Viewpoints on NK cells
Killers and beyond: NK-cell-mediated control of immune responses
Effective immunity requires coordinated activation of innate and adaptive immune responses. NK cells are principal mediators of innate immunity, able to respond to challenge quickly and generally without prior activation. The most acknowledged functions of NK cells are their cytotoxic potential and their ability to release large amounts of cytokines, especially IFN-γ. Recently, it has become clear that NK cells are more than assassins. Indeed, NK cells play critical roles in shaping adaptive immunity.
NK cells: beyond killing
NK cells are an important component of the innate immune system with the ability to eliminate pathogen-infected cells, tumour cells, and to influence the outcome of bone marrow transplantation. NK cells have been traditionally considered effector cells whose rapid activation provides a first line of defence prior to the initiation of an adaptive immune response. However, recent data indicate that, in addition to their role in innate immunity, NK cells have the capacity to shape adaptive immune responses. Indeed, NK cells appear capable of mediating a sophisticated series of activities that impact on the functions of both innate and adaptive immunity. Much however remains to be determined in relation to the mechanisms by which NK cells exert these functions. What has become clear over the last few years is that the impact of NK cells is highly dependent on the subset involved, as well as the site and milieu in which the interactions occur. Importantly, it is worth considering that the activities of NK cells will vary depending on the system being analysed. Although NK cells have been thought to principally promote adaptive immune responses, emerging evidence from in vivo analyses suggest that, in fact, NK cells can prevent, terminate and/or limit adaptive immune responses. The challenge ahead is to define how NK cells exert such activities in as many physiological situations as possible, so that we will eventually be able to harness their potential in therapeutic settings. Here, we will discuss some of the current knowledge about the impact of NK cells on adaptive immunity.
Unlike B and T cells, NK cells have the ability to recognise and kill target cells without prior sensitisation. NK cells identify targets expressing reduced levels of MHC class I and/or “stress-induced” ligands. The precise mechanisms governing NK-cell activation have been extensively reviewed elsewhere 1, 2. In general terms, NK-cell activity is regulated by integrated signals provided by an array of activating and inhibitory cell surface receptors. Typically, activation of NK cells is achieved via the triggering of NK-cell activating receptors in combination with pro-inflammatory cytokines. Once activated, NK cells mediate their functions through the direct killing of target cells and/or the release of cytokines (reviewed in 3). The ability to identify and destroy defective cells is the best-characterised function of NK cells. However, NK cells are known to interact with various components of the immune system, and therefore have the potential to function as regulatory cells. Indeed, recent evidence suggests that NK cells can enhance or restrict adaptive immune responses, with factors such as the tissue microenvironment, the nature of the triggering stimulus and the subtype of NK cell all influencing the final outcome.
Consistent with their role in innate immunity NK cells are broadly distributed. However, rather than consisting of a homogeneous population, NK-cell subsets with distinct functional characteristics have been defined in both mice and humans. In humans, NK cells can be subdivided based on the expression of CD56 and CD16 4. The majority of NK cells in the blood and spleen are CD56dimCD16+; these cells are highly cytotoxic and produce IFN-γ upon stimulation. In contrast, most NK cells found in the lymph node are CD56hiCD16−, display poor cytotoxic capacity, but release large amounts of cytokines 4. Differences in the expression of NK-cell receptors and chemokine receptors between the NK-cell subsets have also been noted 5. While it has been considered that NK-cell sub-populations may belong to separate lineages, recent studies have shown that the two human NK-cell subsets represent different stages of sequential maturation, with the CD56dim NK cells originating from CD56hi NK cells 6–8.
Like human NK cells, murine NK cells comprise functionally distinct subsets. In mice, three subsets of NK cells can be differentiated based on the differential expression of CD27 and CD11b 9. Developmental analysis has demonstrated that upon commitment to the NK-cell lineage, NK cells undergo a maturation process during which they are CD27hiCD11blo in the most immature stage; they then mature to a double positive phenotype, before finally becoming CD27loCD11bhi in the most mature, senescent stage. The CD27hiCD11bhi NK-cell subset elicits the strongest effector functions 9. Beside this hierarchy, a population of NK cells with the phenotype NK1.1+CD11c+B220+ has been identified in lymphoid organs 10–13. These cells produce more IFN-γ than conventional CD11c− NK cells; they also display a defined receptor repertoire and tissue localisation, and are proposed to be the murine equivalent of human CD56hi NK cells 10.
The concept that NK cells develop exclusively from precursors in the bone marrow has been challenged by the recent identification of NK cells that develop in the thymus. These NK cells are CD127+ and their development strictly depends on GATA-3; they have reduced cytotoxic function, but can produce significant quantities of cytokines. These features are shared with human CD56hiCD16lo NK cells, and thus these cells may represent their murine counterpart 14.
While the exact link between the various human and murine NK-cell subsets remains uncertain, it is clear that NK cells in both human and the mouse are not a homogeneous population. Therefore, in attempting to understand the impact of NK cells on other arms of the immune response, the subtype(s) of NK cells involved should be considered.
NK-cell subsets have differential homing properties
NK-cell subsets express a series of chemokine receptors and adhesion molecules that enable specific tissue localisation (reviewed in 15). Under inflammatory conditions, mouse NK cells are recruited from the blood to the lymph nodes through high endothelial venules in a CD62L-dependent manner 16, 17. In humans, CD56hi NK cells express CD62L and this allows their recruitment from the bloodstream into the lymph nodes 16. In response to IL-18, CD56hi NK cells also express CCR7 and respond to lymph node associated chemokines, which direct them to the T-cell area of lymph nodes 18. Absence of these chemokine receptors on CD56lo NK cells is believed to be responsible for their homing to spleen and blood. Similarly, in mice, secondary lymphoid tissues contain mainly the less mature CD11blo NK-cell subset. Murine CD127+ thymus-derived NK cells and NK1.1+CD11c+B220+ NK cells preferentially home to the lymph nodes. On the other hand, more mature CD11bhi NK cells are found in blood, spleen, lung and liver. Under resting conditions, NK cells reside primarily in the red pulp of the spleen and in parafollicular and medullar lymph node areas, excluded from T- and B-lymphocyte zones, but in close proximity to DC and macrophages. Resting NK cells are also found in blood and lymphatic vessels and sinuses where they can encounter tissue DC that are migrating to draining lymph nodes 19, 20. CXCR3 also plays a role in NK cells chemoattraction following an inflammatory stimulus 21. Recently, NK-cell maturation has been found to correlate with acquisition of the lysophospholipid sphingosine 1-phosphate receptor S1P5, which is essential for recruitment into inflamed tissues 20. In addition, other chemokine receptors, including CCR2, CCR5 and CX3CR1, have been involved in NK-cell recruitment (reviewed in 15). Thus, the migratory capacity of NK-cell subsets should be considered when defining their potential impact on localised immune responses.
NK cells: more than just killers
In addition to their classic role as mediators of cytotoxicity, several studies have suggested that NK cells may also have regulatory functions. For example, NK cells were shown to be required for the generation of anti-viral cytotoxic T-cell responses and enhanced activation of B cells in vitro22, 23. Similarly, in vivo administration of anti-neutral glycolipid ganglio-N-tetraosylceramide (asialo-GM1), to remove NK cells, was found to inhibit the generation of T-cell responses induced by allogeneic challenge or by viral infection 24, 25. These data were interpreted as evidence that NK cells are required for the generation of T-cell responses in vivo. However, since expression of asialo-GM1 is not restricted to NK cells, these findings cannot be considered as definitive proof for the role of NK cells in immune regulation. Subsequent depletion studies have utilised the PK136 anti-NK1.1 antibody to more precisely examine the role of NK cells in the regulation of immune responses. In vivo depletion of NK cells using the PK136 antibody was found to inhibit the generation of anti-tumour and anti-viral CTL 26, 27. Although it is likely that loss of NK cells was responsible for the effects observed in these reports, a role for other cells expressing the NK1.1 determinant, such as NKT cells, cannot be excluded. Using recently characterised NKp46 reagents, including antibodies and mice lacking Nkp46 28, it may be possible to more precisely investigate the impact of NK cells on immune responses.
In addition to promoting T-cell responses, under some circumstances NK cells inhibit adaptive immunity. For example, NK cells have been reported to inhibit the severity of disease in models of EAE. Mice or rats depleted of NK cells were found to develop a more severe form of EAE 29–31. Since EAE is primarily mediated by CD4+ T cells, NK cells have been presumed to inhibit the generation of autoreactive T cells in this setting. In a setting of viral infection, depletion of NK cells after murine cytomegalovirus (MCMV) infection was found to result in enhanced proliferation of CD8+ T cells and increased production of IFN-γ by both CD4+ and CD8+ T cells 32. Similarly, depletion of NK cells was shown to enhance the CTL response to MHC class I positive lymphomas 33. Thus, various reports implicate NK cells as having a regulatory role, but little is known as to the mechanism(s) through which NK cells influence adaptive immunity.
NK-cell-dependent regulation of DC function
DC are antigen-presenting cells that initiate and regulate immune responses. In humans, both immature and mature human DC can induce resting NK-cell activation 34. Indeed, numerous studies have found that DC are capable of activating NK cells (reviewed in 35). However, DC–NK cell interactions are not a one-sided affair, but rather, involve reciprocal interactions whereby NK cells can influence the function of DC and vice versa. That cross-talk between NK cells and DC is required for the generation of an appropriate immune response is implied by the findings that NK cells are found in close association with DC in both the lymph node and in inflamed skin 17, 19, 36, 37. Furthermore, depletion of NK cells has been found to affect both the number and activation state of DC in the lymph nodes 17, 38. Studies have attempted to define how activated NK cells influence the function of DC. The ability of activated NK cells to lyse immature DC (iDC) has been documented in a number of settings 34, 39–44. NK-cell-mediated killing of iDC is proposed to function as an editing mechanism. This theory suggests that only mature DC that have appropriate levels of MHC and co-stimulatory molecules, and are thus able to prime an effective immune response, will survive an encounter with an activated NK cell. The ability of NK cells to kill iDC, at least in vitro, is limited to a subset of cells expressing CD94/NKG2A, but lacking killer Ig-like receptors 45. Therefore, killing of iDC may not be mediated by all NK cells, but is potentially limited to a specialised subset of NK cells. The mechanism by which NK cells eliminate iDC is another important issue that still requires clarification. Killing of DC in a transplantation model 17 and in vitro42 is dependent on perforin. By contrast, in vivo, adoptively transferred iDC are eliminated by NK cells in a TRAIL-dependent manner 44. Thus, while the hypothesis that NK cells fine-tune immune responses by eliminating iDC is intriguing, definitive evidence that this process operates in vivo and how is still lacking.
NK cells regulate T-cell priming
Early studies revealed that NK cells can promote the generation of TH1 responses 46–48. In mice, NK cells are rapidly recruited to lymph nodes following Leishmania major infection and are a source of the IFN-γ required for the induction of TH1 polarisation 21. A similar effect has been observed in humans, where NK-cell-derived IFN-γ was found to enhance the activation of CD4+ T cells 49. Importantly, human tonsilar, but not peripheral NK cells were required for the expansion of IFN-γ producing CD4+ T cells 49. The specificity observed here is due to the fact that the cytokine producing CD56hiCD16− NK-cell subset is enriched in secondary lymphoid organs, such as the tonsils. These results reinforce the notion that NK cells are not homogeneous, and that the nature of the NK-cell subset involved can profoundly influence the outcome of an immune response. The pro-inflammatory cytokines produced by NK cells might promote a TH1 response via a number of mechanisms. Naïve T cells require an exogenous source of IFN-γ for TH1 polarisation, which can be produced by NK cells in vivo21. In addition, NK cells may indirectly promote TH1 polarisation by enhancing the maturation of DC. In vitro, NK-cell-mediated maturation of DC requires cell–cell contact and the production of TNF-α and IFN-γ by NK cells 43, 50, 51. The ability of NK cells to activate DC may also be essential for the initiation of immune responses to tumours or pathogens that do not directly activate DC. Some support for this theory comes from the observation that recognition of MHC class I low tumour cells by NK cells activates DC resulting in the induction of a CD8+ T-cell response 52.
In human and mouse bone marrow transplantation systems, donor NK cells have been shown to play a protective role in graft outcome by killing the allogeneic recipient antigen-presenting cells responsible for priming alloreactive T cells and initiating GVH disease 53. In an allogeneic cardiac graft model, long-term graft survival was achieved by inhibiting NK cells in a setting where CD28 co-stimulation was lacking (CD28−/− mice) 54. Interestingly, neither intervention alone was sufficient to improve graft survival. These findings led to the suggestion that NK cells might deliver help to T cells. Thus, after infiltrating the grafts, NK cells synthesise cytokines that circumvent the CD28 deficiency and provide the critical help required for CD8 T-cell priming. Since these studies were conducted using the anti-NK1.1 antibody to remove NK cells, the possibility that the observed effects are mediated by other cells carrying this determinant, particularly NKT cells, needs to be taken in consideration.
In skin graft models, TH2 polarisation can be achieved by the numbers of DC that accumulate in the absence of NK-cell activation. In contrast, the regulation of donor DC by blood-borne NK cells recruited in the lymph nodes has been shown to favour a TH1 response 17, 55. Thus, in autologous systems, the DC maturation state is crucial in determining whether the DC will survive the encounter with NK cells, while TH1 polarisation appears to depend mainly on cytokine production. On the other hand, in transplantation settings, NK-cell activation following interaction with allogeneic DC seems to occur regardless of DC maturation. This is likely due to failure to engage inhibitory NK-cell receptors specific for self-MHC I by the allogeneic DC. The duration of DC persistence will then control the strength of the priming and the subsequent polarisation of the T-cell response.
The impact of NK cells on DC functionality during immune responses has been largely inferred from in vitro studies. Perhaps, the best evidence that NK cells influence the function of DC in vivo has come from studying MCMV infection. Resistance to MCMV in C57BL/6 mice is mediated by Ly49H+ NK cells that recognise the virally encoded m157 protein 56, 57. During MCMV infection maintenance of the CD8α DC population is dependent on Ly49H+ NK cells 58. A recent report has suggested that the ability of Ly49H+ NK cells to maintain splenic DC populations is mediated by an indirect mechanism. The report proposes that the Ly49H+ NK-cell-mediated early control of MCMV replication in the spleen of resistant mice prevents the release of immunosuppressive levels of IFN-αβ 59. Administration of exogenous IFN-α to resistant mice was found to induce loss of DC from the spleen, and a slight and very transient delay in the activation of antigen-specific T cells 59. Alternatively, it has been proposed that the rapid control of viral replication by Ly49H+ NK cells may promote the maintenance of splenic DC by preventing the destruction of the splenic architecture 60.
In addition to potentially influencing the function of DC, NK cells have recently been shown to induce the differentiation of CD14+ monocytes into DC 61. This process was found to require the production of GM-CSF by CD56bright NK cells and direct cell–cell contact. While the process was proposed to contribute to the maintenance of chronic inflammatory diseases, it is conceivable that it could also operate to expand the pool of DC during immune responses to pathogens and thereby impact on the outcome of subsequent T-cell responses.
Together the published data provide evidence that NK cells can indirectly influence DC-induced T-cell priming, however, evidence that NK cells directly influence the functions of DC in vivo remains elusive.
Regulation of effector cells by NK cells
In addition to their potential role in regulating antigen presentation, NK cells may influence the outcome of the immune response by acting directly on effector cells. As mentioned previously, activation of naïve T cells is dependent on IFN-γ produced by NK cells 21. NK cells have also been reported to stimulate autologous CD4+ T cells, an effect that is dependent on the expression of OX40 ligand and CD86 by activated NK cells 62, 63. A role for NK cells in the activation of B cells and the promotion of isotype class switching has also been noted 64–66.
The ability of NK cells to restrain the immune response has also been observed in a number of settings. NKG2D-dependent killing of activated T cells by syngeneic NK cells has been reported 67. Furthermore, expression of Qa-1–Qdm by activated CD4+ T cells is required to prevent lysis by NKG2A+ NK cells 68. An implication of these results is that NK cells may be crucial for the termination of an immune response and consequently prevent the development of immunopathology. Direct evidence for this proposition comes from studies of mice deficient in either perforin or granzymes. Replication of MCMV is enhanced in mice deficient in either perforin or granzymes AB 69. However, granzyme AB-deficient mice survive infection while perforin-deficient mice develop a fatal haemophagocytic lymphohistiocytosis-like syndrome 69. The haemophagocytic lymphohistiocytosis-like syndrome observed in perforin-deficient mice was induced by the accumulation of TNF-α producing CD11b+F4/80+ mononuclear cells and T cells 69. In wild-type mice NK cells were found to prevent immunopathology by eliminating the TNF-α producing cells in a perforin-dependant manner.
A protective role of NK cells has also been reported in autoimmune diseases. In Fas-deficient mice, NK cells can suppress autoreactive B lymphocytes, while NK-cell depletion increases the severity of an autoimmune disease with features similar to those of systemic lupus erythematosus 70. NK cells have also been shown to play a protective role in diabetes; treatment of NOD mice with CFA prevented the disease in an NK-cell-dependent manner 71.
Collectively, the available data indicate that NK cells serve a dual purpose in that they can provide help and promote the initiation of an immune response, but can also curb the activity of immune effectors and thereby prevent immune-mediated damage to the host.
As discussed above, various studies have unveiled the regulatory functions of NK cells, but a crucial issue yet to be addressed is by what mechanisms NK cells influence the outcome of immune responses. The involvement of Treg cells is an intriguing possibility that has been discussed by Zimmer and colleagues in this NK viewpoint. On the other hand, the ability of NK cells to influence DC functions may be the key to how these cells regulate many immune responses. Owing to their powerful ability to induce immune responses, DC have been targeted in many immunotherapy protocols. Since NK cell–DC cross-talk clearly influences innate immune responses and can also impact on adaptive immunity, a better understanding of the mechanisms involved is critical and necessary if the ultimate aim is to develop protocols that will provide better immunity following vaccination, cancer immunotherapy and in transplantation settings.
Conflict of interest: The authors have declared no financial or commercial conflict of interest.