset of cells within the NK cell compartment are reduced. 30,69,70 There is no consensus on the expression of activating and inhibitory receptors by peripheral NK cells from chronic HCV patients, although the slim majority of studies report an overall increase in activating receptors. 41,43,67,69,70 This lack of consensus could potentially be attributed to differences in patient cohorts. Hepatic NK cells in chronic HCV also seem to be broadly activated, with one study reporting that intrahepatic NK cells express increased levels of activating receptors compared with peripheral NK cells 64 whereas another found increased expression of activating receptors by NK cells from HCV-infected liver, compared with control liver. 71 Natural killer cells in chronic HCV infection display reduced cytokine production, but sustained cytotoxicity, similar to the HBV-specific CD8 þ T cells upregulate TRAILR2, becoming susceptible to NK cell killing via the TRAIL pathway. Meanwhile HBV-specific CD4 þ T cells upregulate the NKG2D ligand MICA, which activates NK cells leading to degranulation and cytotoxicity. situation in CHB. Cytotoxic function of NK cells has been reported either to be increased 43 or unaltered in HCV patients compared with healthy controls. 30,70 The HCV-infected liver is enriched in NK cells expressing high levels of NKp46, which are highly cytotoxic, produce IFNγ, and block HCV replication more efficiently than NKp46 lo cells. The frequency of NKp46 hi NK cells in the liver correlates inversely with viral load and fibrosis, 72 in line with NK cells having an antifibrotic 29 and antiviral activity. 65 On the other hand, NK cells from HCV-infected patients have an impaired ability to produce IFNγ. 43,64,73 Interferon-α is a major regulator of NK cell activity during HCV infection, and a key determinant of outcome. In vitro, treatment with IFNα leads to a shift toward activation with increased TRAIL expression on both peripheral and intrahepatic NK cells 64,65,68 and increased degranulation in peripheral NK cells. 64 IFNα stimulation is necessary for NK cells to effectively recognize HCV-replicating hepatoma cells and IFNαstimulated NK cells challenged with these targets degranulate, produce cytokines, and mediate cytotoxicity via TRAIL. 65,74 Until recently, the administration of pegylated IFNα was the standard treatment for HCV. The examination of NK cell responses in patients undergoing therapy sheds light on the importance of IFNα in NK cell responses to HCV in vivo, with TRAIL emerging as a critical effector of virus control. NK cells in acute HCV patients have elevated TRAIL expression, which increases during IFNα therapy. Patients who cleared chronic HCV infection during IFNα treatment showed a greater upregulation of TRAIL compared with patients with ongoing chronic HCV infection. Furthermore, in patients with chronic HCV infection, TRAIL upregulation during the first 48 hours of IFNα treatment correlated with decreases in viral load. 65 Others studies have shown an upregulation of CD69, NKp30, NKp46, and NKG2A and downregulation of CD16, perforin and NKG2D during IFNα treatment. 66 There is also a decreased frequency of NK cells expressing CXCR3 and CCR7, suggesting recruitment to the liver. 68 More recently, studies of blood and liver samples from HCV patients treated with the new IFN-free regimes (direct-acting antivirals) have demonstrated normalization of the disease-associated NK cell phenotype and function.

Hepatitis A and E
The role of NK cells in hepatitis A virus (HAV) has not been investigated in detail. One study suggested NK cells have the potential to control HAV infection, although this was performed in HAV-infected fibroblasts rather than hepatocytes. 75,76 There is controversy over changes in NK cell frequency during hepatitis A, 77,78 but elevated NK cell NKG2A expression correlates with ALT, suggesting a possible role for NK cells in acute liver damage. 78 In patients with acute hepatitis E virus (HEV) infection, NK cells produce proinflammatory cytokines in response to stimulation faster than in patients with resolved infection. 79 NCR, NKG2D, and CD69 expression increase during acute and resolved HEV infection compared with healthy controls. 79 These findings suggest that NK cells are activated in HEV infection. However, NK cell degranulation is impaired, 79 and NK cell numbers decrease and remain low in resolved patients. 80

Natural Killer Cells in Alcoholic and Nonalcoholic Fatty Liver Disease
Alcoholic liver disease (ALD) and nonalcoholic fatty liver disease (NAFLD) are histologically similar disorders, both of which encompass a spectrum from simple steatosis to steatohepatitis to cirrhosis and finally hepatocellular carcinoma (HCC). Hepatocyte ballooning and immune cell infiltration are features of both ALD and NAFLD and the two disorders are likely to have several immunological characteristics in common, including the role of NK cells.

Alcoholic Liver Disease
The role of NK cells in ALD has not yet been extensively investigated. In mice, chronic ethanol consumption reduces NK cell number and function by inhibiting their maturation. This in turn accelerates the development of liver fibrosis in the presence of other injurious stimuli. 81,82 In human peripheral blood, a similar reduction in NK cell frequency and cytotoxic function is observed in ALD patients, compared with controls. 83

Nonalcoholic Fatty Liver Disease
The role of NK cells in NAFLD has been considered in more depth. To the extent that NAFLD is regarded as the hepatic manifestation of the metabolic syndrome, NK cells in adipose tissue are likely to be implicated in NAFLD through their involvement in altering glucose homeostasis. [84][85][86] Similar to ALD patients, NK cell frequency and cytotoxic function is decreased in the peripheral blood of obese patients, compared with lean controls. 87 Nonalcoholic steatohepatitis (NASH) patients have increased NK cell numbers in their liver, compared with those with simple steatosis or healthy controls, suggesting that the decrease in peripheral blood NK cell frequency may be a result of their recruitment to the liver. 88 Notably, this study identified NK cells using CD57, demonstrating that the increase is specifically in the CD57 þ liver-infiltrating, and not the CD57liver-resident population, and supporting the idea that these NK cells have been recruited from the blood. The study also reported increased expression of NKG2D ligands and apoptotic hepatocytes in NASH patients, suggesting increased NKG2Ddependent activation of these NK cells. 88 An increase in NK cells in the liver during NASH was also observed in a mouse model of the disease, and again this was attributable to an increase in liver-infiltrating, rather than liver-resident NK cells. 89 Intriguingly, in this study depletion of NK cells aggravated liver disease by allowing macrophages to become polarized to an M2 phenotype. Therefore, although several associations between NK cells and NAFLD have been reported, it is not yet clear if NK cells are helpful or deleterious in the disease.

Natural Killer Cells in Autoimmune Diseases of the Liver
Natural killer cells are implicated to some extent in the pathology of all three main autoimmune diseases of the liver: autoimmune hepatitis (AIH), primary biliary cirrhosis (PBC, formerly primary biliary cholangitis), and primary sclerosing cholangitis (PSC).

Autoimmune Hepatitis
Autoimmune hepatitis is characterized by hepatocyte necrosis associated with T-cell and plasma cell infiltrates. NK cells have been observed to interact with hepatocytes in AIH. 90 Different human leukocyte antigen (HLA) alleles have been associated with AIH in different ethnic groups, 91,92 and recently in a Sardinian cohort the activating receptor KIR2DS1 has been shown to be a risk factor for AIH, while the presence of the inhibitory receptor-ligand pairs KIR2DL3/ HLA-C1 and KIR3DL1/HLA-Bw4 were protective, 93 implicating NK cell activation in disease.

Primary Biliary Cirrhosis
In PBC, there is destruction of the small intrahepatic bile ducts, with contributions from both adaptive and innate cells. NK cells are more numerous in both liver and blood in PBC, with increased perforin expression and cytotoxicity. 94 This is likely to reflect recruitment of peripheral blood NK cells to the liver via CXCR3 and CX3CR1. 95,96 In vitro, NK cells can be induced to kill autologous biliary epithelial cells via TRAIL, by cross-talk with TLR-3 and -4 stimulated macrophages. 97 Further, IFNγ produced by NK cells after coculture with biliary epithelial cells can induce cytotoxicity in autoreactive CD4 þ T cells. 98

Primary Sclerosing Cholangitis
Primary sclerosing cholangitis is an autoimmune disorder with inflammation of all parts of the biliary tree and associated T-cell infiltrates. It is associated with increased NK cell numbers in the peripheral blood but not the liver, 99 and these NK cells are functionally impaired due to the action of TNFα. 100 Several lines of immunogenetic evidence point to the involvement of NK cell receptors and ligands, and by extension NK cells themselves, in the progression of PSC. The development of the disease is associated with certain alleles of the NKG2D ligands MICA and MICB, 101,102 and also with the absence of HLA-Bw4 and HLA-C2, which are ligands for inhibitory killer-cell immunoglobulin-like receptors (KIRs), an effect that seems to require homozygosity for MICA5.1. 103 Furthermore, two single nucleotide polymorphisms (SNPs) in the NKG2D gene are associated with development of cholangiocarcinoma in the context of PSC. 104

Natural Killer Cells in Liver Cancer
Primary liver cancer is the third most common cause of cancer death worldwide, and the liver is also a major site of metastatic spread for many other malignancies. One major role of NK cells throughout the body is tumor surveillance: NK cells have the potential to play crucial roles in the immune response to both primary and metastatic liver cancers.

Hepatocellular Carcinoma
Hepatocellular carcinoma arises from hepatocytes in the context of pre-existing liver disease, most commonly viral hepatitis, ALD or NAFLD in the developed world, and is strongly associated with aflatoxin exposure in the developing world. Apart from HBV, which can be directly carcinogenic, HCC arises only in cirrhotic livers and is associated with inflammation in all settings, with systemic inflammation carrying a poor prognosis. 105 As in other malignancies, NK cells are functionally impaired in HCC. Tumor-infiltrating NK cells have reduced expression of granzymes and perforin and reduced cytotoxic potential compared with healthy controls, and are also reduced as a proportion of infiltrating lymphocytes compared with nontumor liver tissue. 106 Tumor environments can subvert NK cell immunity through a variety of mechanisms, including increased numbers of immunosuppressive cells such as regulatory T cells and myeloid-derived suppressor cells and increased expression of immunoregulatory cytokines such as TGFβ. 107 Changes in NK receptors or their ligands within the tumor milieu can also downregulate immunity. For example, loss of expression of the NKG2D ligands ULBP1 and MICA by HCC is associated with a poor prognosis. 108,109 The SNP at rs2596542 is associated with MICA shedding in HCC. Among hepatitis C patients, those with the AA genotype, which is associated with low soluble MICA, are at higher risk of HCC, whereas among hepatitis B patients, those with the GG genotype are at higher risk. [110][111][112] The differential effects of this SNP may be due to differences in induction and processing of MICA between the two viruses. The high-affinity allele MICA-129Met is associated with an increased risk of HCC in the context of HBV, 112 providing further evidence to suggest that NKG2D may play a role in both HCC surveillance and control.
Conversely, there is some immunogenetic evidence to suggest that high basal NK cell activity may be deleterious. KIR-HLA interactions that allow the development of highly responsive ("licensed") NK cells and activating KIR in the presence of appropriate HLA ligands are associated with increased progression to HCC in the context of HBV. 113 Similarly, different murine models have pointed to tumorigenic and well as protective roles for the NKG2D pathway in HCC. 114 Therefore, it remains to be seen whether increasing NK cell activation and/or promoting NKG2D will be helpful or harmful in HCC.
Attempts to overcome tumor evasion strategies and harness NK cells for immunotherapy of HCC are now being tested in the clinic. One strategy is enhancing the function of endogenous NK cells; an example is the use of an antibody that blocks engagement of the inhibitory receptors KIR2DL-1/2/3, which is currently being tried, in combination with PD-1 blockade, in patients with advanced solid tumors (NCT01714739). Alternatively, adoptive cell therapy can be performed, using infusions of autologous or allogeneic NK cells, expanded, activated or genetically modified to optimize their antitumor capacity. 107 Because NK cells may function better to prevent metastases than to penetrate established solid tumors, several ongoing trials are using allogeneic NK adoptive immunotherapy once the HCC has been resected or explanted (NCT01147380, NCT02008929). Previous in vitro studies suggested NK cells extracted from donor liver perfusates may be a promising source for such Seminars in Liver Disease Vol. 37 No. 3/2017 adoptive therapy because of an enhanced capacity to kill HCC following IL-2 activation. 7 However, peripheral NK cells can also be rendered effective by expansion using the NK cellstimulating cell line K562-mb15-41BBL or by transduction with the chimeric receptor NKG2D-CD3z-DAP10. 115 Cholangiocarcinoma Cholangiocarcinoma is a primary malignancy of the biliary tree, and is closely associated with inflammation of the biliary system, primarily due to PBC, PSC, or infection with liver flukes. In mice, cytokine-induced killer cells and IFNγ are associated with control of subcutaneously injected cholangiocarcinoma cells. 116,117 In humans, two SNPs in the NKG2D gene are associated with the development of cholangiocarcinoma in the context of PSC, 104 but it remains to be seen whether this effect is NK cell-mediated.

Liver Metastases
In most major malignancies including those of the lung, breast, and gastrointestinal tract, spread to the liver is common. In mice, NK cells are well-known to have a role in the control of tumor metastases and intrahepatic NK cells control metastases via TRAIL. 118,119 The role of NK cells in metastases to the liver in humans is best studied in colorectal cancer, where tumors resected from the liver are available for investigation. In one study, infiltrating T cells, but not NK cells, were associated with a good prognosis, 120 and in other studies tumor infiltrating CD8 þ T cells have been shown to contribute to good prognosis. 121,122 In many solid tumors, NK cells become functionally impaired and unable to lyse tumor cells. These autologous NK cells may represent an effector population that could be recruited therapeutically to target metastases in vivo, although cytokine infusion approaches have so far been limited by clinical toxicity. 123 Combining cytokine activation of NK cells with ADCC using monoclonal antibodies that engage CD16, such as trastuzumab for HER2þ metastatic breast cancer, may be a more promising approach. 124 Lenalidomide, which decreases the activation threshold of NK cells via an effect on actin remodeling, 125 did not induce clinical response in metastatic renal carcinoma including in patients with liver metastases as monotherapy, but could be combined with ADCC-inducing antibodies, as in various hematological malignancies. 126

Natural Killer Cells in Liver Transplantation
Liver transplantation is the only definitive treatment for patients suffering from end-stage liver disease, but its success is heavily dependent on several immune-mediated factors, which may involve the action of NK cells. One early complication of transplantation is ischemia-reperfusion (IR) injury, in which hypoxic organ damage is exacerbated following the re-establishment of blood supply and oxygen delivery. Acute rejection occurs days to weeks after a transplant and is primarily mediated by alloreactive T cells, whereas chronic rejection can occur up to years later, is characterized by occlusion of the blood vessels leading to ischemia, and is a result of mixed mechanisms, including low-grade T-cell responses and antigraft antibody.
The role of NK cells in IR injury during liver transplant has not yet been investigated in humans, but there is evidence from mouse models that NK cells are involved in exacerbating this disease. Treatment of mice with the NK cell depleting antibody NK1.1 reduces hepatic IR injury, 127 and CD39deficient mice, in which NK cells are subfunctional, are less susceptible to hepatic IR injury than wild-type controls. 128 Natural killer cell-mediated skin graft rejection occurs in mice, 129 so it is possible that some component of liver transplant rejection may be NK cell-mediated. In rats, recipient NK cells are rapidly recruited to the allogeneic transplanted liver. These cells are the main producers of IFNγ in the transplanted liver, and antibody depletion of NK cells can prolong the survival of the allograft. 130 During human liver transplantation, recipient NK cells are also rapidly recruited to the graft. 19 Several immunogenetic studies have attempted to determine whether recipient NK cells may be involved in rejection of the transplanted liver, with mixed results. One study of 100 transplants showed that matching of the main HLA ligand recognized by NK cells, HLA-C, between donor and recipient, reduced the likelihood of acute rejection, with recipients who expressed no HLA-C group 2 alleles further protected from acute rejection. 131 However, later, larger studies were unable to replicate these results. 132,133 A study of HLA-C effect on chronic rejection of 595 liver transplants showed that the presence of at least one HLA-C group 2 allele in the donor was protective, 132 but again this finding could not be replicated in a later study. 133 It is unclear, then, if NK cells are important mediators of either acute or chronic rejection in liver transplantation.

Future Directions
Natural killer cells are one of the most prevalent immune cell types in the human liver and are implicated as players in many liver diseases (►Fig. 2). However, in the healthy liver and most disease settings, their dominant roles remain to be clarified. Murine models are an important adjunct, but not a substitute, for studies in humans, especially given some notable differences in NK cells between mouse and man. 16 Researchers in this field are fortunate to have access to human liver tissue from diagnostic biopsies, along with transplant perfusates, metastatic resections and cirrhotic explants, giving more access to NK cells at the site of disease than is possible for most human organs. Nevertheless, the demanding logistics of obtaining such samples has meant that the majority of studies are conducted on peripheral blood; this is becoming more of an issue as noninvasive diagnostic methods start to replace liver biopsies. Recent studies revealing large liver-resident NK cell populations that do not recirculate 19 underscore the importance of continued access to liver sampling, exemplified by a recent study tracking the kinetics of NK cell responses by an alternative approach of repeated fine-needle aspiration. 134 Future studies will need to analyze liver-resident and nonresident NK subsets separately because prior analyses of the whole intrahepatic fraction may have masked important disease-associated changes. Another factor that may have confounded the interpretation of results in previous studies is cytomegalovirus serostatus; it is now clear that prior cytomegalovirus infection is a major driver of NK cell adaptive differentiation 135 that needs to be controlled for in studies of diverse infections and liver diseases. In mice, the liver is a site where "memory" NK cells accumulate. 21,22 Whether this is the case in humans remains to be established, but elegant work in HCV has already revealed another "adaptive" feature of NK cells: their capacity for mediating fine specificity through altered KIR binding to subtle variants of peptide/MHC class I complexes, impairing NK cell inhibition. 136,137 It will be interesting to consider precise NK cell target selectivity in other settings.
In addition to analyzing freshly isolated hepatic NK cells, enhanced technology for multiparameter in situ immunostaining will facilitate improved studies of their anatomical localization and interactions in liver sections. More in-depth mechanistic studies of how NK cells kill their targets in the liver, and how other liver cell types influence hepatic NK cells, will take advantage of emerging three-dimensional models.
Such scaffold systems constitute exciting progress in the attempt to recapitulate the complexity of the liver microanatomy and to allow analysis of NK cell cross-talk with, for example, primary hepatocytes and hepatic stellate cells, rather than immortalized cell lines. 138 Ongoing studies are starting to address the profound influence of local factors such as hypoxia and nutrient metabolism on NK cells in the unique liver niche. Initial work suggests that the exposure of NK cells to hypoxic conditions mimicking those present in the liver (which receives more than 70% of its blood supply from the portal vein) has divergent effects on function: while low oxygen tensions prevent activating receptors driving target cell killing, they maintain ADCC by NK cells. 139 As such, it may be necessary to revisit ideas about NK cell function in the liver that are based on experiments performed at atmospheric oxygen tensions. Hypoxia may also interact with the unique cytokine milieu of the liver because TGFβ is able to inhibit the mTOR-dependent metabolic reprogramming central to IL-15-induced NK cell responses. [140][141][142] It will also be necessary to examine interactions between hypoxia and infection. Coculturing NK cells from HCV patients with HCV-replicating hepatocytes or with stellate cells revealed that hypoxia constrained their antiviral capacity, but not their antifibrotic activity 143 and one of the metabolic adaptations of NK cells exposed to hypoxia, HIF-1α-induced glycolytic gene transcription, 144 appears to be defective in HCV. 143 Studies addressing these novel aspects of NK cell biology in state-of-the art in vitro and in vivo models will enhance our understanding of their diverse roles in liver disease and the scope for immunotherapeutic modification. Meanwhile NK cells are already reaching the clinic for primary and secondary liver cancers and these trials will provide powerful insights into their potential as therapeutic tools.