Correspondence: Dr Jenny Mjösberg, Centre for Infectious Medicine, F59, Karolinska Institutet, Karolinska, University Hospital Huddinge, S-14186 Stockholm, Sweden. Email: email@example.com
Senior author: Jenny Mjösberg
Innate lymphoid cells (ILCs) is the collective term for a group of related innate lymphocytes, including natural killer (NK) cells and the more recently discovered non-NK ILCs, which all lack rearranged antigen receptors such as those expressed by T and B cells. Similar to NK cells, the newly discovered ILCs depend on the transcription factor Id2 and the common γ-chain of the interleukin-2 receptor for development. However, in contrast to NK cells, non-NK ILCs also require interleukin-7. In addition to the cytotoxic functions of NK cells, assuring protection against tumour development and viruses, new data indicate that ILCs contribute to a wide range of homeostatic and pathophysiological conditions in various organs via specialized cytokine production capabilities. Here we summarize current knowledge on ILCs with a particular emphasis on their tissue-specific effector functions, in the gut, liver, lungs and uterus. When possible, we try to highlight the role that these cells play in humans.
Significant progress has been made recently in understanding the complexity of the growing family of innate lymphoid cells (ILCs) and their role in tissue homeostasis and inflammation. The ILCs include natural killer (NK) cells, originally described in the 1970s for their ability to exert cytotoxicity against tumour cells and soon after also found to be important in viral infections.[1, 2] Similar to NK cells, ILCs are RAG-independent lymphocytes that require the transcriptional repressor Id2 and the common γ-chain of the interleukin-2 receptor (IL-2Rγc) for development (reviewed in ref. ). However, whereas IL-15 is crucial for NK cell development, non-NK ILCs instead rely on IL-7 and, hence, are absent in IL-7Rα−/− mice (reviewed in ref. ; Fig. 1). One population of ILCs, developing under the control of the transcription factor retinoic acid receptor-related orphan receptor γt (RORγt), is mainly found in lymph nodes and mucosal tissues, and has the capacity to produce IL-22 and/or IL-17 in both mice and humans.[4-10] Another subset, devoted to production of interferon-γ (IFN-γ), yet distinct from NK cells, was described in the human gut. Lastly, a population of IL-7-dependent ILCs dedicated to production of type 2 cytokines such as IL-5 and IL-13 has been identified.[12-16] Type-2-cytokine-producing ILCs rely on the transcription factors GATA3 and RORα, and are found in human lung and gut.[12, 13, 17, 18] A complicating factor for the ILC research field was the absence of an accepted nomenclature. However, recently, efforts were made to reach consensus on this topic. It was proposed that, in analogy with the T helper cell family nomenclature, NK cells and IFN-γ-producing ILCs would be designated ILC1s. The term ILC2s was chosen for RORα/GATA3-dependent ILCs producing IL-5/IL-13 whereas ILC3s describes RORγt-dependent ILCs producing IL-22 and/or IL-17. In this review, we discuss the current knowledge on ILCs with a focus on tissue-specific functions, more specifically in the gut, liver, lungs and uterus where these cells have been the most extensively studied. When possible, we put particular emphasis on information obtained in the human system, especially for NK cells where such data exist to a greater extent than for the other ILC lineages.
Natural killer cells
Natural killer cells are recognized for their ability to control tumours and viruses.[2, 20, 21] Data indicate that NK cells are not only present in lymphoid tissues and peripheral blood but also at high numbers in peripheral tissues where they exhibit both direct effector functions and indirect immunoregulatory roles..
During fetal development, NK cells are among the first functional immune cells to appear. Transcription factors important for NK cell development and functional maturation include Id2, E4BP4, PU.1, Ets-1, Mef, Irf-2, GATA3 and T-bet. It is generally accepted that the initial steps of NK cell development occur within the bone marrow. However, a complete, in situ, bone marrow NK cell development pathway has not been described in the human. This raises the question whether additional developmental niches exist. Recent work has substantiated such a hypothesis where Lin− CD34+ CD45RA+ NK cell precursors were identified in human secondary lymphoid organs.[33-37] These cells can, through four discrete stages, develop into CD56bright NK cells. Finally, CD56bright NK cells mature into CD56dim NK cells, whereby they up-regulate the low-affinity Fc-receptor CD16, acquire inhibitory killer cell immunoglobulin-like receptors (KIRs), and the capacity to release perforin and granzyme-containing granules.
In humans, two major mature NK cell subsets exist, CD56bright and CD56dim NK cells. CD56dim NK cells dominate in blood, bone marrow and spleen, representing around 90% of all NK cells.[39-41] Tonsils, secondary lymphoid organs, gut and the uterus are enriched for CD56bright NK cells whereas the liver contains equal proportions of CD56dim and CD56bright NK cells.[42-45] With respect to function, CD56bright NK cells were for a long time considered to be prominent cytokine producers with little capacity to perform cytotoxicity, whereas CD56dim NK cells were efficient killers but poor at producing cytokines.[42, 44] However, recent work shows that both CD56bright and CD56dim NK cells are efficient in producing cytokines and performing cytotoxicity upon target cell recognition.[46, 47] In the past, CD56dim NK cells were considered an ‘end stage’ of human NK cell maturation. However, it has now been demonstrated that CD56dim NK cells undergo a continuous differentiation process, characterized by the gradual shift in phenotype from NKG2A+ CD62L+ CD57− KIR− cells towards more terminally differentiated NKG2A+ CD62L− CD57+ KIR+ cells, and an associated functional polarization towards cytotoxicity.[39, 48]
Natural killer cell effector functions, including perforin-dependent and granzyme-dependent cellular cytotoxicity, cytokine (IFN-γ and tumour necrosis factor-α), and chemokine (CCL3-5) production, are tightly regulated processes.[2, 49] The NK cells express a wide repertoire of germline-encoded activation and inhibitory receptors that do not undergo somatic recombination.[50, 51] The net signalling input from these arrays of receptors determines whether NK cells are activated to kill target cells or produce cytokines and chemokines.[51, 52] The NK cell receptor repertoires have evolved to ensure the proper detection of pathogen-affected and transformed cells, but preserve tolerance towards self in accordance with the ‘missing-self’ hypothesis. Besides the signalling input via activation and inhibitory receptors, the NK cell threshold of activation can be influenced in positive and negative manners by cytokine priming and through cross-talk with regulatory immune cells such as dendritic cells.[54, 55] Furthermore, recognition of self-MHC class I molecules by inhibitory receptors, such as KIRs and NKG2A, potentiates the NK cell's ability to respond functionally, a process referred to as ‘education’ or ‘licensing’.[56, 57]
Non-NK innate lymphoid cells (ILC1–ILC3)
Populations of non-NK ILCs, which are developmentally and phenotypically related to NK cells, have been known in mice for approximately a decade. Although the function of these cells is still being explored, it is clear that different subsets of cells exist that are dedicated to production of IL-17/IL-22 (ILC3), IL-5/IL-13 (ILC2), and IFN-γ (ILC1). They are ruled by specific transcription factors; T-bet for ILC1, GATA3 for ILC2, and RORγt for ILC3 (reviewed in refs [58, 59]) (Fig. 1). From these studies, it soon became obvious that there are intriguing parallels between the ILC and T helper cell families, with seemingly similar transcriptional programmes governing the development and function of the various subsets. Hence, ILC1, ILC2 and ILC3 seem be the innate equivalents of T helper type 1 (Th1), Th2 and Th17/Th22 cells, respectively. Although we still have a limited understanding of the consequences of having two such seemingly parallel systems, the realization of these parallels has facilitated the investigations of this family of innate lymphoid cells, which is now rapidly being unravelled.
Lymphoid tissue inducer (LTI) cells are the prototypic members of the non-NK part of the ILC family and were also the first to be discovered in mice. Physiologically, LTI cells have the crucial task to, together with mesenchymal stromal cells, establish fetal lymph node structures. LTI cells are dependent on the transcription factors Id2 and RORγt for development as mice with deletions in these genes lack LTI cells and fail to generate full lymph node capacity.[3, 4, 60] Furthermore, signalling via the IL-2Rγc is required for development of LTI cells as IL-2Rγc−/− animals show reduced numbers of these cells. Interestingly, cells exhibiting an LTI phenotype also persist after birth and have been implicated in restoration of secondary lymphoid organs following acute viral infections. These cells are also present in adult spleen and gut where they are important producers of IL-22 and IL-17, consistent with their expression of the Th17-associated transcription factor RORγt. However, the exact relationship between fetal and adult LTI cells still has to be elucidated.
The capacity to produce IL-22 is associated with surface expression of NK cell-associated receptors belonging to the family of natural cytotoxicity receptors (NCRs), NKp44 in humans and NKp46 in mice. Emphasizing their similarity to NK cells, in humans, IL-22-producing ILCs have been referred to as NK22 cells or as ILC22 cells. Furthermore, IL-22-producing ILC3s are strikingly similar to so-called stage 3 immature NK cells, as these cells too are characterized as Lin− cells expressing c-kit (CD117) and producing IL-22. However, in humans, CD127 and lymphocyte function-associated antigen-1 (LFA-1) staining can be used to distinguish between precursors of NK cells and IL-22-producing ILC3s because Lin− CD117+ CD127+ LFA-1− cells express RORγt, produce IL-22 and fail to generate NK cells when cultured in IL-15.[7, 64] This observation is consistent with the notion that NK cells do not express RORγt during their development, and hence must be a lineage distinct from ILC3s.
The main physiological function of IL-22-producing ILC3s seems to be as defenders of mucosal barrier function, especially in the gut. For instance, several reports have identified these cells as a major source of IL-22 following infection with the mouse pathogen Citrobacter rodentium.[5, 65, 66] Although IL-23 is not needed for development of ILC3 cells, IL-22 production is potently induced by IL-23 and IL-1β, also in humans.[63, 67, 68] Recently, dietary compounds, signalling via the transcription factor aryl hydrocarbon receptor (AHR) expressed by ILC3s, were shown to regulate IL-22 production from ILC3s[69-71] but this has to be confirmed in humans.
Much effort has recently been aimed at addressing the developmental relationship between NK cells, LTI cells and IL-17/IL-22-producing ILC3s. It has been suggested that NK cells and LTI cells represent precursors of IL-22-producing cells. On the other hand, it has also been advocated that IL-22-producing NCR+ ILC3s are a separate lineage with developmental independence from NK cells and LTI cells. In mice, a population of CD3− NKp46+ cells, which is heterogeneous in its composition, produces IL-22. The population was shown to consist of NK1.1+ Ly49+ RORγt− NK cells and NK1.1− CD127+ RORγt+ ILC3s, where the latter subset possessed the capacity to produce IL-22. Although both of these populations depend on Id2 for development, fate-mapping experiments, tracking RORγt+ cells, revealed that NK1.1− CD127+ RORγt+ ILC3s are not precursors of NK1.1+ Ly49+ RORγt− NK cells, as the former had no history of RORγt expression. Similar conclusions were reached using an NKp46-reporter system to track NKp46+ RORγt− and NKp46+ RORγt+ ILCs.
When the progeny of RORγt+ CD3− cells were tracked, one group concluded that IL-22-producing RORγt+ NKp46+ cells were not derived from LTI cells but from a RORγt+ liver precursor that could give rise to all RORγt+ ILC subtypes present in mice (LTI and NKp46+ ILCs). These observations contrast with another report showing that LTI cells, but not NK cells, adoptively transferred to mice generated NKp46+ RORγt+ ILCs capable of producing IL-22. Interestingly, RORγt was not stably expressed in these cells, as a proportion of the adoptively transferred LTI cells lost RORγt expression. RORγt loss was accompanied by decreased IL-22 production and increased capacity to produce IFN-γ. This raised the idea of an ILC1 population that, similar to Th1 cells, would be dedicated to production of IFN-γ rather than IL-22.
Fate mapping experiments in mice have convincingly shown that NK cells are clearly distinct from RORγt+ ILC3s, with respect to both development and function. Interestingly, several studies suggested the existence of a non-NK ILC population with an effector phenotype resembling that of NK cells, namely the capacity to express T-bet and produce IFN-γ. The existence of an IL-12/18-responsive CD127+ ILC population dedicated to the production of IFN-γ was identified in the mouse gut. RORγt+ ILC3s could, after IL-12/18 stimulation, develop into a population of cells that gained the capacity to produce IFN-γ, a process accompanied with loss of RORγt expression. A similar ILC1 population has also been observed in humans. Recent work substantiated these findings, where a detailed characterization revealed the ILC1 cells to be a distinct subset of cells that were, at least partly, derived from a RORγt+ precursor. The ILC1s expressed the Th1-associated transcription factor T-bet and produced significant levels of IFN-γ upon IL-12/IL-18 stimulation. Interestingly, these mucosal-tissue-associated ILC1s do not exist in the human fetus. Instead, they accumulate during intestinal inflammation, suggesting that they require both microbial colonization and inflammatory signals for development and differentiation, a finding recently confirmed in the mouse.
The first indications of the presence of an ILC population dedicated to type 2 cytokine production came in the year 2001, when it was shown that IL-25 treatment in mice elicited a population of non-T, non-B cells that produced IL-5 and IL-13. However, it was not until almost a decade later that the phenotype and function of these cells was fully characterized as a population of IL-25- or IL-33-inducible Lin− CD127+ Sca1+ ST2+ cells dedicated to the production of IL-5 and IL-13. In mice, these cells proved essential for expulsion of helminthic parasites from the gut.[14-16] Hence, they were referred to as type 2 ILCs (ILC2s) in analogy with Th2 cells having a similar cytokine profile. ILC2s were not only important mediators of type 2-mediated inflammation in the gut, but also in the airways where they could mediate tissue repair after virus-induced airway hyper-reactivity.[13, 80] Later reports established that a similar population of cells, ILCs dedicated to type 2-cytokine production, was present also in humans and characterized as Lin− CD127+ cells expressing the prostaglandin D2 receptor chemoattractant receptor homologous molecule expressed on Th2 cells (CRTH2).[12, 13]
Besides having a unique cytokine production profile within the ILC family, ILC2s were also shown to be developmentally and phenotypically distinct from the previously described RORγt-dependent ILC populations (ILC1 and ILC3) as they were present in RORγt−/− mice. Instead, mouse ILC2s were demonstrated to rely on the related transcription factor RORα and developed from a bone marrow-derived Lin− CD127+ Flt3+ common lymphoid progenitor under the influence of Notch1, IL-7 and IL-33 signalling. As a parallel to the other ILC subsets, it would make sense if ILC2s, in analogy with Th2 cells, depend on the Th2 master transcription factor GATA3 for their development and function. Indeed, this was recently shown to be the case both in humans and mice.[17, 18] In mice, a bone marrow Lin− Sca1+ Id2+ GATA3high population proved to be an early precursor of mature ILC2s. This was corroborated by results obtained in the human setting where GATA3 over-expression in a population of CRTH2− ILCs converted these cells into CRTH2+ ILC2s with the capacity to produce high levels of IL-5 and IL-13 in response to IL-33 and thymic stromal lymphopoietin. Collectively, these studies established that, as for Th2 cells, GATA3 functions as the master transcription factor of the ILC2 fate.
Tissue-specific effector functions of ILCs
The recent discovery of non-NK ILCs complicated the interpretation of older literature on NK cells in tissues because these studies sometimes fail to distinguish NK cells from non-NK ILCs. Nevertheless, NKp46+ NKp44− RORγt− lymphocytes with a transcriptional profile similar to splenic NK cells can be found throughout the human small and large intestine (Fig. 2).[82, 83] These gut NK cells display a phenotype similar to peripheral blood CD56bright NK cells with high expression of NKG2A, intermediate levels of intracellular effector molecules, and low to absent levels of CD16 and KIRs. Interestingly, in humans, a lamina propria Lin− c-kit+ precursor population, expressing Id2 and PU.1, could upon culture up-regulate CD56, start to express effector molecules, and gain the capacity to perform cytotoxicity, suggesting that the human gut contains NK cell precursors. The role of NK cells in gut pathology has been examined. In mice, Listeria monocytogenes infection induced gut NK cell IFN-γ and this contributed to control bacterial dissemination. Furthermore, patients with inflammatory bowel diseases (IBD), such as Crohn's disease and ulcerative colitis, exhibit phenotypic alterations in their gut NK cell compartment.[83-85] The significance of these correlative results is further corroborated by population genetic studies indicating that certain NK cell receptors are linked to IBD susceptibility. Whereas less is known with respect to the interplay between gut NK cells and the intestinal microbiota, non-mucosal NK cells are dysfunctional in germ-free mice. While the broader implications of this finding still remain unclear, the relationship between gut NK cell function, inflammation and the intestinal microbiota merit further investigation.
ILC1s and ILC3s – yin and yang of the gut?
In mice, NKp46+ RORγt-dependent ILC3s were shown to reside in the gut lamina propria and in cryptopatches, but less in the intraepithelial layer.[5, 63, 88, 89] In the gut, these cells are important sources of IL-22, which is indispensable for homeostatic maintenance of epithelial barrier function in mice. This is exemplified by the notion that mice lacking NKp46+ RORγt-dependent ILC3s (RAG2−/− IL-2Rγc−/−) or IL-22 (IL-22−/−) rapidly succumb to infection with Citrobacter rodentium, which typically triggers IL-23-dependent IL-22 production from ILC3s.[5, 65, 66] The mechanism behind the gut epithelial barrier protective role of IL-22-producing ILC3s has been extensively studied in several mouse models of colitis and recently also in a model of graft-versus-host disease.
These gut ILC3s are in constant interaction with the microbiota and the net signalling from dietary and epithelium-derived factors as well as IL-23 probably determine the outcome of ILC3 effector functions. In mice, it was recently demonstrated that the source of IL-23 following systemic flagellin administration was a subset of gut CD103+ dendritic cells, which, in turn, promoted IL-22-dependent RegIIIγ production from gut epithelial cells. However, during homeostatic conditions, the levels of IL-23 are likely to be low. Here, dietary compounds, such as AHR ligands, might instead trigger IL-22 production from gut-residing ILC3s.[69-71] Recently, AHR was also shown to be important for IL-23 responsiveness as AHR−/− mice failed to respond to IL-23 production induced by Citrobacter rodentium and rapidly succumbed to infection. In addition to induction of anti-microbial peptides, such as RegIIIa and RegIIIb, IL-22 also has an important role in maintaining barrier function in the gut via its interaction with colonic epithelial cells where signalling through IL-22 receptors, via signal transducer and activator of transcription 3, promotes epithelial proliferation and gut wound healing.[65, 88, 91]
In addition to maintaining gut mucosal barrier function via IL-22 production, ILC3s have been implicated in colitis as demonstrated by work performed in several mouse models. For instance, it was suggested that IL-17 could drive Helicobacter hepaticus-induced colitis where induction of colitis elicited a population of ILCs producing IL-17 and IFN-γ. Interestingly, neutralization of IL-17 had little effect on colitis severity, whereas ablation of IFN-γ restored the phenotype, suggesting that IFN-γ is important in this particular model. In line with these data, colitis could be induced via anti-CD40 injection into RAG2−/− mice promoting a population of IFN-γ-producing NKp46+ ILCs that were distinct from NK cells. Adoptive transfer of such cells was enough to induce colitis, whereas depletion ameliorated the gut inflammation, again suggesting that IFN-γ-producing ILC1s can drive colitis. These findings are further corroborated by results from the human setting where ILC1s with a Th1 phenotype accumulate in the inflamed ileum of patients with Crohn's disease. These ILC1s were distinct from NK cells because they lacked expression of prototypic NK cell markers, including cytotoxic molecules, and did not develop into NK cells in vitro. Intriguingly, although RORγt− precursors cannot be ruled out, these ILC1s could be generated from RORγt+ IL-22-producing ILC3s under the influence of IL-12/18. Interleukin-12/18 stimulation led to down-regulation of RORγt, up-regulation of T-bet, and the capacity to produce IFN-γ. Furthermore, in the absence of T-bet, using Tbx21−/− RAG2−/− (TRUC) mice, colitis can be driven by ILC1s producing excessive amounts of IL-23-triggered and tumour necrosis factor-triggered IL-17A. Hence, there seems to be a redundancy in the system, with IL-17 and IFN-γ both being able to drive gut inflammation. Given that blocking of these two cytokines in human IBD have largely generated disappointing results,[94, 95] more information is needed regarding the role of these cytokines and of ILCs in the pathogenesis of IBD.
In summary, ILC3s seem to be important for the maintenance of gut barrier function and homeostasis (Fig. 2). However, similar to T helper cells, these cells are plastic and may be dysregulated to participate in inflammation through reprogramming by the surrounding pro-inflammatory microenvironment.
The ILC2s, dedicated to production of type 2 cytokines, were the third population of ILCs to be described in the mouse. Several groups established that these cells were critical for gut parasite expulsion (Fig. 2),[14-16] although this still remains to be verified in the human setting. In RAG−/− mice, a population of Lin− IL-7Ra+ ckit+ Sca1+ T1/ST2+ ILC2s, associated with the mesenteric fat, mediated helminthic parasite expulsion from the gut. This population produced type 2 cytokines upon parasite infection or after treatment with IL-33. The ILC2s could induce all the classical features of type 2-mediated inflammation in the gut, including goblet cell hyperplasia and support of B1-cell-mediated IgA production. Absence of ILC2s, as in RAG−/− IL-2Rγc−/− mice, was associated with failure to handle parasitic infection, whereas adoptive transfer of ILC2s to these mice rescued them from A. brasiliensis-associated gut pathology. Corroborating these observations, additional reports, using either IL-4 or IL-13 reporter mice, pinpointed the source of IL-13 in the gut. A population of IL-13-producing Lin− lymphocytes, termed nuocytes by the authors, was induced by IL-25, or IL-33, and was largely absent in IL-25R-deficient animals (IL-17BR−/−). As a consequence, IL-17BR−/− mice failed to clear gut parasitic infection, whereas this function was restored upon adoptive transfer of IL-13-producing ILC2s. Similarly, with IL-13 reporter mice, the presence of a Lin− cell population elicited by IL-25 administration or Nippostrongylus brasiliensis infection was demonstrated. Interestingly, in this model, IL25−/− mice developed severe IFN-γ- and IL-17-mediated infection-induced gut inflammation. This suggests that ILC2s in gut might not only be mediators of type 2 immunity, but might also have immune regulatory functions limiting gut inflammation.
In the human setting, ILC2s have been reported in fetal gut, as well as in the adult gut of IBD and non-IBD patients. However, the frequencies are low, and so far no definite physiological or inflammatory role has been ascribed to these cells. Future studies, aimed at unravelling the role of these cells in human gut parasite infections, which is a major health burden in many low-income countries, are warranted.
Natural killer cells are highly accumulated in the liver where they represent 30–50% of all hepatic lymphocytes, and exist in an environment under constant exposure to bacterial products and gut-derived antigens. Recently, attention has been focused on the role of NK cells in liver diseases (reviewed in refs [98-100]). This is in part a result of population genetics studies in patients with diseases such as hepatitis C, hepatocellular carcinoma, sclerosing cholangitis and biliary cirrhosis, which have all identified susceptibility genes directly, or indirectly, linked to NK cell function.[101-104]
In the human fetus, functional liver NK cells can be found as early as gestational week 6, at a time in development when the liver is the main haematopoietic organ. Interestingly, studies suggest that the liver continues to represent a developmental niche for NK cells long after the bone marrow has assumed the task of haematopoiesis because all five known sequential stages of NK cell development can be found in the adult human liver. In addition to de novo development in the liver, NK cells can, under pathological conditions via CCL2, CCL3, CXCL10 and CXCL16 produced by liver-resident cells,[105-107] be actively recruited to the organ. Distinct from peripheral blood, where CD56dim NK cells predominate, the liver contains an equal proportion of CD56dim and CD56bright NK cells. The intrahepatic CD56dim NK cell subset is similar in phenotype and function to CD56dim NK cells from peripheral blood. Interestingly, although inhibitory KIRs and NKG2A are found on intrahepatic CD56dim NK cells, a recent report suggests that the liver contains a reduced fraction of ‘educated’ (or ‘licensed’) cells. Whether this is important to preserve self-tolerance in the liver, or to avoid an inflammatory response secondary to the constant gut-antigen exposure, still remains to be investigated. Nevertheless, upon cross-talk with activated Kupffer cells, CD56dim NK cells become potent killers and cytokine producers.[43, 108]
Compared with CD56dim NK cells, less is known about CD56bright NK cells in the liver. However, it has been shown that CD56bright NK cells express high levels of the death receptor TRAIL. As both hepatocytes and cholangiocytes are sensitive to dead receptor-induced apoptosis,[109, 110] it has been suggested that NK cells contribute to liver injury via TRAIL.[109-111] In contrast to this, TRAIL also functions in controlling hepatitis C virus infection.[47, 112] Clearly, more studies focusing on the functions of intrahepatic CD56bright NK cells during liver disease are warranted.
The presence and potential roles for the various non-NK ILC subpopulations in the liver are still largely unknown. In mice, the fetal liver is an important site for LTI cell development from an Id2+ RORγt+ liver precursor expressing the integrin α4/β7. Notably, progenitor expression of integrin α4/β7 is likely to discern between development towards an LTI (derived from α4/β7+ precursors) or an NKp46+ LTI fate, the latter cell type most probably representing the IL-22 producers involved in gut homeostasis.[74, 113] Of stage 3 immature NK cells found in the adult liver, a small fraction express RORγt+, potentially representing IL-22-producing ILCs (Fig. 2). Future studies will have to determine the function of these cells and their potential role in liver pathology. Interestingly, IL-22 has a crucial role in the protection against experimental acute liver inflammation by enhancing hepatocyte proliferation and survival and thereby limiting liver damage. Hence, it is tempting to speculate that ILCs are an important source of tissue protective IL-22 in the liver, and as such might pose as an attractive future treatment target for liver damage.
Lung and airways
Natural killer cells constitute approximately 10% of all tissue-residing lymphocytes in the human lung. Pulmonary epithelial cells directly support the resident NK cells by constitutive production of IL-15, a production that increases during inflammation. Interestingly, although IL-15 supports NK cell development and maturation, local NK cell maturation does not seem to take place in mice because immature NK cells are rarely found in the lung. In humans, lung NK cells have only been studied in detail in tissue obtained from patients with lung cancer. Here, lung NK cells have a similar receptor expression profile to that in peripheral blood and mature CD56dim NK cells dominate. Although displaying phenotypical similarities with peripheral blood NK cells, the cytotoxic capacity of lung NK cells is reduced compared with that of peripheral blood or splenic NK cells.[119, 120] One explanation for this might be that a higher proportion of lung NK cells express the inhibitory NK cell receptor NKG2A, indicating that lung NK cells have a higher threshold for activation. Furthermore, the reduced killing activity of lung NK cells might be maintained by prostaglandins and immunosuppressive cytokines such as IL-10 and transforming growth factor-β secreted by the alveolar macrophages.[121-123]
The role of NK cells in respiratory infections and other inflammatory disorders of the pulmonary tract have been reviewed previously. Briefly, studies of fatal cases of influenza or respiratory syncytial virus infection in humans have revealed an impaired NK cell function that could be associated with a more severe course of infection.[125, 126] Depletion of NK cells with monoclonal antibodies in experimental models of influenza virus or respiratory syncytial virus also resulted in increased mortality.[127-129] In contrast to this, recent evidence from IL-15−/− mice shows that complete loss of NK cells reduced morbidity and mortality in experimental influenza virus infection.[127, 129, 130] Furthermore, conventional lung NK cells were the major producers of IL-22 during influenza virus infection and this promoted epithelial regeneration. These IL-22-producing NK cells were negative for CD127 and RORγt, separating them from IL-22-producing ILC3s. Similar to in influenza, NK cells might have dual roles during RSV infection because genetic removal of NK cell IFN-γ responses resulted in a type 2 immune response with asthma-like bronchial hyper-reactivity. This dual role for NK cells in lung infections is corroborated by recent work in the lymphocytic choriomeningitis virus model where NK cells were found to act as rheostats modulating CD4 T-cell responses.
ILC2s and ILC3s – yin and yang of the lung?
The ILC2s are present in both mouse and human lung tissue (Fig. 2),[12, 13] as well as in inflamed nasal polyps of patients with chronic rhinosinusitis where these cells are accumulated. Given the potency of IL-5 and IL-13 to drive type 2-mediated immunity, including many of the clinical features of allergy and asthma, it is plausible that ILC2s would also play a role in type 2-mediated immunopathologies of the airways, such as in AHR. Indeed, several studies support this notion.[80, 135-139] It was shown that ILC2s mediated virus-induced lung inflammation via their responsiveness to macrophage-derived IL-33. Depleting ILC2s significantly ameliorated AHR whereas adoptive transfer of ILC2s into IL-13−/− mice re-established AHR. In contrast to these findings, using a different influenza virus strain, depletion of ILC2s severely worsened virus-induced AHR. The tissue-protective effect of ILC2s was explained by production of amphiregulin, a protein belonging to the epidermal growth factor family. Administration of amphiregulin restored epithelial cell integrity, airway remodelling and lung function.
Although ILC2s seem to have both damaging and tissue protective roles in the lung following virus infection, data on experimental allergic asthma are more consistent. In asthma caused by house-dust mite or by the protease allergen papain, epithelium-derived factors such as thymic stromal lymphopoietin, IL-25 and IL-33 promote airway inflammation and accumulation of ILC2s in the lung.[137, 139] Compared with Th2 cells, ILC2s proved to be major contributors to type 2 cytokine production as IL-25- and IL-33-induced asthma and subsequent ILC2 activation occurred also in RAG2−/− mice.[135, 136, 139] In summary, the role of ILC2s in asthma seems to be as early propagators of stromal-derived signals upon allergen encounter with subsequent induction of type 2 cytokine production. This might be the early trigger of type 2-mediated pathology, which subsequently propagates to also include the adaptive immune system. However, further research is needed to reveal the exact mechanisms behind the role of ILC2s in allergen-induced asthma, especially in the human setting.
Limited information is available regarding the role of RORγt-dependent ILC3s in the lung. However, a recent report demonstrated that IL-22 acts as a negative regulator of experimental airway inflammation as IL-22−/− mice exhibited exacerbated ovalbumin-specific asthma whereas IL-22 treatment before allergen exposure protected against airway inflammation. In this model, IL-22 was mainly produced by RORγt+ ILC3s, which acted on lung epithelial cells to reduce IL-13-induced CCL17 production, thereby limiting Th2 cell recruitment into the lung. Intriguingly, this finding parallels those obtained from the gut, where IL-22-producing ILC3s act as mucosal gatekeepers mediating tissue protection and supporting mucosal homeostasis. Furthermore, whereas in the gut, inflammation seems to be mainly mediated by IFN-γ-producing ILC1s, in the lung, pathology is predominantly caused by ILC2s. Future studies will reveal whether ILC1s are present in lung, and if they contribute to type 1-mediated disease, e.g. that observed in chronic obstructive pulmonary disease.
A somewhat unexpected role for NK cells has been revealed in studies of human pregnancy (Fig. 2). NK cells are found in the uterus throughout the menstrual cycle, but increase in frequency (often representing up to 70% of all lymphocytes) both in the uterine mucosa as well as in the decidua of the placenta after implantation. Interestingly, the uterus was recently suggested to represent a developmental niche for NK cells since stage 3 immature NK cells are present and can differentiate in vitro into mature uterine NK cells. Uterine NK cells exhibit a CD56superbright CD9+ NKG2A+ KIR+ CD16− phenotype, which is distinct compared with both conventional CD56bright and CD56dim NK cells.
During pregnancy, decidual NK cells regulate trophoblast invasiveness and remodel uterine spiral arteries via production of cytokines, chemokines and angiogenic factors, such as CXCL8, CXCL10, IFN-γ, vascular endothelial growth factor, and placental growth factor.[142, 143] Mice lacking uterine NK cells and mice deficient in IFN-γ-signalling fail to initiate pregnancy-induced modifications of their uterine spiral arteries. Whereas many differences in placentation exist between mice and humans, population genetic studies support a role for NK cells in human pregnancy disorders. Pre-eclampsia, fetal growth restriction and recurrent miscarriages have all been linked to certain maternal and fetal KIR–HLA combinations.[144, 145] This suggests that a common uterine NK cell-mediated mechanism underlies these disorder. However, the exact explanation for why these clinical conditions occur still remains an outstanding question.
As discussed above, CD56superbright CD16− NK cells are the major cell population present in the uterus and decidua both before and during pregnancy. In addition, the uterine mucosa harbours stage 3 immature NK cells. However, non-NK ILCs also seem to be present in the non-pregnant uterine mucosa as well as in the decidua during the second trimester (Fig. 2). These cells were characterized as CD127+ RORγt+ with the capacity to produce IL-22. Although expression of the IL-22 receptor could not be verified in uterine mucosa, IL-22 could potentially have pro-proliferative effects on trophoblasts in the decidua, as observed for stromal cells in gut and lung, and so contribute to trophoblast invasion. However, studies are required to establish the role for these cells in reproduction as well as in pregnancy disorders.
Conclusions and future perspectives
ILC1, ILC2 and ILC3s are now emerging as important cell populations regulating tissue homeostasis and inflammation. However, much of the current knowledge on ILCs stems from experimental models and still requires confirmation in humans. Several other challenges lie ahead in understanding this family of cells and particular efforts are needed in some specific areas. One such topic concerns how the subpopulations of ILCs are developmentally related; at which developmental stage of haematopoiesis the different ILC subsets are separated, what identifies the progenitors of each subset, and which signals determine the fate of these progenitors. In addition, the plasticity between the different subsets needs to be unravelled, particularly between ILC3 and ILC1 cells.
Furthermore, although NK cells have been known for decades for their ability to provide protection against viruses and tumour cells, this knowledge now has to be placed in the context of the more recently discovered ILC populations, which share developmental and phenotypical characteristics with NK cells. Along the same lines of reasoning, the roles for the non-NK ILC populations in the defence against viruses and tumour cells needs to be investigated.
The authors thank Hans-Gustaf Ljunggren, Jakob Michaelsson, and Martin Ivarsson, all from the Centre for Infectious Medicine, Karolinska Institutet, for helpful comments. We offer our sincere apologies to colleagues whose work could not be adequately discussed or cited owing to space limitations. Our work is supported by the Swedish Research Council, the Swedish Cancer Society, the Karolinska Institutet, the Alex and Eva Wallström Foundation, the Swedish Society of Medicine, the Jeansson Foundation, and the Bengt Ihre Foundation.