Natural Killer Cell Education and Tolerance

Abstract

Natural killer (NK) cells play a key role in the immune response to certain infections and malignancies by direct cytolysis of infected or transformed cells and by secretion of potent immune mediators. NK cells express an array of activating receptors that recognize self-molecules. If not restrained by inhibitory receptors recognizing major histocompatibility complex (MHC) class I proteins on the surface of self cells, NK cells are able to kill normal healthy cells. Not all NK cells express inhibitory receptors for self-MHC class I; thus, other tolerance mechanisms are necessary to prevent NK cell-mediated autoimmunity. Here we review the major mechanisms of NK cell education and tolerance.

Natural killer (NK) cells are the third population of lymphoid cells that unlike T and B cells do not express receptors that require somatic gene rearrangements to generate receptor diversity and specificity. Instead, NK cell functions are controlled by a wide array of germline-encoded inhibitory and activating receptors, many of which are expressed in a stochastic, variegated pattern, resulting in many subsets of functionally distinct NK cells. Although normally classified as part of the innate immune system due to the lack of receptor gene rearrangement, it has recently been appreciated that NK cells exhibit many features normally associated with adaptive immunity. These include the expansion of pathogen-specific cells, the generation of long-lasting “memory” cells that persist after cognate antigen encounter, and the ability to mount an enhanced secondary recall responses to re-challenge (Sun et al., 2009). Although NK cell-associated receptors have been implicated in certain autoimmune diseases (Gur et al., 2010; Martin et al., 2002; Namekawa et al., 2000; Yen et al., 2001), and NK cells may alter T cell-mediated autoimmunity (Feuerer et al., 2009; Lu et al., 2007; Poirot et al., 2004; Shi et al., 2000), as yet there is no evidence that NK cells alone are able to directly cause autoimmunity. Even on autoimmune-prone genetic backgrounds such as NOD, autoimmune disease is not observed in these mouse strains in the absence of B cells and T cells. Thus, NK cell education and tolerance appear to be more effective than either the central or peripheral tolerance mechanisms governing T cells and B cells that can fail, causing autoimmunity.

Many of the activating or co-activating NK cell receptors expressed by mouse and human NK cells recognize self-antigens (Table 1). For example, the activating NKG2D receptor recognizes numerous self-ligands in the host. Several of the activating Killer-cell Immunoglobulin-like Receptor (KIR), Ly49, and CD94-NKG2C receptors are capable of recognizing self-major histocompatibility complex (MHC) class I proteins, and members of the “natural cytotoxicity receptors” group (such as NKp30, NKp44, and NKp46) appear able to bind as yet undefined self-ligands in the host (Lanier, 2008; Moretta et al., 2001). In addition, several “co-activating” receptors have been identified on NK cells, such as CD2, LFA-1, CD244 (2B4), and CD226 (DNAM-1), which also recognize self-ligands that are broadly distributed on many tissues in the host (Lanier, 2008). Thus, NK cells express many activating receptors for self that could potentially drive auto-reactivity.

Table 1.

Activating and Co-activating NK Cell Receptors

Gene	Common name	Species	Ligand
Klra4	Ly49D	Mouse	H-2Da Hamster MHC I
Klra8	Ly49H	Mouse	MCMV-m157
Klra16	Ly49P	Mouse	MCMV
Klrb1c	NK1.1, NKR-P1C	Mouse	Unknown
Klrb1f	NKR-P1F	Mouse	Clrg
Pilrb1	PILRβ	Mouse	O-glycosylation CD99
CD244	2B4	Human	CD48
ITGAL	LFA-1, CD11a	Mouse, Human	ICAM-1, 2, 3
KLRD1-KLRC2, 3	CD94-NKG2C, E	Mouse, Human	Mouse Qa-1b Human HLA-E
KLRK1	NKG2D	Mouse, Human	Mouse Rae-1,H60, MULT1 Human MICA,MICB, ULBP1-6
CD2	LFA-2	Human	CD58
KLRF2	NKp65	Human	KACL
KLRF1	NKp80	Human	AICL
NCR1	NKp46	Mouse, Human	hemagglutinin
NCR2	NKp44	Human	Unknown
NKC3	NKp30	Human	B7-H6
KIR2DS	CD158	Human	HLA class I
KIR3DS	CD158	Human	HLA class I
FCGR3	CD16	Mouse, Human	IgG
CD226	DNAM-1	Mouse, Human	CD112, CD155
SLAMF7	CRACC	Human	CRACC
SLAMF6	NTB-A	Human	NTB-A

NK Cell Receptors for MHC Class I

Although NK cells express a wide variety of inhibitory receptors that recognize diverse self-molecules to prevent auto-aggression, inhibitory NK receptors recognizing self-MHC class I are considered to be the predominant mechanism responsible for NK cell tolerance to self. This is most dramatically demonstrated by the ability of NK cells in wildtype, healthy mice to rapidly reject splenocytes or bone marrow from syngeneic mice lacking the MHC component β2-microglobulin (β2m, an MHC component), MHC class I heavy chains, or TAP-1 (Bix et al., 1991; Dorfman et al., 1997; Liao et al., 1991; Ljunggren et al., 1994). This finding implies that healthy hematopoietic cells from wildtype mice constitutively express ligands for activating NK receptors that can initiate a strong NK cell response, if not restrained by the expression of MHC class I on healthy lymphoid or myeloid cells (Dong et al., 2009).

Three families of NK receptors recognize MHC class I: the primate KIR, the murine Ly49 receptors, and the CD94-NKG2 receptors in both rodents and primates. All three gene families encode both inhibitory and activating family members (Long, 2008). The activating receptors in the human KIR and murine Ly49 families arose by gene duplication and conversion from inhibitory receptors (Abi-Rached and Parham, 2005), such that the extracellular domains of these activating receptors are highly homologous to their inhibitory counterparts (Figure 1A). Therefore, several of the activating KIR and Ly49 receptors have the ability to bind with low affinity to MHC class I ligands.

Figure 1. Evolutionary Origins of Non-autoreactive Activating Receptors.

(A) New activating receptors for natural killer cells are generated by duplication of an inhibitory receptor gene and the subsequent conversion to an activating receptor. This conversion occurs when the codons for the transmembrane (TM) and intracellular domains are exchanged with those of an activating receptor by crossover events. If the new activating receptor is specific for self-major histocompatibility complex (MHC) class I, it might cause autoimmunity and be selected against. (B) Inhibitory receptors that are specific for both MHC class I and pathogen-encoded ligands have also been converted to activating receptors. Subsequent mutations have eliminated the specificity for MHC class I and retained specificity for the pathogen-encoded ligand.

All of the inhibitory receptors in humans and rodents contain one or more intracellular immunoreceptor tyrosine-based inhibitory motifs (ITIM), which are characterized by the signature sequence, (I/L/V/S)xYxx (L/V) (where x represents any amino acid, and slashes separate alternative amino acids that may occupy a given position). Many of the activating receptors lack intracellular signaling motifs but instead non-covalently associate via a charged residue in the receptor transmembrane domain with the immunoreceptor tyrosine-based activating motif (ITAM)-containing adapters DAP12, FcεRIγ or CD3ζ, which recruit and activate the Syk or ZAP70 tyrosine kinases, or in some cases the DAP10 adapter, which induces PI3-kinase signaling (Figure 1A). With a few exceptions, human KIR contain either two (KIR2D) or three (KIR3D) immunoglobulin (Ig)-like domains in the extracellular domain, and are designated as KIR2DL or KIR3DL, respectively, if they possess a long cytoplasmic domain containing ITIM. KIR2DS or KIR3DS have short cytoplasmic domains lacking ITIM, but possess charged residues in their transmembrane region to allow for association with DAP12 or FcεRIγ. KIR2D receptors typically recognize human leukocyte antigen -C (HLA-C) alleles, whereas KIR3D receptors recognize HLA-B, or some HLA-A alleles. The NKG2 family contains one inhibitory family member, NKG2A, and two activating members, NKG2C and NKG2E. The inhibitory CD94-NKG2A and the activating CD94-NKG2C receptors both recognize MHC class Ib molecules, either HLA-E in humans or its ortholog Qa-1 in mice (Lanier, 2005). In humans, a subset of NK cells express another KIR-related inhibitory receptor, LILRB1 (also designated LIR-1, CD85j, or ILT2), which recognizes a shared epitope present in all human MHC class I proteins (Chapman et al., 1999).

Chronic Exposure to Activating Ligands Tolerizes NK Cells

Given the existence of numerous activating receptors that can interact with host-encoded ligands, how do NK cells avoid causing autoimmunity? In mice, the Rae-1 family of NKG2D ligands is constitutively expressed in the embryo, but is silenced before birth and is not expressed or is expressed in only low amounts in healthy adult tissues. Immunocompetent adult mice reject transplanted tumors expressing NKG2D ligands and eliminate virus-infected cells in which NKG2D ligands are induced (Cerwenka et al., 2001; Champsaur and Lanier, 2010; Diefenbach et al., 2001; Lodoen et al., 2003; Lodoen et al., 2004). NK cells efficiently reject adoptively transferred syngeneic hematopoietic cells from mice constitutively expressing a Raet1 transgene, encoding a ligand for the activating NKG2D receptor. These findings indicate that NK cells have the capacity to recognize and attack otherwise normal, healthy cells if they express NKG2D ligands (Ogasawara et al., 2005; Oppenheim et al., 2005). In contrast, NK cells in Rae-1 transgenic mice do not attack Rae-1-bearing tissues and are unable to reject transplanted tumors expressing NKG2D ligands (Champsaur and Lanier, 2010). Thus, when NK cells develop in the continuous presence of NKG2D ligands they tolerate cells expressing these ligands, both due to the down-regulation of the NKG2D receptor on the NK cells as well as hypo-responsiveness to NKG2D signaling (Champsaur and Lanier, 2010). Studies examining NK cells from transgenic mice constitutively expressing Rae-1 have found that activating NK receptors other than NKG2D may or may not be impaired depending on the particular ligand expressed, its distribution, and its abundance (Champsaur et al., 2010; Ogasawara et al., 2005; Oppenheim et al., 2005; Wiemann et al., 2005). In vitro co-culture of mouse NK cells with NKG2D ligand-bearing tumors can also desensitize other activating receptors on the NK cells (Coudert et al., 2008).

The activating Ly49D receptor recognizes H-2D^d, and NK cells from mice lacking H-2D^d are able to kill H-2D^d-bearing targets via Ly49D. Conversely, NK cells from H-2D^d-expressing mice are unable to kill H-2D^d target cells via Ly49D, demonstrating that NK cells with this self-reactive activating receptor are functionally tolerant when the NK cells develop in the presence of the ligand (George et al., 1999). Although the molecular mechanisms of this tolerance or receptor anergy are undefined, it has been suggested that it is mediated by the co-expression of the H-2D^d-reactive inhibitory receptors Ly49A and Ly49G2 (George et al., 1999). Similarly, there is no direct evidence that human NK cells expressing activating KIR with the ability to recognize self-HLA-C ligands cause autoimmunity in these individuals, although a genetic association between activating KIR and certain autoimmune diseases has been reported, but might be caused by expression of activating KIR on T cells (Kulkarni et al., 2008).

Another activating Ly49 receptor, Ly49H, recognizes a mouse cytomegalovirus (MCMV)-encoded glycoprotein, m157, but does not recognize host MHC class I ligands (Arase et al., 2002; Smith et al., 2002). MCMV m157 also binds the inhibitory Ly49I receptor and thus likely evolved to inhibit NK cell responses to MCMV infection. The gene encoding Ly49H arose from the duplication of the gene encoding Ly49I, and conversion to an activating receptor by swapping the ITIM-bearing domain for a charged transmembrane domain (Figure 1B). Although beneficial for controlling MCMV, this Ly49H precursor might have potentially been autoreactive due to its specificity for MHC class I. Further selection events eliminated this self-reactivity, while retaining MCMV specificity (Figure 1B). Whereas Ly49H-expressing NK cells from wildtype mice respond robustly when they encounter MCMV-infected cells, ubiquitous transgenic or retroviral expression of m157 in mice renders the Ly49H-expressing NK cells unresponsive to Ly49H stimulation (Sun and Lanier, 2008b; Tripathy et al., 2008). This is caused, in part, by down-modulation of the Ly49H receptor on the NK cells encountering m157 during development, but also caused by hypo-responsiveness of receptor signaling. Whether or not chronic stimulation through Ly49H affects signaling through other activating receptors may depend on the amount of ligand expressed or by the anatomical location of the expressed ligand (Sun and Lanier, 2008b; Tripathy et al., 2008). In these systems, anergy is induced in NK cells transferred from non-m157 expressing hosts into the transgenic animal, indicating that anergy can be induced in mature NK cells (Tripathy et al., 2008). The m157 did not need to be expressed by the developing NK cells themselves as expression of m157 in trans is sufficient to impair Ly49H functions (Sun and Lanier, 2008b). Collectively, these studies demonstrate that NK cells, like T cells, achieve a state of tolerance when they are chronically exposed to endogenous, as well as foreign, ligands recognized by their activating receptors. Unlike T cells, which only express a single dominant activating receptor, chronic exposure to one activating ligand may tolerize NK responses only to the cognate receptor or alternatively to other activating receptor on the same cells, depending on the system. The mechanisms of this receptor-proximal tolerance and whole cell anergy are not yet defined.

Do NK Cells Undergo Selection of Inhibitory Receptor Repertoires?

The inhibitory KIR and Ly49 receptors recognize polymorphic variants on their MHC class I ligands. Within the total NK cell population, the inhibitory receptors for MHC class I are expressed in a variegated manner on overlapping subsets of NK cells. Moreover, the inhibitory KIR and Ly49 receptors themselves are highly polymorphic and have differing affinities for their MHC class I ligands (Carlyle et al., 2008; Hanke et al., 1999; Parham, 2008). These features, as well as the fact that the NK receptors and their MHC class I ligands are not genetically linked, result in a situation where some NK cells in the host do not express an inhibitory receptor for any self-MHC class I protein. Indeed, we now appreciate that in some individuals a significant proportion of NK cells fail to express inhibitory receptors reactive with the polymorphic MHC class I molecules that they have inherited (Fernandez et al., 2005; Yawata et al., 2008). Yet, these individuals do not suffer from autoimmune attack by their NK cells, raising the question of how NK cells lacking self-MHC class I inhibitory receptors achieve tolerance.

Some strains of rodents possess 20 or more Ly49 genes and some humans possess 15 KIR genes, providing for considerable receptor diversity within the NK cell population. Expression of these receptors appears stochastic and NK cells can express many KIR or Ly49 receptors that do not bind to self-MHC class I ligands in the host. Therefore, the presence of a MHC class I ligand is not required for NK cells to stably express a given Ly49 or KIR on a substantial subset of cells. In addition, NK cells expressing an activating KIR or Ly49 receptor that is able to bind MHC class I, in general, are not deleted or preferentially expanded in a host possessing a reactive MHC class I allele. Therefore, unlike in T cells, there is no evidence for either strong positive or negative selection of NK cells with respect to the activating KIR or Ly49 receptors.

Nonetheless, the expression of self-ligands for the inhibitory NK receptors for MHC class I does influence the development of NK cells. The frequency of NK cells expressing multiple inhibitory receptors can be approximately predicted by multiplying the frequency of NK cells expressing each inhibitory receptor (Raulet et al., 1997; Valiante et al., 1996). Although this “product rule” roughly holds for MHC class I-deficient mouse models, expression of two or more inhibitory receptors for self-MHC class I occurs less frequently than would be expected in MHC class I-sufficient mice. In fact, the frequency of NK cells expressing MHC class I-reactive inhibitory receptors is higher in MHC class I-deficient animals than in MHC class I-sufficient animals (Held et al., 1996; Salcedo et al., 1997). Thus, it appears that there is a selective or educational event that limits the expression of two or more inhibitory receptors for self-MHC class I on a given NK cell. The expression of MHC class I on both hematopoietic and non-hematopoietic cells influences the development of the inhibitory Ly49 repertoire (Roth et al., 2000; Sykes et al., 1993). Transgenic expression of an inhibitory Ly49 receptor limits the expression of endogenous Ly49 receptors in vivo, suggesting that engagement with self-MHC class I prevents subsequent expression of new inhibitory receptors during development (Held and Raulet, 1997). Therefore, there is an as yet undefined mechanism to ensure that NK cells do not express too many inhibitory receptors for self-MHC class I (Hanke et al., 2001; Held and Raulet, 1997; Raulet et al., 1997). Similarly, human NK cell interactions with self-HLA class I ligands shape the inhibitory KIR repertoire in an individual (Shilling et al., 2002; Yawata et al., 2006; Yawata et al., 2008).

NK Cell Education Mediated by Self-MHC Class I and Inhibitory Receptors

An early study reported that all human NK cell clones from two individuals expressed at least one inhibitory receptor that recognized a self-MHC class I protein, suggesting that a selective process shaped the NK cell repertoire to maintain tolerance (Valiante et al., 1997). However, as mentioned above, recent studies have detected a significant number of mature NK cells in humans and mice in vivo that do not express any inhibitory receptors for MHC class I or only express receptors that recognize non-self MHC class I (Fernandez et al., 2005; Yawata et al., 2008). Additionally, mice lacking surface expression of MHC class I due to genetic ablation of β2-microglobulin, TAP-1, and/or MHC class I heavy chains H-2K and H-2D, have normal numbers of mature NK cells and do not exhibit overt NK cell-mediated autoimmunity.

NK cells developing in the absence of MHC class I are unable to kill MHC class I-deficient tumor cell lines in vitro and fail to reject MHC class I-deficient bone marrow in vivo (Bix et al., 1991; Jonsson and Yokoyama, 2009; Liao et al., 1991). In MHC class I-sufficient humans or mice, the subset of NK cells lacking inhibitory receptors for self-MHC class I are hypo-responsive to in vitro stimulation through several activating receptors and fail to reject MHC class I-deficient bone marrow in vivo (Anfossi et al., 2006; Fernandez et al., 2005; Kim et al., 2005). Thus, expression of self-MHC class I-reactive inhibitory receptors enhances the responsive potential of NK cells (Anfossi et al., 2006; Fernandez et al., 2005; Kim et al., 2005) (Figure 2A).

Figure 2. Models of NK Cell Tolerance.

(A) Under homeostatic conditions interactions between activating natural killer (NK) receptors and their endogenous ligands render NK cells “disarmed” or anergized. Engagement of inhibitory receptors by MHC class I oppose this activation, resulting in “armed” or “licensed” NK cells. (B) Upon infection with pathogens such as MCMV, all NK cells become activated and previously anergized NK cells efficiently control infection. Responding NK cells expressing inhibitory receptors are inhibited by self-MHC class I, preventing optimal responses to infection. (C) Activated NK cells lacking MHC class I inhibitory receptors may kill uninfected bystander cells that express endogenous activating ligands or alternatively, inhibitory receptors for self-ligands other than MHC class I may prevent this auto-aggression.

The Basis for NK Cell Inhibition, Education, and Tolerance

NK cell education requires signaling via the inhibitory receptors because mutations within the ITIM of inhibitory receptors render NK cells hypo-responsive even in the presence of the cognate MHC class I ligand of the mutant receptor (Kim et al., 2005). Two mechanisms have been proposed to explain this tolerance of NK cells lacking self-MHC class I. The first hypothesis postulates that NK cells are initially unresponsive or “unlicensed” and that MHC class I engagement of inhibitory receptors during development licenses or “arms” these cells to become competent effector cells (Kim et al., 2005). Alternatively, the “disarming” hypothesis suggests that all NK cells are initially responsive, but if unopposed by MHC class I-specific inhibitory receptors chronic stimulation by normal cells renders NK cells anergic (Gasser and Raulet, 2006).

In mixed bone marrow chimeras generated with MHC class I-sufficient and β2m-deficient bone marrow all of the NK cells were hypo-responsive (Wu and Raulet, 1997). C57BL/6 mice expressing a H-2D^d transgene are able to reject C57BL/6 bone marrow and kill C57BL/6 lymphoblasts (Johansson et al., 1997). Although not formally proven, H-2D^d likely educates NK cells expressing an H-2D^d specific inhibitory receptor such as Ly49A or Ly49G2 that would otherwise not be educated by H-2^b. Presumably, NK cells in these transgenic mice that lack inhibitory receptors for H-2^b, but express inhibitory receptors for H-2D^d, are responsible for rejection of the H-2^b C57BL/6 graft. Interestingly, NK cells from a transgenic H-2D^d line that demonstrated mosaic expression of H-2D^d (that is H-2D^d was only expressed on 10 – 80% of hematopoietic cells) were unable to reject C57BL/6 bone marrow or kill C57BL/6 lymphoblasts (Johansson et al., 1997). Thus, NK cell education requires the educating MHC class I to be expressed on all cells or alternatively hypo-responsiveness is dominantly induced by lack of MHC class I, as would be predicted by the disarming hypothesis. The SHP-1 and SHIP-1 phosphatases associated with inhibitory receptor signaling are dispensable for NK education, suggesting that inhibitory signals are not needed for NK cell education, supporting the licensing model (Kim et al., 2005; Orr et al., 2010). However, these findings are not incompatible with the disarming model because many inhibitory receptors also associate with SHP-2, so there might be signaling redundancy between the different phosphatases recruited by the inhibitory receptors.

A third “rheostat” model proposes that NK cell reactivity is tuned by the number of self-MHC class I inhibitory receptors an NK cell expresses and by the affinity of each receptor for self-MHC class I (Brodin et al., 2009; Raulet and Vance, 2006). NK cells that express two or more inhibitory receptors for self-MHC class I respond more frequently and possess stronger effector functions than NK cells with only one inhibitory receptor for self-MHC class I (Brodin et al., 2009; Joncker et al., 2009). The affinity of the interaction between the inhibitory receptor and its MHC class I ligand also influences the NK cell education process (Jonsson et al., 2010). Thus, NK cell education is likely to be a quantitative process whereby NK cell responsive capacity is determined by the frequency and strength of engagement of inhibitory receptors with self-MHC class I, either opposing chronic activating receptor stimulation (disarming) or by transmitting undefined activating signals themselves (licensing or arming). Unlike T cells, there is no evidence for clonal deletion of potentially autoreactive NK cells either during development in the bone marrow or in the periphery. Rather, NK cells lacking self-MHC class I reactive inhibitory receptors are maintained in an unresponsive state – at least in the absence of inflammation or infection.

Inhibitory Receptors can Bind to MHC Class I on the Same Cell Surface

In mice, inhibitory receptors appear to be expressed at lower levels if cognate MHC class I is expressed. This has been interpreted as calibrating of the inhibitory receptors to a useful level (Sentman et al., 1995). Inhibitory Ly49 receptors can, however, bind to MHC class I expressed on the same cell surface (binding in cis) as well as binding to MHC class I on other cells (in trans) (Back et al., 2009; Doucey et al., 2004; Scarpellino et al., 2007). Disrupting cis interactions restores apparent levels of these inhibitory receptors to that detected on MHC class I-deficient NK cells, providing evidence against the calibration model, rather implying that cis interactions simply mask the antibody-binding site necessary to detect the inhibitory Ly49 receptor (Doucey et al., 2004). These cis interactions have been proposed to mediate NK cells licensing, because mutant inhibitory receptors incapable of cis interactions are unable to engender NK cell responsiveness (Chalifour et al., 2009). However, results from mixed wildtype and β2m-deficient bone marrow chimeras demonstrate that disarming of wildtype NK cells is dominant in the chimeras and is induced in trans, thus arguing against licensing in cis (Wu and Raulet, 1997). Cis interactions do prevent inhibitory receptor recruitment to the immunological synapse, thus reducing the inhibitory potential of these receptors (Andersson et al., 2007; Doucey et al., 2004).

Breaking NK Cell Anergy and Tolerance

Just as anergized or exhausted T cells can often regain functionality by ex vivo cytokine stimulation or by blocking the tolerizing signal (Barber et al., 2006; Teague et al., 2006), functionality of unlicensed or disarmed NK cells can be restored. This restoration of function can be accomplished in a number of ways, including in vitro culture with interleukin-2 (IL-2) or with IL-12 and IL-18, strong stimulation via activating receptors in vitro, or by infection with Listeria monocytogenes or MCMV in vivo (Fernandez et al., 2005; Johansson et al., 1997; Kim et al., 2005; Orr et al., 2010; Sun and Lanier, 2008a; Yokoyama and Kim, 2006). After MCMV infection, licensed and unlicensed NK cells, as defined by self-MHC class I inhibitory receptor expression, both express the early activation antigen CD69 and equivalently upregulate effector molecules such as interferon-γ (IFN-γ) and granzyme B (Orr et al., 2010). This early activation is driven by the inflammatory environment associated with infection and is independent of cognate activating receptor engagement. During MCMV infection NK cells in C57BL/6 mice undergo a proliferative burst, driven by the activating Ly49H receptor engaging its cognate ligand, the MCMV-encoded glycoprotein m157, which is expressed on the surface of infected cells (Dokun et al., 2001). Expression of inhibitory receptors for self-MHC class I restrain the proliferation of licensed NK cells, whereas in NK cells lacking self-MHC reactive inhibitory receptors proliferation proceeds unimpeded (Orr et al., 2010). This restraint of proliferation by inhibitory receptors is mediated by SHP-1 phosphatase signaling. NK cells are critical for controlling MCMV viral titers and pathogenesis, yet depletion of licensed NK cells does not negatively affect control of MCMV. Similarly, adoptive transfer of unlicensed, but not licensed NK cells, is sufficient to protect neonates against MCMV challenge (Orr et al., 2010). Thus, when NK cells need to directly interact with virus-infected cells to control the infection, inhibitory receptors for MHC class I impair, rather than license, NK cell functions (Figure 2B).

In mixed bone marrow chimeric mice containing β2m-deficient and wildtype hematopoietic cells, the NK cell tolerance demonstrated by the wildtype NK cells can also be broken during viral infection. In these chimeric mice, β2m-deficient hematopoietic cells, which were largely tolerated for weeks or months, are rapidly eliminated by NK cells after viral infection (Sun and Lanier, 2008a). Rejection of the β2m-deficient cells in these mixed chimeric mice does not require that they be infected with the virus. Similarly, when NK cells lacking expression of H-2D^d are removed from the transgenic mice displaying mosaic expression of H-2D^d and cultured in vitro in IL-2, tolerance to H-2D^d is lost and these NK cells are capable of attacking and killing target cells expressing H-2D^d (Johansson et al., 1997). NK cell anergy in response to chronic stimulation to m157 could be reversed if the Ly49H-expressing NK cells are removed from the m157 transgenic mice, indicating that anergy requires sustained exposure to the activating ligand (Tripathy et al., 2008). Collectively, these findings reveal that an NK cell tolerance or anergy is reversible and broken in the presence of inflammatory cytokines or infection.

NK Cell Inhibitory Receptors and Control of Human Leukemias

In human leukemia patients receiving irradiation therapy followed by bone marrow transplantation, donor NK cells can play a critical role in eliminating residual leukemic cells and preventing rejection of the donor bone marrow by killing recipient antigen-presenting cells that normally prime allogeneic T cell responses. In T cell-depleted, HLA-matched, but KIR-mismatched bone marrow transplants (in which not all donor-derived NK cells are inhibited by the recipient HLA or educated by donor or recipient HLA), the number of inhibitory KIR – HLA mismatches (that is, missing inhibitory ligands in the recipient) correlates with positive outcomes, including fewer leukemic relapses and improved graft acceptance (Clausen et al., 2007; Hsu et al., 2005; Sobecks et al., 2007; Symons et al., 2010). Donor-derived NK cells in the bone marrow recipient that are unlicensed (that is, do not express KIR reactive with donor or recipient HLA class I) ex vivo express high levels of IFN-γ and are competent to kill target cells, indicating that they are activated effector cells (Yu et al., 2009). Reciprocally, in HLA-mismatched, T cell-depleted bone marrow grafts, positive outcomes correlate with lack of HLA ligands in the recipient for inhibitory KIRs on donor-derived NK cells (Miller et al., 2007; Ruggeri et al., 2002). Thus, in the case of NK cell-mediated graft-versus-leukemia, expression of inhibitory KIR on donor-derived NK cells that are reactive with recipient HLA limited the effectiveness of donor NK cells in clearing leukemias and promoting graft acceptance (Yokoyama et al., 2010).

NK Cells Inhibitory Receptors in Immune Defense

In the absence of infection or inflammation, NK cells that possess inhibitory receptors for self-MHC class I are more responsive to stimulation than NK cells lacking self-MHC class I inhibitory receptors. Under what circumstances might these NK cells with self-MHC class I receptors provide benefit to the host? One scenario would be when host cells become transformed and downregulate MHC class I expression due to chromosomal instability that results in the loss of β2-microglobulin, TAP-1, or MHC class I heavy chain genes. The ability of healthy mice to reject transplanted cells from syngeneic β2m-deficient mice clearly demonstrates that as yet undefined ligands for activating NK receptors are present on non-transformed host cells, but NK cell attack is being prevented by the inhibitory receptors for self-MHC class I. Although not yet experimentally demonstrated, the NK cells that possess inhibitory receptors for self-MHC class I might provide for “missing-self” detection and elimination of nascent tumors.

Another scenario whereby NK cells that possess inhibitory receptors for self-MHC class I might be beneficial is in sparing MHC class I-expressing dendritic cells from attack by NK cells during an infection. In the absence of NK cells or in mice lacking the Ly49H receptor, a stronger T cell response is mounted against MCMV infection. This may be due to a robust early NK cell response that eliminates MCMV-infected dendritic cells, thus decreasing priming of MCMV-specific T cells (Andrews et al., 2010; Su et al., 2001). Although not tested in this study, it is possible that dendritic cells are preferentially killed by Ly49H⁺ NK cells lacking self-MHC class I inhibitory receptors. Thus, by restraining NK cell responses, inhibitory receptors might contribute to better overall control of infection by permitting more robust T cell priming.

What Prevents Autoimmunity Mediated by NK Cells?

In steady-state conditions, NK cells lacking self-MHC class I inhibitory receptors might simply be anergized by chronic stimulation through their unknown activating receptors engaging ligands on normal cells and tissues. However, both in vivo and in vitro these NK cells can be activated by cytokines or infection and mediate potent effector functions. What then might regulate their attack against normal, healthy cells or tissues? In addition to inhibitory receptors for MHC class I, NK cells express several other inhibitory receptors that recognize ligands not encoded by MHC genes (Kumar and McNerney, 2005) (Table 2). Many of these inhibitory receptors are expressed constitutively on most NK cells. For example, mouse 2B4 (CD244) recognizes the ubiquitously expressed CD48 glycoprotein, human NKR-P1A (CD161) recognizes the LLT1 ligand on activated B cells and dendritic cells, and human and mouse LAIR-1 (CD305) recognizes collagen. Other receptors such as KLRG1 that recognizes cadherins and gp49B1 that recognizes the α_vβ₃ integrin are upregulated on activated NK cells during viral infection (Robbins et al., 2004; Wang et al., 2000). The role of these inhibitory receptors in shaping NK cell tolerance and functionality is largely unexplored, particularly in regulating the subset of NK cells lacking self-MHC class I inhibitory receptors. Thus, these inhibitory receptors might prevent NK cell-driven autoimmunity. Alternatively, elimination of uninfected cells (bystander killing) by NK cells lacking MHC class I-reactive inhibitory receptors might be an evolutionarily acceptable trade-off for the enhanced control of infections (Figure 2C).

Table 2.

NK Cell Inhibitory Receptors

Gene	Common Name	Species	Ligand
Klra1, 3, 7, 9	Ly49A, C, G2, I	Mouse	Various H-2 class I
Klrb1b, d	NKR-P1B, D	Mouse	Ocil (Clr-b)
Lilrb4	gp49b1	Mouse	avβ3
Pilra	PILRα	Mouse	O-glycosylated CD99
SIGLEC-E		Mouse	Sialic acid
CD244	2B4	Mouse	CD48
KLRG1	Mafa	Mouse, Human	E-, N-, R- cadherins
KLRD1-KLRC1	CD94-NKG2A	Mouse, Human	Mouse Qa-1b, Human HLA-E
KLRB1	NKR-P1A, CD161	Human	LLT1 (CLEC2D)
LAIR1	LAIR-1, CD305	Human, Mouse	Collagen XVII
LILRB1	ILT2, LIR1, CD85j	Human	HLA class I
KIR2DL1–3, 5	CD158	Human	HLA-C
KIR3DL1, 2	CD158	Human	HLA-Bw4, some HLA-A
CEACAM1	CD66a	Human	CD66
SIGLEC7	CDw328	Human	Ganglioside GD3
SIGLEC9		Human	Sialic acid

Outstanding Questions

Although NK cells have been studied for more than 30 years and many of their receptors and signaling pathways have been identified, the molecular basis of NK cell tolerance remain undefined. Strikingly, we still do not know the physiological relevance of “missing self” surveillance in the host. Another major unsolved question is how NK cells lacking self-reactive MHC class I inhibitory receptors can functionally mature and acquire lytic effector molecules, yet remain tolerant of their environment. Finally, the mechanisms inducing tolerance upon chronic exposure to activating ligands are undefined. These and many other questions need to be answered before we can fully understand NK cell development, education, and tolerance.