|Home | About | Journals | Submit | Contact Us | Français|
Natural killer (NK) cells mediate innate defense against viral infections, but the mechanisms in place to access their functions as needed during diverse challenges while limiting collateral damage are poorly understood. Recent molecular characterization of effects mediated through infection-induced inhibitory/activating NK receptor-ligand pairs and cytokines are providing new insights into pathways regulating their responses and revealing unexpected consequences for NK cell subset effects, maintenance, proliferation and function through times overlapping with adaptive and long-lived immunity. The observations define flexible pathways for experience-induced “conditioning” and challenge narrowly defined roles for NK cells and innate immunity as first responders with prescribed functions. They suggest that individual experiences as well as genes influence the innate immune resources available to fight off an infection.
Natural killer (NK) cells were originally described as lymphocytes of the innate immune system capable of eliminating tumor and infected cells without prior antigen exposure . This was quickly followed by the demonstration of the first NK cell activating receptor, the receptor for Fc portions of immunoglobulin molecules, CD16, and the ability of NK cells to interface with these components of adaptive immunity to mediate antibody-dependent cellular cytotoxicity (ADCC) . NK cells are now known to be proficient at many important tasks  but are still best appreciated for their non-redundant roles in defense against some viruses. Plentiful studies in the mouse and human demonstrate NK cell-dependent protective effects during infections with coxsackievirus, human immunodeficiency virus (HIV), hepatitis C virus (HCV), influenza virus, and poxvirus, but most importantly herpesviruses [3-5]. The mechanisms for delivery of their direct antiviral effects include killing of infected target cells and production of interferon γ (IFNγ), and absence of the cells prior to infection increases early replication of viruses sensitive to NK cell-mediated defense [5,6]. There is compelling evidence that their antiviral effects are regulated by a repertoire of germ-line encoded NK cell receptors (NKRs) recognizing ligands on virus-infected cells and by innate cytokine responses induced during infections. Despite extensive work in many laboratories, the molecular details of the pathways controlling NK cell responses have been elusive because of the many closely related inhibitory and activating receptors as well as the pleotropic and sometimes paradoxical effects of cytokines.
Exciting new molecular characterization of the NK cell receptor-ligand pairs and cytokine effects modified in response to viral infections are filling in significant gaps in knowledge. In addition to providing details on the pathways controlling NK cell contributions to defense and those protecting from potential immune-mediated pathology, the new information is revealing unexpected flexibility in NK cell responses based on experience-induced “conditioning”. In particular, much progress has been made in identifying infection-induced changes in the expression of ligands for NK cell activating and inhibitory receptors, in the role for activating receptors in expanding and sustaining NK cell subsets for extended periods, and in their responses to innate cytokines. The observations have implications for understanding the regulation of NK cell functions not only during early innate responses to viruses but also during periods overlapping with adaptive and long-term immunity. Moreover, they suggest that individual variations in resistance to infection are based on experiences as well as genes.
The NK cell receptors (NKRs) delivering inhibitory signals were first appreciated because of their ability to receive negative “self” signals from class 1 major histocompatibility (MHC1) molecules expressed on normal tissues [7,8]. It is now clear that there are complex groups of NKRs with different genetic and structural characteristics, representatives of both inhibitory and activating receptor in each group, and individual NK cells expressing combinations of both inhibitory and activating receptors (Fig. 1). In total, the NKRs are thought to provide NK cells with a system for surveying the extracellular environment such that the integration of resulting positive and negative signals determines the overall nature of their responses with a balance towards inhibition under normal conditions but a shift to activation under conditions of infection or stress [8,9].
The NKRs can be segregated into two main classes based on their distinct extracellular structure: the Ig-like and the C-type lectin-like receptors . The majority of Ig-like receptors, including the killer Ig-like receptors (KIRs) in humans, map to the leukocyte receptor cluster (LRC) present on human chromosome 19p and mouse chromosome 7. Families of C-type lectin-like receptor genes, including Ly49 in mouse and CD94, NKG2A/C/E, NKRP1 and NKG2D receptors in human and mouse, reside in a region known as the natural killer gene complex (NKC) on human chromosome 12p or mouse chromosome 6 . The NK cell repertoire, and to a certain extent the ability of NK cells to recognize specific target cells, stems from genetic polymorphism as well as different combinations and expression levels of NKRs. The KIR receptors in humans subserve the function of the murine Ly49 receptors. Both sets of receptors belong to highly polygenic and polymorphic families with different individuals or strains inheriting genetic variation in the individual genes as well as different numbers of genes. These receptors are expressed on overlapping subsets of NK cells in a stochastic fashion [12,13]. By contrast, CD94 and NKG2 genes have limited polymorphism and are orthologous genes in mouse and human [14,15]. Essentially all NK cells express NKG2D whereas CD94/NKG2A and CD94/NKG2C receptors are expressed on overlapping subsets of NK cells. Unlike the KIR and Ly49 receptors that are stably maintained once expressed, CD94/NKG2 receptors can be modulated by cytokines in the environment . With few exceptions, NKRP1 genes are conserved and expressed in the majority of NK cells .
Inhibitory KIR and Ly49 NKRs have been more extensively characterized than their activating counterparts, especially with regard to ligand-binding specificity and function. These receptors individually possess a high binding affinity for particular polymorphic MHC1 molecules (HLA-A/B/C in humans; H2-D/K/L in mice). The CD94/NKG2A heterodimers are also inhibitory and bind to non-classical MHC1 molecules (human HLA-E and mouse Qa1) . Inhibitory receptors signal via immunoreceptor tyrosine-based inhibition motifs (ITIMs) located in their cytoplasmic tails. Inhibitory signals are predominant under normal conditions, when the majority of target cells express MHC1 molecules. When target cells are transformed or infected, surface expression of MHCI molecules can be abrogated; the absence of inhibitory NKR ligation allows the NK cell to kill its target (“missing-self” recognition).
In addition to a requirement for release from dominant inhibitory signaling, delivery of effective NK cell killing requires the ligation of activating NKRs. Activating NKR lack ITIMs, but they associate with membrane-bound adaptor molecules, such as DAP12 (Ly49, KIR, mouse NKG2D, NKG2C/E), FcεRγ, and CD3ζ (CD16, NKp46), that bear immunoreceptor tyrosine-based activation motifs (ITAMs) [8,13,17]. The phosphorylation of ITAM leads to the recruitment of the tyrosine kinases, Zap70 and Syk. In addition, human NKG2D, and a mouse NKG2D length variant, associate with DAP10. The signaling motif in DAP10 is a YINM sequence, with phosphorylation leading to the binding of PI3K and Grb2 for signaling . In the human, a KIR2DL4 receptor can have both inhibitory and stimulating activity. KIR2DL4 has a cytoplasmic tail containing a single ITIM, and a charged amino acid in the transmembrane domain that interacts with FcεRγ . Some activating NKRs have no or attenuated affinity for MHC1 molecules. Others bind stress-induced molecules or non-self peptides thought to be associated with MHC1 molecules. Remarkably, the activating NKG2D receptors are evolutionary conserved and inherited by all individuals in a species. Likewise, activating CD94/NKG2C and NKp46 receptors are conserved and broadly inherited. Thus, in contrast to the diversity of receptors inherited for the Ly49 and KIR genes, the NKG2D, CD94/NKG2C, and NKp46 receptors provide opportunities for accessing responses in different individuals under different conditions of infections.
Some activating NKRs are able to directly recognize virus-infected cells. For example, Ly49H from C57BL/6 resistant mice was the first NKR demonstrated to be a viral resistance gene (cmv1) and mediates its effects by recognizing the mouse cytomegalovirus (MCMV) m157 viral protein, which is highly homologous to MHC1 and expressed only on MCMV-infected cells [20-23] (Fig. 2; Table 1). Ligation of Ly49H by m157 results in the release of cytolytic granules, cytokines, and chemokines, as well as a robust NK cell proliferation , and although only short of half of the NK cells in strains of mice having the Ly49h gene basally express the molecule, the proportions are dramatically increased during the infection . Investigation of this activating receptor-ligand pair has made a number of key discoveries concerning the role for activating receptors in accessing NK cell responses to viral infection possible, but combinations of activating receptors stimulated directly by viral products may be rare because of their negative effects on viral replication and the ability of viruses to be selected with modified genetic material under immune pressure. Another example might be the influenza hemagglutinin (HA) glycoprotein binding to NKp46 .
Recent studies have extended the characterization of stimulatory receptor-ligand pairs during MCMV infections to activating NKRs recognizing changes induced by infection and restricted by MHC1 molecules. NK cells have a protective effect in MCMV-infected mice of the MA/My strain, which lacks Ly49H. Under these conditions, resistance is dependent on both the H-2Dk and the NKC loci [27,28], with an activating Ly49P receptor recognizing MCMV-infected cells and in vivo defense depending on the presence of both the viral m04/gp34 protein and H2-Dk [27,29**]. In H2-Dk transgeneic mice, MHC1 controls early viral replication and robust NK cell proliferation; however, the specific activating NKR responsible for these effects has not yet been identified [30*,31*]. Other activating Ly49 receptors recognizing MCMV-infected cells in vitro in an m04-specific and H2-dependent manner include the Ly49P1 receptor from NOD/Ltj mice, Ly49D2 from PWK/Pas mice, and Ly49L from BALB/c mice [32**]. Of these, only Ly49LBALB/c has been validated in vivo with infection of BALB mice having both NK cell and H2k-dependent control of viral spread as compared to H2d or H2b animals. Here, the antiviral effects correlate with expansion of Ly49L+ NK cells and IFNγ secretion, both of which are abrogated during infection with Δm04 MCMV. Thus, there is a developing picture of different combinations of the highly polygeneic and polymorphic Ly49 activating receptors in the mouse with resistance to MCMV depending on the presence of particular polymorphic MHC1 molecules and viral products. The interactions between the viral products and/or their peptides with MHC1 remain to be defined, but the observations fit a model of particular MHC1 molecules presenting peptides to particular activating receptors. They suggest that the interactions may have provided important evolutionary pressure for the diversification of these receptor-ligand combinations.
Another family of receptors, NKG2, has recently been implicated in resistance to ectromelia virus (mouse pox). The CD94/NKG2E heterodimer specifically triggers NK cells for activation via the recognition of the non-classical class 1 molecule, Qa-1b on virus-infected cells, and the combination results in improved survival. Again, whether the virus modifies Qa-1b expression or provides a peptide for Qa-1b presentation remains unknown, but the combination in this case is between a conserved activating receptor and an inherited polymorphic host molecule influenced by the infection [33**]. Another conserved activating receptor, NKG2D, is broadly expressed on all NK cells. It recognizes multiple stress ligands (MICA-C and ULBP1-6 in humans; Raeα-ε, H60a-c, and MULT1 in mice) that are induced during CMV infection . The importance of these conserved receptor-ligand pairs is validated by the demonstration that both mouse and human CMV have mechanisms for blocking the expression of NKG2D ligands on the cell surface to escape NK cell detection . In short, current data indicate that numerous activating receptors allow NK cells to recognize MCMV infection either directly (Ly49H:m157) or indirectly (Ly49P/L:m04:H2-Dk; NKG2D:stress molecules).
Like their activating counterparts, inhibitory NKRs can recognize infected cells in numerous ways. The viral product m157 is recognized by an inhibitory Ly49 receptor . Several others are triggered (directly or indirectly) in the presence of m04/gp34. Most susceptible strains, including BALB, show improved viral control when infected with Δm04 MCMV [36**]. One reason for this could be the role of m04/gp34 in preventing “missing-self” recognition by NK cells because the protein is known to escort MHC1 molecules to the surface of infected cells to maintain a basal level of expression throughout the infection. This basal level of MHC1 expression is sufficient to trigger several inhibitory receptors (e.g., Ly49A). During infections with Δm04 MCMV, MHC1 ligands are depleted and viral titers significantly decrease. Conversely, the presence of m04/gp34 during MCMV infection impairs NK cell in vivo expansion and ex vivo killing of cognate target cells as compared to Δm04 MCMV infection. Effects of the inhibitory Ly49C receptor can also be observed during MCMV infections. Ly49H+ NK cells bearing Ly49C have decreased IFNγ secretion and proliferation compared to Ly49H+Ly49C- NK cells . In studies of the H2q allele impact on MCMV resistance in H-2Dk transgenic mice, H-2Dk animals are only marginally more resistant than their non-transgenic counterparts in the presence of H2q. Indeed, a gene dosage effect is observed on the inhibitory action of H2q . This suggests that inhibitory NKR triggering by H2q dominates over activating NKR triggering by H2-Dk on MCMV-infected cells. However, it is also possible that inhibitory receptors benefit proper NK cell responses. For example, in the context of MA/My resistance Ly49G+ NK cells are found to preferentially expand and secrete IFNγ upon MCMV infection, while their depletion increases viral titers. It remains to be clarified if the improved NK cell antiviral response of Ly49G+ NK cells results from MHC1-dependent education [30,38] or from MHC1-independent mechanisms at play during infection .
Human viral infections are infrequently studied in the acute phase due to logistical difficulties in obtaining samples. However, there are some informative studies. Peripheral blood NK cell numbers are reduced in the acute phase of influenza H1N1 infection and in acute hepatitis C virus (HCV) infection the numbers of cytotoxic CD56dim NK cells are also reduced [40,41*]. Conversely, in acute Puumala hantavirus infection, NK cell numbers are elevated up to two-fold [42**]. The expansion is associated with increased proliferation, increased plasma levels of IL-15, and increased expression of NKG2C. Hantavirus infected endothelial cells upregulate HLA-E, and both IL-15 and HLA-E are required to drive NK cell proliferation. The increases in NK cells are long-lived and more prominent in CMV-seropositive individuals. Because CMV seropositivity is associated with increased levels of NKG2C-positive NK cells , CMV may “prime” NKG2C NK cells to respond to hantavirus. The expansion may occur by upregulating HLA-E through expression of an HLA-E binding peptide , and/or an as of yet unidentified virus-induced peptide [42**]. Upregulation of NKG2C has also been observed in chronic hepatitis B virus (HBV) and HCV infections [45,46*]. This may therefore be a common theme of viral imprinting on the NK cell repertoire through NKG2C activation.
Human NK responses can be driven by either conserved activating receptors such as NKG2C or by the polymorphic KIR system. The activating KIR3DS1 alone or in combination with its HLA-Bw480I ligands is associated with a beneficial outcome of HIV infection [47-49]. This immunogenetic observation is supported by in vitro assays of KIR3DS1-positive NK cell effects for limiting viral replication in HIV-infected HLA-Bw480I-positive targets . Taken together, the studies in the human indicate that activating receptor-ligand pairs provide protection against acute infections, stimulate increases in NK cell numbers or particular subsets, and result in long-term benefit to the host. The extended periods of increases in NK cells expressing the conserved NKG2C receptor following different viral infections suggest that this receptor might provide protection to most members of a species if its ligands are induced. Conversely, because the KIR activating receptors are polymorphic and polygenic and the MHC1 molecules are polygenic, any requirement for these receptor-ligand pairs would allow opportunities for viruses to pass through “holes” in the genetic repertoire of subgroups in a diverse population while others mount effective defense.
NK cell activation occurs when target cells upregulate ligands for activating receptors that are present. In addition to those discussed above, the list of human NK activating receptors associated with recognition of viruses is growing, and candidates are in place for influenza virus [51,52], Ebola , herpes simplex , Epstein Barr , vaccina virus , and Kaposi's sarcoma-associated viruses (KSV) . NK cell activation, however, also may occur when constitutive inhibitory signals are lost, such as by MHC1 down-regulation on infected cells [58,59]. An important role for inhibitory receptors has been shown in HIV infection in which an allelic hierarchy of KIR3DL1:HLA-Bw4 interactions can determine the rate of HIV disease progression [60,61]. Similarly in HCV infection, KIR2DL3 and its group 1 HLA-C (HLA-C1) ligands are protective against chronic infection and are associated with a beneficial response to IFNα based therapies [62-65]. The KIR2DL3:HLA-C1 combination is a relatively weak inhibitory interaction when compared to other inhibitory KIR:HLA assortments. Critically, it has been shown that HLA class 1 down-regulation is not necessary to modulate this inhibitory interaction, and that changing the peptide content of the HLA molecule is a more efficient mechanism of removing an inhibitory signal on an NK cell [66**] (Fig. 3). This may be important for viral infections without specific mechanisms to downregulate MHC1, i.e. “missing self”, and likened to sensing “altered-self”. The DRiP hypothesis suggests that peptides loaded onto MHC1 are in a large part derived from unfolded recently synthesized polypeptides or defective ribosomal products “DRiPs” . Thus, viral infection may change the MHC1 peptide repertoire more rapidly than would be anticipated from degradation of mature viral proteins, thereby permitting a rapid NK cell response. This may be important for HIV infection, as a peptide variant that escapes T cell recognition also abrogates KIR3DL1 binding, and could thus lead to NK cell activation . Furthermore HIV can adapt to the NK cell response by mutating to increase the inhibition of KIR2DL2+ NK cells [69**]. Hence, resistance might be linked to an inhibitory receptor, and this would require particular combinations of inhibitory receptors and MHC1 molecules.
Synergy between innate receptor mediated and cytokine responses may be an important factor in the ability of the host to eradicate viral infections. With the exception of CD16, activating NK cell receptors function as co-receptors and require stimulation of at least two receptors or one receptor plus cytokine stimulation [70*,71]. Moreover, cytokines can influence expression of target cell ligands to regulate NK cell delivery of antiviral effects with interferons (IFNs) having profound effects on protein/peptide synthesis and MHC1 expression . In HCV infection, a SNP upstream of the IFNλ3 gene (IL-28B), a type III IFN, is strongly associated with clearance of infection [73**,74*,75*,76*,77*]. KIR2DS3 and IFNλ3 synergize on a genetic level to render individuals particularly susceptible to chronic HCV infection [78*]. However, KIR2DL3:HLA-C1 and this polymorphism do not provide synergistic protective effects . The paradox of HCV infection is that the protective IFNλ3 SNP and clearance of HCV infection are both associated with lower levels of interferon stimulated genes . This, combined with the observation that successful treatment of HCV infection in liver transplants assorts with the donor liver IFNλ3 genotype , implies that instead of synergizing at the level of the NK cell, these two genetic mechanisms act at discrete places during the innate immune responses; IFNλ3 acting at a direct intracellular level to inhibit virus replication within infected cells and KIR acting at the level of NK cell activation to enhance this arm of immune defense.
Early interest in NK receptors was driven by efforts to understand the regulation of innate NK cell-mediated killing of virus-infected target cells, but delivery of activating signals can also lead to proliferation. Because NK cells are relatively abundant in many tissues, and high proportions basally express or are induced by cytokines to express different activating receptors, the advantages of NK cell proliferative responses have remained obscure. Interestingly, however, NK cells numbers or subset frequencies can be decreased during particular conditions of viral infections. As noted above, reductions are observed during acute infections with influenza or HCV [40,41], but they have also been reported during symptomatic primary infections with Epstein Barr virus (EBV)  and during chronic infections with HCV and HIV [83-87]. Thus, NK cell loss can accompany viral infection. Given that NK cells have the potential to mediate a wide range of important biological effects [1,6], such loss can significantly alter the resources available for fighting off viruses. Interestingly, activating receptor-ligand pairs, depending on the repertoire of highly polygeneic and polymorphic KIR receptors and a particular MHC1 molecule inherited by an individual, are associated with recognition of EBV and HIV [50,55]. In the case of HIV, the combination is coupled with NK cell proliferation and increased cell yields  as well as with beneficial long-term outcomes, including protection against opportunistic infections [47-49]. The ability of activating receptors to stimulate proliferation suggests that they might have important biological consequences for maintaining NK cells during an infection as well as delivering direct antiviral effects.
Mouse studies examining the consequences of Ly49H deficiency as compared to the absence of cytotoxicity due to a deficiency in perforin 1 have separated the biological importance of an activating receptor's role in proliferation from any function it has in killing [89**]. The ligand for Ly49H, m157, is induced, and resistance to infection is elevated with NK cell proliferation stimulated during MCMV infections of mice expressing Ly49H. The absence of either Ly49H or perforin results in elevated levels of early MCMV replication. There are, however, profound differences in NK cell proliferation and yields, with the populations disappearing under Ly49H-deficient, but Ly49H+ NK cells dramatically expanding under perforin-deficient, conditions. The critical role for Ly49H in supporting proliferation in the absence of cytotoxicity is shown under double deficient conditions. These observations indicate that if an activating receptor is missing and/or if the virus does not induce a ligand for an activating receptor, NK cells would fail to limit early infection such that under the resulting circumstances of sustained viral replication, maintenance of and access to NK cells would not be sustained for extended periods of time (Fig. 4). By protecting from events paralleling those of “exhaustion” in T cell responses to viral infections , inhibitory receptors might help balance the magnitude of the proliferative response to obtain optimal numbers and frequencies of cells, but this putative role in maintaining NK cell subsets has yet to be examined. Thus, although there is much to be learned about controlling the magnitude of NK cell proliferation, the results to date demonstrate that signaling through activating receptors is important for sustaining the populations into later periods of infection.
What do NK cells do if they are maintained? In addition to mediating defense that can be strictly assigned to periods of innate immunity, NK cells help bridge innate and adaptive immunity in part by producing cytokines, particularly IFNγ, or mediating cytotoxicity to shape downstream adaptive responses. In the case of acute HCV infection of humans, there is NK cell activation of cytotoxic and IFNγ production that does not correlate well with control of viremia , but a detected degranulation does correlate with the magnitude of HCV-specific T cells . Surprisingly, the sustained Ly49H+ NK cells that proliferate into periods overlapping with adaptive immunity during MCMV infections of perforin-deficient mice make IL-10 to regulate the magnitude of adaptive immune responses and protect from CD8 T cell-mediated pathology . Increased proportions of NK cells with particular activating receptors producing IL-10 are also observed during chronic HCV infections in human . The regulatory effects of the NK cell IL-10 production can protect from T cell-mediated pathology but have also been suggested to interfere with protective T cell responses against the virus. Thus, a consequence of NK cell expansion and maintenance is providing access to the populations for mediating immunoregulatory as well as direct antiviral effects.
Finally, there are indications that NK cells activated in responses to viral infections and/or vaccination with viral ligands can become long-lived cells with the memory characteristics of specificity and heightened recall during secondary challenges [93**,94**,95]. The first and best studied of these long-lived NK cells are the Ly49H+ cells elicited in response to MCMV infection [93**]. Their importance as memory populations under immunocompetent conditions remains to be determined, as most of the work has relied heavily on immunodeficient conditions to demonstrate the trait, and as the fitness of the populations in the context of memory T cells has not been tested . Nevertheless, the demonstration of the induction of such long-lived populations through activating receptor-ligand pairs, may provide an explanation for the HCMV conditioning, with expansion of NKG2C NK cells, for enhanced resistance to hantavirus infections in humans [42**,43]. It also provides evidence for the influence of experience on an individual's innate immune resources.
The importance of innate cytokines in regulating NK cell functions was first realized with the demonstration that the antiviral type 1 IFNs, IFNαβ, were potent inducers of elevated NK cell-mediated lysis . Since that time, it has becomes clear that many viral infections induce complex innate pro-inflammatory cytokine cascades, including in addition to type 1 IFNs, biologically active IL-12p70, IL-18 and IL-15 . In terms of overall kinetics of production, cellular sources, and receptor-sensing for production, these responses have been best characterized following MCMV challenges of mice or mouse cells [98-104]. Not all viruses are the same, however, and infections of mice with lymphocytic choriomeningitis virus (LCMV), an agent inducing high systemic levels of type 1 IFN production for expended periods of time, elicit little to no detectable IL-12 [98,105-107]. In general innate cytokine responses are conserved in the human, but they differ substantially amongst extant human viral infections . In acute HIV infection, there is a rapid increase in IFNα and IL-15, with a more delayed increase in serum IL-18 and IL-12p70 [108*]. The cytokine response to HCV infection is delayed following infection and is attenuated further in response to HBV; as compared to hepatitis A virus (HAV), HBV stimulates very weak IFNα and IFNλ responses [109**]. Experimental infections in chimpanzees, with comparisons of HBV and HCV [110,111] or HAV and HCV , suggest that there may be a hierarchy of viruses in terms of their ability to result in “stealth” infections below detection for induction of type 1 IFN responses, with HBV and HAV being “stealthier” than HCV.
Interestingly, three of the innate cytokines commonly induced by viruses, type 1 IFNs, IL-12 and IL-15, bind to different receptors using subsets of Janus kinases (Jaks) and signal transducers and activators of transcription (STATs) to stimulate biological responses in cells [113,114]. There are seven total STAT molecules, 1 through 6, with two 5s. Once activated by phosphorylation, STAT complexes translocate to a cell's nucleus, bind to DNA regulatory sequences in promoters of genes, and modify transcriptional control of the target genes. Much of the early work defining cytokine-signaling pathways focused on identifying individual STATs preferentially used by the different cytokines to assign their roles in eliciting specific responses. The type 1 IFNs, IFNαβ, use STAT1 and 2 as a major pathway for induction of antiviral effects, but they have a wide range of other functions. Some of these are immunoregulatory and some are paradoxical . Their activation of STAT1 also stimulates elevated STAT1 expression [115,116] and promotes anti-proliferative effects [117,118] as well as elevated NK cell cytotoxicity and IL-15 expression . The type 1 IFNs appear to be unusually promiscuous in their choice of STATs and have been shown to conditionally active all STATs including STAT4 . STAT4 is well characterized as an important intermediary in IL-12 induction of IFNγ.
Mouse studies dissecting the roles of individual cytokines for eliciting NK cell responses to viral infections have shown that although the type 1 IFNs are key for inducing elevated NK cell cytotoxicity, and the cytokines induce IL-15 to support modest levels of NK cell proliferation at very early times after infection , biologically active IL-12 is required for a strong systemic NK cell IFNγ response [99,119] (Fig. 5A). The ability to assign individual cytokines to the early NK cell responses observed during viral infections of mice, i.e., type 1 IFN for cytotoxic function, IL-15 for proliferation, and IL-12 for IFNγ, is surprising given the overlapping effects that can be induced by each of these factors. In addition, it can be contrasted to the linked stimulation of all three responses through activating receptors , the role of activating receptor in supporting extended proliferation  and maintenance [89**], as well as the observed synergism between activating receptor stimulation for IL-15-dependent  and under immunodeficient conditions, IL-15-independent NK cell proliferation . The observations suggest that the NK cells themselves and/or the conditions of infections are limiting the responses that each of the cytokines can induce.
Studies evaluating the role of type 1 IFN signaling pathways in regulating NK cell responses have shown that in comparison to other immune cells from uninfected mice, NK cells basally express high levels of STAT4 and respond to type 1 IFN exposure with activation of STAT4 . In the absence of STAT1, type 1 IFNs elicit systemic production of IFNγ by NK cells through a STAT4-dependent pathway. The type 1 IFN use of the STAT1-dependent pathway to STAT1 induction, however, is important for protection against deregulated induction of IFNγ and a resulting cytokine-mediated disease. The effect occurs at the level of competition for access to the type 1 IFN receptor with its preference for STAT1 over STAT4. Thus, the basal expression of high STAT4 in NK cells allows access to STAT4 activation by type 1 IFNs immediately after exposure whereas STAT1 induction by type 1 IFN provides a built-in regulatory pathway to limit the availability of STAT4 and protect from immune-mediated pathology. Interestingly, type 1 IFN access to STAT4 is inversely correlated with total STAT1 protein levels in healthy donors as compared to those chronically infected with HCV [122**]. Thus, the conditioning of NK cell cytokine responses through regulation of STAT levels is likely to be common to both species. The results suggest that the strict requirement for IL-12 in NK cell IFN-γ responses is a consequence of needing a different cytokine to access STAT4 once STAT1 levels are increased.
Although systemic NK cell IFNγ responses require IL-12, type 1 IFNs can access STAT4 and induce NK cell IFNγ production at the first site of infection, i.e., the peritoneal cavity, and this contributes to promoting an antiviral state [123**]. The pathway appears to be available at this site because of tightly regulated type 1 IFN production over a narrow period of time and delayed induction of STAT1 in NK as compared to other cells in the compartment. The regulation of type 1 IFN access to STAT4 (Fig. 5B) is an example of how fine-tuning of NK cell responses is important for shaping their effects as needed during the changing conditions of infection and of how the immune system is equipped to do this. It is another example of how “experience” modifies innate immunity. In this case, the changes are happening under acute conditions during periods of innate responses. The longer-term consequences for NK cell “conditioning” have yet to be examined, but may play a role in the polarization of NK cells to cytotoxicity following IFNα-treated of individuals chronically infected with HCV  as well as the development of NK cell subsets in the human and mouse preferentially mediating cytotoxicity as compared to cytokine production [124,125].
In summary, although NK cells were first proposed to be key mediators of innate immune responses to viral infections, it is now clear that there are many mechanisms in place to dramatically regulate their functions at the very earliest times after infections as well as to support their continued availability through times overlapping with adaptive immunity. A profound flexibility in responding to virus-induced cytokines and changes in the expression of ligands for activating and inhibitory receptors equips previously unexposed individuals or subsets of individuals in a population with the ability to protect against a wide range of potential infections, but it also results in conditioning of the NK cell populations to reflect experience. In the case of the earliest pathways to activation of NK cell responses elicited by type 1 IFNs, there are predisposed responses in place to allow IFNγ production for early antiviral defense and possibly immuneregulation but rapid induction of molecules limiting the response to protect from continued IFNγ production and resulting cytokine-mediated disease. These changes are important acutely. The long-term consequences of this “conditioning” by experience remain to be determined but may have roles in the development of NK cell subsets preferentially mediating particular functions. Alternative pathways are in place to access certain overlapping responses, including the use of different innate cytokines and/or ligands for activating receptors. In contrast, the pathways to NK cell-mediated effects delivered via the activating receptors not only make killing of virus-infected target cells possible, but they also help maintain NK cells through periods of extended viral infections and change the proportions of NK cells expressing particular activating receptors. This “conditioning” has significant consequences for the resources available to fight off an infection and responses to subsequent primary viral infections. Taken together, the results force the consideration of NK cells in particular and the innate immune system in general as being influenced both stably though genetic variations and dynamically by experience.
Here we review recent data contributing to our understanding of how NK cells mediate innate defense against viral infections. The molecular mechanisms mediated through infection-induced inhibitory/activating NK receptor-ligand pairs and cytokines as well as their effects onto NK cell subset function, maintenance and proliferation through times overlapping with adaptive and long-lived immunity are highlighted.
The authors apologize to their colleagues whose work was not cited due to space limitations. Research in the authors' laboratories is funded by the Canadian Institutes of Health Research, Canada, The Wellcome Trust, UK, and the National Institutes of Health, USA. They thank their many current and past laboratory members for their contributions to the work reviewed here.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Papers of particular interest, published within the period of review, have been highlighted as:
* of special interest
** of outstanding interest